Macroscopic Composition Gradient

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 51

Progress in Materials Science 57 (2012) 1010–1060

Contents lists available at SciVerse ScienceDirect

Progress in Materials Science


journal homepage: www.elsevier.com/locate/pmatsci

Development of ‘‘Macroscopic Composition Gradient


Method’’ and its application to the phase transformation
Toru Miyazaki ⇑
Nagoya Institute of Technology, Nagoya 466-8555, Japan

a r t i c l e i n f o a b s t r a c t

Article history: A new characterization method, ‘‘Macroscopic Composition


Received 30 June 2009 Gradient (MCG) Method’’ is proposed to investigate the phase
Received in revised form 3 May 2011 transformations near the phase boundaries. The MCG method is a
Accepted 10 November 2011
new technique to investigate the phase transformations in various
Available online 2 December 2011
composition alloys by utilizing a single specimen having the
macroscopic solute composition gradient. Since the macroscopic
composition gradient in the MCG alloy is so prepared as to cross
over the phase boundary, the morphological transition of critical
phenomena at the phase boundary can continuously be
investigated by means of analytical transmission electron
microscopy. By utilizing the MCG method, the various kinds of
phase transformation, such as the coherent and incoherent
precipitation boundaries, the order/disorder phase transition and
the morphological change at the spinodal line have successfully
been investigated. Furthermore, to an important thing, the critical
size of precipitate-nucleus and the nucleation rate near the
solubility limit can be experimentally obtained for respective
nucleus. The phase decomposition of supersaturated solid solution
progresses by a mechanism of spinodal decomposition even in
the N-G region of phase diagram. On the basis of experimen-
tal results, the application limit of the conventional nucleation
theory is investigated, and hence the failure of Boltzmann–Gibbs
free energy becomes obvious in the early stage of phase
decomposition.
It is noteworthy that the present experiment is systematically
conducted for the alloy composition range very close to the solubil-
ity limit. Such critical phenomena of phase transformation have

⇑ Address: 973-223, Minamiyama, Komenoki-cho, Nisshin 470-0111, Japan. Tel.: +81 561 73 7688; fax: +81 561 73 2242.
E-mail address: tmiyazaki@mub.biglobe.ne.jp

0079-6425/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.pmatsci.2011.11.002
T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060 1011

been scarcely examined in the past. The MCG method proposed


here is considered to open a new way to investigate the critical
phenomena in the phase boundary.
Ó 2011 Elsevier Ltd. All rights reserved.

Contents

1. The outline of macroscopic compositional gradient (MCG) method . . . . . . . . . . . . . . . . . . . . . . . . . . . 1011


1.1. General Idea of MCG method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1011
1.2. Preparation of MCG alloy specimen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1014
1.3. Experimental procedures and a typical example of TEM microstructure in MCG Alloys . . . . 1015
1.3.1. Experimental procedures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1015
1.3.2. A typical example of TEM microstructure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1015
2. Influences of macroscopic composition gradient (MCG) on the phase decomposition . . . . . . . . . . . . 1015
2.1. Profile change of MCG specimen during aging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1016
2.2. Influences of existence of MCG on phase transformation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1017
2.2.1. The macroscopic composition gradient energy, E grad macro . . . . . . . . . . . . . . . . . . . . . . . . . 1018
2.2.2. The gradient strain energy of microstructure, E stgrad str .......................... 1018
2.3. Computer simulations of microstructure formation based on phase field method. . . . . . . . . 1021
2.3.1. Theoretical basis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1021
2.3.2. Computer simulated microstructures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1022
2.4. Experimental verification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1023
3. Transition of microstructures at phase boundary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1023
3.1. Determination of coherent precipitation line of Ni–V and Ni–Mo alloy systems . . . . . . . . . . 1023
3.1.1. Ni–V alloy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1024
3.1.2. Ni–Mo alloy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1026
3.2. Difference of the solubility Limits between the coherent and incoherent precipitation . . . . 1026
3.3. Microstructure continuity between spinodal and N-G phase decompositions . . . . . . . . . . . . 1027
3.4. A2/B2 order/disorder transition in Fe–Al ordering alloy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1029
4. Precipitate-nucleation near the edge of miscibility gap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1030
4.1. Microstructure changes with aging in varies alloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1030
4.2. Evaluations of nucleus-stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1032
4.3. Calculation of the critical stable nucleus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1038
4.4. Thermodynamic discussion on the basis of conventional nucleation theory . . . . . . . . . . . . . 1042
5. Kinetic investigations on the nucleation process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1046
5.1. Experimental results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1046
5.2. Theoretical basis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1047
5.3. Kinetic investigation on the experimental results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1051
5.4. Pre-nucleation phenomena . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1053
5.5. High speed growth of big nucleus near the solubility limit . . . . . . . . . . . . . . . . . . . . . . . . . . . 1054
5.6. Problems of nucleus formation in the N-G region . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1055
6. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1056
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1057
Appendix A. Method of ‘‘composition vs. distance’’ curve in the MCG specimen . . . . . . . . . . . . . . 1057
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1059

1. The outline of macroscopic compositional gradient (MCG) method

1.1. General Idea of MCG method

The comprehensive description of phase transformations should be realized in a form of three


dimensional diagram consisting of the temperature (T), time (t) and composition (c) axes [1–4] as
shown in Fig. 1. The section parallel to the temperature (T) and the composition (c) axes is well known
1012 T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060

Nomenclature

cA composition of A atom
cB composition of B atom
ca local average composition
ce equilibrium solute concentration at the interface of particle
ce(r) equilibrium solute concentration at interface of particle with radius r
ce(1) equilibrium solute concentration at interface of particle with the infinite size
cp composition of precipitate
Egrad
macro macroscopic composition gradient energy (see Eq. (1))
Estgrad
str gradient elastic strain energy (see Eq. (12))
X volume of particle
eTij eigenstrain
hriIðHÞ internal stress in precipitate particle
r1 ij ðXÞ internal stress of a particle existing in the infinite matrix
V volume of alloy
Vm molar volume of precipitate
N number of particle
Eincl elastic strain energy of a single particle in infinite matrix (see Eq. (4))
Ehom
str elastic strain energy per unit volume of the composition flat alloy (see Eq. (5))
Gsystem total free energy of microstructure (see Eq. (13))
Gc chemical free energy
Esurf interfacial energy
Estr elastic strain energy
ci(r) composition parameter of ith component at position r
sj(r) structure parameter of jth component at position r
lc chemical potential
lsurf surface potential
lstr strain potential
v diffusion potential
M mobility of atoms
e
D inter-diffusion coefficient
j gradient energy coefficient
g expansion coefficient with respect to the solute concentration
Y elastic constant
r radius of precipitates
r minimum radius of a critical stable precipitate (radius of stable nucleus)
cs interfacial energy density between particle and matrix
R gas constant
T temperature
Qd activation energy for diffusion of solute atoms
DG(c) nucleation energy barrier that must be overcome to form nucleus
U frequency of nucleation
U1 time to nucleation (see Fig. 42 and Table 4)
Dc degree of super-saturation
a0 lattice parameter
DF m energy barrier due to the convex part of free energy–composition curve (see Fig. 46)

as the phase diagram, and the section parallel to the temperature (T) and the time (t) axes is the TTT
diagram. A section parallel to the composition (c) and time (t) axes is also important, nevertheless it
has not attracted attention yet. In the T–t–c diagram, the experiments whose variables are the T-axis
and t-axis are well known as the thermal analysis and the isothermal aging, respectively. Since the
T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060 1013

T-c diagram (Phase diagram)

Temperature,T
am am
) α1 + α 2
i a gr iagr
d
td T
T- -T-
(T

0 1.0
c-t diagram

e, t
tim
Composition,c

Fig. 1. A comprehensive description of phase transformation in a form of three dimensional diagram with temperature (T), time
(t) and composition (c) axes.

c-axis is also an important element in the T–t–c diagram, an experiment whose variable is composition
c should take more attention.
The other view point of phase transformation should be taken into consideration. It has widely
been accepted that the researchers pay attention to a typical phenomenon in the realistic complex
phenomena and try to linearize it in order to understand it simply. However, since the most of phase
transformations may include the non-linear part, the conventional treatment based on the linear
theory often gives insufficient results. The non-linearity is very important to understand the phase
transformation in the vicinity of phase boundary. Nevertheless, the critical phenomena have hardly
been examined on the materials science, particularly in the phase transformation phenomena. A char-
acterization method to investigate the critical phenomena in the vicinity of phase boundary should be
developed.
On the basis of two considerations above described, we proposed a new characterization method
of phase transformation, that is ‘‘Macroscopic Composition Gradient (MCG) Method’’ [4–6]. This is a
new characterization method of microstructure by utilizing a macroscopic composition gradient
introduced into alloys. In the MCG method the solute atom can be prepared so as to cross over
the phase boundary, so the various critical phenomena caused by the phase transformation near
the phase boundary are visually investigated by means of analytical transmission electron
microscopy.
In the conventional research, the alloy characteristic has been investigated for the individual alloy
whose composition was fixed. Inevitably there is no continuity in composition. Therefore, the change
of characteristics with composition has hardly been investigated systematically. However by using the
MCG method, we can continuously observe the change of phase transition with composition. We can
study the phase transformation based on a new viewpoint.
By utilizing the MCG method, various kinds of phase transformation such as the coherent and inco-
herent precipitations, the order/disorder phase transition, the spinodal/nucleation-growth phase
decomposition and the other many phase transformations can successfully be investigated, as pre-
sented in Section 3. Furthermore, to more important thing, the precipitate-nucleation very close to
solubility limit is experimentally investigated by utilizing the MCG method (Section 4). On the basis
of these important experimental results, the validity of conventional nucleation theories is discussed
statistically and kinetically (Sections 4 and 5) and hence the failure of Boltzmann–Gibbs extensive en-
tropy becomes clear for the early stage of phase decomposition.
These observations have succeeded for the first time by MCG method and therefore the experimen-
tal facts obtained here give us many new knowledge of phase transformation. The MCG method is very
useful for investigating various composition-dependent phenomena, especially the non-linear phe-
nomena in the vicinity of a phase boundary.
1014 T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060

1.2. Preparation of MCG alloy specimen

There are many preparation methods to create the macroscopic composition gradient in alloys, i.e.,
a diffusion coupling method, an imperfect arc melting method of sandwiched alloys and an imperfect
homogenization method of the coarse discontinuous precipitates and so on [1–6].
In the present investigation, the Ni–Si, Ni–Al, Cu–Co and Cu–Ti MCG binary alloy systems and Ni–
Al–Co ternary MCG alloy were investigated. At first, the preparation of Ni–Si MCG alloy was briefly
explained in Fig. 2. A Ni–14 at.%Si alloy, prepared by arc melting, was coupled with pure Ni. The sur-
faces of coupled alloy should be chemically polished to the mirror surface. The coupled alloy was
imperfectly homogenized at high temperature for a suitable duration or imperfectly arc-melted for
a short time. The heating time or arc-melting time was controlled so as to realize the MCG region over
several lm at least. The specimen was cut to thin plate of 0.5–1.0 mm in thickness, along the compo-
sitional graduation of solute atoms, as illustrated with a thin solid rectangular in Fig. 2. The specimen
must be collected from the region not including the Kirkendall interface.
Similarly, the Ni–15 at.%Al and Cu–20 at.%Co alloy were also coupled with pure Ni and Cu, respec-
tively, and then imperfectly arc-melted. The Cu alloy plates were annealed at 1373 K for 30 ks and the
Ni alloys were annealed at 1423 K for 40 ks to stabilize the macroscopic composition gradient and
then quenched into the iced brine. By this treatment the supersaturated solid solutions having MCG
were produced.
In the case of arc-melting or couple diffusion method, the crystallographic directions of coupled
two mother alloys are usually different, hence the high angle grain boundary may be formed in the
MCG portion. This boundary is usually formed at the initial interface (Kirkendall interface), where
the lattice defects and impurity atoms are concentrated. The MCG alloy specimen must be a single
crystal which does not contain the grain boundary, because the continuity of microstructure is neces-
sary in the MCG method. Thus, the MCG alloy should be extracted from the portion of no-grain bound-
ary in the front side or back side of Kirkendall interface, as illustrated in Fig. 2b. The best way is to use
two single crystals whose coupled faces are coherent in crystallography.
The macroscopic composition gradient for Cu–Ti alloy was prepared by utilizing the imperfect
homogenization of coarse discontinuous precipitates of Cu3Ti-stable phase. A Cu–4 at.%Ti alloy was
firstly prepared in a vacuum induction furnace, and then forged and rolled to a thin plate of about
1 mm in thickness, and solution treated at 1173 K for a suitable duration. After homogenization at
1173 K, the specimens were aged at 973 K for a long duration enough to produce the large
discontinuous Cu3Ti precipitates whose inter-lamella distance is over 5 lm. These specimens were
again heated at 1173 K for a short duration (10–15 min) to make imperfectly homogenized solid solu-
tion and then directly quenched into the aging temperatures. By this heat treatment the macroscopic
composition gradient of the solute atom (Ti) was formed between the lamellar Cu3Ti precipitates.

Fig. 2. Schematic illustrations of typical preparation method of Ni–Si MCG alloy. (a) Coupled two alloys having different
compositions. (b) MCG Ni–Si alloy produced by imperfect arc melting or imperfect annealing at high temperature. The specimen
is cut to thin plate of 0.5–1.0 mm in thickness, perpendicular to the equi-concentration direction. As shown by a thin solid
rectangular. The specimen must be collected from the region not including the Kirkendall interface.
T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060 1015

In the ternary alloy system, the MCG should be performed along the tie-line of phase decomposition.
The Ni–Cr–Al MCG alloy used was prepared by coupling Ni–24 at.%Cr with Ni–25 at.%Al alloys, so that Al
content in the MCG alloy varied along the tie line of c/c0 (see the phase diagram of Fig. 32 in Section 4.1).
The macroscopic composition gradient of these solid solutions should be controlled to be less than
10 at.%/l, because 10 at.%/l is the critical value of composition gradient which does not give a wrong
influence on the phase transformation. The value of 10 at.%/l is very large compared with the
usual macroscopic composition gradient prepared in the usual MCG alloys. The detail is described
in Section 2.2.

1.3. Experimental procedures and a typical example of TEM microstructure in MCG Alloys

1.3.1. Experimental procedures


The specimens were aged at various temperatures for suitable durations and then quenched into
iced brine. The aging should be undertaken at the lower temperatures than the stabilizing tempera-
ture at least 300–400 K, for instance, at 973 K for the Ni–Al and Ni–Si alloys and 873 K for the
Cu–Co and Cu–Ti alloys. The specimens were quenched into the iced brine after aging. The thin foils
for TEM were prepared by electro-polishing after aging in an electrolyte, H2SO4:CH3OH = 1:9 at 240 K
for the Ni–Al, Ni–Si and Ni–Al–Co alloys and HNO:CH3OH = 1:3 at 240 K for the Cu–Ti and Cu–Co
alloys. Since the solute concentration varies at the portion of MCG specimen, the condition of
electro-polishing may often encounter the difficulty. The careful control of polishing-voltage is partic-
ularly requested. The platinum mesh with a single hole method should be used for this case.
The microscopic observation was performed by the analytical transmission electron microscope
(JEOL 2000FX), and the solute concentration analysis by the energy dispersive X-ray spectroscopy
(EDS) (the detector: the Tracor Northern Company TN-5500) was concurrently performed at several
locations in the same thin foil. The electron microscope was operated at 200 kV. The LaB6 filament
was used at an accelerating voltage of 200 kV and the beryllium mesh is used. The K-factor defined
by the Cliff–Lorimer method [7] is estimated in the limit of a thin film specimen and determined
by the EDS measurement on the standard samples whose chemical compositions are already known.
The spurious X-ray was detected for correction of the characteristic X-ray Since the size of the incident
electron beam is operated so as to be large enough to cover the area containing several precipitates,
the measured values of solute concentration indicates a locally averaged composition ca. The measure-
ment was performed 5–7 times on the same place and then these were averaged The measurement
error of each averaged value was within ±0.1 at.%. The identical measurements were performed at
the many places in the thin foil. In the MCG method the chemical composition at any place can be esti-
mated from the ‘‘composition vs. distance curve’’ calculated by the least square method of the error
function for the many measured values. By utilizing this method, one order higher accuracy than that
of as-measured value can be obtained (see Appendix A).

1.3.2. A typical example of TEM microstructure


The precipitation behavior and the microstructure formation in the MCG alloys may be unfamiliar
for the most researchers. Therefore, the precipitation of Ni3Si particles in the Ni–Si MCG alloy is dem-
onstrated, as an example. Fig. 3 shows a 100 dark field TEM image of microstructure formed by aging
at 973 K for 7.2 ks. The white small cuboids in Fig. 3 show Ni3Si precipitates. The gray solid circles in
the photograph indicate the EDS measuring points of local average composition, whose values are cor-
respondingly plotted in the small figure of Fig. 3. The solid line in the small figure is determined by
using the least square method of the error function. A virtual line connecting two black arrows in
the photograph is named as ‘‘precipitation front’’ whose composition is given from the inserted figure.

2. Influences of macroscopic composition gradient (MCG) on the phase decomposition

The MCG method is a technique to investigate the phase transformations in various composition
alloys by utilizing a single specimen having the macroscopic solute composition gradient. Namely,
the macroscopic composition gradient method is assumed to be identical with that of ‘‘the plywood
of the thin plates of alloy’’ whose solute composition changes little by little. In order to materialize
1016 T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060

Fig. 3. A 100 dark field TEM image of the microstructure of Ni–Si MCG alloy aged at 973 K for 7.2 ks. The white cuboids are Ni3Si
precipitates. The gray solid circles in the photograph indicate the measuring points of locally average solute composition of
which values are correspondingly plotted in the small figure inserted. A virtual line connecting two black arrows in the
photograph is defined as a ‘‘precipitation front’’ whose composition is given by the inserted figure to be 10.92 at.%Si for this
aging condition.

the assumption, the following conditions are indispensable, i.e., the thin plates are thermodynamically
independent with each other and are never affected from other surrounding plates during phase trans-
formation. Therefore, it is important to make clear the threshold value of composition gradient for
safety use of MCG method.
In the present section, we make clear the following two points in order to judge the validity of MCG
method: the first point is to make clear that the profile of macroscopic composition gradient does not
change during aging at lower temperature. The second point is to make clear the influence of MCG it-
self on the phase transformation, because the existence of MCG may give some influences on the
microstructure formation even if the profile of MCG does not change during aging. Therefore, it is
important to know the critical lower limit of MCG which does not give any influence on the micro-
structure formation.

2.1. Profile change of MCG specimen during aging

In order to avoid the profile change of MCG during aging, the aging temperature, TA, must be fairly
lower than the stabilizing temperature, TS. The temperature difference between the stabilizing and
T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060 1017

Fig. 4. The composition profile of Mo atoms calculated for a Ni–Mo MCG alloy. A solid curve is for as-stabilized specimen and a
dotted line is for the aged specimen after stabilization. The both line are perfectly consistent.

Fig. 5. The macroscopic composition profile measured by XPMA for Fe–Si MCG alloy, where solid circles and squires indicate Si-
concentration before and after a long time aging (1023 K for 2.42Ms), respectively. Any significant difference is not recognized
between the specimens before and after aging.

aging temperatures should be over 400 K. The diffusion coefficient of solute atoms varies approxi-
mately 10 times per each 50 K in the alloy system whose activation energy of diffusion is about
250 kJ/mol. Therefore, the temperature difference, TS  TA, of 400 K results in about 108 times-differ-
ence between both diffusion coefficients. Such big difference between the diffusion coefficients is con-
sidered to give hardly change in MCG profile even if the MCG specimen is aged for a long duration.
Fig. 4 shows theoretically calculated MCG profile of Mo atoms in two Ni alloys; one is the alloy just
stabilized at 1373 K for 7.2 ks and the other is that of the alloy aged at 923 K for 864 ks after the sta-
bilization. These lines are theoretically calculated on the basis of the Fick’s 2nd law of diffusion equa-
tion. The both lines are perfectly coincident with each other and any difference cannot be recognized
between them. The same behavior is also experimentally confirmed. Fig. 5 shows the macroscopic
composition profile measured by XPMA for Fe–Si MCG alloy, where solid round circles and squares
indicate Si-contents before and after aging for a long time, respectively. Any difference is also not
recognized between them, as is theoretically predicted. Thus, it is concluded that the MCG profile does
not change during aging, so long as the MCG profile is stabilized at a higher temperature of 350–400 K
than the aging temperature.

2.2. Influences of existence of MCG on phase transformation

Here, the direct influence of MCG on the phase transformation is discussed. Even if the macroscopic
composition profile does not change during aging, the existence itself of MCG may give some influences
1018 T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060

on the microstructure formation. Therefore, it is important to know the critical limit of MCG which does
not give any influence on the microstructure formation. In this section, the threshold value is evaluated.
The MCG alloy aged has a gradient of microstructure consisting of many precipitate particles,
namely the volume fraction of precipitates inclines. Therefore, the gradient energy of the microstruc-
ture is considered to consist of the following two gradient energies, i.e., the first is a macroscopic com-
position gradient energy, Egrad
macro , caused by the macroscopic composition gradient itself (hereafter,
called as ‘‘macroscopic compositional gradient energy’’ Egrad
macro ) and the other is an additional elastic
strain energy arising from the gradient of volume fraction of precipitates (hereafter, called as ‘‘micro-
structure gradient strain energy’’ Estgrad
str ).

2.2.1. The macroscopic composition gradient energy, Egrad


macro
According to Cahn and Hilliard [8,9], the chemical energy change due to the composition fluctua-
tion has been given by the Taylar expansion with respect to composition c. Similarly in the MCG alloy
the macroscopic gradient energy Egrad
macro is expressed by
Z  2
dc
Egrad
macro ¼ j dx ð1Þ
dx macro
where (dc/dx)macro is the macroscopic composition gradient in the alloy and j is the gradient energy
coefficient. The typical value of (dc/dx)macro for the macroscopic composition gradient is the order of
105/nm since the composition change is roughly 0.5 at.% per 1 lm distance, as is known from
Fig. 3. On the other hand, a microscopic composition gradient energy Egrad micro arises from the interface
of precipitate-particle newly produced by aging. The (dc/dx)micro at the precipitate interface is consid-
ered to be order of 101/nm because the composition change of several tens of percents occurs at the
precipitate interface whose width is about 1 nm. Therefore, the value of ðdc=dxÞ2macro due to the
macroscopic composition gradient is extremely small, compared with that of precipitate interface,
i.e., approximately, (dc/dx)macro/(dc/dx)micro = 109 [6]. Thus, the effect of (dc/dx)macro is so small that
it can be disregarded.

2.2.2. The gradient strain energy of microstructure, Estgrad


str
Here, we evaluate the elastic strain energy of microstructure Estgrad
str caused by the gradient of vol-
ume fraction of precipitates. In Fig. 6a the particles are distributed uniformly in the matrix, whereas
in Fig. 6b the distribution of particles varies with change of the macroscopic composition. The bold
rectangulars in Fig. 6a and b show the region where the volume fraction of precipitates are equal.
The total elastic strain energies in the rectangular are evaluated here for the two cases. The elastic
strain energy in (b) is possibly different from that of (a), because the additional elastic interaction
energy among particles may be generated from the gradient of volume fraction of precipitates. Thus,
the additional elastic strain gradient energy Estgrad
str may occur in the case of (b).
The Estgrad
str is evaluated here. Firstly we consider the case of composition flat alloy in which many
particles distribute uniformly. Each particle volume is X and the eigenstrain is eTij . The particles are
assumed to be ellipsoidal revolution in shape and same in size. The volume fraction of particle is f.
The mean internal stress of a particle in an infinite matrix is given by [10,11]
hrij iIðHÞ ¼ r1 1
ij ðXÞ  f rij ðXÞ ð2Þ

Eq. (2) means that hriIðHÞ is sum of the internal stress of precipitate particle existing in an infinite
matrix, r1ij ðXÞ, with the mean internal stress affected by the surrounding particles. Therefore, the elas-
tic strain energy per unit volume of alloy Ehom
str is expressed by
Z
1 1 NX 1
Ehom
str ¼  rij eTij dx ¼  ð1  f Þr1 T 1 T
ij ðXÞeij ¼  f ð1  f Þrij ðXÞeij ð3Þ
2V H 2 V 2
where N is number of precipitate particles, which is related with volume fraction f as f = NX/V. V is a
volume of alloy considered. The elastic strain energy of single particle existing in an infinite matrix
Eincl is given by [10,11]
T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060 1019

Fig. 6. Schematic illustrations of the microstructure of (a) macroscopically composition flat alloy and (b) composition gradient
alloy (MCG alloy). The solid circles represent the coherent precipitates. The bold rectangles in the center indicate the region
having the same volume fraction of precipitates in the both alloys. The stability of microstructure is compared for this region.

1 1
Eincl ¼  r ðXÞeTij ð4Þ
2 ij
Hence Ehom
str of Eq. (3) is simplified as shown in the following equation:

Ehom
str ¼ f ð1  f ÞEincl ð5Þ
Next, the case of MCG alloy shown in Fig. 6b is discussed. In this case, an additional strain energy
term caused by non-uniform distribution of particles should be added to the second term of Eq. (2).
The heterogeneity of volume fraction is obtained by utilizing the Taylar expansion of f as a function
of distance r.
1 2 2
f inhom ¼ f þ r r f ð6Þ
2!
By using Eq. (6), the elastic strain energy per unit volume in the rectangular region of Fig. 6b Einhom
is given by
 
1NX 1 1 2
Einhom
str ¼ 1  f  r 2 r2 f r1 T 2
ij ðXÞeij ¼ f ð1  f ÞEincl  r fEincl  r f ð7Þ
2V 2 2
The Taylar expansion of Einhom
str with respect to the volume fraction f and its derivative terms (df/dx),
(d2f/dx2), gives
2
Einhom
str ðf ; rf ; r2 f ; . . .Þ ¼ Ehom 2
str ðf Þ þ e1 r f þ e2 ðrf Þ þ    ð8Þ
The terms over third power are omitted here, and the term of 5f is disregard because of symmetry
of energy. Here, the integral sign is omitted for simplifying the expression of formula. By applying
Gaussian divergent theorem to the second term in the right side of Eq. (8), Eq. (8) is transferred to
2
Einhom
str ðf ; rf ; r2 f ; . . .Þ ¼ Ehom
str ðf Þ þ eðrf Þ ð9Þ
where e ¼ de1 =df þ e2 .
Comparing Eq. (7) with Eq. (8), the following equations are known; e1 ¼  12 r2 Eincl f and e2 = 0.
Therefore, e is given by
1
e ¼ r2 Eincl ð10Þ
2
1020 T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060

Defined to be Estgrad
str ¼ Einhom
str  Ehom
str and compared Eq. (9) with Eq. (10), the gradient elastic strain
stgrad
energy Estr is consequently given by

 2
1 2 1 df 1
Estgrad
str ¼ r ðrf Þ2 Eincl ¼ r 2 Eincl ¼ ðdf Þ2 Eincl ð11Þ
2 2 r 2

where df is a variation of the volume fraction with respect to distance r.


The gradient elastic strain energy is numerically evaluated on the basis of Eq. (11). The macroscopic
composition gradient in the actual MCG specimen is approximately 0.5 at.%/lm [1,6], and the compo-
sition difference between precipitate and matrix is about 10 at.% [4]. Using these values and the fol-
lowing equation: @x@c
¼ @c  @f ¼ 0:5 at:%=lm and @c
@f @x
@f @f
¼ 10 1at:%, we obtain @x ¼ 0:5  103 =lm.
Therefore, taking the interparticle
. distance to be 5 nm, we get df ¼ 2:5  104 .
Thus, the ratio of Estgrad
str Ehom
str is evaluated for f = 0.5 as follows:

. 1
 104 Þ2 Eincl
ð2:5
Estgrad
str Ehom
str ¼
2
¼ 1:25  107 ð12Þ
ð0:5  ð1  0:5ÞÞEincl
In the case of f = 0.5, Estgrad
str is very small compared with the elastic strain energy Ehom str for composi-
tion flat alloy. However, in the alloy whose composition is close to the edge of miscibility gap (f is
small), Eq. (12) has a fairly large value, for example 4.0  105 for f = 0.01%. Therefore, Estgrad str is not
all ways negligibly small but may be effective for the alloy having a small f.
Using Eq. (12), we estimate roughly the critical limit of the macroscopic composition gradient
which does not give any influence on the microstructure formation. We assume that, when the
Estgrad
str is less than 1/1000 of the elastic interaction energy Ehom stgrad
str , Estr does not affect the microstructure
formation. The elastic interaction energy is about 5% of Ehom str , the necessarily condition for safety use of
MCG alloys is approximately given by

1
Estrgrad
str ¼ Ehom ð120 Þ
20000 str
After repeating a trial and error, we determined the critical limit of macroscopic composition gra-
dient approximately to be 10 at.%/lm, detail of which is described in the next section. We call this va-
lue the allowable maximum composition gradient MCGmax.
It is noteworthy that, since the Fe–Mo alloy is a very strong elastically constrained system, the
MCGmax value evaluated for Fe–Mo is available for the most alloy systems. As described above, the
heterogeneity of microstructure is influenced by the gradient elastic strain energy Estgrad str , which is
determined with the elastic strain energy Eincl (see Eq. (11)). Eincl is approximately determined
 2  2 by
the square of expansion coefficient with respect to the solute concentration g2 ¼ 1a @a @c
(a is lattice
constant) of alloy system. Thus, the alloys having larger g tend to form a strong heterogeneous micro-
structure. Table 1 lists up g-values for several alloy systems [12–16] where the phase decomposition
takes place. An Au–Ni alloy system has so extraordinarily large g-value which leads to breakdown the
lattice coherence at the particle interface on the mid-way of phase decomposition [34]. The Fe–Mo
alloy has a very large g except Au–Ni alloy. Therefore, the MCGmax estimated for Fe–Mo alloy is valid
enough to be effective for the most alloy systems. Furthermore, it should be noted that the
macroscopic composition gradient of 10 at.%/lm is fairly large than the gradient of the typical MCG
alloys used in the present work (about 0.5 at.%/lm (see Fig. 3)). On the basis of these simulations it
is satisfactory to consider 10 at.%/lm as the MCGmax.

Table 1
g-Value of various alloy systems.
Alloy Cu–Co [12] Al–Zn [12] Ni–Al [13] Ni–Si [13] Fe–Mo [16] Au–Ni [14]
g 0.019 0.026 0.043 0.057 0.083 0.15
1@a
g¼ a @c . Brackets [] show the reference number.
T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060 1021

2.3. Computer simulations of microstructure formation based on phase field method

On the basis of consideration of Section 2.2, the time-development of microstructure in the area of
bold rectangle of Fig. 6 is simulated on the basis of phase field method.

2.3.1. Theoretical basis


The theoretical basis of the phase field method has already been described in our papers [17,18]. In
the phase field method, the total free energy of microstructure Gsystem is expressed in terms of various
order parameters prescribing the microstructure, i.e., composition, degree of order, crystal structure
and so on
Z
Gsystem ¼ ½Gc fci ðrÞ; sj ðrÞ; Tg þ Esurf fci ðrÞ; sj ðrÞ; Tg þ Estr fci ðrÞ; sj ðrÞ; Tgdr ð13Þ
r

where ci and sj are a composition parameter and a structure parameter, respectively. Gc is the chemical
free energy, Esurf is the interfacial energy and Estr is the elastic strain energy. The gradient strain energy
Estgrad
str given by Eq. (11) is included in the elastic strain energy Estr.
The time-dependencies of c and s are evaluated by using following two kinetic Eqs. (14a) and (14b)
in the phase field method. Eq. (14a) is available for the conservative parameter such as a solute
composition, while Eq. (14b) is available for the non-conservative order parameter such as a long-
range order parameter. Eq. (14a) is usually called as Cahn–Hilliard equation [8] and Eq. (14b) is called
as Allen–Cahn equation [19].
  
@ci ðr; tÞ dGsystem
@sj ðr; tÞ
¼ r M ci fci ðr; tÞ; T g r þ nci ðr; T; tÞ þ K c ci ðr; tÞ; sj ðr; tÞ; T ð14aÞ
@t dci ðr; tÞ @t
 
@sj ðr; tÞ
dGsystem
@ci ðr; tÞ
¼ Lsj sj ðr; tÞ; T þ nsj ðr; T; tÞ þ K s ci ðr; tÞ; sj ðr; tÞ; T ð14bÞ
@t dsj ðr; tÞ @t
The ci(r, t) and sj(r, t) are functions of position r and time t in three dimensional. The interaction be-
tween two order parameters ci(r, t) and sj(r, t) proceeds through the total free energy of microstructure
Gsystem shown in Eq. (13). Mci{ci(r, t), T} and LSj{sj(r, t), T} are the mobility of order parameters, ci(r, t) and
sj(r, t), respectively, and are assumed to be a function of temperature T. v(r, t) is the diffusion potential
and n(r, T, t) is a so-called Gaussian noise term for the order parameters ci and sj. The final terms in Eqs.
(14a) and (14b) are a dynamic coupling terms in the phase field, i.e., a dynamic feed back term. The
coupling coefficients Kc and Ks in Eqs. (14a) and (14b) are usually zero for most phase transformations
not so much deviating from the equilibrium state. The phase transformation where the dynamic cou-
pling term cannot be ignored is only for a few phenomena highly deviated such as dendrite growth in
solidification and fractal pattern formation.
The precise evaluation of diffusion potential v(r, t) is most essential and important for the present
simulation. Since Gsystem is the total free energy of microstructure consisting of the chemical free
energy Gc, the interfacial energy Esurf and the elastic strain energy Estr, the each potential is given by
the following equations, respectively.
     
dGc dEsurf dEstr
lc fcðr; tÞg  ; lsurf ðr; tÞ  ; lstr ðr; tÞ  ð15Þ
dc r¼r0 dc r¼r0 dc r¼r0
Thus, the diffusion potentials v(r, t) are given by

dGsystem c cp c
vcp ðr; tÞ  ¼ lcp ðr; tÞ þ lsurf ðr; tÞ þ lstrp ðr; tÞ
dcp ðr; tÞ
ð16a; 16bÞ
dG
vsq ðr; tÞ  system ¼ lscq ðr; tÞ þ lssurf
q s
ðr; tÞ þ lstrq ðr; tÞ
dsq ðr; tÞ
Consequently, the diffusion potential v(r, t) at position r’ in microstructure is obtained from Eqs. (16a)
and (16b), and then @c/@t and @s/@t are evaluated from Eqs. (14a) and (14b). Thus, we can calculate
changes in c and s with progress of phase transformation by repeating Eq. (17a,17b).
1022 T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060

Table 2
The numerical values used for the calculation of microstructure formation in a MCG alloy. The values
are corresponding to the Fe–Mo binary alloy system.

Temperature, T/K 773


Alloy composition, c0 0.3
Composition gradient, Dc/lm 0, 0.1, 0.5
Elastic stiffness, Cij/104 MN m2
C Fe Mo
11 C 11
23.3, 46.3
C Fe Mo
12 C 12
13.5, 16.1
C Fe Mo
44 C 44
11.8, 10.9
Expansion coefficient with respect to the solute concentration, g 0.083
Calculation area, L/109 m 120
Interaction distance, d1/1010 m 2.86
Number of Fourier wave, N 256  256

cp ðr; t þ DtÞ ¼ cp ðr; tÞ þ f@cp ðr; tÞ=@tgDt


ð17a; 17bÞ
sq ðr; t þ DtÞ ¼ sq ðr; tÞ þ f@sq ðr; tÞ=@tgDt

2.3.2. Computer simulated microstructures


On the basis of Section 2.2.1, the time-development of microstructures for MCG alloy are demon-
strated. The numerical values used for calculation are listed in Table 2, each of which corresponds to
the thermodynamic data of Fe–Mo alloy system [16,18].
Fig. 7 shows time-developments of microstructures calculated for the flat composition alloy A(a–c)
and two MCG alloys B(d–f) and C(g–i) of which macroscopic composition gradient are 10 at.%/lm. and

Fig. 7. The time development of microstructure formation of Fe–30 at.%Mo aged at 773 K, calculated on the basis of the phase
field method. The microstructures A (a–c) are composition flat alloy. The B (d–f) and C (g–i) are for Fe–30 at.%Mo MCG alloy
whose composition gradient are 10 at.%Mo/lm and 50 at.%Mo/lm, respectively. The small arrows in (d) and (g) indicate the
solute concentration at the both ends of microstructure of the MCG alloys.
T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060 1023

50 at.%/lm, respectively. The calculating area corresponds to the bold rectangular area of Fig. 6a and b.
The macroscopic composition gradient decreases linearly from the right hand site to the left side, and
the Mo contents are just same at the centre of rectangular box of A, B and C alloys in Fig. 6. The initial
composition fluctuations were given by the same Fourier waves generated by the same seeds of the
random function. These are added to the macro composition profile. Therefore, the difference of initial
composition profiles between the composition flat alloy and MCG alloy is only for the macroscopic
composition profile.
It is obvious from Fig. 7 that the microstructures do not show any difference between the alloy A
and alloy B through the all aging time. Hence, it is clearly known that the macro composition gradient
of 10 at.%/lm does not give any influence. Thus, It approved that the macroscopic composition
gradient less than 10 at.%/lm does not give any influence on microstructure formation. When the
MCG becomes larger, the influence appears clearly. The alloy C in Fig. 7 demonstrates the simulated
microstructure of the alloy having a steeper composition slope, i.e. 50 at.%/lm. In early stages of aging,
the remarkable difference is not recognized between the both alloys, but in the later stage the micro-
structure of MCG alloy C becomes different from that of composition flat alloy, as is clearly known by
comparing the long aging time (1600s’). The precipitate particles in Fig. 7i are extended to the [0 1 0]
direction. This means that the atom diffusion along [1 0 0] direction becomes predominant than [0 1 0]
direction by existence of macroscopic composition gradient.

2.4. Experimental verification

Here, we perform experimental verification on the basis of the theoretical evaluations. The Fe–15–
20 at.%Mo alloys are well known to decompose spinodally so as to make h1 0 0i modulated structure
[18].
Fig. 8 shows changes in wavelength of modulated structure with progress of aging, experimentally
obtained for the compositionally flat Fe–19.2 at.%Mo alloy and the Fe–Mo MCG alloy whose macro
composition gradient is 8.9 at.%/lm which is less than the MCGmax 10 at.%/lm. The wavelength in
the MCG alloy was detected from the composition area of 19.2 at.%Mo. The wavelengths of composi-
tionally gradient alloy coincide well with that of flat alloy for all aging time. Thus, it is experimentally
proved that the macro composition gradient does not affect the microstructure formation so long as
the gradient is less than 10 at.%/lm. The wavelengths of MCG alloy are well consistent with that of
composition flat alloy for all aging time. Thus, it is experimentally proved that the macro composition
gradient does not give any influence on the microstructure formation so long as the gradient is less
than 10 at.%/lm.
Consequently, on the basis of theoretical and experimental investigations it is confirmed that the
MCG method does not give any influences on the phase transformation and microstructure formation
unless the MCG is over the critical composition gradient 10 at.%/lm.

3. Transition of microstructures at phase boundary

The MCG method is very useful to investigate the composition-dependence of phase transforma-
tion, particularly phase transformations in the vicinity of phase boundary, because of the drastic
change of chemical free energy at the edge of miscibility gap, compared with other energies such as
elastic strain energy and interfacial energy. In the present section the several representative micro-
structure transformations which demonstrate superiority of MCG method are demonstrated as fol-
lows; the precipitation limit of coherent and incoherent precipitates, the order/disorder phase
transition, the morphological transition of the microstructure from the spinodal to the N-G type phase
decomposition and so on.

3.1. Determination of coherent precipitation line of Ni–V and Ni–Mo alloy systems

It is examined for Ni–V and Ni–Mo MCG alloys whether the coherent solubility limit experimen-
tally obtained by the MCG method is precisely consistent with the conventional phase diagram or
1024 T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060

Fig. 8. Change in wavelength of the modulated structure with aging time for the Fe–19.2 at.%Mo composition flat alloy (usual
composition-fixed alloy) and Fe–Mo MCG alloy. The average composition (rectangular region of Fig. 5b) in MCG alloy is
19.2 at.%Mo and the average composition gradient of (oc/ox)MCG is 8.9 at.%Mo/lm. The wavelength of modulated structure is
examined by means of the satellite of 200 electron reflection spot.

Fig. 9. A TEM microstructure formed by aging at 873 K for 1.38 Ms. in the Ni–V MCG alloys, indicating that the solubility limit
for coherent Ni3V precipitates is 15.5 at.%V.

not. A Ni–34 at.%V and a Ni–20 at.%Mo alloys were coupled with pure Ni-plate, respectively and then
the coupled specimens were imperfectly arc-melted for a short time. After that, the alloy was heated
at 1373 K for 3.6 ks for stabilization of the macroscopic composition gradient and then quenched into
the iced brine. By this treatment the macroscopic composition gradient were formed in the Ni–V and
Ni–Mo supersaturated solid solutions. The alloys were isothermally aged at several temperatures of
873–1173 K for very long duration so as to reach the equilibrium solubility limit.

3.1.1. Ni–V alloy


Fig. 9 shows a TEM bright field image of Ni–V MCG alloy aged at 873 K for a very long time, 1.38 Ms.
It is clear that the strain contrasts due to the Ni3V coherent precipitates are uniformly observed in the
high V-concentration area, while in the lower concentration area the precipitates cannot be seen ex-
cept the dislocations. The microstructure is obviously separated into two parts by the virtual line con-
necting with two white arrows in the micrograph. The precipitation front is determined from the
inserted figure to be 15.5 at.%Mo for 1.38 Ms aging at 873 K. The similar investigations are performed
T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060 1025

Fig. 10. A coherent solubility limit for the Ni3V precipitation, which is firstly proposed by utilizing the MCG method. Our
experimental results are consistent precisely with Moreen’s coherent precipitation line.

Fig. 11. A microstructure of TEM dark field image of Ni–Mo MCG alloy aged at 923 K for 864 ks shows that the precipitation
front is 13.0 at.%Mo.

at other temperatures. The solubility limits experimentally determined are plotted in Fig. 10 for sev-
eral aging temperatures. Our experimental results are well consistent with the reliable coherent pre-
cipitation line which was obtained by the work of Moreen et al. [20]. The Pearson’s result [21] is
contradictory to the coherent precipitation line, possibly because of inaccurate analysis.
It is noteworthy that the precipitation front shifts gradually to the lower concentration region with
longer time aging and finally stops at the equilibrium phase boundary. (The experimental data of this
phenomenon is described in detail and discussed in Section 4.) Therefore, in order to investigate the
equilibrium state, we should adopt a long aging time enough to realize the equilibrium solubility limit.
1026 T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060

Fig. 12. A part of phase diagram of Ni–Mo system, showing the coherent solubility limit of Ni4Mo precipitates obtained by
several researchers, and our experimental results of Ni–Mo MCG alloy.

3.1.2. Ni–Mo alloy


Fig. 11 shows a dark field TEM image of Ni–Mo MCG alloy aged at 923 K for a long time aging,
864 ks. The Ni4Mo particles are observed uniformly in the Mo-high composition area, but the number
of precipitates decreases gradually in the lower composition area and finally no precipitates can be
seen. The composition of precipitation front is determined to be 13.0 at.%Mo. The similar experiments
were performed at the other aging temperatures (923–1123 K). The solubility limits experimentally
determined are shown in Fig. 12. The incoherent solubility lines proposed by Heiwegen and Rieck
[22] and by Gust et al. [23] and Casselton and Hume-Rothery [24] are also plotted, which were inves-
tigated by means of the diffusion-coupling [22] and by the X-ray measurement of lattice change with
composition [23,24]. These experimental results are adopted by Massalski [25] into the Ni–Mo equi-
librium phase diagram. Our coherent solubility line is slightly shifted to low temperature compared
with the incoherent solubility lines of Heiwegen’ work and Gust’ work, because the elastic strain en-
ergy due to the coherency forces down the miscibility gap to the lower temperature.

3.2. Difference of the solubility Limits between the coherent and incoherent precipitation

Fig. 13 shows an incoherent precipitate heterogeneously nucleated on the dislocation in the Cu–Ti
alloy, which is indicated by a large arrow. This photograph clearly shows the difference in solubility
limit between the coherent and incoherent precipitates. The solubility limit of incoherent precipita-
tion is given by the virtual line connecting with two paired arrows which are numbered as No. 2, while
the coherent precipitate is indicated by the two arrows numbered as No. 1. From the virtual lines of 1
and 2, the coherent and incoherent precipitation limits are determined to be 1.75 at.%Ti and
2.15 at.%Ti, respectively, at 873 K. Such solubility-difference is also observed in other MCG alloy sys-
tem. Fig. 14 shows the homogeneous coherent Ni3Al precipitates and the heterogeneous incoherent
Ni3Al formed on a dislocation. The composition difference between both is 0.4 at.%Al from the inserted
figure.
It has been generally conceptualized that the elastic strain energy forces down the miscibility gap
to the lower temperature [26–28]. Figs. 13 and 14 show the clear evidence to verify the concept, and
may give the quantitative evaluation of the elastic strain energy.
T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060 1027

Fig. 13. Coherent Cu4Ti precipitates and incoherent precipitates heterogeneously nucleated on a dislocation. The each
precipitation front is indicated by the arrow numbered 1 (2.15 at.%Ti) and 2 (1.75 at.%Ti), respectively. The incoherent
precipitate is isolated apart from the coherent precipitation front.

Fig. 14. Homogeneously dispersed coherent Ni3Al precipitates and predominant precipitates on the dislocation line.

3.3. Microstructure continuity between spinodal and N-G phase decompositions

On the basis of Boltzmann–Gibbs’s free energy [29], the process of phase decomposition has been
considered to be divided into the two types of phase decomposition, i.e., spinodal decomposition and
the Nucleation-Growth type decomposition, The theoretical research on the spinodal decomposition
started from the linear theory of Cahn–Hilliard [8,9], and has resulted in the dynamical investigations
based on the nonlinear diffusion equation. In the linear theory the spinodal and the Nucleation-
Growth are separate type phase decomposition. However, by recent theoretical considerations, it is
predicted that the both are continuously shifted gradually [30]. There have been a lot of experimental
investigation of spinodal decomposition for many alloys, i.e., Fe–W [31], Fe–Cr [32], Fe–Mo [16],
Nb–Zr [33], Au–Pt [34], Cu–Ni–Fe [35], Cu–Ni–Si [36] and so on. Nevertheless the empirical proof of
the continuity of microstructure has not be seen in a TEM photograph, so long as author know. The
MCG alloy is suitable to verify the continuity of microstructure.
By utilizing the Fe–Mo MCG alloy the morphological transition from the spinodal region to N-G re-
gion is investigated. The Fe–Mo alloy system is well known to form the typical h1 0 0i modulated struc-
ture by the spinodal decomposition [16]. Fig. 15 is the phase diagram of Fe–Mo alloy system, showing
1028 T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060

Fig. 15. A Equilibrium phase diagram of the Fe–Mo binary system. The dotted and chain lines are meta-stable coherent bimodal
and spinodal lines of bcc A2 phase, respectively.

Fig. 16. A TEM microstructure and the electron reflection satellites around 200 spot for Fe–Mo MCG alloy aged at 823 K for
10.8 ks, showing the continuous transition of microstructure from the h1 0 0i modulated structure to the mottled structure
which is considered to arise from a so-called Nucleation-Growth phase decomposition. The two arrows indicate the area of
spinodal line (19.0 at.%Mo) given by the Fe–Mo phase diagram [16].

the equilibrium phase diagram and a meta-stable miscibility gap of Fe–Mo supersaturated solid solu-
tion [16]. The spinodal and binodal lines are asymmetric with composition c = 0.5, because of asymmet-
ric change of elastic stiffness with Mo content [37]. The spinodal point of meta-stable miscibility gap at
823 K is 19.0 at.%Mo. Fig. 16 shows a TEM microstructure in the vicinity of spinodal line of Fe–Mo MCG
alloy aged at 823 K.. Many contrasts caused by the h1 0 0i lattice modulation can be observed in the high
Mo concentration area, but gradually decrease in contrast with decrease of Mo concentration. The
microstructure of spinodal point is marked by two arrows, i.e. 19.0 at.%Mo. The microstructural mor-
phology seems to change gradually from the typical h1 0 0i modulated structure in the right side to
the mottled structure. The 100 satellites around the 200 electron reflection spot are also obtained in
the modulated structure area, whereas the satellites becomes diffuse and finally cannot be recognized
in the left side of photograph. The extinction of satellite signifies disappearance of the periodic distri-
bution of solute atom, that takes place usually in a so-called N-G type phase decomposition. Thus, the
microstructure changes continuously from the h1 0 0i modulated structure to the non-periodic mottled
structure without a distinct boundary. The experimental fact shown here is a clear proof of continuity
of microstructure from spinodal to N-G phase decompositions [30,38,39] signify.
T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060 1029

Fig. 17. A part of phase diagram of Fe–Al alloy system, showing a heat treatment temperature for Fe–Al MCG alloy.

Fig. 18. A TEM 100 dark field image of Fe–Al MCG alloy aged at 1023 K for 86.4 ks in the vicinity of the B2/A2 2nd order
transition lines, showing the microstructure change of ordered phase with composition. The bright region in the figure
corresponds to the B2 ordered domain and the black smoothly curved lines show the anti-phase boundary. The A2/B2 transition
line is determined to be 24.7 at.%Al at 1023 K, which is good agreement with the phase diagram of Fe–Al binary alloy system.

3.4. A2/B2 order/disorder transition in Fe–Al ordering alloy

In the present section, the microstructural change accompanying with the A2/B2 2nd order transi-
tion is demonstrated. Fig. 17 shows a part of Fe–Al equilibrium phase diagram.[40,41] The 2nd order
transition A2/B2 can be seen at the high temperature, although the coexistence with two phase
A2 + B2 and A2 + D03 exist at the lower temperature. The change of microstructure with 2nd order
transition was investigated at 1023 K.
1030 T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060

Fig. 18, showing a dark field 100 TEM image of Fe–Al MCG alloy aged at 1023 K for 86.4 ks [19],
demonstrates the morphological change of domain of anti-phase boundary (APB) near the A2/B2 tran-
sition point. In the region of high Al concentration the ordered domain has a large size with bright con-
trast. However, with decreasing Al content, the size of ordered domain becomes gradually smaller,
while the volume fraction of APB domain (black part) increases and finally reaches to 100%, as is
shown in the left edge of Fig. 18. As a matter of course the degree of long range order gradually re-
duces. It was experimentally confirmed that the domain size of microstructure does not vary even
if the aging time and quenching rate are changed. Thus, the microstructure in Fig. 18 is considered
to have been in equilibrium state. The A2/B2 transition in Fe–Al alloy system has been well known
as the 2nd order transition. Therefore, the contrast of B2 ordered domain ought to decrease gradually
without change of ordered domain size, because the brightness is related to the degree of order. How-
ever, Fig. 18 shows clearly that the ordered domain keeps the contrast still bright even at the region
close to A2 boundary. The morphological behavior shown in Fig. 18 does not imply the feature of the
2nd order transition but the 1st order transition [42]. This experimental fact may suggest the necessity
of reconsidering over the 2nd order transformation of Fe–Al alloy system.

4. Precipitate-nucleation near the edge of miscibility gap

A huge number of researches have theoretically and experimentally investigated the precipitate-
nucleation of alloys in the past 100 years. The first development of nucleation theory was performed
by Volmer and Weber [43]. They introduced a concept that the local composition fluctuation brought
to the embryo and critical stable nucleus. On the basis of Volmer’s theory, Becker and Döring [44–46]
evolved largely the kinetic theory of nucleation. They, assuming that the solute composition and struc-
ture of embryo are same as nucleus, proposed that the free energy change due to nucleation is only
caused by the embryo size On the other hand, Borelius [47,48], assuming the embryo to keep at con-
stant size, proposed a new kinetic theory that the free energy change due to nucleation was estimated
only by solute composition. Furthermore, Höbstetter [49,50] proposed the change of free energy for a
case where size and composition of embryo are changeable.
A big theoretical development on the spinodal decomposition was performed by Cahn and Hilliard
[8,51]. They introduced a new concept of diffuse interface changeable continuously from the precipi-
tate to matrix, and described the interfacial energy as a function of solute composition gradient
@c=@x. The kinetic theories of nucleation and coarsening were largely developed by Wagner [52] and
Lifshiz–Slyozov [53]. A comprehensive research has been performed on the homogeneous nucleation
by Binder and Stauffer [54]. A large number of detailed experimental investigations have been per-
formed [55–58]. Furthermore, excellent comprehensive reviews and books on the nucleation have been
also proposed [59–61].
However, such previous investigations have been examined on the alloy specimens whose solute
atom distribution is macroscopically flat. Hence, the alloy composition of each specimen was discontin-
uous. As described in Section 1.1, the phase transformation should be expressed by the three dimensional
space consisting of three axes of composition, time, and temperature. Thus the continuity on the compo-
sition axis should be desired. However, it is very difficult to prepare many alloy specimens each of which
have a little composition intervals such as less than 0.01 at.%. The discontinuity of composition may give
a large error to the composition-dependence of phase transformation, particularly phase transformation
in the vicinity of phase boundary where the thermodynamic variables change drastically.
In the present chapter, by utilizing the MCG method, it is possible to obtain the following experi-
mental data; composition-dependence of the size of precipitate-nucleus, the time to nucleate and the
equilibrium solute concentration at the particle interface for each nucleus. These are very important to
understand the stability of precipitate-nucleus and mechanism of nucleation. The MCG method will
contribute greatly to our knowledge of the phase transformation of alloys.

4.1. Microstructure changes with aging in varies alloys

Firstly, the precipitation of Ni3Al particles in the Ni–Al MCG alloy is examined. Fig. 19 shows a 100
dark field TEM image of Ni3Al particles formed near the solubility limit by aging at 973 K for 10.8 ks.
T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060 1031

Fig. 19. A 100 dark field image of TEM microstructure of Ni–Al MCG alloy aged at 973 K for 10.8 ks. The small white cuboids
show Ni3Al precipitates. The white circles in the photograph indicate the measuring points of the locally averaged solute
composition, whose values are correspondingly plotted in the small figure inserted.

The white small cuboids show Ni3Al precipitates. The gray solid circles in the photograph indicate the
EDS measuring points of the solute composition. The solid line in the inserted figure is evaluated by
using the least square method for the error function. It is obvious that the Al concentration decreases
gradually from the right side of photograph to the left over several lm. Many coherent Ni3Al particles
can be seen in the high concentration region, but the number of particles decreases in the lower con-
centration area and finally the particles are extinguished. Thus, a line connecting the two white arrows
of photograph is defined as the ‘‘precipitation front’’ of coherent particles for aging time of 10.8 ks at

Fig. 20. A 100 dark field TEM image of the Ni–Al MCG alloy aged at the same temperature with Fig. 19 but for a longer time,
86.4 ks. The precipitation front for 86.4 ks, indicated by the gray big arrows, is 11.7 at.%Al. The precipitation front is clearly
moves to the lower Al side by the longer time aging.
1032 T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060

973 K. The alloy composition at the precipitation front is determined to be 12.0 at.%Al from the in-
serted figure. It is remarkable that the particle size gradually increases large and then becomes a max-
imum size at the precipitation front.
The further aging gives a forward movement of the precipitation front to the lower composition side.
In Fig. 20 the precipitation front is given to be 11.7 at.%Al for 86.4 ks aging. It is obviously known by com-
paring Fig. 20 with Fig. 19 that the precipitation front shifts toward the lower Al side and the particle size
becomes larger. The similar behavior of precipitation is also recognized in other alloy systems. Fig. 21,
showing 100 dark field TEM images of Ni–Si MCG alloys aged at 973 K for various aging times, the pre-
cipitation front shifts gradually toward the lower Si composition side with progress of aging.
The characteristic of microstructures described above is not only in Ni–Al and Ni–Si alloys but also
general in the alloy systems of Cu–Co and Cu–Ti. Fig. 22 shows a TEM bright field image in the vicinity
of coherent phase boundary of the Cu–Ti MCG alloy aged at 873 k for 30s. Many coherent Cu4Ti precip-
itates disperse in the high Ti region, but the particle gradually decreases in number toward the left side.
Since the photograph was taken under the condition of g = 111, a so-called butterfly-contrast is observed
around the coherent precipitates. Such contrast may lead to a fallacious particle size. Therefore, the
square contrast taken under the multi beam condition was also used to estimate the accurate particle
size. Fig. 22B shows a microstructure taken under a multi-beam condition of B = 100 for same specimen
with Fig. 22A. The square contrast in Fig. 22B shows a real particle shape and size. Thus, the solute con-
centration of precipitation front of Fig. 22 is given to be 2.30 at.%Ti and the particle size is r = 15 nm.
The dark field TEM image of Ni–Cr–Al ternary MCG alloy aged at 973 K for 173 Ks is represented in
Fig. 23. In a case of ternary system, the tie-line of phase decomposition should be taken into consid-
eration. The direction of tie-line and phase compositions will be discussed later (see Fig. 32). The rea-
son why the shape of precipitate of Ni–Cr–Al alloy system is spherical comes from very small elastic
strain energy, i.e., g = 0.008[15]. Thus, the identical behavior of microstructure is recognized not only
in the binary MCG alloy system but also in the ternary MCG alloy system.
Here, it is noteworthy that the movement of precipitation front results from the composition
dependence of incubation time for nucleation, not caused by the change of macroscopic gradient
profile during aging (see Section 2). The composition dependence of incubation time is schematically
represented in Fig. 24. According to Darken [62], the mobility of atoms MðcA ; cB Þ in the interdiffusion
coefficient D e for A–B binary system is given by

MðcA ; cB Þ  ðM B cA þ M A cB ÞcA cB ð18Þ


By assuming that MA = MB  M0, the mobility M(cA, cB) is expressed as

MðcA ; cB Þ ¼ M 0 cð1  cÞ ð19Þ


Therefore, the diffusion of atoms is faster in the center part of miscibility gap, whereas becomes
slower in the lower solute composition area and remarkably slow in the edge of miscibility gap. Thus,
the time-lag of precipitation arises as a result. Therefore, when the MCG alloy is aged, the precipitation
front moves gradually to the lower composition side and stops eventually at the equilibrium solubility
limit, although the attainment has need of a long time aging.

4.2. Evaluations of nucleus-stability

By utilizing the changes of microstructure represented in Figs. 19–21, we are able to obtain
relationships between the equilibrium solute concentration ce at the particle interface and the particle
radius r. Fig. 25 shows a schematic illustration to explain how to get them. In Fig. 25 the local average
compositions ca, the schematic microstructures formed for different aging times t1, t2, t3 (t1 < t2 < t3)
and the solute composition profiles around the particle locating at positions r and s are described
in the top, middle and bottom parts, respectively. The volume fraction of precipitate f with respect
to the matrix is given by
ca  ce
f ¼ ð20Þ
cp  ce
T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060 1033

Fig. 21. Dark field TEM images of Ni–Si MCG alloys aged at 973 K for (a) 0.3, (b) 1.8 and (c) 7.2 ks, respectively. The precipitation
fronts are 11.30, 11.11 and 10.92 at.%Si, respectively.

Here, a case of aging time t3 is taken into consideration. The composition profile of solute atom in
the high composition area must be as that of s in Fig. 25. In this case the local average composition ca
is not equal to the equilibrium composition ce. This profile is usual for the microstructure consisting of
two phases. However, at the precipitation front the volume fraction is nearly equal to zero, i.e., f ; 0, so
1034 T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060

Fig. 22. TEM microstructures formed in Cu–Ti MCG alloy aged at 873 K for 30sec, showing many coherent Cu4Ti precipitates.
(A) The dark circles show the measuring points of average solute composition, whose values are correspondingly plotted in the
inserted figure. The photograph (A) was taken under a condition of g = 111, showing the butterfly contrast. The photograph (B)
was taken under g = 100 and multi-beam condition, showing cubic contrasts.

Fig. 23. A 100 dark field TEM micrograph of Ni–Cr–Al MCG ternary alloy aged at 973 K for 173 ks. The white particles are
c0 -precipitate of which composition is Ni–11.32 at.%Cr–13.70 at.%Al (see Fig. 32). The shape of precipitate particle is spherical
and the distribution of particles is in random, because of small lattice mismatch, g = 0.008%. The composition gradient is along
the tie-line shown by a thick solid line.

the average composition ca should be approximately equal to ce, as is clearly known from Eq. (20). The
composition profile is given by r. Thus, the relationship between the equilibrium composition ce (ca
at the precipitation front) and the diameter of a precipitate particle 2r are experimentally determined
by means of the analytical transmission electron microscope.
T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060 1035

Fig. 24. A schematic representation of change of incubation time for the precipitate-nucleation with alloy composition,
indicating that the nucleation time depends on the alloy composition.

Fig. 25. A schematic illustration how to get the equilibrium composition at the particle interface ce whose diameter is 2r,
where the local average composition ca, the microstructures with different aging time t1, t2, t3 (t1 < t2 < t3) and the solute
composition profiles around the particle locating at positions r and s, are schematically represented in the top, middle and
bottom parts, respectively. The particle r at the precipitation front is called as a critical stable precipitate hereafter.

Such relationship between the precipitate radius r and the equilibrium concentration ce holds al-
ways at the precipitation front, because the thermodynamic equilibrium is always maintained at
1036 T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060

the precipitation front. Therefore, by conducting aging of various times, we can obtain the systematic
values of r and ce for various particle size.
It is noteworthy that this relation is only available for the precipitation front. In the
higher composition area than the precipitation front, the competitive growth among particles
has already elapsed, so that the distribution of particle size is different from that of just nucleated
particles.
The experimental facts above described bring up a following rule that the precipitate particle
whose diameter is 2r is only able to nucleate at the precipitation front. The particles smaller than
the critical diameter 2r are unstable because ce > ca, thereby dissolve into the matrix. On the other
hand, the larger particles than 2r are produced as a result of particle coarsening starting from the size
of 2r. The competitive growth among the particles is very slow in the precipitation front. Therefore,
the particles in the precipitation front show the size as just nucleated. Thus, the particle observed at
the precipitate front corresponds to the thermodynamically equilibrium size.
The precipitation front shifts gradually to the lower concentration side with aging time and a new
nucleus is generated in a new front.
Fig. 26 shows the experimentally determined relationship between the radius of precipitate r at
the precipitation front and the equilibrium composition ce (ca) for the Ni–Si MCG alloy. A thick so-
lid curve, described along the lower limit of measuring points, shows the boundary of stability of
Ni3Si particle. The upper side of the curved line shows the thermodynamically stable region of Ni3Si
precipitates, whereas the lower side shows the unstable region. The ‘‘precipitate-nucleus’’ is defined
to be the ‘‘minimum stable precipitate-particle’’. Therefore, the solid curve in Fig. 26 represents the
change of nucleus size with alloy composition. The vertical broken line in Fig. 26 indicates the equi-
librium solubility limit for the particle of infinite radius, i.e., the equilibrium phase boundary. It is
clearly known from the figure that the size of stable particle steeply increases up to several ten
nm with approaching to the phase boundary and such a rapid increase takes place in an extremely
narrow composition range less than 0.3 at.% from the equilibrium phase boundary. The similar rela-
tionships are also obtained for the Ni–Si binary alloys aged at different temperatures, 923 K and
873 K, as represented in Fig. 27. Furthermore, the identical behavior is found in Ni–Al, Cu–Co,
Cu–Ti binary MCG alloys and Ni–Cr–Al ternary MCG alloy. Figs. 28 and 29 show the nucleus size
changes with composition Ce for Ni–Al and Cu–Co MCG alloys, respectively. Both MCG alloys rapidly
increase their particle size in the area of very vicinity of solubility limit, although the equilibrium
solubility limit is separate each other.

Fig. 26. A relationship between the critical radius of precipitate r and the equilibrium solute composition ce for the Ni–Si alloy
aged at 973 K. The solid circles in the figure indicate the radii of the particles observed in the vicinity of precipitation front for
various aging times and a thick solid curve (stability boundary) is described along the lower limit of these measured values.
T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060 1037

Fig. 27. The relationships between the critical radius of precipitate r and the equilibrium solute composition ce for the Ni–Si
MCG alloys aged at 923 and 823 K. The solid circles in the figure indicate the radii of the particles observed in the vicinity of
precipitation front for various aging times and the stability boundary (thick solid line) is described along the lower limit of these
measured values.

Fig. 30a shows a composition-dependence of nucleus size in Cu–Ti MCG alloy aged at 873 K for var-
ious times, clearly indicating rapid increase of size near the solubility limit. The experimental results
at 873 K and other aging temperatures are plotted on Cu–Ti phase diagram as shown in Fig. 30b. The
spinodal point of the alloy has been reported to be 2.77%Ti for 873 K [16], as indicated in the figure.
Any discontinuity in nucleus size is not recognized at the spinodal point, as seen in Fig. 30a.
Fig. 31 shows composition-dependence of r on the tie-line in the Ni–Cr–Al ternary MCG alloy aged
at 973 K. The rapid increase of particle nucleus is also recognized in ternary alloy system. This alloy
was prepared by coupling Ni–24 at.%Cr and Ni–25 at.%Al binary alloys, so that the phase decomposi-
tion proceeds along the tie line of c/c0 described with a thick solid line in the phase diagram of Fig. 32.
The composition of c-matrix is Ni–21.13Cr–3.42Al(at.%) and the precipitate is Ni–11.32Cr–
13.70Al(at.%).
Here, we again argue the advantage of the MCG method. As described above, the MCG method can
systematically detect the drastic increase of nucleus size in the very narrow composition range near
the solubility limit. If we try to get the same result by using the usual alloy specimen whose compo-
sition is macroscopically flat, we have to prepare many specimens whose compositional interval
changes precisely less than 0.05 at.% at least. Such precise control of alloy composition is actually
impractical. By contrast, in the MCG method the microstructure change is detectable even for extre-
mely small composition interval. Thus, it becomes possible for us to observe in detail the nucleation
behavior at the very vicinity of phase boundary.
1038 T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060

Fig. 28. A relationship between the critical radius of stable particle r and equilibrium solute composition Ce for Ni–Al MCG
alloy aged at 973 K.

Fig. 29. A relationship between the critical radius of stable particle r and equilibrium solute composition Ce for Cu–Co MCG
alloys aged at 823 K.

4.3. Calculation of the critical stable nucleus

The critical nuclei experimentally obtained for several MCG alloys are theoretically evaluated in the
present section. The Ni–Al MCG alloy is adopted for calculation.
The solute composition profile of a globular shaped precipitate in the matrix is expressed by using
several parameters which are shown by Eq. (21) and Fig. 33.
T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060 1039

Fig. 30. (a) A relationship between the critical radius of stable particle r and equilibrium solute composition Ce for Cu–Ti MCG
alloy aged at 873 K. (b) A part of phase diagram showing meta-stable coherent binodal and spinodal lines.

cðrÞ ¼ cp ð0 r 6 r 2 Þ
c  c 2ðr  r Þ n
p e 2
cðrÞ ¼  ¼ cp ðr 2 6 r 6 r3 Þ
2 r1  r2
c  c 2ðr  r Þ n ð21Þ
p e 2
cðrÞ ¼   2 ¼ ce ðr 2 6 r 6 r 3 Þ
2 r1  r2
cðrÞ ¼ ce ðr 1 6 r 6 r0 Þ

Fig. 31. Change of the critical radius of precipitate nucleus r with composition ce for Ni–Cr–Al ternary MCG alloy system.
1040 T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060

Fig. 32. A partial equilibrium phase diagram of Ni–Al–Cr ternary alloy system, indicating the tie line (a thick solid line) of phase
equilibrium of c/c0 . The MCG alloy is prepared by the imperfect arc melting on the interface of stacked Ni–25 at.%Al and Ni–
24 at.%Cr binary alloys, so that the phase decomposition proceeds on the tie line connecting with both mother binary alloys.

Fig. 33. A Solute composition profile of three-dimensional spherical precipitate particle in the matrix, which is used for the
theoretical estimation of particle stability.

The term n is a parameter determining the slope of composition gradient at the particle surface
(odd number). The total free energy change due to the nucleation DG is given by Eq. (22), proposed
by Cahn and co-workers [8,9,51], where the first, second and third terms are a chemical, gradient
and coherent elastic strain energies, respectively.

Z "  2 #
1 @cðrÞ
DGðca ; ce ; cp ; r 0 ; r 1 ; r2 ; nÞ ¼ DGc þ j þ g2 YðcðrÞ  ca Þ2 dr ð22Þ
V r @r

where V is the volume of precipitate, j is the gradient energy coefficient, g is the expansion coefficient
with respect to the solute concentration c, and Y is the elastic constant. The chemical free energy of
Ni–Ni3Al alloy at 973 K is proposed by Fig. 34 according to Hultgren et al.’s research [63]. The equilib-
rium compositions of matrix and precipitate phases given by the free energy curves are Ni–
11.68 at.%Al solid solution (c) and Ni–24.54 at.%Al (c0 ), which precisely correspond to the actual phase
boundaries [63]. The chemical free energy DGc, the gradient energy j(oc(r)/or)2 and the elastic strain
energy g2Y(c  c0)2 are functions of the parameters ca, ce, cp, r0, r1, r2 and n. Therefore, the parameters
are determined so as to give the minimum DG for a given particle size r1. If the particle of size r1 is
thermodynamically stable, the calculated composition profile has a suitable finite value. On the other
hand, when the particle is unstable, the composition profile should be flat. The critical size of stable
precipitate r1 is obtained by calculating the composition profile for various particle sizes r1, The
composition profiles calculated for various sizes of c0 -precipitate are represented in Fig. 35a for the
Ni–12.35 at.%Al and (b) for Ni–11.75 at.%Al. The numerical values used for calculation are listed up
T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060 1041

Fig. 34. The chemical free energy of c and c0 phases in Ni–Al alloy system (Hultgren et al. [63]).

Fig. 35. The solute composition profiles of particle with various size, calculated for a Ni3Al particle in (a) Ni–12.35 at.%Al and (b)
Ni–11.75 at.%Al alloys aged at 973 K, showing that the critical radii of stable particle r1 are 3.0 nm and 22.0 nm, respectively.
The r1 values show the outer radius defined by Fig. 33.

in Table 3, It is clearly known from Fig. 35a that the particles larger than 3.0 nm in radius r1 P 3.0 nm
are energetically stable, whereas the particles smaller than r1 = 2.9 nm are unstable. Therefore, the
critical radius of stable precipitate r 1 is determined to be 3.0 nm for the Ni–12.35 at.%Al alloy at
973 K. On the other hand, the composition profiles of Ni–11.75 at.%Al alloy described in Fig. 35b,
which is very close to the solubility limit, shows that the particles larger than r1 = 22.0 nm are stable,
while the particle of r1 = 21.0 nm is unstable. The critical radius r 1 ¼ 22:0 nm is very large compared
with that of Ni–12.35 at.%Al alloy, because the chemical driving force DGc is extremely small in the
1042 T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060

Table 3
Numerical values used for the calculation of stable precipitate-nucleus in Ni–Al
system.

Molar volume, VM/105 m3 mol1 2.791 [18,19]


Elastic stiffness, Cij/104 MN m2
C Ni Ni Ni 20.21, 14.79, 9.74 [21]
11 ; C 12 ; C 44
Lattice mismatch, g 0.0434 [20]
Interfacial energy density, cS/J m2 0.014 [16]

Fig. 36. Theoretical and experimental critical size of stable Ni3Al precipitate in Ni–Al MCG alloy.

vicinity of edge of miscibility gap. Thus, the critical radii of precipitate nucleus are theoretically eval-
uated for various alloy compositions. The calculated values are described by open circles in Fig. 36. The
calculated curve almost coincide with the solid line experimentally determined (see Fig. 28). Conse-
quently, the critical sizes of precipitate nucleus experimentally obtained are theoretically proven to
be proper on the basis of energetic evaluation. Thus, it becomes experimentally and theoretically clear
that the solid curves of Figs. 26–28 and 30 represent the boundary of thermodynamic stability of pre-
cipitate-nucleus. Here, it should be noted that such discussion is restricted only for the stability of
formed nucleus, not for embryo on the way of nucleation. The energy barrier for the nucleation is
not taking into consideration at all in this calculation.

4.4. Thermodynamic discussion on the basis of conventional nucleation theory

The thermodynamic stability of big nucleus formed near the phase boundary is discussed in this
section.
According to the conventional nucleation theories [43], the critical nucleus radius, i.e. the maxi-
mum embryo radius r em is given by Eq. (23).
r em ¼ 2cs =DGvc ol ð23Þ
v ol
The chemical free energy change due to the nucleation DGc is given by Eq. (24) in the dilute solid
solution.

DGev ol ¼ ð1=V m RT lnðce ðrÞ=C e ð1ÞÞ ð24Þ


where cs is the interfacial energy density between the particle and matrix, Vm is the molar volume of
precipitate, R is the gas constant and T is the temperature. The ce(r) and ce(1) are the equilibrium sol-
ute compositions at the interface of particle of radius r and infinite size, respectively.
T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060 1043

Substituting Eq. (23) to Eq. (24), Eqs. (25) and (26) are introduced
rem RT=2cs V m ¼ 1= lnðce ðrÞ=ce ð1ÞÞ ð25Þ
ce ðrÞ ¼ ce ð1Þ expð2cs V m =r em RTÞ ð26Þ

Eq. (26) is linearized for a large radius to the well known Eq. (27) [64].

2c V m
ce ðrÞ ¼ ce ð1Þ 1 þ s ð27Þ
r em RT
Since the particle size obtained experimentally in the present work is that of the critical stable nu-
cleus r, Eq. (27) should be replaced by Eq. (28) where the critical size remb is replaced by r.

3c V m
ce ðrÞ ¼ ce ð1Þ 1 þ s ð28Þ
r RT
where r = 1.5rem Eqs. (26)–(28) are called as the Gibbs–Thomson equation. Eq. (28) shows a linear
relationship between the inverse particle size 1/r and equilibrium solute concentration ce(r). The
Gibbs–Thomson relations are demonstrated for Ni–Si alloys whose particle size are fairly large and
Cu–Ti alloy systems whose particle size are small. Fig. 37 represents the relationships between ce(r)
and r1 for the Ni–Si alloys aged at three temperatures. The experimental values of ce(r) are obviously
straight for the three lines. The equilibrium compositions for the infinite particle size ce(1) are given
to be 10.77 for 973 K, 10.48 for 923 K and 9.91 for 823 K, which are precisely consistent with the sol-
ubility limits shown by vertical dotted lines in Figs. 26 and 27. By giving Vm = 2.596  105 m3/mol
and R = 8.314 J/K mol, the interfacial energy density cs is evaluated to be 0.012 J/m2 for 973 K,
0.010 J/m2 for 923 K and 0.011 J/m2 for 823 K, respectively. The interfacial energy density has been
reported to be the order of 0.01 J/m2 for the coherent L12 type precipitate particle [64,65], so that
the interfacial energies cs evaluated here are considered to be proper.
Next, we show the case of Cu–Ti alloy whose nucleus radius is fairly smaller than that of Ni–Si al-
loy. Fig. 38 shows the G–T relations between ce(r) and r1 for the Cu–Ti alloy aged at 873 K and 823 K.
The experimental values of ce(r) are not straight but bend in the small particle area, as predicted by
Eq. (26). For the large particle region the straight lines are recognized, and so the equilibrium
concentration ce(1) is given by the intercept of ordinate of Fig. 38 to be 2.12 at.%Ti for 873 K, which
is perfectly consistent with the solubility limit 2.12 at.%Ti at 873 K (see Fig. 30a). By giving
Vm = 7.12  106 m3/mol and R = 8.314 J/K mol, the interfacial energy cs is estimated from the slope
of the straight line to be 0.11 J/m2 for 873 K and 0.10 J/m2 for 823 K. Since the interfacial energy den-
sity of the coherent precipitate particles cs has been known to be the order of 0.1 J/m2 [65], the eval-
uated values are considered to be proper.

Fig. 37. The Gibbs–Thomson’s relations for Ni3Si precipitates of Ni–Si MCG alloy aged at the three temperatures. The three
numerical values of the vertical axis indicate the equilibrium solute concentration at the interface of infinite particle, i.e. the
equilibrium solubility limit.
1044 T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060

Fig. 38. The Gibbs–Thomson’s relationships between the particle radius r and equilibrium composition at the particle interface
ce for Cu4Ti particles in Cu–Ti MCG alloys. The deviation from the straight line for the small particles is rationalized by Eq. (27) in
the text. The interfacial energies evaluated for 873 K and 823 K are 0.11 and 0.10 J/m2, respectively.

Fig. 39. A relationship between the normalized nucleus size and the equilibrium solute composition at the particle interface.
These are scaled for different aging temperatures of Ni–Ni3Si MCG alloys, based on the Gibbs–Thomson’s equation given by Eq.
(27).

On the basis of Eq. (26), the critical nucleus size r can be scaled with respect to the supersaturation
of solute atom. Fig. 39 demonstrates that the Ni3Si precipitate particles nucleated are well scaled by
the Gibbs–Thomson’s equation even though the aging temperature is different. Similarly, the nucle-
ation for different alloy system is well scaled as represented in Fig. 40.
The consideration above explained is also verified by Fig. 41 which shows the relationship between
critical nucleus size r and the solute supersaturation Dc for the Cu–Co alloys. The nucleus size of Cu–
Co alloy was obtained by Wagner [66] and Wentd and Haasen [67]. They determined the smallest radii
in several Cu–Co alloys aged at 783–833 K. Fig. 41 shows the experimental results (open marks) deter-
T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060 1045

Fig. 40. The critical particle sizes scaled for all alloy systems with respect to the super-saturation of solute atom. This figure
clearly shows that the big precipitate-nuclei in the vicinity of the miscibility gap are well explained by the Gibbs–Thomson’s
equation, i.e. the conventional nucleation theory.

Fig. 41. The relationship between the critical nucleus size r and the solute supersaturation Dc in Cu–Co alloys. The solid circles
are experimentally obtained by present work for Cu–Co MCG specimens and the open marks are obtained by P. Haarsen and R.
Wagner [58] and arranged by H.I. Arronson and F.K. LeGoues [59] for the composition flat alloys. The solid straight line is
theoretically given by Wagner (see [52]) on the basis of Eq. (24).

mined for the usual composition flat Cu–Co alloys which were obtained by Haarsen and Wagner and
arranged by Aaronson and LeGoues [59]. The solid straight line shows the changes of critical nucleus
radius with composition change, estimated on the basis of Eq. (23) by Wentd and Haasen [67]. The
solid circles in Fig. 41 show our experimental results for the Cu–Co MCG alloy. The our results of
Cu–Co MCG alloy are clearly on the extension of the theoretical line of nucleation. In the usual com-
position flat specimen, the detectable lowest limit of alloy composition remains at about 0.1 at.%, but
by utilizing the MCG method the detectable limit is improved up to the order of 0.01 at.%.
Consequently, the experimental results of the nucleus size formed in the region of solubility limit
are summarized as follows; the nucleus size increases rapidly up to several hundreds of nanometers in
a very narrow composition range less than 0.3 at.% from the equilibrium solubility limit. The diameter
of nucleus size can reach over 500 nm in the edge composition very close to the precipitation limit.
Such the nuclei are energetically predicted by the conventional nucleation theory and the Gibbs–
Thomson’s equation.
1046 T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060

However, it is noteworthy that, though the stability of nucleus is thermodynamically rationalized,


it is impossible that the conventional nucleation theory explains kinetically why the big nucleus is
formed for so short duration of aging. The kinetic investigation is dealt with in the next Section 5.

5. Kinetic investigations on the nucleation process

As described in Section 4, the thermodynamic stability of the big nucleus is rationalized by the clas-
sical nucleation theory. However, it seems difficult for the conventional nucleation theory to rational-
ize that the big nucleus appears within a fairly short aging time, for instance, t = 86.4 ks for the nucleus
radius r = 86 nm (see Fig. 20).
The nuclei smaller than the stability-boundary (see Figs. 27–30) are thermodynamically unstable,
as described in Section 4. Thus, the nucleus, starting from the supersaturated solid solution, must pass
through the unstable region and then reach to the stable region. We discuss the kinetics of nucleation
in the present section on the basis of experimental results.

5.1. Experimental results

The kinetic data on the nucleation are summarized firstly. Fig. 42 represents the radius of critical
stable nucleus r, the equilibrium composition ce and the time to nucleate Ni3Si nucleus in Ni–Si MCG
alloy aged at 973 K. The upper and lower regions of the solid line show the stable and unstable regions,
respectively. Hence, the line means the critical minimum size of stable nucleus. ‘‘Aging time’’ in Fig. 42
is identical with ‘‘time for nucleation’’, i.e., the incubation time for each nucleation. The representative
experimental results obtained for three aging temperatures are listed up in Table 4. It is clear that the
nucleation time needs more time with lower composition for all temperatures. Fig. 43 and Table 5
show the radius of critical stable nuclei r, the equilibrium composition Ce and the time to nucleate
Ni3Al nucleus in Ni–Al MCG alloy aged at 973 K for various times. The similar results are demonstrated
for Cu–Ti MCG alloy in Fig. 44 and Table 6.
In this manner, the nucleus size and the formation time were measured to each nucleus. Such the
results have obtained first by utilizing the MCG method. These experimental results are very impor-

Fig. 42. A composition dependence of the critical radius of stable particle r, the equilibrium solute concentration ce and the
aging time required to nucleate Ni3Si-nucleus in the Ni–Si alloy system aged at 973 K. A solid curved line, described along the
lower limit of experimental particle radius, shows the boundary for nucleus stability. The upper and lower regions of the solid
line show the stable and unstable regions, respectively. Hence, the line means the critical minimum size of stable nucleus.
‘‘Aging time’’ is identical with ‘‘time to precipitate’’, i.e., incubation time for nucleation.
T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060 1047

Table 4
The equilibrium solute concentration at nucleus interface ce, the nucleus size r and the incubation time to nucleation t,
experimentally obtained for Ni–Si MCG alloy aged at 823, 923 and 973 K.

Ni–Si MCG alloy


T = 973 K T = 923 K T = 823 K
Dce (at.%Si) r⁄ (nm) t (ks) Dce (at.%Si) r⁄ (nm) t (ks) Dce (at.%Si) r⁄ (nm) t (ks)
1.13 2.1 0.06 0.932 5.1 1.02 1.02 5.1 18.62
0.74 2.6 0.13 0.466 11.1 3.83 0.851 7.0 21.54
0.53 8.0 0.30 0.325 17.6 7.08 0.53 10.1 48.64
0.29 19.5 1.15 0.205 22.2 22.4 0.30 20.0 146.78
0.23 25.1 1.80 0.150 35.2 35.48 0.183 34.3 301.30
0.135 40.2 7.20 0.096 43.1 89.12 0.100 61.5 681.24
0.068 61.5 20.81 0.060 45.0 237.14 0.067 82.0 1467.57
0.041 85.5 61.89 0.049 78.7 442.38

Fig. 43. A composition dependence of the critical radius of stable particle r, the equilibrium solute concentration ce and the
aging time required to nucleate Ni3Al-nucleus in the Ni–Al alloy system aged at 973 K. The upper and lower regions of the solid
line show the thermodynamically stable and unstable regions, respectively. Aging time is identical with the incubation time for
nucleation.

tant to analyze the process of nucleation. We investigate the kinetics of nucleation on the basis of
these experimental results.

5.2. Theoretical basis

The composition-dependence of the incubation time has been qualitatively well known. However,
the quantitative evaluation, particularly in the vicinity of solubility limit, has not been investigated,
although the feature of nucleation comes out most clear in the region of phase boundary, because
of the drastic change of energy barrier associated with composition- change.
According to the kinetic equation for nucleation, the frequency of nucleation U is well known to be
expressed by Eq. (29) [42].
   
Qd DG ðcÞ
U ¼ Ad MðcÞ exp   Af exp  ð29Þ
kT kT

where Qd is the activation energy for the diffusion of solute atoms and DG(c) is the activation barrier
for nucleation, i.e., the nucleation energy barrier that must be overcome to form a nucleus. Ad and Af
1048 T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060

Table 5
The equilibrium solute concentration at nucleus interface ce, the nucleus
size r and the incubation time to nucleation t, experimentally obtained
for Ni–Al MCG alloy aged at 973 K (ce = 11.65 at.%Al).

Dce (at.%Al) r⁄ (nm) t (ks)


0.71 2.5 0.22
0.56 3.2 0.64
0.38 5.1 1.57
0.18 12.0 4.53
0.081 28.0 10.22
0.040 40.0 36.00
0.040 49.1 54.11
0.025 62.0 100.0

Fig. 44. A composition dependence of the critical radius of stable particle r, the equilibrium solute concentration ce and the
aging time required to nucleate Cu4Ti-nucleus in the Cu–Ti alloy system aged at 873 K. A solid curved line shows the boundary
for nucleus stability. The upper and lower regions of the solid line show the thermodynamically stable and unstable regions,
respectively. Aging time is identical with the incubation time for nucleation.

are constant parameters. The mobility for atom-diffusion M(c) has been proposed by Darken [62] as
M(c) = M0c(1  c) (see Eq. (19)).
The first term of Eq. (29) is rewritten here as Eq. (30).
 
Qd
Ud ¼ Ad MðcÞ exp  ð30Þ
kT

Eq. (30) is available for the phase decomposition phenomenon of which rate-controlling process is the
atomic diffusion, i.e., the spinodal decomposition, the microstructure coarsening such as Ostwald rip-
ening of precipitate-particles.
On the other hand, Eq. (31) which is the second term of Eq. (29), is caused by the energy barrier for
nucleation, and is available for the phase decompositions of ‘‘nucleation-growth’’ type.
 
DG ðcÞ
Uf ¼ Af exp  ð31Þ
kT
Firstly, Eq. (30) is taken into discussion. At a constant temperature, Ud is only a function of com-
position c and expressed by a quadratic function of composition, because M(c) = M0c(1  c). Therefore,
the time to nucleation U1 d is expressed by a hyperbolic curve which increases with decrease of super-
T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060 1049

Table 6
The equilibrium solute concentration at nucleus interface ce, the nucleus size r and
the incubation time for nucleation t, experimentally obtained for Cu–Ti MCG alloy
aged at 873 K.

Dce (at.%) (ce = 2.12 at.%Ti) r⁄ (nm) t (s)


0.755 3.00 3
0.615 4.55 5
0.501 4.70 10
0.205 5.01 15
0.125 7.50 20
0.065 15.3 30
0.050 26.0 60

Fig. 45. Typical two types of nucleation process; (a) size growth of small particle (Becker type) and (b) increase of solute atom
concentration (Borelius type).

saturation Dc and diverges at Dc = 0. Thus, the relationship between log U1 d and log. Dc is given by a
straight line in the region of low solute concentration. Even if the aging is carried out at different tem-
perature, the straight line moves only in parallel.
Next, we consider Eq. (31) which arises from the energy barrier for nucleation. Two typical energy
barriers have been commonly known in a mechanism of nucleation, as illustrated by Fig. 45. One is a
case where a small embryo with a high solute concentration and a sharp interface is initially formed
and then widens its size to the stable size. Another is a case where an embryo increases its solute con-
centration with progress of nucleation. The former is well known as Becker type nucleation [44,45]
and the latter is called as Borelius type nucleation [47,48]. In either way, the nucleation frequency
Uf is given by Eq. (31), because the nucleation progresses by a mechanism of thermal activation
process.
The energy barrier DG(c) is given by Eqs. (32) and (33) for the Becker type and the Borelius type,
respectively.
4pr 2 cs
DG ðcÞ ¼ ð32Þ
3
 3
r
DG ðcÞ ¼ 4p DF v ð33Þ
a0
1050 T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060

Free Enregy
Composition
Fig. 46. A schematic illustration of the concave part of free energy–composition curve, which is essentially caused by the
Boltzmann–Gibbs extensive entropy.

Table 7
Increments of the energy barrier DG(c) with nucleus size for Ni–Si alloys.

Nucleus radius (nm) Becker type DG(c) (kJ/mol) Borelius type DG(c) (kJ/mol)
2
1 1.0  10 0.8  102
5 2.5  103 1.0  104
10 1.0  104 8.0  104
20 4.0  104 6.4  105
30 9.0  104 2.2  106
50 2.5  105 1.0  107

Fig. 47. A schematic illustration of the nucleation times, which are theoretically given by Eqs. (29)–(31) for the alloy whose
composition is very close to the solubility limit. It is noteworthy that the line of U1 must be bend near the solubility limit so
long as based on the Gibbs–Boltzmann free energy.

where r is the radius of nucleus, a0 is the lattice constant and DF m is the energy barrier arising from a
concave part of the free energy–composition curve (see Fig. 46). Since formulas are the functions of
nucleus size r, DG(c) depends greatly on the size. Table 7 shows the energy barriers estimated for var-
ious nucleus size of Ni3Si nucleus. When the nucleus radius is small (1 nm), the energy barrier is
about 80 kJ/mol, that is consistent with the conventional activation energy for nucleation, e.g.,
86 kJ/mol [68]. However, DG(c), increasing exponentially with size increase, reaches to an incredibly
huge value, for instance, 107 kJ/mol for r = 50 nm.
In the two mechanisms, the Becker’s type barrier is artificially generated from the assumption that
the interface of nucleus is sharp at the starting point of nucleation. Since the interfacial energy is a
function of composition gradient (oc/or)2 [8,9], it is not necessary that the nucleus has a sharp interface
of high energy from the beginning of nucleation. This means that the premise of Becker’s type energy
barrier collapses. Therefore, it is not necessary to take the Becker type barrier into consideration. On
the other hand, the Borelius type energy barrier is caused by the concave part of the free energy–com-
position curve. Since the concave part originates in the Boltzmann–Gibbs free energy essentially, the
T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060 1051

Borelius type barrier is inevitable. The non-classical nucleation theory proposed by Cahn–Hilliard [8,9]
is consistent with the classical nucleation theory in the vicinity of the edge of miscibility gap. Conse-
quently, any nucleation theory cannot avoid the Borelius type energy barrier so far as the Boltzmann–
Gibbs free energy is taken into the consideration.
As shown in Table 7 the nucleation energy barrier DG(c) is enormously high for the big nucleus in
either nucleation mechanism. Hence, the composition-dependence of Eq. (31) must be enormously
large than that of Eq. (30) for the big nucleus. Accordingly, the line of log U1 must be strongly bent
in the vicinity of edge of miscibility gap, as is schematically drawn in Fig. 47.

5.3. Kinetic investigation on the experimental results

Nevertheless, the experimental results never show such bending. Fig. 48 shows composition-
dependences of the nucleation time U1 1
d , Uf and U1, which are calculated for Ni3Si nucleus in Ni–
Si MCG alloy aged at 823 K. In Fig. 48 the two solid lines are theoretically given by Eqs. (30) and
(31), respectively, and the experimental data are plotted by solid circles whose numerical values
are given in Table 3. The experimental data are perfectly compatible with the straight line theoreti-
cally given by Eq. (30), whereas never show any consistence with Eq. (31). The identical experimental
facts are also obtained in other alloys examined in the present work, as represented in Fig. 49a–c.
Therefore, the experimental facts described above clearly prove that the first term of Eq. (28) is only
effective, whereas the 2nd term does not contribute at all. These straight lines are parallel each other.
The activation energies estimated by means of Ahrenius plot of them are 221 K J/mol for Ni–Si alloy
system, 234 K J/mol for Ni–Al system and 190 K J/mol for Cu–Ti alloy system. These experimental acti-
vation energies approximately consist with the activation energies for solute atom diffusion [69];
250 K J/mol for Ni alloys and 204 K J/mol for Cu–Ti alloys. Thus, it is proved that the nucleation pro-
cesses shown in Fig. 49a–c are resulting from the solute atom diffusion without energy barrier for
nucleation, i.e. spinodal-like phase decomposition.
Thus, it is necessarily approved that the phase decomposition of supersaturated solid solution pro-
gresses by a mechanism of spinodal even in the N-G region. Fig. 50 illustrates a schematic phase dia-
gram showing that the phase decomposition proceeds spinodally in whole area of miscibility gap. The

Fig. 48. A composition dependence of the time to nucleation U1, experimentally obtained for Ni3Si precipitates. A straight
solid line and a curved solid line are theoretically given by the 1st and 2nd terms of Eq. (29), respectively, and the experimental
results are represented by solid circles. The theoretical line is fit with the experimental data at the point of star mark.
1052 T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060

Fig. 49. The relationships between the time to nucleation U1 and the super-saturation of solute atoms Dc for (a) Ni–Si, (b) Cu–
Ti and (c) Ni–Al alloys, respectively.
T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060 1053

experimental fact denies inevitably the existence of concave part in the free energy–composition
curve described in Fig. 46. The concave arises originally from the Boltzmann–Gibbs’s extensive entro-
py. Therefore, the denial of concave part is synonymous with denial of the Boltzmann–Gibbs free en-
ergy consisting of the extensive entropy. Thus, the behavior of phase decomposition demonstrated in
Fig. 50 is never rationalized by the Gibbs–Boltzmann free energy. A comprehensive consideration on
the nucleation is opened in Section 5.5.

5.4. Pre-nucleation phenomena

As described above, the big nucleus appears at the precipitation front, and never be observed in the
lower composition region than the precipitation front. Namely, the precipitation front is the front line
of nucleus precipitation.
When the further aging is progressed, the bigger nucleus appears newly in the lower region. There-
fore, it is inferred that a pre-nucleation phenomenon is progressing in the lower composition region.
The pre-nucleation phenomenon is experimentally investigated.
Fig. 51 is schematic illustrations of predictable microstructures when the MCG alloy is heat-treated
by the two step aging. When the MCG specimen is aged at high temperature T2, the precipitate par-
ticles are formed till the precipitation front, as illustrated in Fig. 51a. In the usual sense, when the
specimen is rapidly quenched from the high temperature T2 to the lower temperature T1 and then hold
at T1 for a short duration, very fine precipitates ought to appear newly in the lower composition area
for the cause of difference of solubility limit between the two temperatures, as illustrated in Fig. 51b.
However, in the actual microstructure the several big precipitates are coexistent with fine precipitates,
as schematically illustrated in Fig. 51c. Fig. 52 shows a 100 dark field TEM image of Ni–Si MCG alloy
aged at 823 K for 6 ks after aging at 973 K for 54 ks. The precipitation front of 973 K aging is indicated
by the two arrows in Fig. 52. It is obviously recognized in Fig. 52 that the big precipitates mingle with
the fine precipitates in the lower composition side. The similar microstructure is also recognized in the
other alloy system. Fig. 53 shows a dark field image of Ni–Al MCG alloy formed by two step aging,
namely aged at 823 K for 180 s after aged at 973 K for 54 ks. Those micrographs certainly show the
mingled structure of very big particles with the fine particles. The side planes of large square particles
seem to dent in the side planes, which are possibly the beginning of particle-splitting [70,71].
These experimental facts imply that big clusters inhere in the lower composition area, although the
cluster is not recognized by TEM observation. Such the solute rich clusters have already been reported.

Fig. 50. A schematic phase diagram showing that the phase decomposition proceeds spinodally in whole area of miscibility gap.
A solid and dotted lines are the binodal and spinodal lines which are described on the basis of Boltzmann–Gibbs free energy.
The solid circles in the figure indicates the qualitative change of nucleus size with c and T.
1054 T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060

Fig. 51. A schematic illustration of expected microstructures when the Ni–Si MCG alloy is heat treated by the two-step aging.

Precipitation front of aging at 973K for 54ks.

Fig. 52. A TEM dark field image of Ni–Si MCG alloy aged at 823 K for 6 ks after aging at 973 K for 54 ks, showing the non-
uniform large particles produced in the lower composition area from the precipitation front, probably arising from the pre-
nucleation phenomenon.

On the basis of investigation by the small angle scattering, Wagner [72–74] argued that the Co-rich
cluster exists in the matrix before formation of nucleation. Furthermore, Rzdilsky et al. [75] and Blavet
et al. [76] found the Co-concentrated region existing in the Cu matrix by means of AP-FIM observation.
Consequently, it is considered that the solute-rich diffuse droplet has been formed in the matrix as a
pre-stage of nucleus, in which the solute atoms may be chained, not dispersed randomly. However, it
should be noted that such large droplets are also not thermodynamically unstable, so far as based on
Boltzmann–Gibbs free energy. The comprehensive discussion is in Section 5.6.

5.5. High speed growth of big nucleus near the solubility limit

We discuss whether the formation-rate of big precipitate particle is rationalized or not by the dif-
fusion equation. We concentrate a big Ni3Al particle observed in the Ni–Al MCG alloy.
Fig. 20 shows that a particle with radius of r⁄ ; 50nm was formed in the Ni–Al MCG alloy aged at
973 K for 86.4 ks. According to Zener [77] and Zener and Wert [78], the growth of precipitate particle
due to the atom diffusion is given by
 
@r @c
ðcp  ce Þ ¼D ð34Þ
@t @r r¼r
T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060 1055

Fig. 53. A TEM micrograph of Ni–Al MCG alloy aged at 823 K for 1 ks after aging at 973 K for 54 ks, showing that the non-
uniform large particles mingle with very fine particles formed during 823 K aging.

The curvature effect of particle surface is taken into consideration, so this formula will turn into
 
D c0  ce
r  r0 ¼ ðt  t 0 Þ ð35Þ
2r D cp  ce
The numerical values used for calculation are listed up in Table 8, which are corresponding to the
microstructure in Ni–Al MCG alloys shown in Figs. 19 and 20. we get the particle radius r = 6.47 nm as
by calculation. This value is very small, compared with actual radius r = 50 nm. Namely, even if the
fastest condition of particle growth is assumed, Eq. (35) cannot rationalize the formation of big parti-
cle for a short aging time. The diffusion process is supposed to be accelerated for some mechanism.

5.6. Problems of nucleus formation in the N-G region

In the present section, the problems of nucleation process are comprehensively reconsidered on the
basis of the experimental facts obtained by MCG method. As described in Section 4, the thermody-
namic behavior of ‘‘nucleus created’’ is explained well by the conventional thermodynamic theories
based upon the Boltzmann–Gibbs free energy, such as the equilibrium phase diagram, the Gibbs–
Thomson relationship, and the composition dependence of nucleus size. However, it is impossible
for the conventional nucleation theory to rationalize the kinetics of formation-process, particularly
for the big nucleus in the vicinity of solubility limit.

Table 8
Numerical values used for calculation of Eq. (44).

Temp (K) T = 973 K


Composition (at.%Al) ca = 11.75 ce = 11.70 cp = 23.00
Diffusion coeff. D (m2/s) 1.87  104 exp(268,000/RT)
Time t, t0 t = 86.4 ks t0 = 10.8 ks
1056 T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060

The problem of the nucleate process is to find out how the nucleus passes through the unstable
region and reach the stability region, as demonstrated in Figs. 42–44. In the bottom area than the
stability boundary the nucleus is thermodynamically unstable. The critical nucleus size depends on
the alloy composition ce, that is, the nucleus size is small at the deep area of miscibility gap, whereas
in the vicinity of the solubility line the big nucleus of hundreds nm is only stable. Thus, the embryo
cannot grow up to the stable big nucleus so far as based upon the conventional free energy of
Boltzmann–Gibbs. The Boltzmann–Gibbs free energy accompanies necessarily with the concave
portion, as explained in Section 4. Even if any modification is introduced into the Boltzmann–Gibbs
extensive entropy, the concave part of free energy must appear in the Boltzmann–Gibbs free energy,
because the differential coefficient of extensive entropy becomes infinite in the very vicinity of pure
metals. Hence, the energy barrier resulting from the concave portion appear always and should be
enormously high for a big nucleus, so that the thermal fluctuation is not able to overcome the extre-
mely high energy barrier, as demonstrated in Sections 5.2 and 5.3.
There has been a general concept that the nucleation in the N-G region is progressed through the
compositional fluctuation resulting from the thermal energy. The experimental nuclei are continu-
ous from a small size up to a huge size, as demonstrated in Figs. 42–44. Therefore, it is necessary
that the nucleation mechanism must be unified for all nucleus size including the big nucleus. There-
fore, the formation of the big nucleus must be also explained by the compositional fluctuation.
However, it is impossible that the compositional fluctuation creates such big nucleus in the extre-
mely low concentration alloy, because the nucleus size is so large that the solute atoms in the very
wide range must gather up to one place by the fluctuation. Consequently, it is impossible to cross
over the unstable domain by the usual nucleation mechanism and also impossible by the thermal
fluctuation.
However, the big nucleus is actually formed within a comparatively short aging time. Since all nu-
clei including a big nucleus are produced by mechanism of spinodal decomposition, a free energy
curve having no concave portion can only rationalize the process of phase decomposition. It is note-
worthy that the behavior of nucleus in the equilibrium state is rationalized by the conventional Boltz-
mann–Gibbs free energy. Hence it is supposed that the separate free energies should be applied for the
equilibrium and non-equilibrium states respectively. A free energy having such multiplicity has not
been established currently.
It is considered that the conventional Boltzmann–Gibbs free energy has a problem in estimation of
the entropy of atom-configuration. The extensive entropy of atom configuration has been introduced
into the Boltzmann–Gibbs free energy. The extensive entropy defines that a small atom-ensemble
consisting of several atoms represents the solute atom arrangement of whole system. However, in a
case of very low concentration alloy, since the solute atoms disperse very sparsely in the matrix,
the size of the small domain may be changed with the solute concentration [79]. Such free energy con-
sisting of the non-extensive entropy has been proposed by Tsallis [79,80]. However, the evaluation of
Tsallis’s theory has not been decided yet for the phase transformations of alloy. The nucleation mech-
anism based on the Tsallis theory will be reported in near future.

6. Summary

A new characterization method, ‘‘Macroscopic Composition Gradient (MCG) Method’’ was pro-
posed to investigate the critical phenomena of phase transformation. The distinctive feature of MCG
method is to investigate systematically the phase transformations in the various composition alloys
by utilizing a single specimen which has a macroscopic composition gradient. Since the macroscopic
composition gradient in the alloy is prepared so as to step over the phase boundary, the morphological
observation of critical phenomena at the phase boundary can be realized by means of analytical trans-
mission electron microscopy.
The Macroscopic Composition Gradient (MCG) Method was theoretically and experimentally con-
firmed that it never affects on the phase transformation and microstructure formation.
T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060 1057

By using this method, various kinds of phase transition such as the coherent and incoherent pre-
cipitation lines, the order/disorder phase transition and morphological change at the spinodal line,
have successfully been evaluated.
Furthermore, the critical size of precipitate nucleus and the nucleation rate near the solubility limit
are experimentally investigated for several binary alloy systems.
The nucleus size shows a steep increase up to several tens of nm in a very narrow composition
range less than 0.3 at.% from the phase boundary. The diameter of nucleus size can reach over
500 nm in the composition very close to the precipitation line. The Gibbs–Thomson relation and the
conventional nucleation theory thermodynamically rationalize such composition-dependence of nu-
cleus size change.
However, the kinetics of nucleation is never explained by the conventional nucleation theories. The
kinetic experimental results show distinctly that the nucleation time is only controlled by the atom
diffusion and the phase decomposition of supersaturated solid solution is progresses spinodally with-
out energy barrier for the nucleation, even in the so-called Nucleation-Growth region.
On the basis of experimental results the application limit of conventional nucleation theory is dis-
cussed, and hence the failure of Boltzmann–Gibbs’s extensive entropy becomes clear for the early
stage of phase decomposition. The N-G region is brought from the artificial effect arising from the
Boltzmann–Gibbs’s extensive entropy and appears artificially only in the equilibrium phase diagram.
It is noteworthy that the experiments presented here have not been performed in the past. The MCG
method proposed here is considered to open a new way to study the microstructure evaluation, par-
ticularly for the critical phenomena near the phase boundary.

Acknowledgements

We have been constructing a new characterization method of the phase transformation, namely,
‘‘Macroscopic Composition Gradient (MCG) Method’’ with many colleagues for over 15 years. We have
developed this technique uniquely and believed that the MCG method will open a new way to inves-
tigate the phase transformation, particularly the critical phenomena in materials science. The author is
very grateful to many colleague for their cooperation and assistance, particularly Dr. Sengo Kobayashi
in the Ehime University, Prof. Toshiyuki Koyama in the National Institute of Materials Science (now
Nagoya Institute of Technology), Prof. Takao Kozakai in Nagoya Institute of Technology and many
postgraduate students over 60 members of my laboratory. I am also grateful to Dr. Claudio G. Schön
in the University of Sao Paolo in Brazil for his partly cooperation in discussion. The author is grateful to
Prof. Tetsuo Mohri in Hokkaido University and Dr. John W Cahn in NIST of USA for their encourage-
ment to promote the investigation.

Appendix A. Method of ‘‘composition vs. distance’’ curve in the MCG specimen

The MCG specimen is prepared by utilizing the diffusion of solute atoms. Therefore, ‘‘the composi-
tion-distance’’ curve (see the inserted figure in Fig. 3, for instance) ought to be given by the Gaussian
error function
" Z x=2pffiffiffiffi #   
Dx
ce 2 2 ce x
cðx; tÞ ¼ 1  pffiffiffiffi ex dx ¼ 1  erf pffiffiffiffiffiffi ðA1Þ
2 p 0 2 2 Dx

where the initial condition is given as follows; c = ce for x 5 0 and c = 0 for x > 0, when Eq. (A1) is intro-
duced from the Fick’s diffusion equation.
However, since the MCG specimen is usually collected from the portion apart from the Kirkendall
interface (see Fig. 2) it may be difficult to fit the measured data to Eq. (A1), because Eq. (A1) expresses
the solute concentration profile starting from the Kirkendall interface. Hence, by introducing the
parameters p and q, Eq. (A1) is generalized to be applicable for the actual measured data, as described
in the following equation:
1058 T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060

Fig. A1. Changes of the standard deviations of composition error for the independently measured values cm and the evaluated
composition value from the least square curve cfit with increase of composition data.

" q
!#
fit ce p
q
c ðx; t; p; qÞ ¼ 1  erf pffiffiffiffiffiffi ðA2Þ
2 2 Dx
2
X fit 2
m
di ¼ ci  ci ðA3Þ
i

Eq. (A3) shows the square of difference between the experimental solute concentration cm i ðxi Þ and the
estimated value from Eq. (A2) cfiti ðxi Þ. The parameters p and q in Eq. (A2) are so determined as to min-
imize d2i . In the experiment, a computer program for the calculation has been arranged, and the com-
position profile is simultaneously calculated on the occasion of composition measurement. The solute
concentration at any position can be estimated from the composition vs. distance curve. By utilizing
this method, we obtain the more accurate composition than that of as-measured value.
Next, we evaluate a concrete improvement of accuracy of cfit(x) which is given by Eqs. (A2) and
(A3). We assume a true value of composition ctrut(x)e at position x, given by Eq. (A4). The parameters
are fixed to p = 2 and q = 1.
 x 
true ce 2
1
c ðxÞ ¼ 1  erf p ffiffiffiffiffiffi ðA4Þ
2 2 Dx
Thus, we obtain experimentally cm(x) and evaluate ctrue(x) from Eq. (A4). On the basis of these values,
we evaluate the accuracy of cfit(x) by utilizing the following two equations.
 
true 2
ðdm Þ2 ¼ cm x  cx ðA5Þ
 2
ðdfit Þ2 ¼ ctrue
x  cfit
x ðA6Þ
fit pffiffiffiffi
The accuracy of composition profile c (x) is expressed in the form of standard deviation r of the
predicted composition value.
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
uP 2 " #
pffiffiffiffiffiffi u d 1 ðx  xi Þ
rx ¼t i
1þ þP ðA7Þ
n2 2 ðxi  xi Þ2
T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060 1059

pffiffiffiffi
It is clearly understood from Eq. (A7) that the standard deviation r decreases with increase of
fit
number of data n, that is, the accuracy of c (x) ispimproved
ffiffiffiffiffiffiffi with increase of number of data.
pffiffiffiffiffiffiffi
Fig. A1 shows changes of the two standard deviations rm and rfit with number of measuring point,
which are obtained
pffiffiffiffiffiffiffi by computer calculation for the model based on Eq. (A4). In the calculation,pffiffiffiffiffiffiffi theffi
pffiffiffiffiffiffi
average of rm is assumed to be 0.5. In Fig. A1 the p open
ffiffiffiffiffiffiffi and solid circles represents r fit and r m,
pffiffiffiffi
respectively. It is clear from the solid circles that prffiffiffiffiffiffiffi
m is scattered around r ¼ 0:5 even though
the data n increases in number. On the other hand, rfit clearly decreases with increase of number
of composition data n. Consequently, it is obvious that the cfit(x) given by the least square formula
of error function has a higher accuracy than independently isolated data cm.

References

[1] Miyazaki T, Koyama T, Kobayashi S. A new characterization method of the microstructure using the macroscopic
composition gradient in alloys. Metall Mater Trans 1996;27A:949–54.
[2] Miyazaki Toru, Kobayashi Sengo, Koyama Toshiyuki. Determination of the critical nucleus size of precipitates using the
macroscopic composition gradient method. Metall Mater Trans A 1999;30A:2783–9.
[3] Miyazaki Toru. A new evaluation method of phase decomposition by utilizing the macroscopic composition gradient in
alloys. In: Koiwa K, Otsuka K, Miyazaki T, editors. Proc solid-solid phase transformation, 1999 PTM99, vol. 1. JIM Press;
1999. p. 15–22.
[4] Miyazaki Toru, Kozakai Takao, Schoen ClaudioG. Precipitate nucleation near the edge of miscibility gap. In: Howe JM et al.,
editors. Proc solid-to-solid phase transformation in inorganic materials, 2005 PTM05, vol. 2. TMS Press; 2005. p. 271–90.
[5] Kobayashi S, Sumi T, Koyama T, Miyazaki T. Determination of coherent precipitation lines utilizing the composition
gradient method. J Jpn Inst Metals 1996;60:22–8.
[6] Miyazaki Toru, Kobayashi Sengo. Evaluation of microstructures in alloys having a macroscopic composition gradient.
Philos Mag 2010;90:305–16.
[7] Cliff G, Lorimer GW. J Microsc 1975;103:203.
[8] Cahn JW, Hilliard JE. J Chem Phys 1958;28:258.
[9] Cahn JW. Acta Metall 1961;9:795.
[10] Mori T, Cheng PC, Kato M, Mura T. Acta Metall 1978;26:1435.
[11] Mura T. Micromechanics of defects in solids. Kluwer Academic Publishers; 1987.
[12] Miyazaki T, Koyama T. Proc inter conf solid–solid phase transformation in inorganic materials. Scripta Metall
1994;37:365–70.
[13] Miyazaki T, Imamura T, Kozakai T. Mater Sci Eng 1982;54:9–15.
[14] Carpenter RW. Acta Metall 1967;15:1567.
[15] Miyazaki Toru, Doi Minoru. Mater Sci Eng 1989;A110:175–85.
[16] Miyazaki T, Takagishi S, Mori H, Kozakai T. The phase decomposition of iron-molybdenum binary alloys by spinodal
mechanism. Acta Metall 1980;28:1143–53.
[17] Miyazaki T, Koyama T, Kozakai T. Mater Sci Eng 2001;A312:38–49.
[18] Toru Miyazaki. Recent developments and the future of computational science on microstructure formation. In: The forty-
seventh Honda memorial lecture materials transactions 2002;43(6)1266–72.
[19] Allen SM, Cahn JW. Acta Metall Mater 1979;27:1085–93.
[20] Moreen HA, Taggart R, Polonis RH. Metall Trans 1974;5:79.
[21] Pearson WB, Hume-Rothery W. J Inst Metals 1952;80:641.
[22] Heiwegen CP, Rieck GD. Z Metallkd 1973;64:450.
[23] Gust W, Nguyen-Tat T, Predel B. Z Metallkd 1979;70:241.
[24] Casselton REW, Hume-Rothery W. J Less-Common Metals 1964;7:212.
[25] Massalski TB. Binary alloy phase diagrams. 2nd ed. ASM International; 1990.
[26] Cahn JW. Acta Metall 1962;10:907.
[27] Williams RO. Metall Trans 1980;11A:247.
[28] Hornbogen E. Aluminium 1967;43:115.
[29] Kampmann R, Wagner R. Decomposition of alloys; early stages. In: Haasen P, Gerold V, Wagner R, Ashby MF,
editors. Oxford Pergamon Press; 1984. p. 91–103.
[30] Binder K. Stochastic nonlinear systems in physics, chemistry and biology. In: Arnold L, Lefever R, editors. Berlin: Springer;
1981. p. 62–72.
[31] Kozakai T, Aihara H, Doi M, Miyazaki T. Trans ISIJ 1985;25:161.
[32] Kubaschewski O, Chart TG. J Inst Metals 1964;93:329.
[33] Flewitt PEJ. Acta Metall 1974;22:47.
[34] Carpenter RW. Acta Metall 1967;15:1297.
[35] Buther EP, Thomas G. Acta Metall 1970;18:347.
[36] Miyazaki T, Murayama H, Mori H. Trans JIM 1977;18:697–706.
[37] Miyazaki T, Koyama T. Trans JIM 1998:169–78.
[38] Mirold P, Binder K. Acta Metall 1977;25:1435.
[39] Binder K, Billotet C, Mirold P. Z Phys 1978;B30:183.
[40] Kubaschewski O. Iron-banary phase diagrams. New York: Springer-Verlag; 1982.
[41] Oki K, Sagane H, Eguchi T. Proc order and disorder in solids, CNRS inter symposium, Paris, 1977. J Phys C
1977;7(Suppl.):414.
1060 T. Miyazaki / Progress in Materials Science 57 (2012) 1010–1060

[42] Mebed AM, Miyazaki T. A new concept of the A2/B2 phase transformations based upon recent experimental investigations.
In: Proc of the inter conf on solid-solid phase transformations, PTM’99(JIMIS-3); 1999. p. 61.
[43] Volmer M, Weber A. Z Physik Chem 1925;119:277.
[44] Becker R, Döring W. Ann Phys 1935;24:719.
[45] Becker R. Z Metallkd 1937;29:245.
[46] Becker R. Proc Phys Soc 1940;52:69.
[47] Borelius G. Ann Phys 1937;28:507.
[48] Borelius G. Ark Met Ast Fys 1945;32:1.
[49] Höbstetter JN. Trans Met Soc AIME 1949;180:121.
[50] Hobstetter N. Metals Technol 1948:2447.
[51] Cahn JW, Hilliard JE. J Chem Phys 1959;31:688.
[52] Wagner C. Z Electrochem 1961;65:581.
[53] Lifshitz IM, Slyozov VV. Phys Chem Solids 1961;19:35.
[54] Binder K, Stauffer D. Adv Phys 1976;25:343.
[55] Sundquist BE, Oriani RA. J Chem Phys 1962;36:2604.
[56] Heady RB, Cahn JW. J Chem Phys 1973;58:896.
[57] Howland RG, Wong NC, knobler CM. J Chem Phys 1980;73:522.
[58] Haasen P, Wagner R. Metall Trans A 1992;23A:1901.
[59] Aaronson HI, LeGoues FK. Metall Trans A 1992;23A [9115-1945].
[60] Khachaturyan AG. Theory of structural transformations in solids. New York: Dover Publications, Inc.; 2008.
[61] Wagner R, Kampmann R. Homogeneous second phase precipitation. In: Haasen P, et al., editors. Phase transformations in
materials, vol. 5. Weinheim (Germany): VCH; 1991.
[62] Darken LS. Diffusion in metal accompanied by phase change. Trans AIME 1942;150:157–70; Trans AIME 1948;175:184;
1949;180:430.
[63] Hultgren R, Desai PD, Hawkins DT, Gleiser M, Kelley KK. Selected values of the thermodynamic properties of binary
alloys. Metals Park (OH): American Society for Metals; 1961.
[64] Martin JW, Doherty RD. Stability of microstructure in metallic system. Cambridge (UK): Cambridge University Press; 1976.
[65] Ardell AJ, Nicholson RB, Acta Metall 1966;14:1295; J Phys Chem Solids 1996;27:1793.
[66] Wagner W. J Phys F: Metal Phys 1986;16:1239.
[67] Wendt H, Haasen P. Scripta Metall 1985;19:1053.
[68] Burke J. The kinetics phase transformations in metals. Pergamom Press; 1965.
[69] Pelleg J. Acta Metall 1966;14:229.
[70] Miyazaki Toru, Sekio Kazuhiro, Doi Minoru, Kozakai Takao. Mater Sci Eng 1986;77:125–32.
[71] Miyazaki Toru, Doi Minoru. Mater Sci Eng 1989;A110:175–83.
[72] Wagner W. Z Metallkd 1989;80:873.
[73] Jiang X, Wagner W, Wollenberger H. Z Metallkd 1989;80:873.
[74] Wagner W. Acta Metall 1990;38:2711.
[75] Rzdilsky I, Cerezo A, Smith GDW. Solidification 1998;1998:83.
[76] Blavet D, Deconihout B, Bostel A, Sarrau JM, Bouet M, Morrand A. Rev Sci Instrum 1993;64:2911.
[77] Zener C. J Appl Phys 1944;20:962.
[78] Zener C, Wert C. J Appl Phys 1945;21:5.
[79] Tsallis CT. In: Abe S, Okamoto Y, editors. Nonextensive statistical mechanics and its application. Springer; 2000. p. 3.
[80] Tsallis CT. J Stat Phys 1988;52:479.

You might also like