You are on page 1of 21

Int. J. Mater. Res.

2022; 113(5): 351–371

Liang Zhong, Yuling Liu*, Huixin Liu, Shiyi Wen, Fei Wang, Changfu Du, Qianhui Min,
Milena Premovic, Zhoushun Zheng, Jieqiong Hu and Yong Du*

Diffusivities and atomic mobilities in the Ni-rich


fcc Ni–Al–Cu alloys: experiment and modeling
https://doi.org/10.1515/ijmr-2021-8426
Received June 20, 2021; accepted November 25, 2021;
1 Introduction
published online April 28, 2022
As typical SMAs (shape memory alloys), Ni–Al–Cu alloys
Abstract: The diffusivities along the whole diffusion path have an interesting characteristic, viz., “remembering”
in one diffusion couple of Ni-rich fcc Ni–Al–Cu alloys
the shape they kept before pseudoplastic deformation.
at 1273, 1333 and 1433 K are obtained by means of the
Like other SMAs, Ni–Al–Cu alloys present outstanding
numerical inverse method. In total 12 diffusion couples
machinability [1], excellent stability of mechanical proper-
are used to obtain the diffusivities at the intersections
in the diffusion paths by the Matano–Kirkaldy method. ties at elevated temperatures [2], and high thermal stability
Two sets of atomic mobilities are assessed using the [3]. These properties at elevated temperatures applied
CALTPP program based on the diffusivities calculated in sensors, high damping material and actuators [4–6]
by the numerical inverse method and Matano–Kirkaldy can be remarkably improved by heat treatment, during
method, respectively. The validity of the obtained atomic which diffusion plays a key role. Therefore, a systematical
mobilities is confirmed by comparing the model-predicted study of the composition- and temperature-dependent
diffusivities, concentration profiles and diffusion paths interdiffusivities as well as the atomic mobilities for the
with the measured ones. The whole calculation by the
Ni–Al–Cu alloys is of fundamental importance for these
numerical inverse method in CALTPP with higher efficiency
applications.
can achieve the same accuracy as the Matano–Kirkaldy
The interdiffusivities of the Cu-rich Cu–Ni–Al alloys
method. The present study demonstrates that one diffusion
couple combined with the numerical inverse method can have been investigated according to just two semi-infinite
reliably assess the atomic mobilities for fcc Ni–Al–Cu and one finite diffusion couples at 1273 K by Liu et al. [7, 8].
alloys. More experimental investigations on the temperature- and
composition-dependent interdiffusivities and reliable dif-
Keywords: Atomic mobility; CALTPP program; Diffusivity;
Experimental measurements; Ni–Al–Cu alloy. fusion kinetic parameters are required for the fcc Ni–Al–Cu
alloys. Therefore, the Ni-rich fcc Ni–Al–Cu alloys are
chosen as the target to study the interdiffusivities in the
present work. It should be noted that after the present
authors established the atomic mobilities of the Ni-rich
fcc Ni–Al–Cu alloys, most recently the atomic mobilities
*Corresponding authors: Yuling Liu and Yong Du, State Key Lab- of the Cu-rich Cu–Ni–Al alloys were reported by Zhang
oratory of Powder Metallurgy, Central South University, Changsha, et al. [9]. They have performed eight diffusion couples of
Hunan, 410083, P. R. China, E-mail: liu.yuling@csu.edu.cn (Y. Liu). the Cu-rich Cu–Ni–Al alloys and established the kinetic
https://orcid.org/0000-0002-5898-567X (Y. Liu) and
parameters. However, the diffusion behavior in the Cu-rich
yong-du@csu.edu.cn (Y. Du). https://orcid.org/0000-0002-4187-
3989 (Y. Du) alloys is different from that in the Ni-rich alloys, which will
Liang Zhong, Huixin Liu, Shiyi Wen, Fei Wang and Qianhui Min, be demonstrated in Section 4. Therefore, it is necessary to
State Key Laboratory of Powder Metallurgy, Central South University,
study the diffusivities and atomic mobilities in the Ni-rich
Changsha, Hunan, P. R. China
Changfu Du and Zhoushun Zheng, School of Mathematics and fcc Ni–Al–Cu alloys by experiment and modeling.
Statistics, Central South University, Changsha, Hunan, P. R. China As is well-known, the traditional Matano–Kirkaldy
Milena Premovic, Faculty of Technical Science, University of
method is the frequent way to calculate the diffusivities in
Prishtina, Kosovska Mitrovica, Serbia
Jieqiong Hu, Sino-Precious Metals Holding Co., Ltd., Kun Ming, a ternary system [10–12]. Although the Matano–Kirkaldy
Yunnan, P. R. China method is of great reliability and accuracy, its efficiency
352 | L. Zhong et al.: Diffusivities and atomic mobilities in the Ni-rich fcc Ni–Al–Cu alloys

is low since two diffusion couples with an intersection Table 1: Terminal compositions of the diffusion couples in the
in the diffusion paths are indispensable to calculate the Ni-rich Ni–Al–Cu system.
diffusivities. To improve this situation, a novel numerical
inverse method was proposed by our group [13] which can Couple Composition (at.%)

obtain the composition-dependent diffusion matrix along C1 Ni/Ni-10.2Al-7.0Cu


the whole diffusion path of a single diffusion couple with C2 Ni/Ni-6.0Al-19.8Cu
a high efficiency. The novel numerical inverse method C3 Ni-5.2Al/Ni-21.1Cu
C4 Ni-9.4Al/Ni-10.1Cu
and Matano–Kirkaldy method were implemented in the
CALTPP (CALculation of ThermoPhysical Properties) [14].
CALTPP has been successfully utilized to calculate the
interdiffusivities and optimize the atomic mobilities of were ground to move surface contamination and polished
various alloys [13, 15–19]. to get a smooth surface, and then bound together with
In the present work, just one diffusion couple at three Mo clamps to form diffusion couples, as listed in Table 1.
temperatures is applied to obtain the interdiffusivities After that, each diffusion couple was sealed into evacuated
by the numerical inverse method in CALTPP. To confirm quartz tubes and annealed at 1273 K for 96 h, 1333 K for
the results from the numerical inverse method, three 36 h and 1433 K for 6 h, respectively, followed by quenching
new diffusion couples at each temperature are designed in cold water. After applying standard metallographic
to create intersections. Then the interdiffusivities of five techniques, the concentration profiles were obtained by
intersections in diffusion paths of all diffusion couples are means of electron probe microanalysis with wavelength
calculated using the Matano–Kirkaldy method. Based on dispersive X-ray spectroscopy (EPMA-WDX, JXA-8530F,
the obtained interdiffusivities, two sets of atomic mobilities JEOL, Japan; accelerating voltage 20 kV, probe current 2.0
are evaluated, separately. × 10−8 A) on a polished section, parallel to the diffusion
The major purposes of this work are (i) to measure direction. The software for EPMA measurements was
the interdiffusivities of Ni-rich fcc Ni–Al–Cu alloys along Phi-Rho-Z Quantitative Analysis Program developed by
the whole diffusion path of one diffusion couple using the JEOL with embedded ZAF method to perform corrections.
numerical inverse method and at intersections in diffusion Variations in alloy compositions were measured to be
paths of four diffusion couples using the Matano–Kirkaldy within ±0.5 at.% in the present work.
method, (ii) to evaluate two sets of atomic mobilities in
CALTPP based on the interdiffusivities obtained from the
numerical inverse method and Matano–Kirkaldy method, 3 Evaluation of ternary
respectively, and (iii) to demonstrate the reliability of two
sets of atomic mobilities by comprehensively comparing
indiffusivities and atomic
the calculated diffusivities, model-predicted concentration mobilities
profiles, and diffusion paths with the experimental ones
and validate the accuracy of the numerical inverse method.
3.1 Matano–Kirkaldy method for obtaining
ternary diffusivities
2 Experimental procedure
The Boltzmann–Matano method has been applied to cal-
Based on the Ni–Al–Cu isothermal sections at 1273, 1333 culate diffusion coefficients in ternary and even high-order
and 1433 K from Wang et al. [20], samples with the desired systems [10, 11]. Alternatively, with assumption of constant
compositions listed in Table 1 were prepared by arc melting molar volume over the diffusion zone, interdiffusion fluxes
of appropriate mixtures of pure Ni (99.99 wt.%), Cu (99.99 of component i can be represented in the view of Kirkaldy
wt.%) and Al (99.99 wt.%). The ingots were re-melted four [11] and Fick’s first law as follows:
times to ensure the homogeneity. Then the buttons were ci

linearly cut into blocks of approximate dimensions of 5 × 1


J̃i = (x − xM ) dci , (i = Al, Cu) (1)
5 × 3 mm3 by wire-electrode cutting. After mechanically 2t ∫
c−
i
or c+
i
removing the surface material, the buttons were sealed
into evacuated quartz tubes, and homogenized at 1433 ± 5 where c− i
and c+i
denote terminal concentrations at the
K for 10 days to obtain large grains and minimize the effect left and right side of diffusion couples, respectively. ci the
of grain-boundary diffusion. Subsequently, all the samples concentration of element i at the certain position, x the
L. Zhong et al.: Diffusivities and atomic mobilities in the Ni-rich fcc Ni–Al–Cu alloys | 353

distance and t the diffusion time. xM represents the Matano As regards the standard deviation of interdiffusivities,
plane for the diffusion couple. Lechelle et al. [23] introduced a scientific method to
Theoretically, the positions of Matano planes of calculate it, which is expressed as:
solutes Cu and Al should be the same. However, Matano

planes of different solutes deviate from each other to √ ( )2
√ ∑ 𝜕f
a certain extent generally. Therefore, Whittle and Green u ( f (A, B, …)) = √ (u (𝛼 ))
2
(5)
𝛼=A,B,…
𝜕𝛼
[21] introduced the normalized concentration parameter
( ) ( + )
Yi = ci − c−i
∕ ci − c−i to exclude the parameter xM from
calculation. After that, the following formula could be where A and B are the correlation quantities of function f ,
derived from Equation (1): while u (𝛼 ) (𝛼 = A, B, …) is the uncertainty in the measure-
ments of variable 𝛼 including concentration, interdiffusion
x
⎡ flux and the gradient of the composition.
̃ Ni
D ̃ Ni dcAl = − 1 dx ⎢(1 − YCu )
CuCu + DCuAl YCu dx The thermodynamic stability of the diffusivities is
dcCu 2t dYCu ⎢ ∫
⎣ −∞ examined by the following constraints [24]:
+∞

+ YCu (1 − YCu ) dx⎥ ̃Ni
D ̃Ni
CuCu + DAlAl > 0
∫ ⎥
x ⎦ ̃Ni ̃Ni ̃Ni ̃Ni
D CuCu DAlAl − DCuAl DAlCu ≥ 0 (6)
x
⎡ ( )2
̃ Ni + D̃ Ni dcAl 1 dx ⎢ ̃Ni ̃Ni ̃Ni D ̃Ni ≥ 0
D =− (1 − YAl ) Y dx D CuCu − DAlAl + 4D
AlAl CuAl dcCu 2t dYAl ⎢ ∫ Al CuAl AlCu
⎣ −∞
+∞

+ YAl (1 − YAl ) dx⎥ (2)
∫ ⎥ 3.2 Numerical inverse method for assessing
x ⎦
ternary diffusivities
where D ̃ Ni and D̃ Ni (i, j = Al, Cu) are the interdiffusion
iCu iAl As mentioned before, the Boltzmann–Matano method can
coefficients. The coefficients D ̃ Ni and D̃ Ni are the
CuCu AlAl only obtain the interdiffusivity at intersection points. Con-
main coefficients, which represent the influences of the
sequently, the numerical inverse method was developed
concentration gradients of elements Cu and Al on their
[13, 14, 25] to calculate interdiffusivity more efficiently.
own fluxes, respectively, while the cross interdiffusion
In the Ni–Al–Cu ternary system, the interdiffusivity
coefficients D̃ Ni and D̃ Ni reflect the influences of the
AlCu CuAl can be described by Equation (7) on the basis of the
concentration gradients of elements Cu and Al on the
extended Fick’s law [26].
fluxes of elements Al and Cu, respectively.
The interdiffusion flux of component i (i = Al, Cu) in ( ) ( )
𝜕 ci 𝜕 ̃Ni 𝜕 cCu 𝜕 ̃Ni 𝜕 cAl
the Ni–Al–Cu ternary alloy can be expressed as follows = DiCu + DiCu , (i = Al, Cu)
𝜕t 𝜕x 𝜕x 𝜕x 𝜕x
[22]: (7)
+
c− − cAl
J̃Al (z∗ ) = Al
̃Ni (i, j = Al, Cu) can be obtained
The interdiffusivity D
2t ij
z ∗
+∞ by the following formula [13]:
⎡( ) ⎤
= ⎢ 1 − YAl
∗ ∗
YAl dz + YAl (1 − YAl ) dz⎥ (3) ∑
3 ( )
⎢ ∫ ∫ ⎥ Φk
⎣ −∞ z∗ ⎦ ̃Ni =
D (𝛿ki − ci ) exp 𝜙Ni , (i, j = A1, Cu) (8)
ij RT kj
k=1
c− − c+Cu
J̃Cu (z∗ ) = Cu
2t where Φk and 𝜙Ni kj
denote the atomic mobility parameter
z∗
+∞ and thermodynamic factor, respectively. R is the gas
⎡( ) ⎤
⎢ constant, T temperature and 𝛿 ki the Kronecker delta (𝛿 ki

= 1 − YCu ∗
YCu dz + YCu (1 − YCu ) dz⎥ (4)
⎢ ∫ ∫ ⎥ = 1 if k = i, otherwise 𝛿 ki = 0). The numerical procedure
⎣ −∞ z∗ ⎦
is carried out by the integration of the finite element
where t is the diffusion annealing time, the normalized method (FEM) [27] and simple genetic algorithm (SGA) [28].
composition variable Yi∗ of element i (i = Al, Cu) at section FEM and SGA are utilized to solve forward problems and
ck (z∗ )−c−
z∗ is the value of c+ −c−k
k
. optimize parameters, respectively.
k
354 | L. Zhong et al.: Diffusivities and atomic mobilities in the Ni-rich fcc Ni–Al–Cu alloys

3.3 Modeling of atomic mobility and were designed to locate in a single fcc Ni-rich Ni–Al–Cu
diffusivity phase region. The concentration profiles perpendicular
to the interface direction for all diffusion couples were
Based on the absolute-reaction rate theory [29], the atomic acquired by EPMA-WDX. Only the experimental concentra-
mobility of element k expressed as M k could be described tion profiles of C1 diffusion couple at three temperatures
with frequency factor Mk0 and activation enthalpy Qk . were taken directly to calculate the interdiffusivities along
Andersson and Ågren [30] reported that atomic mobility the diffusion path using the numerical inverse method in
can be shown as Equation (9). CALTPP.
( ) ( ) For calculating the interdiffusivities by the Matano–
RTlnMk0 −Qk 1
Mk = exp exp (9) Kirkaldy method, the Boltzmann or additive Boltzmann
RT RT RT
function is utilized to fit all the measured concentration
Generally, Qk and RT ln Mk0 can be merged into profiles of C1, C2, C3 and C4 diffusion couples which is
one composition, temperature and pressure-dependent expressed as:
parameter, i.e., Φk . In consideration of the CAL-
p p
PHAD (CALculation of PHAse Diagram) method, the ci = p1 + (( 2 ) )+ ( 5 )
1 + exp x − p3 ∕ p4 1 + exp (x − p6 )∕ p7
composition dependency can be represented by the
(13)
Redlich–Kister–Muggianu polynomial [31, 32] as follows:
where p1 –p7 are the parameters to be fitted. While for
[ ]
∑ i
∑∑ ∑r i, j ( )r relatively symmetrical curves, only p1 –p4 are to be fitted.
Φ k = xi Φ k + xi x j Φ k xi − x j Note that the composition profiles for the solvent Ni are
i i j>i r
[ ] obtained by 1–c(Al)–c(Cu). The fitting results are listed
∑∑∑ ∑ i, j,l in Table 2. Here, the estimation for the error of the main
r s
+ xi x j xl vi jl Φk , (s = i, j, l) (10) ̃Ni at the intersection of diffusion
i j>i l> j i
diffusion coefficients D A1A1
couples C1 and C3 is given in detail as an example.
i, j i, j,l
where Φik is the value of Φk for pure i. r Φk and r Φk are According to the Matano–Kirkaldy method, D ̃Ni can be
A1A1
interaction parameters for the mobility of element k, which calculated using the following function:
could be represented by a polynomial of temperature and
( ) ( )
pressure if necessary. The parameter visjl is expressed as ̃J C3 𝜕 cCu |
− ̃J C1 𝜕 cCu |
|
𝜕 z |C1
|
𝜕 z |C3
follows: ̃Ni = (
D )(
Al
)
Al
( )( ) (14)
𝜕 cAl | 𝜕 cCu | 𝜕 cAl | 𝜕 cCu |
− ̃J C1
AlAl
1 − xi − x j − xl |
𝜕 z |C1
|
𝜕 z |C3 Al
|
𝜕 z |C3
|
𝜕 z |C1
visjl = xs + , (s = i, j, l) (11)
3
where xs , xi , x j and xl are mole fractions of elements s, i, j where ̃J C3 Al
and ̃J C1
Al
are the fluxes of Al in C3 diffusion
and l, respectively. couple and C1 diffusion couple at the intersection ( ) of
|
The relationship between atomic mobility M i and diffusion couples A1 and A3, respectively; 𝜕𝜕czA1 | and
( ) |C1
interdiffusion coefficients D ̃n is described with the follow- 𝜕 cCu |
kj |
𝜕 z |C1
are the concentration gradients of Al and Cu
ing equation [33]:
in C1 diffusion couple at the intersection( ) of ( diffusion )
( ) | |
∑n
𝜕𝜇i 𝜕𝜇i couples C1 and C3, respectively; 𝜕𝜕czA1 | and 𝜕𝜕czCu |
̃n =
D 𝛿
( ik − x x M − (12) |C3 |C3
k) i i
kj
i=1
𝜕 x j 𝜕 xn are the concentration gradient of Al and Cu element
of C3 diffusion couple at the intersection of diffusion
where xi and 𝜇 i denote the mole fraction and chemical couples C1 and C3, respectively. ( Therefore,
) the error of the
potential of element i, respectively, ̃
main diffusion coefficient, u DA1A1 contains the contri-
Ni

butions of individual variables and could be expressed as


4 Results and discussion follows:

( )2 ( )2
( )2 ̃ ( )2 ̃ ( )2
̃Ni 𝜕D 𝜕D
u D = u ̃J C3 + u ̃J C1
4.1 Interdiffusion coefficients at 1273, 1333 AlAl ̃
𝜕 J C3 Al ̃
𝜕 J C1 Al
Al Al
and 1433 K 2
⎡ ( )2
𝜕D̃ ⎤ 𝜕 cAl ||
Based on the isothermal sections from Reference [20], + ⎢( | )⎥ u
⎢ 𝜕 cAl | ⎥ 𝜕 z ||C1
the terminal compositions of all the diffusion couples ⎣ 𝜕 z |C1 ⎦
L. Zhong et al.: Diffusivities and atomic mobilities in the Ni-rich fcc Ni–Al–Cu alloys | 355

Table 2: The fitting results for the experimentally measured composition profiles by the Boltzmann function.

Couple p1 p2 p3 p4 p5 p6 p7

A1-Al 0 4.265 329.1 27.96 5.986 382.7 16.09


A1-Cu 0.0389 6.961 360.8 16.69 / / /
A2-Al 0.0140 2.259 254.8 25.94 3.569 316.4 11.53
A2-Cu 0 19.80 294.1 13.39 / / /
A3-Al 5.095 −3.187 371.2 18.26 −1.905 323.3 27.00
A3-Cu 0 21.25 353.8 16.60 / / /
A4-Al 9.370 −9.370 354.1 18.22 / / /
A4-Cu 0.0323 10.22 352.8 12.24 / / /
B1-Al 10.40 −5.526 388.5 26.82 −4.864 337.9 13.38
B1-Cu 6.960 −6.921 362.1 16.52 / / /
B2-Al 0 4.209 363.1 14.36 1.930 278.8 27.20
B2-Cu 0.0037 19.81 337.6 15.948 / / /
B3-Al 5.121 5.080 313.2 28.96 / / /
B3-Cu 0 21.13 307.5 19.36 / / /
B4-Al 0 9.426 350.9 29.51 / / /
B4-Cu 10.16 −10.09 350.2 18.26 / / /
C1-Al 10.16 −4.848 363.0 29.01 −5.306 309.4 15.30
C1-Cu 7.011 −6.976 332.9 17.01 / / /
C2-Al 6.204 −2.504 389.6 29.97 −3.685 318.2 13.55
C2-Cu 19.38 −19.38 346.4 16.38 / / /
C3-Al 0 2.118 375.4 22.75 3.216 330.2 16.20
C3-Cu 20.12 −20.12 350.1 17.03 / / /
C4-Al 0 7.080 360.1 21.32 2.475 313.1 22.33
C4-Cu 10.02 −9.980 352.2 17.19 / / /

( ) to be zero (i.e., u (t) = 0, u (z) = 0).


2
⎡ ( )2 two terms are assumed
𝜕D̃ ⎤ 𝜕 cCu ||
+ ⎢( | )⎥ u Consequently, u ̃J is estimated as follows:
⎢ 𝜕 cCu | ⎥ 𝜕 z ||C1
⎣ 𝜕 z |C1 ⎦
( ) 2 ( ̃ )2
𝜕J
u ̃J =
2
⎡ ⎤ ( )2 u (c)2 (18)
𝜕D̃ ⎥ 𝜕 cAl || 𝜕c
+ ⎢( ) u
⎢ 𝜕 cAl || ⎥ 𝜕 z ||C3
⎣ 𝜕 z |C3 ⎦ where u (c) is the uncertainty of concentration, which
2 is associated with the error in EPMA measurement and
⎡ ( )2
⎢ 𝜕D̃ ⎤⎥ 𝜕 cCu || fitting function. In this work, the double iterations of
+ ( ) u (15)
⎢ 𝜕 cCu || ⎥ 𝜕 z ||C3 Boltzmann function Equation (13) were used to smooth
⎣ 𝜕 z |C3 ⎦
the experimental data, so the error calculation for fitting
()
concentration is given by the following equation:
where u ̃J is the uncertainty of flux at the intersection
( )2
point. According to Equations (3)–(5), ̃J (z∗ ) and u ̃J can c = c (z, p1 , p2 , … , pn )
n ( )2
be calculated as follows: ∑ 𝜕c (19)
u (c)2 = 𝜆 ⋅ ⋅ u ( pk )2
z ∗ +∞ 𝜕 p k
− − c+ ⎡( ⎤ k=1
c )
̃J (z ) =
∗ ⎢ 1−Y ∗
Y dz + Y ∗
(1 − Y ) dz ⎥
2t ⎢ k ∫ k k∫ k
⎥ where the parameter 𝜆 is estimated as 1.02 in view of
⎣ −∞ z ∗ ⎦
composition measurement error (about 0.02) from EPMA.
(16)
According to present calculations, the uncertainty of calcu-
( )2 ( ̃ )2 ( )2 ( )2 lating the ternary interdiffusion coefficients are estimated
𝜕J 𝜕̃J 𝜕̃J
u ̃J = u (c)2 + u (t)2 + u (z)2 as: 60% uncertainty from interdiffusion flux and 40%
𝜕c 𝜕t 𝜕z
(17) uncertainty from concentration.
Since the diffusion annealing time t and diffusion Then, the fitted results are then applied to obtain
̃Ni , D
D ̃Ni , D
̃Ni and D ̃Ni at five intersections in diffu-
distance z can be measured very accurately, the errors for CuCu A1A1 CuA1 A1Cu
356 | L. Zhong et al.: Diffusivities and atomic mobilities in the Ni-rich fcc Ni–Al–Cu alloys

sion paths of all the diffusion couples at each temperature 4.2 Assessment of atomic mobilities
by the Matano–Kirkaldy method. The obtained diffusivi-
ties in fcc Ni–Al–Cu alloys together with their standard The thermodynamic descriptions of the Ni–Cu, Cu–Al
error calculated by Equation (5) are listed in Table 3. and Ni–Al binary systems were obtained from Miettinen
To make a quantitative comparison, the diffusivities of [34], Chen et al. [35] and Ansara et al. [36], respectively,
the same compositions obtained by the numerical inverse and the Ni–Al–Cu ternary system was reported by Wang
method are also listed. As shown in the table, both negative et al. [20]. These thermodynamic parameters are applied in
and positive values for cross interdiffusivities can be found, this work. The adopted atomic mobility parameters of fcc
which indicates an attractive and a repelling interaction Cu–Ni, Cu–Al and Ni–Al binary systems are taken from
between Cu and Al atoms, respectively. Furthermore, Zhang et al. [37], Liu et al. [38] and Zhang et al. [39],
all these diffusivities meet the thermodynamically stable respectively.
constraints i.e., Equation (6), and thus they are thermody- The optimization of the ternary atomic mobilities
namically stable. Moreover, the results calculated by the for fcc Ni–Al–Cu alloys was carried out using CALTPP
Matano–Kirkaldy method listed in Table 3 show that the [13, 14] using the presently obtained concentration pro-
value of the main interdiffusion coefficients D ̃Ni is larger files or interdiffusivities combined with binary kinetic
A1A1
than D̃Ni at each temperature, indicating that Al diffuses parameters from References [37–39]. The atomic mobil-
CuCu
faster than Cu in Ni-rich Ni–Al–Cu alloys. ities of version 1 were obtained during the calculation
Table 3: Diffusion coefficients in fcc Ni–Al–Cu alloys obtained by the Matano–Kirkaldy method (denoted by ‘‘M-K’’) and numerical inverse
method (denoted by ‘‘N-I’’).

Temperature (K) Diffusion couple Composition (at.%) Interdiffusion coefficient ( ×10−15 m2 s−1 )a
Cu Al ̃Ni (SD)
D ̃Ni (SD)
D ̃Ni (SD)
D ̃Ni (SD)
D method
CuCu CuA1 A1Cu A1A1

1273 C1/C3 1.92 4.20 1.40 ± 0.04 0.16 ± 0.37 0.035 ± 1.14 2.26 ± 0.02 M-K
1.10 0.07 0.30 2.23 N-I
C1/C4 3.32 5.63 0.83 ± 0.01 0.32 ± 0.03 0.61 ± 0.04 2.27 ± 0.01 M-K
1.03 0.10 0.42 2.63 N-I
C2/C3 6.57 3.01 0.96 ± 0.01 1.04 ± 0.08 0.022 ± 1.40 2.22 ± 0.07 M-K
1.31 0.29 0.22 2.36 N-I
C2/C4 7.40 3.16 0.71 ± 0.01 0.12 ± 0.14 0.12 ± 0.16 1.65 ± 0.02 M-K
1.32 0.32 0.22 2.45 N-I
C3/C4 8.05 2.74 1.26 ± 0.02 0.84 ± 0.07 0.23 ± 0.05 1.14 ± 0.03 M-K
1.38 0.37 0.20 2.43 N-I
1333 C1/C3 1.91 4.07 4.84 ± 0.05 −1.31 ± 0.18 −1.05 ± 0.12 6.68 ± 0.02 M-K
3.43 −0.21 −0.92 6.84 N-I
C1/C4 3.52 5.71 3.34 ± 0.01 0.52 ± 0.09 0.26 ± 0.54 9.00 ± 0.02 M-K
3.20 0.32 1.33 8.30 N-I
C2/C3 6.07 3.09 3.60 ± 0.01 −3.23 ± 0.07 0.20 ± 0.32 7.98 ± 0.04 M-K
3.94 −0.80 0.71 7.21 N-I
C2/C4 7.40 3.38 3.00 ± 0.01 1.16 ± 0.08 0.18 ± 0.80 9.24 ± 0.02 M-K
3.96 0.96 0.79 7.66 N-I
C3/C4 8.36 2.69 4.05 ± 0.01 0.029 ± 7.96 0.18 ± 0.30 10.03 ± 0.03 M-K
4.22 1.14 0.62 7.47 N-I
1433 C1/C3 2.17 4.32 22.67 ± 0.03 −2.83 ± 0.25 3.88 ± 0.12 37.68 ± 0.02 M-K
17.97 −1.22 5.54 38.06 N-I
C1/C4 3.45 5.58 19.05 ± 0.01 −0.58 ± 0.29 3.93 ± 0.16 48.64 ± 0.02 M-K
17.08 −1.62 7.02 43.44 N-I
C2/C3 6.06 3.13 18.83 ± 0.01 −3.38 ± 0.27 1.96 ± 0.26 39.27 ± 0.05 M-K
20.52 −4.09 3.93 38.08 N-I
C2/C4 6.87 3.31 18.00 ± 0.01 −2.29 ± 0.16 2.40 ± 0.23 41.59 ± 0.02 M-K
20.50 4.54 4.18 39.37 N-I
C3/C4 7.54 2.80 18.51 ± 0.02 −2.40 ± 0.28 1.31 ± 0.26 39.70 ± 0.02 M-K
21.41 5.25 3.51 38.64 N-I

a SD = standard deviation, which was evaluated using the method considering the error propagation [23].
L. Zhong et al.: Diffusivities and atomic mobilities in the Ni-rich fcc Ni–Al–Cu alloys | 357

of interdiffusivities using the measured concentration manifests a consistency between the diffusivities calcu-
profiles of C1 diffusion couple at three temperatures, as lated by the numerical inverse method and the experimen-
expressed in Equation (8). It should be noted that the tal ones. From the figure, it can be inferred easily that
calculation speed for CALTPP is very fast. For the three ̃Ni and D
the values of main interdiffusivities D ̃Ni are
CuCu A1A1
diffusion couples in the present work, it only takes 40 s affected by temperature and concentration. With tempera-
for the calculation with a basic desktop computer (CPU: ̃Ni and D
ture rising, both D ̃Ni increase. Furthermore,
CuCu A1A1
Intel Core i7-8700HQ). The algorithms for the optimization the increases of Cu and Al concentrations contribute
of atomic mobilities based on the interdiffusivities in to the slight decreases of D ̃Ni and the increases of
CuCu
CALTPP contain two types: deterministic algorithm and ̃Ni in general. The comparisons between the calculated
D A1A1
heuristic algorithm, such as the Levenberg–Marquardt main diffusivities based on the two versions of atomic
method and Simulated Annealing algorithm. Principally, mobilities and the ones measured by the Matano–Kirkaldy
the Levenberg–Marquardt method is advisable to adopt method in fcc Ni–Al–Cu alloys at 1273, 1333 and 1433 K
first [15] and was utilized in the present work to optimize are shown in Figure 2. Dashed lines refer to the gen-
the atomic mobilities of version 2. In addition, the initial eral acceptable deviation of model-predicted diffusivities,
parameters to be optimized and the iteration stop criteria i.e., with a factor of 2 or 0.5. As illustrated in the
are automatically selected in CALTPP. Using the above algo- figure, the calculated main diffusivities D ̃Ni and D ̃Ni
CuCu A1A1
rithms, CALTPP can obtain the optimal atomic mobility from two versions can fit well with the experimental
values automatically. The summary of the two versions ones.
of atomic mobility parameters for fcc Ni–Al–Cu alloys is The comparisons between the model-predicted con-
shown in Table 4. centration profiles using the present two versions of
With the aim of confirming the reliability of the kinetic parameters and the measured ones are illustrated.
obtained atomic mobilities, the application to predict Figures 3–5 exhibit the model-predicted concentration
the diffusion characteristics is further performed by the profiles of Cu and Al for fcc Ni–Al–Cu alloys, compared
CALTPP program. Figure 1 shows the interdiffusivity with the measured data for the diffusion couples annealed
planes at 1273, 1333, 1433 K calculated by the numerical at 1273 K for 96 h, 1333 K for 36 h, 1433 K for 6 h,
inverse method. The interdiffusivities at the intersections respectively. The solid and dashed lines represent the
in diffusion paths obtained through the Matano–Kirkaldy calculated results utilizing the kinetic parameters from the
method are overlaid by symbols in Figure 1, which version 1 and version 2, respectively. As revealed by these
Table 4: Summary of the atomic mobility parameters for fcc Ni–Al–Cu alloys assessed in the present work as well as those from the
literatures (all in SI units).

Mobility Parameters (J mol−1 ) Reference

ΦNi
Ni
= −271 377.6 − 81.79∗ T [39]
ΦNi = −229 936.8 − 78.83∗ T
Cu
[37]

Mobility of Ni ΦA1
Ni = 144 600.0 − 64.85 T [39]
,Cu ∗
ΦNi
Ni
= 39 620.8 − 24.19 T [37]
,A1
ΦNi
Ni
= − 29 571.8 [39]

ΦA1
A1 = −123 111.6 − 97.34 T [39]

ΦA1 = −181 583.4 − 99.8 T
Cu
[38]

ΦNi
A1 = −268 381.0 − 71.04 T [39]
,Cu ∗
Mobility of Al ΦA1
A1
= − 183 094.3 − 159.04 T [38]
A1,Ni ∗
ΦA1 = −308 067.5 − 111.52 T [39]
,Cu
ΦNi
A1
= −32 688.2 This work (version 1)
,Cu
ΦNi
A1
= −45 329.1 This work (version 2)
ΦCu
Cu = −205 872.0 − 82.52∗ T [38]
ΦCu = −263 689.7 − 77.04∗ T
Ni
[37]

ΦA1
Cu = −133 184.4 − 83.65 T [38]
,Ni ∗
Mobility of Cu ΦCu
Cu
= 14 204.2 − 4.98 T [37]
,A1
ΦCu
Cu
= −31 461.4 + 78.91∗ T [38]
Ni,A1
ΦCu = −207 745.5 This work (version 1)
,A1
ΦNi
Cu
= −226 583.1 This work (version 2)
358 | L. Zhong et al.: Diffusivities and atomic mobilities in the Ni-rich fcc Ni–Al–Cu alloys

Figure 1: Interdiffusivity planes predicted by the numerical inverse method (denoted as surface) at (a) 1273 K, (b) 1333 K and (c) 1433 K
compared with the results obtained by the Matano–Kirkaldy method (denoted in symbols).

figures, the model-predicted concentration profiles from of Al and Cu atoms. From the figure, it can be concluded
version 1 and version 2 present nearly the same profiles that two sets of the assessed atomic mobilities of Ni–Al–Cu
and agree well with the corresponding measured ones. alloys are quite reliable since the model-predicted diffusion
Also, all of the concentration profiles of Al and Cu have an paths can attach a general agreement with the measured
“S” shape and symmetry is obvious to be seen in Cu and Al ones.
diffusion. The diffusion distances of Al are about 100 μm As mentioned in Section 1, the diffusion behaviors in
longer than the ones of Cu. Cu-rich Cu–Ni–Al alloys at 1073, 1173, 1273 K were studied
Figure 6 illustrates the model-predicted diffusion and the corresponding atomic mobilities were reported by
paths utilizing presently obtained parameters at 1273, 1333 Zhang et al. [9]. The individual experimental investigations
and 1433 K, compared with the experimental ones. Almost at the common temperature of 1273 K are performed in the
all the diffusion paths are observed as the serpentine present work focusing on Ni-rich alloy and in Reference [9]
shape, which results from the difference in diffusion rates on Cu-rich alloy. To further analyze the differences in the

Figure 2: Comparisons between the calculated main interdiffusivities using the kinetic parameters of the present (a) version 1 (based on the
numerical inverse method), and (b) version 2 (based on the Matano–Kirkaldy method) and the measured ones by the Matano–Kirkaldy
method in fcc Ni–Al–Cu alloys at 1273, 1333 and 1433 K. Dashed lines refer to the diffusivities with a factor of 2 or 0.5 from the
model-predicted diffusivities.
L. Zhong et al.: Diffusivities and atomic mobilities in the Ni-rich fcc Ni–Al–Cu alloys | 359

Figure 3: Comparisons between the model-predicted concentration profiles using the kinetic parameters of the present version 1 (based on
the numerical inverse method) and version 2 (based on the Matano–Kirkaldy method) in diffusion couples of (a) Ni/Ni-10.2Al-7.0Cu, (b)
Ni/Ni-6.0Al-19.8Cu, (c) Ni-5.2Al/Ni-21.1Cu, and (d) Ni-9.4Al/Ni-10.1Cu (at.%) annealed at 1273 K for 96 h, compared with the measured data
in this work.

diffusion behaviors and atomic mobilities of Cu-rich and the atomic mobilities from the present work and Reference
Ni-rich alloys, the model-predicted concentration profiles [9], together with the measured data in the Cu-rich alloys
at 1273 K together with the measured data are displayed in from Reference [9] annealed at 1273 K for 48 h. It is obvious
Figures 7 and 8. Figure 7 presents the concentration profiles that the concentration profiles simulated by the atomic
predicted by the atomic mobilities from the present work
mobilities from Reference [9] can fit the experimental data
and Reference [9], together with the presently measured
well for Cu-rich alloys since the experimental data are used
data in Ni-rich alloys annealed at 1273 K for 96 h. It is obvi-
in the optimization of the atomic mobilities. It is noted
ous that the concentration profiles simulated by the atomic
that the presently obtained atomic mobility parameters in
mobilities from the present work can fit the experimental
data well for Ni-rich alloys. The concentration profiles Ni-rich alloys can predict the concentration profiles in Cu-
calculated by the atomic mobilities from Reference [9] lead rich alloys reasonably although the experimental data in
to noticeable discrepancies from the experimental ones. Cu-rich side from Reference [9] are not used in the present
Figure 8 presents the concentration profiles predicted by optimization.
360 | L. Zhong et al.: Diffusivities and atomic mobilities in the Ni-rich fcc Ni–Al–Cu alloys

Figure 4: Comparisons between the model-predicted concentration profiles using the kinetic parameters of the present version 1 (based on
the numerical inverse method) and version 2 (based on the Matano–Kirkaldy method) in diffusion couples of (a) Ni/Ni-10.2Al-7.0Cu, (b)
Ni/Ni-6.0Al-19.8Cu, (c) Ni-5.2Al/Ni-21.1Cu, and (d) Ni-9.4Al/Ni-10.1Cu (at.%) annealed at 1333 K for 36 h, compared with the measured data
in this work.

As shown in Figures 7 and 8, the concentration profiles regarding the establishment of the universal atomic mobil-
in the Ni-rich alloys show symmetrical characteristics, ity parameters in the Cu-rich and Ni-rich alloys will be
while the concentration profiles in the Cu-rich alloys proposed.
display the complex feature that obvious uphill diffusion In summary, comparing the diffusion behaviors, viz.,
appears in the concentration profile of Al. The different main interdiffusivities, concentration profiles and diffu-
diffusion characteristics in the Ni-rich and Cu-rich alloys sion paths adopting two versions of atomic mobility
lead to the incompatibility of the established atomic parameters with the experimental ones, it can be con-
mobilities of different side alloys. Therefore, it is necessary cluded that these atomic mobility parameters obtained
to study the diffusivities and atomic mobilities in the through both methods can predict the diffusion behav-
Ni-rich fcc Ni–Al–Cu alloys. It is worth mentioning that the iors of fcc Ni–Al–Cu alloys well. What is more, the
diffusion experiment of four diffusion couples annealed diffusion characteristics predicted through the numerical
at three temperatures for Cu-rich Ni–Al–Cu alloys has inverse method are nearly the same as those predicted
been finished by the present authors and further work by the Matano–Kirkaldy method. However, the numerical
L. Zhong et al.: Diffusivities and atomic mobilities in the Ni-rich fcc Ni–Al–Cu alloys | 361

Figure 5: Comparisons between the model-predicted concentration profiles using the kinetic parameters of the present version 1 (based on
the numerical inverse method) and version 2 (based on the Matano–Kirkaldy method) in diffusion couples of (a) Ni/Ni-10.2Al-7.0Cu,
(b) Ni/Ni-6.0Al-19.8Cu, (c) Ni-5.2Al/Ni-21.1Cu, and (d) Ni-9.4Al/Ni-10.1Cu (at.%) annealed at 1433 K for 6 h, compared with the measured data
in this work.

inverse method possesses two advantages in comparison inverse method is that the intersection in diffusion paths is
with the Matano–Kirkaldy method. One advantage is not a necessity in its application and the atomic mobilities
that the numerical inverse method does not need to thus obtained utilizing the experimental concentration
fit the concentration profiles with functions. Regarding profile of just one diffusion couple can reproduce the
the Matano–Kirkaldy method, the adoptions of different accurate diffusion behaviors of the other three diffusion
fitting functions of concentration profiles may result in couples in the present work. It is noted that the choice of
different diffusion coefficients and thereby influence the C1 diffusion couple is arbitrary and the optimized results
establishment of atomic mobilities. On the other hand, the of the other three diffusion couples are presented in the
numerical inverse method is devoted to applying the exper- Appendix. Thus, fewer diffusion couples combined with
imental concentration profiles directly to the assessment the numerical inverse method can reliably determine the
of interdiffusivities and atomic mobilities, without fitting atomic mobilities for fcc Ni–Al–Cu alloys in an effective
periods. Another advantage associated with the numerical way.
362 | L. Zhong et al.: Diffusivities and atomic mobilities in the Ni-rich fcc Ni–Al–Cu alloys

Figure 6: Comparisons between the model-predicted diffusion paths using the kinetic parameters of the present version 1 (based on the
numerical inverse method) and version 2 (based on the Matano–Kirkaldy method) in fcc Ni–Al–Cu alloys annealed at (a) 1273 K for 96 h,
(b) 1333 K for 36 h, and (c) 1433 K for 6 h, compared with the experimental data in this work.

Figure 7: Comparisons between the model-predicted concentration profiles using the kinetic parameters of the present version 1 (denoted in
solid line) and in Reference [9] (denoted in dashed line) in diffusion couples of (a) Ni/Ni-10.2Al-7.0Cu, (b) Ni/Ni-6.0Al-19.8Cu,
(c) Ni-5.2Al/Ni-21.1Cu, and (d) Ni-9.4Al/Ni-10.1Cu (at.%) annealed at 1273 K for 96 h, compared with the measured data in this work.
L. Zhong et al.: Diffusivities and atomic mobilities in the Ni-rich fcc Ni–Al–Cu alloys | 363

Figure 8: Comparisons between the model-predicted concentration profiles using the kinetic parameters of the present version 1 (denoted in
solid line) and in Reference [9] (denoted in dashed line) in diffusion couples of (a) Cu-5.07Al/Cu-16.53Ni, (b) Cu-5.56Al/Cu-7.62Al-8.92Ni,
(c) Cu/Cu-3.83Al-21.57Ni (at.%) annealed at 1273 K for 48 h, compared with the measured data in Reference [9].

5 Conclusions – Two sets of atomic mobilities were obtained through


an intelligent model using the diffusivities from the
In the present work, a total of 12 diffusion couples of fcc numerical inverse method and the Matano–Kirkaldy
Ni–Al–Cu alloys were prepared and the concentration method implemented in CALTPP.
profiles after annealing at 1273, 1333 and 1433 K were – The validity of the present two versions of atomic
measured by EPMA-WDX. The main conclusions are as mobilities was confirmed by comparing the assessed
follows: diffusion characteristics including diffusivities, con-
– The numerical inverse method and the Matano– centration profiles and diffusion paths of four diffu-
Kirkaldy method were utilized to obtain the diffusion sion couples with the measured ones, which showed
coefficients along the C1 diffusion path and at the good agreement. These diffusion behaviors predicted
intersections of all the diffusion couples at each through the numerical inverse method and the
temperature, respectively. Matano–Kirkaldy method achieved good consistency.
364 | L. Zhong et al.: Diffusivities and atomic mobilities in the Ni-rich fcc Ni–Al–Cu alloys

The numerical inverse method combined with fewer Province 2019ZE001-1 and the Research Fund for foreign
diffusion couples can obtain accurate diffusion coef- young scholars of National Natural Science Foundation of
ficients and thus reliably assess the atomic mobility China (No. 51950410600) are gratefully acknowledged.
parameters for the Ni-rich fcc Ni–Al–Cu alloys. Conflict of interest statement: The authors declare no
conflicts of interest regarding this article.
Author contribution: All the authors have accepted respon-
sibility for the entire content of this submitted manuscript
and approved submission. Appendix
Research funding: The financial support by grant from
Genetic Engineering Project of Materials in Yunnan Figures A1–A8

Figure A1: Comparisons between the model-predicted concentration profiles using the kinetic parameters of the Model 1, Model 2, Model 3
and Model 4 in diffusion couples of (a) Ni/Ni-10.2Al-7.0Cu, (b) Ni/Ni-6.0Al-19.8Cu, (c) Ni-5.2Al/Ni-21.1Cu, and (d) Ni-9.4Al/Ni-10.1Cu (at.%)
annealed at 1273 K for 96 h, compared with the measured data in this work.
L. Zhong et al.: Diffusivities and atomic mobilities in the Ni-rich fcc Ni–Al–Cu alloys | 365

Figure A2: Comparisons between the model-predicted concentration profiles using the kinetic parameters of the Model 1, Model 2, Model 3
and Model 4 in diffusion couples of (a) Ni/Ni–10.2Al–7.0Cu, (b) Ni/Ni–6.0Al–19.8Cu, (c) Ni–5.2Al/Ni–21.1Cu, and (d) Ni–9.4Al/Ni–10.1Cu
(at.%) annealed at 1333 K for 36 h, compared with the measured data in this work.
366 | L. Zhong et al.: Diffusivities and atomic mobilities in the Ni-rich fcc Ni–Al–Cu alloys

Figure A3: Comparisons between the model-predicted concentration profiles using the kinetic parameters of the Model 1, Model 2, Model 3
and Model 4 in diffusion couples of (a) Ni/Ni–10.2Al–7.0Cu, (b) Ni/Ni–6.0Al–19.8Cu, (c) Ni–5.2Al/Ni–21.1Cu, and (d) Ni–9.4Al/Ni–10.1Cu
(at.%) annealed at 1433 K for 6 h, compared with the measured data in this work.

Figure A4: Comparisons between the model-predicted diffusion paths using the kinetic parameters of the Model 1, Model 2, Model 3 and
Model 4 in fcc Ni–Al–Cu alloys annealed at (a) 1273 K for 96 h, (b) 1333 K for 36 h, and (c) 1433 K for 6 h, compared with the experimental
data in this work.
L. Zhong et al.: Diffusivities and atomic mobilities in the Ni-rich fcc Ni–Al–Cu alloys | 367

Figure A5: The main interdiffusivity versus the diffusion distance for C1 at (a) 1273 K, (b) 1333K, (c) 1433 K, which are calculated by the
Matano–Kirkaldy method (denoted by symbols) and Model 1 (denoted by lines).
368 | L. Zhong et al.: Diffusivities and atomic mobilities in the Ni-rich fcc Ni–Al–Cu alloys

Figure A6: The main interdiffusivity versus the diffusion distance for C2 at (a) 1273 K, (b) 1333 K, (c) 1433 K, which are calculated by the
Matano–Kirkaldy method (denoted by symbols) and Model 2 (denoted by lines).
L. Zhong et al.: Diffusivities and atomic mobilities in the Ni-rich fcc Ni–Al–Cu alloys | 369

Figure A7: The main interdiffusivity versus the diffusion distance for C3 at (a) 1273 K, (b) 1333 K, (c) 1433 K, which are calculated by the
Matano–Kirkaldy method (denoted by symbols) and Model 3 (denoted by lines).
370 | L. Zhong et al.: Diffusivities and atomic mobilities in the Ni-rich fcc Ni–Al–Cu alloys

Figure A8: The main interdiffusivity versus the diffusion distance for C4 at (a) 1273 K, (b) 1333 K, (c) 1433 K, which are calculated by the
Matano–Kirkaldy method (denoted by symbols) and Model 4 (denoted by lines).
L. Zhong et al.: Diffusivities and atomic mobilities in the Ni-rich fcc Ni–Al–Cu alloys | 371

References 19. Wang F., Du Y., Min Q., Du C., Wen S., Liu H., Zhang L.,
Zhang S., Chu M., Liu Y. Calphad 2021. 74, 102316.
20. Wang W., Chen H., Larsson H., Mao H. Calphad 2019, 65,
1. Bhattacharya K., Kohn R. V. Acta Mater. 1996, 44, 529 − 542.
346 − 369. https://doi.org/10.1016/j.calphad.2019.03.011.
https://doi.org/10.1016/1359-6454(95)00198-0.
21. Whittle D. P., Green A. Scripta Metall. 1974, 8, 883 − 884.
2. Xin B., Luo Z., Zhang W., Li Y. Hot Work. Technol. 2011, 40,
https://doi.org/10.1016/0036-9748(74)90311-1.
44 − 47. https://doi.org/10.1631/jzus.A1000150.
22. Dayananda M. A. Metall. Trans. A 1983, 14, 1851 − 1858.
3. Li D., Zhang S., Liao W., Geng G., Zhang Y. Int. J. Miner. Metall.
https://doi.org/10.1007/BF02645555.
Mater. 2016, 23, 928 − 933. https://doi.org/10.1007/s12613-
23. Lechelle J., Noyau S., Aufore L., Arredondo A., Audubert F.
016-1308-y.
Diffus. Fund. 2012, 17, 1 − 39.
4. Gastien R., Corbellani C. E., Villar H., Sade M., Lovey F. Metall. 24. Haase R., Thermodynamisch-phänomenologische Theorie der
Mater. Trans. 2003, 349, 191 − 196. https://doi.org/10.1016/ irreversiblen Prozesse; Springer: Berlin, Heidelberg,
S0921-5093(02)00789-X. 1952.
5. Font J., Cesari E., Muntasell J., Pons J. Metall. Mater. Trans. 25. Bouchet R., Mevrel R. Acta Mater. 2002, 50, 4887 − 4900.
2003, 354, 207 − 211. https://doi.org/10.1016/S0921- https://doi.org/10.1016/S1359-6454(02)00358-0.
5093(03)00003-0. 26. Bocquet J. L. Diffusion in Alloys. In Diffusion in Materials;
6. Gastien R., Corbellani C. E., Sade M., Lovey F. Acta Mater. Laskar, A. L., Bocquet, J. L., Brebec, G., Monty, C., Eds.; NATO
2005, 53, 1685 − 1691. https://doi.org/10.1016/j.actamat ASI Series, vol. 179; Springer: Dordrecht, 1990. https://doi
.2004.12.018. .org/10.1007/978-94-009-1976-1_4.
7. Liu Y., Chen W., Zhong J., Chen M., Zhang L. J. Inst. Eng. 2017, 27. Strang W. G., George J. Math. Rev. 1973, 62, 306. https://doi
23, 197 − 211. https://doi.org/10.30544/308. .org/10.1115/1.3423272.
8. Liu Y., Chen W., Tang Y., Du Y., Zhang L. Acta Metall. Sin. 2016, 28. Goldberg D. E., Genetic Algorithm in Search, Optimization,
52, 1009 − 1016. https://doi.org/10.11900/0412.1961.2015 and Machine Learning; Addison-Wesley: Boston, 1989.
.00541. 29. Eyring H. J. Chem. Phys. 1935, 3, 107 − 115. https://doi.org/10
9. Zhang B., Zhang E., Tang Y., Wu X., Yi W., Zhong J., Zhang L. .1063/1.1749604.
Metall. Mater. Trans. 2021, 52, 2331 − 2343. https://doi.org/ 30. Andersson J. O., Agren J. J. Appl. Phys. 1992, 72, 1350 − 1355.
10.1007/s11661-021-06224-6. https://doi.org/10.1063/1.351745.
10. Kirkaldy J. S. Can. J. Phys. 1957, 35, 435 − 440. https://doi 31. Muggianu Y. M., Gambino M., Bros J. P. J. Appl. Phys. 1975, 72,
.org/10.1139/p58-096. 83 − 88. https://doi.org/10.1051/jcp/1975720083.
11. Kirkaldy J. S., Lane J. E., Mason G. R. Can. J. Phys. 1963, 41, 32. Redlich O., Kister A. T. Ind. Eng. Chem. 1948, 40, 345 − 348.
2174 − 2186. https://doi.org/10.1139/p63-212. https://doi.org/10.1021/ie50458a036.
12. Du Y., Sundman B. J. Phase Equilibria Diffus. 2017, 38, 33. Jönsson B. Z. Metallkd. 1994, 85, 502 − 509. https://doi.org/
601 − 602. https://doi.org/10.1007/s11669-017-0527-x. 10.1515/ijmr-1994-850708.
13. Du C., Zheng Z., Min Q., Du Y., Liu Y., Deng P., Zhang J., Wen 34. Miettinen J. Calphad 2003, 27, 309 − 318. https://doi.org/10
S., Liu D. Calphad 2020, 68, 101708. https://doi.org/10.1016/ .1016/j.calphad.2003.10.001.
j.calphad.2019.101708. 35. Chen H., Du Y., Xu H., Xiong W. Metall. Mater. Trans. 2009, 24,
14. Liu Y., Zhang C., Du C., Du Y., Zheng Z., Liu S. J. Mater. Sci. 3154 − 3164. https://doi.org/10.1557/jmr.2009.0376.
Technol. 2020, 42, 229 − 240. https://doi.org/10.1016/j.jmst 36. Ansara I., Dupin N., Lukas H. L. J. Alloys Compd. 1997,
.2019.12.005. 247, 20 − 30. https://doi.org/10.1016/S0925-
15. Zeng Y., Liu Y., Min Q., Wen S., Wang J. Calphad 2020, 71, 8388(96)02652-7.
102209. https://doi.org/10.1016/j.calphad.2020.102209. 37. Zhang W., Du Y., Zhang L., Xu H., Chen L. Calphad 2011,
16. Liu Y., Fu T., Liu H., Du C., Zheng Z. Calphad 2020, 70, 101780. 35, 367 − 375. https://doi.org/10.1016/j.calphad.2011.04
https://doi.org/10.1016/j.calphad.2020.101780. .009.
17. Wen S., Liu Y., Liu H., Min Q., Du Y., Du C., Zheng Z., Zhou P., 38. Liu D., Zhang L., Du Y., Xu H., Liu S., Liu L. Calphad 2009,
Zhang S., Chu M. Calphad 2021, 72, 102229. https://doi.org/ 33, 761 − 768. https://doi.org/10.1016/j.calphad.2009.10
10.1016/j.calphad.2020.102229. .004.
18. Premovic M., Du Y., Liu Y., Du C., Liu H. J. Phase Equilibria 39. Zhang L., Du Y., Chen Q., Steinbach I., Huang B. Int. J. Mater.
Diffus. 2020, 41, 378 − 389. https://doi.org/10.1007/s11669- Res. 2010, 101, 1461 − 1475. https://doi.org/10.3139/146
020-00793-6. .110428.

You might also like