H. Xu et al. / Journal of Alloys and Compounds 644 (2015) 687–693
689
simplicity, the end-members corresponding to self and impurity
diffusivities are fixed by using the experimental data,
first-principles calculations, molecular dynamic simulation or
some semi-empirical relations. The other coefficients such as
all the 5 diffusion couples measured by EPMA in the present work
are presented in Fig. 2. Based on Eqs. (8) and (9), the corresponding
interdiffusion fluxes in these diffusion couples can be also calcu-
lated and presented in Fig. 2. It should be noted that all the exper-
imental data are denoted in symbols.
1
;2
1;3
2;3
1;2;3
DGi
,
DGi
,
DGi and
DGi
are treated as adjustable parameters.
Based on the measured composition profiles, the
composition-dependent interdiffusivities of fcc Cu–Ag–Sn alloys
at 1073 K along the entire diffusion path of each diffusion couple
were then effectively determined by using the pragmatic numerical
inverse method. During the evaluation of the interdiffusivities, the
thermodynamic descriptions for fcc phase in the Ag–Cu–Sn system
were directly taken from Kattner [23], from which the thermody-
namic factor in Eq. (4) can be provided. While the end-members
Moreover, only one or two of them are needed to be evaluated in
most cases for one diffusion couple. Therefore, by combining Eqs.
(
2)–(6), one can evaluate the composition-dependent interdiffusiv-
ities based on the measured concentration profiles together with
the available thermodynamic description.
During the evaluation of interdiffusivities, the Fick’s second law
was applied to simulate the concentration profiles based on the
computed interdiffusivities, an optimal set of adjustable parame-
1
2
3
1
;2
1;3
2;3
1;2;3
for the three binary systems,
DG , DG and DG , were taken from
ters such as
D
Gi
,
D
Gi
,
D
Gi and/or
DGi
were carefully chosen
i i i
Refs. [24–26], and fixed all the time. Moreover, the vacancy-wind
effect was considered in the present work. One or two of the adjus-
by iteratively fitting until the minimization of the error between
the measured and the simulated concentration profiles is achieved:
1
;2
1;3
2;3
1;2;3
*
+
table parameters,
D
Gi
,
D
Gi
,
D
Gi and
D
Gi , were tried to get the
N
ꢂꢄ
ꢄꢃ
ꢄ
XX
1
N
ꢄ
cal
exp
best fit to the experimental composition profiles for each diffusion
couple. With the final optimal set of adjustable parameters for each
diffusion couple, the ternary interdiffusivities of composition
dependence can be directly calculated by using Eqs. (2)–(6). All
the calculated interdiffusivities were subject to the examination
of the following thermodynamically stable constraints [27],
min < error > ¼ min
ꢄc ꢁ cij
ꢄ
ð7Þ
ij
i¼1;2 j¼1
where c ic jal and cij are the calculated and the experimental concen-
exp
trations of component i at the jth point, respectively, and N is the
total number of the experimental data. With the optimal set of
the fitting parameters, the concentration-dependent interdiffusivi-
ties in the target ternary system can be computed via Eq. (3).
~
Cu
~
Cu
D
þ D
> 0
ð12Þ
ð13Þ
SnSn
AgAg
3.2. Matano–Kirkaldy method
~
Cu
~
Cu
AgAg
~
Cu
~
Cu
AgSn
D
ꢂ D
ꢁ D
ꢂ D
P 0
SnSn
SnAg
Based on the Fick’s first law in Eq. (1), Kirkaldy successfully
extended the Boltzmann–Matano method into ternary and even
higher-order systems [13,14].
Assuming that each component has the same molar volume, the
interdiffusion flux of each component can be determined directly
from the concentration profiles without using the interdiffusion
coefficients [22]. The interdiffusion flux of component i can be
expressed as:
ꢂ
ꢃ
2
~
Cu
~
Cu
AgAg
þ 4 ꢂ D~ Cu ꢂ D
~
Cu
AgSn
D
ꢁ D
P 0
ð14Þ
SnSn
SnAg
Only those interdiffusivities fulfill the above constraints can be
output.
The finally obtained ternary interdiffusivities, D~ Cu , D
and D~
Fig. 3 as a three-dimensional (3-D) illustration. For readers’ conve-
nient usage, all the original experimental data are also provided as
the electronic Supplementary Materials. The errors of the interdif-
fusivities determined by the pragmatic numerical inverse method
were evaluated according to the scientific method proposed by
Lechelle et al. [28], who considered the error propagation via the
following function,
~
Cu
SnAg
~
Cu
AgSn
, D
SnSn
Cu
for the 5 fcc Cu–Ag–Sn diffusion couples are presented in
AgAg
Z
c
i
1
t
~
J
i
¼ 2
ðx ꢁ x
þ1
orci
0
Þdc
i
ð8Þ
ꢁ1
c
i
where t is the diffusion time, x
0
is the position of Matano plane
[
13,14] and can be obtained from the following relation:
Z
þ1
c
i
ðx ꢁ x
0
Þdc ¼ 0
ð9Þ
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u
ꢀ
ꢁ
ꢁ
1
2
c
u X
i
@f
@a
2
t
uðfðA; B . . .ÞÞ ¼
ðuð
a
ÞÞ
ð15Þ
ꢁ
1
þ1
and ci and ci (i = 1, 2) are the terminal compositions at the left
and right sides of the diffusion couple. The initial and boundary
conditions for the semi-infinite diffusion couples are:
a
¼A;B...
Here, A and B. . . are the correlation quantities of function f like
Eqs. (3) and (5), while u(
able like concentration. During the evaluation of the errors, the
differences between the experimental concentration profiles and
the predicted ones propagate to calculation of atomic mobility in
Eq. (5) and then to calculation of diffusivities in Eq. (3). In order
to eliminate the effect of the absolute value, the relative error
a) (a = A, B. . .) is the uncertainty of vari-
c
i
ðꢃx; 0Þ ¼ c
i
ðꢃ1; tÞ ¼ cꢃi 1
ð10Þ
a
Combining Eqs. (1) and (8), one can obtain:
Z
ꢀ
ꢁ
c
i
dc
dx
1
dc
2
dx
~
3
i1
~
3
i2
ðx ꢁ x
0
Þdc
i
¼ ꢁ2t D
þ D
ð11Þ
ꢁ
1
c
ci
i
(
i.e., equals to the uncertainty divided by the absolute value of
With four equations similar to Eq. (11) from two diffusion cou-
ples, the four main and cross interdiffusivities in Eq. (1) of the
intersection point can be then determined.
the interdiffusivity) rather than the uncertainty itself was utilized
in the present work. Moreover, considering that one relative error
can be determined for one diffusivity, numerous errors should be
evaluated, displayed and stored. For a clear display of the
concentration-dependent diffusivities along the entire diffusion
paths, as well as to save the space, only the average error of the dif-
fusivities was provided here for simplification. The average relative
error of the interdiffusivities obtained by this numerical inverse
method was finally evaluated to be 8%. It can be clearly seen from
Fig. 3 that the ternary interdiffusivities for all the diffusion couples
vary apparently along with the composition of Sn and Ag. All the
4
. Results and discussion
Considering that all the 5 diffusion couples are in the same fcc
single-phase region, one typical microstructure of the diffusion
zone is given in Fig. 1, which shows the backscattered electron
image (BEI) of C5 (Cu–1.98Sn/Cu–3.5Ag) diffusion couple annealed
at 1073 K for 36 h. The concentration profiles of each component in