test of darken’s mobility assumption in au-ni & cu-ni-zn ...cecamp/tms2005/tms05... · oak...
TRANSCRIPT
OAK RIDGE NATIONAL LABORATORYU. S. DEPARTMENT OF ENERGY
Test of Darken’s Mobility Assumption in Au-Ni & Cu-Ni-Zn Systems
Nagraj Kulkarni: Oak Ridge National Labs.Robert T. DeHoff: University of Florida
Multicomponent Multiphase Diffusion Symposium in Honor of John E. Morral: Session II
2005 TMS Annual MeetingSan Francisco, Feb. 14, 2005
Overview
1. Introduction: Diffusion Formalisms & Simplifications
2. Intrinsic Diffusion Simulation- Model System
3. Tests of Darken Relations- A: Binary : Au-Ni- B: Ternary: Cu-Ni-Zn
4. Conclusions5. Comments on development of multicomponent
diffusion databases
1. Introduction
• Why is diffusion modeling needed?− Ability to use thermodynamic, diffusion, property databases to
predict microstructural evolutionary processes• Experimental studies: expensive, time-consuming, limited
variability− Assess existing diffusion formalisms (many indeed!)
• Examine inherent assumptions & simplifications− Examine newer formalisms and understand mechanisms
• How is diffusion modeling done?− Atomistic approaches
• Monte Carlo, molecular dynamics, phase field, etc.− Finite difference methods
• Solve Fick’s second law for appropriate boundary conditions and geometry
• Use intrinsic flux expression for different time intervals (here)
Diffusion Formalisms: A. Interdiffusion
Isothermal, isobaric system
Multicomponent version of Fick’s law
There are (n - 1) (n - 1) = (n - 1)2 coefficients
Fluxes defined in the lattice frame
B. Intrinsic Diffusion Formalism
Multicomponent version of Darken’s equation
There are n (n - 1) coefficients
Simplified Version of Formalism Intrinsic flux expression that ignores cross-terms
Mk is known as the “mobility” [Darken 1948]− One unique mobility for each component in a
multicomponent system
Multicomponent Darken theory connects mobilities with tracer diffusion coefficients
Diffusion coefficients in other frames can be obtained from mobilities (or Lkk’s) by suitable transformations
Not proven that Lkk’s or Mk’s are unique functions of composition− Need intrinsic measurements in diffusion couples with
crossing composition paths
Simplified Formalism Questions(used by DICTRA)
Darken relations in binary systems not systematically tested with experimental data− Tests in ternary systems where tracer diffusion
information is available are rare
Objectives
• Develop a versatile simulation capable of treating diffusion for different formalisms
• Test internal consistency of simulation with model systems
• Develop systematic procedure for the assessment of Darken theories− Binary: Ag-Cd, Au-Ni, Cu-Zn, Cu-Ni− Ternary: Cu-Ni-Zn
1. Initial concentration distribution for each component2. Depending upon the formalism, kinetic descriptors as a function
of composition, temperature or other variables3. Thermodynamics of phase, if needed4. Molar volume as a function of composition
Finite difference method for semi-infinite couples
• Simple yet effective algorithm
• Handles variable molar volumes
• Developed in MathCad – user-friendly
• Efficient – practical output for a single diffusion couple in less than a minute
• Versatile – adaptable to various formalisms
2. Single-Phase, Intrinsic Diffusion Simulation for Multicomponent Systems
R = 5Do = 10-10 cm2/s
Model System with Variable Molar Volume• Do = DV = constant: Error function solution for ck (x,t)
•Ratio of intrinsic diffusion coefficients constant:
Simulation Output for Model System
Concentration profiles ck (mol/cc)
Atom fraction Xk = ck V
t = 12 x 3600 s
Lattice velocity (vK)
Velocity in number frame (vN)
Phase diagram Molar volume dependence on composition (almost ideal mixing)
• Classical system often quoted for the success of the Darken relation between the tracer and interdiffusion coefficient
3. A: Tests of Darken Relations in Au-Ni at 900oC
Darken Au-Ni (contd.)
Experimental intrinsic diffusion coefficients[Van Dal et al. [J. Alloys & Comp., 2000]
Interdiffusion coefficient
• Polynomial functions were used to fit the intrinsic diffusion data
Darken Au-Ni: Assessing Experimental Data
Concentration profile (6000 iterations) Lattice shift profile
Quality of experimental intrinsic data appears excellent
Au-Ni, 900oC, 100 h
x (µm)
-200 0 200 400 600 800
Latti
ce s
hift
(µm
)
0
20
40
60
80
100
xK
AuNi
Au-Ni, 900oC, 50 h
x / ot (µm/oh)
-80 -60 -40 -20 0 20
Atom
Fra
ctio
n N
i
0.0
0.2
0.4
0.6
0.8
1.0
data of Dal et al. [Dal00a]simulation
xK
XNi
Au Ni
Darken Au-Ni: Tracer Diffusion
Tracer data for Au [Kurtz et al. , Acta Met., 1955] Tracer diffusion coefficients of Au and Ni
1/T (10-4) K-1
7 8 9 10
D* Au
(cm
2 /s)
10-11
10-10
10-9
10-8
100% Ni 80% Ni65% Ni 50% Ni35% Ni20% Ni0% Ni
100
8065
5035
20
0
0
50
Atom fraction (XNi)0.0 0.2 0.4 0.6 0.8 1.0
Trac
er d
iffus
ion
coef
ficie
nt (c
m2 /s
)
10-12
10-11
10-10
10-9
D*Au (Kurtz et al.) D*Ni (Reynolds et al.) D*Au - fit D*Ni - fit fit by Reynolds et al. fit by Reynolds et al.
D*Au
D*Ni
Darken Au-Ni: Thermodynamics
Redlich-Kister polynomial fits for activity data Thermodynamic factor (M)
Thermodynamic factor is less than unity
Au-Ni, 900oC
Atom fraction (XNi)
0.0 0.2 0.4 0.6 0.8 1.0
Activ
ity (a
Ni)
0.0
0.2
0.4
0.6
0.8
1.0
R-K fit to Bienzle's dataBienzle et al. [Bie95] R-K fit to Sellars and Maak's dataSellars and Maak [Sel66]Seigle et al. [Sei52]
Atom fraction (XNi)
0.0 0.2 0.4 0.6 0.8 1.0
Ther
mod
ynam
ic fa
ctor
(Φ)
0.0
0.2
0.4
0.6
0.8
1.0
Bienzle et al. [Bie95]Sellars and Maak [Sel66]Reynolds et al. [Rey57]
Darken Relations for Au-Ni: Predicted Intrinsic Diffusion Coefficients
Darken: DAu
Problem with predicted Au intrinsic diffusion coefficient
Au-Ni, 900oC
Atom fraction nickel (XNi)0.0 0.2 0.4 0.6 0.8 1.0
Au
diffu
sion
coe
ffici
ent (
cm2 /s
)
10-13
10-12
10-11
10-10
10-9
10-8
DAu (van Dal)DAu
Dark (Φ : Sellars)DAu
Dark (Φ : Bienzle)DAu
Dark by Reynolds (Φ : Seigle)DAu* (Kurtz)
Atom fraction nickel (XNi)0.0 0.2 0.4 0.6 0.8 1.0
Ni d
iffus
ion
coef
ficie
nt (c
m2 /s
)10-13
10-12
10-11
10-10
10-9
10-8
DNi (van Dal)DNi
Dark (Φ : Sellars)DNi
Dark (Φ : Bienzle)DNi
Dark by Reynolds (Φ : Seigle)DNi* (Reynolds)
Darken: DNi
Darken Relations for Au-Ni: Predicted Interdiffusion Coefficient
Darken: Do
Predicted interdiffusion coefficient reasonably close to data
Au-Ni, 900oC
Atom Fraction Nickel (XNi)0.0 0.2 0.4 0.6 0.8 1.0
Inte
rdiff
usio
n C
oeffi
cien
t (cm
2 /s)
10-13
10-12
10-11
10-10
10-9
10-8
van DalReynoldsDAu* DNi* Do Dark (Φ: Sellars)Do Dark (Φ: Bienzle)Do Dark by Reynolds (Φ : Seigle)
Do
Assessing Darken Relations for Au-Ni with Simulation
Concentration profile Lattice shift profile
Differences between predicted and experimental DAu results in large differences in lattice shift profiles
Au-Ni, 900oC
x / ot (µm/oh)
-80 -60 -40 -20 0 20
Atom
Fra
ctio
n N
i
0.0
0.2
0.4
0.6
0.8
1.0
experiment (Dal)intrinsic data (Dal)Darken (Φ: Sellars)Kirkendall location for intrinsic data
xK
XNi
Au Ni
Au-Ni, 900oC, 100 h
x (µm)
-200 0 200 400 600 800
Latti
ce s
hift
(µm
)
0
20
40
60
80
100
experiment (Dal)intrinsic data (Dal)Darken (Φ: Sellars)
xK
AuNi
3. B: Darken Relations in Cu-Ni-Zn at 900oC
Intrinsic fluxes in a multicomponent system (i = 1, 2, … , c):
Ji Lii−x
µidd
⋅ ci− Mi⋅x
µidd
⋅DtiR T⋅
− ci⋅x
µidd
⋅ Dti−XiV
⋅xln ai( )d
d⋅
Dti = tracer diffusion coefficient of component iµi = chemical potential of i; ai = activity of iV = molar volume; Xi = mole fraction of i
Cserhati, Ugaste, van Loo et al., Def. & Diff. For., 2001
Φ ij ln Xj( )ln ai( )d
d
Φij = thermodynamic factors obtained from Gibbs free energy of phase
JiDti−
VXi⋅
Φ iiXi x
Xidd
⋅
1 i≠
c 1−
j
Φ ijXj x
Xjdd
⋅∑=
+
⋅
Darken Relations in Cu-Ni-Zn: contd.
Intrinsic fluxes in a ternary:
Relation between thermodynamic factors
J 1Dt 1−
VX 1⋅
Φ 11X 1 x
X 1dd
⋅Φ 12X 2 x
X 2dd
⋅+
⋅
J 2Dt 2−
VX 2⋅
Φ 22X 2 x
X 2dd
⋅Φ 21X 1 x
X 1dd
⋅+
⋅
J 3Dt 3V
X 1Φ 11X 1 x
X 1dd
⋅Φ 12X 2 x
X 2dd
⋅+
⋅ X 2
Φ 22X 2 x
X 2dd
⋅Φ 21X 1 x
X 1dd
⋅+
⋅+
⋅
Φ 21X 11 X 1−( )
1 X 2−
X 2Φ 12⋅ Φ 22+ Φ 11−
⋅
Cu-Ni-Zn: Thermo Jiang, Ishida et al., J. Phys. Chem. Sol., 2004, in print
Isoactivity curves of Zn in fcc-CuNiZn at 727oC(liq. Zn ref.)
Cu-Ni-Zn: Thermodynamics contd.
∆G_α mix_xs X 1 X 2⋅ L_α 12 X 1 X 2, T,( )⋅ X 1 X 3⋅ L_α 13 X 1 X 3, T,( )⋅+ X 2 X 3⋅ L_α 23 X 2 X 3, T,( )⋅+X 1 X 2⋅ X 3⋅ L_α 123 X 1 X 2, X 3, T,( )⋅+
...
L_α 123 X 1 X 2, X 3, T,( ) X 1 L_α0 123 T( )⋅ X 2 L_α1 123 T( )⋅+ X 3 L_α2 123 T( )⋅+
L_α 12 X 1 X 2, T,( ) L_α0 12 T( ) L_α1 12 T( ) X 1 X 2−( )1⋅+ L_α2 12 T( ) X 1 X 2−( )2⋅+
Redlich-Kister model for Gibbs excess free energy:
∆Gxsk ∆G mix_xs Xk∆G mix_xs
dd
+
1
c
j
Xj Xj∆G mix_xs
dd
⋅∑=
−
Partial molal Gibbs free energy:
-11.0
-10.5
-10.0
-9.5
-9.0
-8.5
-8.0
05
1015
2025
3035
05
1015
2025
3035
log(DC
u ) (cm2/s)
XNi (At.%)
X Zn (A
t.%)
Cu-Ni-Zn: Tracer diffusion
Anusavice & DeHoff, Met. Trans. A., 1972log Dt Cu( ) 3.53− X Ni
1.16⋅ 3.6 X Zn1.02⋅+ 9.46−
log Dt Ni( ) 4.05− X Ni1.3⋅ 3.28 X Zn
1.07⋅+ 9.96−
log Dt Zn( ) 3.2− X Ni⋅ 5.21 X Zn1.25⋅+ 9−
D*Zn > D*
Cu > D*Ni
Similar composition dependence
Ni0 10 20 30 40
Zn
Cu
70
80
90
100
Wan experimentalDarken simulation
Darken relations in Cu-Ni-Zn: Composition paths at 900oC
#4,5,7,9
Wan, Ph.D. Thesis, Univ. of Florida, 1972Wan & DeHoff, Acta Met., 1977
Cu-Ni-Zn: Concentration profilesCouple #9 (Wan 13-00)
x (µm)
-800 -600 -400 -200 0 200 400 600 800
Atom
%
0
20
40
60
80
100
Experiment (Wan 895oC)Darken simulationDarken: Kirkendall composition
Cu
Zn
Ni
xK
Cu-Ni-Zn: Concentration profilesCouple #4 (Wan 00-32)
x (µm)
-600 -400 -200 0 200 400 600 800 1000
Atom
%
0
20
40
60
80
100
Experiment (Wan 897oC)Darken simulationDarken: Kirkendall composition
Cu
Zn
Ni
xK
Ni0 10 20 30 40
Zn
Cu
70
80
90
100
Wan experimentalDarken simulation
Composition paths Cu-Ni-Zn contd.
#6,8,11
Cu-Ni-Zn: Concentration profilesCouple #11 (Wan 02-30)
x (µm)
-600 -400 -200 0 200 400 600 800
Atom
%
0
20
40
60
80
100
Experiment (Wan)Darken simulationDarken: Kirkendall composition
Cu
Zn
Ni
xK
Ni0 10 20 30 40
Zn
Cu
70
80
90
100
Wan experimentalDarken simulation
Composition paths Cu-Ni-Zn contd.
#1,2,3
Cu-Ni-Zn: Concentration profiles
Couple #2 (Wan 33-10)
x (µm)
-600 -400 -200 0 200 400 600 800
Atom
%
0
20
40
60
80
100
Experiment (Wan 900oC)Darken simulation
Cu
Zn
Ni
xK
Ni
Zn
Cu
Kirkendall Shift Comparison
Kirkendall shifts
Experimental (Wan)0 20 40 60 80 100
Pre
dict
ed (D
arke
n)
0
20
40
60
80
100
4. Conclusions• In the binary Au-Ni and the ternary Cu-Ni-Zn system, Darken
relations result in− Reasonable concentration profiles; however− Kirkendall shifts are noticeably underestimated
Manning relations in Cu-Ni-Zn to be explored
• Intrinsic simulation powerful tool for testing quality of experimental data & various diffusion formalisms with their associated assumptions− Need to explore intermetallic phases with broad solubility
ranges where vacancy concentrations & fluxes are likely to be important
Alternate approachA kinetic approach based on tracer jump frequencies has also been found to have the same level of success in predicting diffusion paths in Cu-Ni-Zn using the same simulation
However the Kirkendall shifts are overestimatedBut no thermodynamics necessary; tracer diffusion coefficients sufficient
5. On development of multicomponentdiffusion databases
• Tracer diffusion data vital for high-quality databases
• More experimental efforts in this direction needed− But need not use radioactive isotopes− Stable isotopes (except Al) combined with mass
spectrometry may be used instead (errors?)*
• Explore use of combinatorial techniques for sample preparation and automated analysis of tracer (stable) isotope profiles
* Schmidt, Schmalzried et al., “Self-diffusion of B in TiB2,” J. App. Phys., 2003
John Morral’s other contributions…(thermodynamics, oxidation, solidification,
phase diagrams, coarsening, etc.)• Morral & Davies, “Thermodynamics of Isolated Miscibility Gaps,” J. Chimie Phys.
Physico-Chimie Bio, 1997.• Morral & Purdy, “Thermodynamics of Particle Coarsening,” J. Alloys & Comp.,
1995.• Hennessey & Morral, “Oxidation & Nitridation of Niobium in Air above 1150oC,”
Oxid. Met., 1992. • Morral & Gupta, “Phase-boundary, ZPF, and Topological Lines on Phase
Diagrams,” Scripta Met., 1991.• Gupta, Morral & Nowotny, “Constructing Multicomponent Phase Diagrams by
Overlapping ZPF Lines,” Scripta Met., 1986.• Pati & Morral, “The Effect of Quenching on Mushy Zone Microstructures,” Met.
Trans. A, 1986.• Cameron & Morral, “The Solubility of Boron in Iron,” Met. Trans. A, 1986.• Ohriner & Morral, “Precipitate Distribution in Sub-Scales,” Scripta Met., 1979.• Morral & Asby, “Dislocated Cellular Structures,” Acta Met., 1974.• Morral & Cahn, “Spinodal Decomposition in Ternary Systems,” Acta Met., 1971.• Morral, “Characterizing Stability Limits in Ternary Systems,” Acta Met., 1972.
… and many more