PHYSICAL REVIEW B, VOLUME 64, 174418
Surfactant-mediated growth of giant magnetoresistance multilayers
W. Zou, H. N. G. Wadley, X. W. Zhou, and R. A. Johnson
Department of Materials Science and Engineering, University of Virginia, Charlottesville, Virginia 22903
D. Brownell
Nonvolatile Electronics, Inc., Eden Prairie, Minnesota 55344
共Received 8 March 2001; published 11 October 2001兲
A series of experiments have been conducted to evaluate the magnetotransport properties of rf-diode-sputterdeposited giant magnetoresistive multilayers with either copper or copper-silver-gold nonferromagnetic 共NFM兲
conducting layers. The study revealed that rf-diode-deposited multilayers utilizing Cu80Ag15Au5 as the NFM
conducting layer possess significantly superior giant magnetoresistance to otherwise identical device architectures that used pure copper as the NFM conducting layer. To explore the origin of this effect, copper and
Cu80Ag15Au5 films of varying thickness have been grown under identical deposition conditions and their
surface morphology and roughness have been investigated. Atomic-force microscopy revealed significant
roughness and the presence of many pinholes in thin pure-copper films. The surface roughness of the
Cu80Ag15Au5 layers was found to be much less than that of pure copper, and the alloying eliminated the
formation of pinholes. Using an embedded-atom-method alloy potential, molecular-dynamics simulations have
been used to investigate the role of silver and gold upon the multilayer growth process. The smoother growth
surface of Cu80Ag15Au5 was found to predominantly result from the addition of silver, which acts as a
surfactant during growth. Molecular statistics estimates of atom migration energy barriers indicated that both
silver and gold have significantly higher mobilities than copper atoms on a flat copper surface. However, gold
was found to be incorporated in the lattice whereas silver tended to segregate 共and concentrate兲 upon the free
surface, enhancing its potency as a surfactant. The atomic-scale mechanism responsible for silver’s surfaceflattening effect has been explored. We found that silver, when present at a ledge edge, reduces the EhrlichSchwoebel barrier for copper, promoting a step-flow growth mode. Gold was also found to reduce the EhrlichSchwoebel barrier, but its potency was less than that of silver due to its lower surface concentration. These
observations suggest that small alloy additions can be used to manipulate the energy barriers that fundamentally control atomic assembly during vapor deposition, and provide a potentially powerful means of controlling
the structure of thin films.
DOI: 10.1103/PhysRevB.64.174418
PACS number共s兲: 75.70.Pa
I. INTRODUCTION
Metal multilayers consisting of alternating ferromagnetic
共FM兲 and nonferromagnetic 共NFM兲 metals sometimes exhibit large changes in their electrical resistance when a magnetic field is applied.1–11 The effect results from a change in
their spin-dependent electron scattering when an applied
magnetic field rotates the magnetic moment of one of the
ferromagnetic layers.1–11 The giant magnetoresistance
共GMR兲 effect has quickly become technologically
important.3–5,7 Devices based on it are widely used as
magnetic-field sensors in hard-disk-drive read heads.7 Related devices are being investigated for use as magnetic
random-access memories 共MRAM兲.7 These applications require multilayers with high GMR ratios 共defined as the maximum resistance change divided by the resistance at magnetic
saturation兲 and low magnetic-saturation fields. Both the resistance change and saturation field are sensitive to the materials used for each layer, the thickness of the layers, and the
conditions used for their growth via their effect upon atomicscale morphology and structure.1–11
Current GMR multilayers appear not to have achieved
their performance upper bound. Reducing both the roughness
and chemical diffuseness of the interfaces in GMR multilayers appears to be particularly important.2,9,11 Even moderate
0163-1829/2001/64共17兲/174418共10兲/$20.00
interfacial roughness can result in Neel coupling and undesired electron scattering, which are both deleterious to the
magnetoresistance.2,4,8,9 Extreme interfacial roughness in the
NFM layer can even lead to pinholes in NFM layers. When a
FM metal layer is then overdeposited, the pinhole enables a
connection between the two FM layers and the formation of
a strong ferromagnetic coupling. An applied magnetic field
then causes no differential rotation of magnetic moments and
the GMR effect is lost.5 Synthesizing materials with a large
GMR ratio at a low saturation field therefore requires the use
of materials, layer thickness, and deposition conditions that
minimize interfacial roughness.
Many vapor-deposition approaches are being explored for
controlling the interface morphology, including adatomenergy variations and low-voltage ion assistance.11–14 Here
we use a standard growth process 共rf diode deposition兲 to
investigate the effects of adding silver and gold to a copper
NFM layer in a GMR multilayer. Both the magnetotransport
properties and the surface morphology are explored. Single
films of both Cu and an Cu80Ag15Au5 alloy are also grown
and their surface morphology evaluated. A moleculardynamics 共MD兲 method is then used to investigate the
atomic-assembly mechanisms during film growth for the two
compositions. A dependence of interfacial roughness and
pinhole formation upon the NFM layer composition is iden-
64 174418-1
©2001 The American Physical Society
ZOU, WADLEY, ZHOU, JOHNSON, AND BROWNELL
PHYSICAL REVIEW B 64 174418
FIG. 1. The GMR multilayer
structure studied here. Layer compositions and thicknesses are
shown. White arrows indicate the
magnetic alignment of the
ferromagnetic-alloy layers under
the ambient field.
tified and correlated with the performance of the devices.
The work suggests new directions for the development of
improved GMR materials.
II. EXPERIMENTS
The study has focused upon the influence of the nonferromagnetic conducting-layer composition 共either Cu or
Cu80Ag15Au5兲 on the GMR ratio and the saturation field of
rf-diode-sputter-deposited GMR multilayers. Multilayers
with either Cu or Cu80Ag15Au5 NFM conducting layers as
well as 100-, 500-, and 1000-Å-thick Cu and Cu80Ag15Au5
single-layer films were grown at Non-volatile Electronic,
Inc. 共Eden Prairie, MN兲 using a Randex Model 2400-6J rf
diode system 共see Ref. 12 for details兲. With a target diameter
of 20.32 cm, a substrate diameter of 10.00 cm, and a targetto-substrate distance of 3.81 cm, all the multilayers and the
single-layer films were grown on the silicon-wafer substrates
at an argon chamber pressure of 20 mTorr and a plasma
power of 175 W.
The architecture and compositions of the antiferromagnetically coupled multilayers are shown in Fig. 1. The multilayers contained three repeated exchange-coupled-layer
stacks. Each layer stack was a composite of two ferromagnetic layers 共Ni65Fe15Co20 and Co95Fe5兲 arranged so that the
Co95Fe5 layer was at the interface with either the copper or
the copper-alloy layer. This arrangement results in relatively
good GMR ratios by helping to minimize the Ni and Cu
interdiffusion.3,5,6 The deposition rate for each metal target
had been previously measured.11 The layer thicknesses were
therefore controlled by timing the deposition. The multilayers were grown by placing the substrate sequentially under
each target. In each case a fixed NFM conducting-layer
thickness of 16 Å was used. An amorphous 2000 Å Si3N4
film had been previously deposited on top of each wafer
using a chemical-vapor-deposition method. The Si3N4 film
174418-2
SURFACTANT-MEDIATED GROWTH OF GIANT . . .
PHYSICAL REVIEW B 64 174418
stacks. Both the GMR ratio and the saturation fields using
the Cu80Ag15Au5 layer are reasonably well suited for
magnetic-sensing applications.3,5,6 Many researchers have
successfully used pure Cu as a conducting layer2– 8 in GMR
multilayers. They used a different deposition approach4 and
the Cu layer thickness was usually greater than 20 Å. For the
particular architecture and composition of the multilayer and
rf-diode-deposition method investigated here, Cu thickness
below 20 Å resulted in no GMR.5
To explore the reasons for the loss of the GMR effect,
three different thickness Cu and Cu80Ag15Au5 single-layer
films were grown under the same deposition conditions. The
surface morphology of each film was then evaluated with
tapping-mode atomic-force microscopy 共AFM兲, Fig. 3. The
scan size was 1⫻1 ␮ m2 and scan rate was 1 Hz for all the
films. These results revealed periodic surface roughness with
an asperity width that increased with the film thickness in
both types of films, Fig. 3. However, the Cu films had larger
asperity widths and higher surface amplitudes than their
Cu80Ag15Au5 counterparts. A top view of the films is shown
in Fig. 4. Pinholes 共the dark spots in the image兲 were present
in the 100-Å-thick Cu film, Fig. 4共d兲. AFM measurements
共Fig. 5兲 indicated that the average depth of the pinholes
could be above 40 Å, and their diameters were approximately 800 Å. No pinholes were observed in the CuAgAu
films. Figure 6 shows that the RMS roughness increased with
the film thickness in both the Cu and Cu80Ag15Au5 films, but
the Cu films were always much rougher. The loss of the
GMR effect in multilayers with a pure copper conducting
layer is consistent with ferromagnetic layer coupling through
pinholes in the NFM spacer layer.5
IV. MOLECULAR DYNAMICS SIMULATIONS
FIG. 2. The dependence of the GMR ratio and the saturation
field on the number of repeated multilayer stacks based on both Cu
and Cu80Ag15Au5 NFM conducting layers. In each case a fixed
NFM conducting-layer thickness of 16 Å was used. Multilayers
with the pure Cu conducting layer exhibited no giant magnetoresistance, while multilayers with the Cu80Ag15Au5 conducting layer
possessed significant GMR ratios.
acted as an electrically insulating and diffusion inhibiting
layer between the silicon wafer and the subsequently deposited metal films. The RMS roughness of the Si3N4 film was
2.0 Å. Si3N4 was preferred over SiO2 because the oxygen in
the latter can sometimes react with the magnetic materials
and alter their coercivity.
III. EXPERIMENTAL RESULTS
Figure 2 shows the dependence of the GMR ratio and
saturation field on the number of repeated stacks in the Cu
and the Cu80Ag15Au5 conducting-layer multilayers. The results in Fig. 2 show that multilayers grown with a pure Cu
conducting layer exhibited no GMR effect, whereas multilayers grown with a Cu80Ag15Au5 conducting layer had significant GMR ratios. The saturation magnetic field of the
Cu80Ag15Au5 multilayers also increased with the number of
The results above indicate a significant dependence of
surface morphology upon the layer composition. To explore
the origin of this effect, molecular-dynamics simulations of
vapor deposition were conducted. Details of the method can
be found in Refs. 13 and 14. For the problems of interest
here, an interatomic potential that accurately describes the
interaction forces between all metal atoms in the system was
required. A recently developed embedded-atom-method
共EAM兲 potential has been used for this.13–16 EAM potential
incorporates the effects of local atomic environment, and has
been successfully used to study a variety of surface problems
in fcc metals.13–16 By normalizing the EAM functional form
and generalizing the potential cutoff distance, the model has
been found to realistically describe the interaction between
atoms of 16 different elements.
A molecular-dynamics model of multilayer deposition
was constructed by first defining the atomic coordinations of
a substrate. The simulated substrate has 120 planes in the
关 224̄ 兴 direction, 12 planes in the 关 22̄0 兴 direction, and 3
planes in the 关111兴 direction. To reduce the effects associated
with the small crystal size, periodic boundary conditions
were used in the x and z directions 共see Fig. 7兲. To prevent
crystal shift during adatom impact, the bottom two 共111兲
planes were fixed. Under this condition, atoms above the
fixed region were beyond their interaction range with the
174418-3
ZOU, WADLEY, ZHOU, JOHNSON, AND BROWNELL
PHYSICAL REVIEW B 64 174418
FIG. 3. The topography of Cu80Ag15Au5 and Cu films of different thicknesses. While growth column width increased with film thickness
for both film compositions, Cu films had larger column widths and rougher surfaces compared to Cu80Ag15Au5 films of identical thickness.
atoms below the fixed region and the substrate could then be
viewed as being extended indefinitely into the bulk.13,14 Due
to the time-scale problem typical of molecular-dynamics
simulations, an accelerated deposition rate of 4 nm/ns was
employed while the substrate was held at 300 K.
Figure 7共a兲 shows the morphology of a 60-Å Cu film
deposited on a Ni65Fe15Co20 substrate. To amplify the flux-
shadowing contribution to surface roughness, the incident
flux angle was set at 45° to the copper surface. The incident
atom energy was 0.15 eV, which is similar to that predicted
for the rf-diode growth conditions.11 It can be seen that during Cu deposition, flux shadowing dominated the assembly
process, which resulted in significant surface roughness and
the formation of a pinhole. Figure 7共b兲 shows that when a
174418-4
SURFACTANT-MEDIATED GROWTH OF GIANT . . .
PHYSICAL REVIEW B 64 174418
FIG. 4. Plan view of Cu80Ag15Au5 and Cu films. Pinholes have been identified only in the 100-Å-thick Cu film.
Cu80Ag15Au5 layer was deposited, much smoother 共continuous兲 film was achieved even at the oblique incident angle.
The MD analysis indicated that by adding Ag and Au to Cu,
a film can be prepared without pinhole formation, in agreement with the AFM images.
Figure 7共b兲 also indicates a very strong Ag surface
segregation17 even at the high deposition rate simulated. The
low equilibrium solubility of Ag in Cu at 300 K suggests that
even a more complete Ag segregation could be expected under the lower deposition rates encountered in real
deposition.18 Interestingly, Au atoms did not show any segregation effect, consistent with the high Au solubility and its
intermetallic compound formation in Cu.18
Figure 8 shows the simulated morphology of a GMR
multilayer structure grown using Cu and Cu80Ag15Au5 NFM
conducting layers. It again shows that pinholes were formed
whereas the addition of Ag and Au inhibited their formation,
which resulted in smoother surface and interfaces. These results revealed that silver and perhaps gold are potent surfactants during multilayer growth. It is interesting to note that
174418-5
ZOU, WADLEY, ZHOU, JOHNSON, AND BROWNELL
PHYSICAL REVIEW B 64 174418
FIG. 5. AFM profile analysis shows that the depth of the pinhole
can be above 40 Å. The diameter of the pinhole is approximately
800 Å.
silver is also able to float through the FM layers 共since no
significant silver concentration existed at the FM on NFM
interface兲.
V. DISCUSSION
A growing body of experiments data2,9,10,11 and transport
calculations4,11 indicate that the flattening of interfaces and
reductions in interfacial mixing contribute to increased magnetoresistance in FM/NFM/FM multilayers. Improved devices are anticipated through the controlled fabrication of
smoother and more chemically abrupt interfaces. This has
proven difficult to perfect. In the limit of thermodynamic
equilibrium 共high substrate temperature/low deposition rate兲,
the interface morphology during vapor deposition is determined by a balance between the surface and interfacial energies involved and interfacial mixing is governed by
equilibrium-solubility consideration.18 –20 Neither is conducive to the growth of the desired structure, and these growth
conditions are not used during metal-multilayer deposition.
Instead, a low homologous temperature is used to limit the
extent of interfacial diffusion. Films are then grown under
experimental conditions where the resulting morphology is
determined by kinetic effects.19,20 In this regime, the energy
barriers for surface diffusion 共in combination with processes
for overcoming them兲 control morphology. The barrier for
hopping over an atomic terrace, the Ehrlich-Schwoebel barrier, is particularly important. In thermalized atom deposition, the only activation comes either from thermal activation
provided by the background temperature or the transient
heating associated with the latent heat of each adatom condensation. Smooth films require an Ehrlich-Schwoebel barrier comparable to that for hooping on a flat surface.
Using the three-dimensional EAM as a starting potential,
two-dimensional molecular statics can be used to estimate
the effects of Au and Ag on the activation barriers for atom
migration in different pathways 共see the Appendix for details兲. The barrier for the hopping of a Cu atom on a smooth
close-packed Cu surface was found to be 0.25 eV. The corresponding energies for Ag and Au on such a copper surface
were only 0.14 and 0.11 eV, respectively. If the jump-attempt
frequencies are assumed identical, Ag and Au have a much
higher mobility on a Cu surface than Cu atoms, Fig. 9共a兲.
These molecular-statistics estimates also indicated that the
Ehrlich-Schwoebel barrier for Cu depends on which atom
type is attached to the edges of ledges. When the edge of the
ledge is either Cu, Ag, or Au, the Ehrlich-Schwoebel barrier
is 0.35, 0.28, or 0.33 eV, respectively, Fig. 9共b兲. Silver at the
ledge edge reduces the Ehrlich-Schwoebel barrier for copper
to approximately the barrier for hopping on a close-packed
surface.
Similar molecular-statistics calculations indicated that it is
energetically favorable 共by 0.15 eV/atom兲 for a Au atom on a
copper surface to exchange 共and alloy兲 with a Cu atom, Fig.
10. They also showed that it is unfavorable 共by 0.05 eV/
atom兲 for a Ag atom on the Cu surface to exchange with a
copper atom. As a result, Au is likely to be buried in Cu
whereas Ag trends to segregate to the surface, which is in
good agreement with the MD observations. The segregation
of Ag appears beneficial since it maintains a high concentration of Ag at the surface.21–25 A high Ag surface concentration and a high Ag surface mobility increase the probability
for Ag atoms to attach to ledge edges. This has the effect of
increasing the probability for other atoms to overcome the
‘‘reduced’’ Ehrlich-Schwoebel barriers. This effectively
minimizes flux shadowing and results in pinhole-free films
and smoother surfaces. Gold atoms also reduce the EhrlichSchwoebel barrier, but not as significantly as the silver, and
no Au concentration enhancement at the surface occurs.
Additional MD simulations were carried out to investigate
the best silver concentration. The results indicated that surface roughness decreased as the Ag concentration was initially increased. However, the effects were found to have
saturated at a silver concentration of about 10%. The use of
small additions of elements with a low solid solubility provides a route to control barriers for atomic assembly during
vapor deposition. Synergistic coupling with energetic deposition condition may provide new degrees of freedom to synthesize superior metal multilayer.
VI. SUMMARY
FIG. 6. The RMS roughness increased with the film thickness
for both Cu and Cu80Ag15Au5 films. The pure Cu films were much
rougher compare to the Cu80Ag15Au5 films.
Radio-frequency-diode-sputter deposition has been used
to grow Ni65Fe15Co20 /Co95Fe5 /Cu or Ni65Fe15Co20 /
Co95Fe5 /Cu80Ag15Au5 multilayers. Multilayers utilizing
Cu80Ag15Au5 as the nonferromagnetic conducting layer ex-
174418-6
SURFACTANT-MEDIATED GROWTH OF GIANT . . .
PHYSICAL REVIEW B 64 174418
FIG. 7. 共Color兲 Moleculardynamics simulated pure Cu and
Cu80Ag15Au5 films deposited on a
共111兲 Ni65Fe15Co20 substrate at
300 K. The incident energy of the
atomic flux was 0.15 eV. The incident flux angle was 45°. A pinhole
was formed in pure Cu film, but
not in the Cu80Ag15Au5 film. Rejection of silver 共but not gold兲 to
the growth surface is also evident.
hibited significant giant magnetoresistance whereas otherwise identical device architectures that use a pure copper
conducting layer fail to exhibit any magnetoresistance. The
cause is the AFM-identified presence of pinholes in copper
films but not in Cu80Ag15Au5 films. Molecular-dynamics
simulations revealed that silver acts as a surfactant reducing
the energy barriers for atomic transport into flux-shadowed
regions on the growth surface. The segregation of silver to
the growth surface creates a very high silver coverage and
enhancement of the surfactant effect. Molecular-statistics
analyses confirmed that silver and gold atoms are very mobile and can reduce the Ehrlich-Schwoebel barriers at a copper ledge. These changes both contribute to a smoother surface.
ACKNOWLEDGMENTS
We are grateful to the Defense Advanced Research
Projects Agency 共A. Tsao and S. Wolf兲 and the National
Aeronautics and Space Administration for support of this
work through NASA Grants Nos. NAGW 1692 and NAG-11964.
APPENDIX:
ESTIMATES OF SURFACE KINETIC
EFFECTS
1. Introduction
Kinetic effects associated with the deposition of a coppersilver-gold alloy can be studied using a two-dimensional
174418-7
ZOU, WADLEY, ZHOU, JOHNSON, AND BROWNELL
PHYSICAL REVIEW B 64 174418
FIG. 8. 共Color兲 Moleculardynamics
simulations
of
multilayer deposition showing
that a pinhole was formed when
pure Cu was used as the NFM
conducting layer in GMR
multilayer structure. The use of
Ag and Au alloying promotes
smoother FM on NFM interfaces
and prevents formation of pinholes.
embedded-atom-method 共EAM兲 model.26,27 Numerous cases
can be analyzed in detail to obtain insight into the effects of
the silver and gold additions to copper near its surface. The
magnitudes of the specific activation energies are not significant, but the nature of the changes due to alloying are realistically indicated.
The approach begins by obtaining three-dimensional
EAM models for the pure metals. These potentials are fitted
to their respective equilibrium lattice spacing, cohesive energy, vacancy-formation energy, bulk modulus and average
shear modulus, and cross potentials between different metals
based on an analytic equation in the literature.27 The EAM
parameters are then applied to a two-dimensional tightpacked plane of atoms. Relaxation is permitted only in the
plane, and a two-dimensional equilibrium distance is obtained for the pure metals. Atomistic configurations 共including alloying兲 are studied using molecular statistics. Twodimensional alloy lattice constants are taken as an average of
the two-dimensional lattice constants of the pure metals
weighted by their atomic concentration. With careful use of
constraints, the energy of the configuration is mapped as atomistic migration takes place to provide the requisite activation energies. With the exception of one positional component of the jumping atom, relaxation within the plane is
permitted and an iterative procedure is used to minimize the
energy of each configuration relative to its constraints. The
energy was commonly minimized at 21 configurations to obtain the energy curve for each activation energy.
174418-8
SURFACTANT-MEDIATED GROWTH OF GIANT . . .
PHYSICAL REVIEW B 64 174418
FIG. 10. Energies for exchanges between Ag and Cu atoms and
between Au and Cu atoms on a Cu surface. Ag atoms tend to float
on the Cu surface, while Au atoms tend to stay in the bulk of the Cu
film.
FIG. 9. Alloy effects on atomic assembly. Ag and Au atoms are
relatively more mobile than Cu atoms on a flat Cu surface. EhrlichSchwoebel barrier of Cu atoms is the lowest when the edge of the
ledge is Ag rather than Cu and Au.
2. Analytic model
where r e is the three-dimensional nearest-neighbor equilibrium spacing, r c is a cutoff distance less than the secondneighbor distance at which the two-body potential and the
electron density function have zero value and slope, and ␳ e is
the equilibrium electron density. In terms of the atomic volume ⍀, the cohesive energy E c , the vacancy formation energy E v , the bulk modulus B, and the Voigt average shear
modulus G, the parameters are given by
The energy of an atom ‘‘i’’ in an EAM model is the sum
over neighboring atoms ‘‘j’’ of a two-body potential ␾ an
embedding energy F based on the electron density ␳ at the
site of atom ‘‘i,’’
E i⫽
1
2
兺j ␾ 共 r i j 兲 ⫹F 共 ␳ i 兲 ,
r e⫽
冋 冉
r c ⫽r e 1⫹
共A1兲
where ␳ is a sum of the contribution f from neighboring
atoms j,
␳ i⫽
兺j
f e⫽
冋冉 冊 冉 冊册
冉 冊
冋 冉 冊 册冉 冊
r⫺r c
r e ⫺r c
f 共 r 兲⫽ f e
3
⫺3
r e r⫺r c
r r e ⫺r c
F 共 ␳ 兲 ⫽F e 1⫺ln
␳
␳e
n
r⫺r c
r e ⫺r c
2
,
␳
␳e
2
Ec
,
12⍀
共A9兲
共A10兲
共A4兲
The two-body cross potential between type ‘‘a’’ and ‘‘b’’
is given by
共A5兲
P. M. Levy, J. Magn. Magn. Mater. 140–144, 485 共1995兲.
J. C. S. Kools, J. Appl. Phys. 77, 2993 共1995兲.
3
J. Daughton, J. Brown, E. Chen, R. Beech, A. Pohm, and W.
Kude, IEEE Trans. Magn. 30, 4608 共1994兲.
4
Magnetic Multilayers and Giant Magnetoresistance: Fundamentals and Industrial Applications, edited by U. Hartmanm
共Springer, New York, 1999兲.
5
J. M. Daughton, P. A. Bade, M. L. Jenson, and M. M. M. Rahmati, IEEE Trans. Magn. 28, 2488 共1992兲.
1
共A8兲
n⫽ 冑共 9⍀B⫺15⍀G 兲 / 共 E c ⫺E v 兲关共 r c ⫺r e 兲 / 共 r e ⫹r c 兲兴 .
共A11兲
n
,
共A7兲
,
共A3兲
2
,
冊册
1/2
F e ⫽E v ⫺E c ,
The analytic forms used in the present model are
␾ 共 r 兲 ⫽⫺ ␾ e 2
Ev
2.5⍀G
共A6兲
␾ e ⫽⫺E v /6,
共A2兲
f 共 ri j 兲.
& 共 4⍀ 兲 1/3
,
2
␾ ab 共 r 兲 ⫽
6
冋
册
1 f b共 r 兲
f a共 r 兲
␾ a共 r 兲 ⫹
␾ 共r兲 .
2 f a共 r 兲
f b共 r 兲 b
共A12兲
J. Daughton and J. Brown, in Conference Proceedings of Sensors
Expo Boston, Boston, MA 共Helmers, Peterborough, NH, 1995兲,
pp. 97–105.
7
G. A. Prinz, Science 282, 1660 共1998兲.
8
B. Dieny, V. S. Speriosu, S. Metin, S. S. P. Parkin, B. A. Gurney,
P. Baumgart, and D. R. Wilhoit, J. Appl. Phys. 69, 4774 共1991兲.
9
W. F. Egelhoff, Jr., P. J. Chen, C. J. Powell, M. D. Stiles, and R.
D. McMichael, J. Appl. Phys. 79, 2491 共1996兲.
10
W. F. Egelhoff, Jr., P. J. Chen, C. J. Powell, M. D. Stiles, R. D.
174418-9
ZOU, WADLEY, ZHOU, JOHNSON, AND BROWNELL
PHYSICAL REVIEW B 64 174418
McMichael, J. H. Judy, K. Takano, and A. E. Berkowitz, J. Appl.
Phys. 82, 6142 共1997兲.
11
W. Zou, H. N. G. Wadley, X. W. Zhou, R. A. Johnson, S. Ghosal,
R. Kosut, and D. Brownell, J. Vac. Sci. Technol. A 共to be published兲.
12
W. Zou, Master thesis, University of Virginia, 1999.
13
X. W. Zhou and H. N. G. Wadley, J. Appl. Phys. 84, 2301 共1998兲.
14
X. W. Zhou and H. N. G. Wadley, J. Appl. Phys. 87, 8487 共2000兲.
15
H. N. G. Wadley, X. Zhou, R. A. Johnson, and M. Neurock, Prog.
Mater. Sci. 46, 329 共2001兲.
16
X. W. Zhou, H. N. G. Wadley, R. A. Johnson, D. J. Larson, N.
Tabat, A. Cerezo, A. K. Petford-Long, G. D. W. Smith, P. H.
Clifton, R. L. Martens, and T. F. Kelly, Acta Mater. 共to be published兲.
17
W. C. Johnson and J. M. Blakely, Interfacial Segregation, American Society for Metals Seminar on Interfacial Segregation, Chicago, 1977.
18
J. M. Howe, Interfaces in Materials: Atomic Structure, Thermodynamics and Kinetics of Solid-Vapor, Solid-Liquid and Solid-
Solid Interfaces 共Wiley, New York, 1997兲.
H. Brune, K. Bromann, and K. Kern, in Evaluation of Epitonial
Structure and Morphology, edited by A. Zangwill, D. Jesson, D.
Chambliss, and R. Clarke, MRS Symposia Proceedings No. 399
共Materials Research Society, Pittsburgh, 1996兲, p. 213.
20
M. Iwanami et al., in Thin Films–Structure and Morphology, edited by S. C. Moss, D. Ila, R. C. Cammarata, E. H. Chason, T. L.
Einstein, and E. D. Williams, MRS Symposia Proceedings No.
441 共Materials Research Society, Pittsburgh, 1997兲, p. 145.
21
Y. Liu and P. Wynblatt, Surf. Sci. 240, 245 共1990兲.
22
P. Wynblatt and Y. Liu, J. Vac. Sci. Technol. A 10, 2709 共1992兲.
23
A. Mosser and A. Jouaiti, Surf. Sci. Lett. 304, L427 共1994兲.
24
W. F. Egelhoff, Jr., P. J. Chen, C. J. Powell, M. D. Stiles, R. D.
McMichael, C. L. Lin, J. M. Sivertsen, J. H. Judy, K. Takano,
and A. E. Berkowitz, J. Appl. Phys. 80, 5183 共1996兲.
25
J. M. Daughton, U.S. Patent No. 5,617,071 共1997兲.
26
M. S. Daw and M. I. Baskes, Phys. Rev. B 29, 6443 共1984兲.
27
R. A. Johnson, Phys. Rev. B 39, 12 554 共1989兲.
19
174418-10