JOURNAL OF APPLIED PHYSICS
VOLUME 90, NUMBER 7
1 OCTOBER 2001
Mechanisms of inert gas impact induced interlayer mixing in metal
multilayers grown by sputter deposition
X. W. Zhou and H. N. G. Wadley
Intelligent Processing of Materials Laboratory, Department of Materials Science and Engineering,
School of Engineering and Applied Science, University of Virginia, Charlottesville, Virginia 22903
共Received 20 December 2000; accepted for publication 3 July 2001兲
Control of interfacial roughness and chemical mixing is critical in nanomaterials. For example,
multilayers composed of ⬃20 Å conductive layer sandwiched between two ⬃50 Å ferromagnetic
layers can exhibit giant magnetoresistance 共GMR兲. This property has caused a tremendous recent
increase in hard disk storage capacity, and can potentially result in a new generation of nonvolatile
magnetic random access memories. It has been established that good GMR properties can be
obtained when the interfacial roughness and interlayer mixing of these multilayers are low.
However, flat interfaces in nanoscale multilayers are not thermodynamically stable, and cannot be
obtained using thermal energy deposition processes such as molecular-beam epitaxy. Hyperthermal
energy sputter deposition techniques using either plasma or ion-beam gun are able to create
nonequilibrium flat interfaces, and have been shown to produce better GMR multilayers. In these
processes, however, inert gas ions or neutrals with energies between 50 and 200 eV can impact the
growth surface. This may be a major source for interlayer mixing. By using a molecular dynamics
technique and a reduced order model, the composition profile across the thickness of multiply
repeated Ni/Cu/Ni multilayers has been calculated as a function of the energy and the relative flux
of the inert gas ions or neutrals as well as the layer thickness. The results indicate that the 50–200
eV inert gas impact caused atomic exchange between adjacent atomic layers near the surface. The
probability of exchange increased with impact energy, but decreased with the number of overlayers.
The exchange between Ni overlayer and Cu underlayer atoms was much more significant than that
between Cu overlayer and Ni underlayer atoms. As a result, the Ni on Cu interfaces were much more
diffuse than the Cu on Ni interfaces, in good agreement with experiments. At very high inert gas flux
and impact energy, an increased probability for the underlying Cu atoms to be exchanged to the
surface resulted in significant Cu surface segregation. © 2001 American Institute of Physics.
关DOI: 10.1063/1.1398073兴
low growth rates, and low adatom energies兲.7 Experimental
data have indicated that low-temperature vapor deposition
produces better GMR films because thermally activated interlayer diffusion can be reduced.8 Recent molecular dynamics 共MD兲 simulations9,10 showed that the interfacial roughness and mixing of nanoscale multilayers is also sensitive to
the adatom energy used for the deposition. Under kinetically
constrained growth conditions, these analyses revealed that a
moderate 共1 eV兲 adatom energy can be used to flatten the
interfaces of the multilayers without causing extensive interlayer mixing. This appears to explain experimental observations that sputter deposition processes 共such as rf diode and
magnetron sputtering, and ion-beam deposition兲, which create energetic metal fluxes, result in better GMR multilayers
than thermal adatom energy molecular-beam epitaxy
methods.11 MD simulations further indicated that very high
adatom incident energies could promote interlayer mixing by
an atomic exchange mechanism.9,10 This well accounts for
the observation in rf or magnetron sputter deposition12 and
ion-beam deposition13 experiments that the best GMR multilayers are grown under intermediate energy deposition conditions. These simulation studies have led to a proposal for
modulated adatom energy deposition which appears capable
of further reducing both the interfacial roughness and interlayer mixing.9,10
I. INTRODUCTION
When a thin 共15–30 Å兲 copper layer is sandwiched between a pair of thicker 共say, 50 Å兲 ferromagnetic layers, the
multilayer stack can sometimes display giant magnetoresistance 共GMR兲 manifested as a large 共5%–50%兲 change in
electrical resistance upon the application of an external magnetic field.1 This occurs because a change of the magnetic
moment alignment of the ferromagnetic layers changes the
extent of the electron-spin-dependent scattering.2– 4 GMR effects are of growing technological importance. They are used
for the read heads of hard disk drives.5 They are also being
investigated for use in a class of nonvolatile magnetic random access memories.5 Both classes of applications require a
high GMR ratio 共defined as the maximum resistance change
divided by the resistance at magnetic saturation兲, a low saturation magnetic field, and a small coercivity. These properties are optimized when the multilayers have flat interfaces
with no mixing between the layers.6
Deposition of nanoscale multilayers with flat interfaces
and low interlayer mixing has proved to be challenging.
Thermodynamics analysis indicated that flat interfaces in
nanoscale multilayers are not energetically favored, and cannot be achieved under growth conditions that promote thermodynamic equilibration 共e.g., high substrate temperatures,
0021-8979/2001/90(7)/3359/8/$18.00
3359
© 2001 American Institute of Physics
Downloaded 11 Aug 2003 to 128.143.168.139. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/japo/japcr.jsp
3360
J. Appl. Phys., Vol. 90, No. 7, 1 October 2001
X. W. Zhou and H. N. G. Wadley
significant fraction of ions are reflected as neutrals from the
target surface.15 The reflected neutrals have energies ranging
between 50 and 200 eV,15 and are emitted in an angular
distribution that enables many to reach the substrate surface.
Recent preliminary MD analysis indicated that high-energy
inert gas ion or neutral impacts with a growth surface are
likely to cause significant interlayer mixing during nanoscale
multilayer deposition.16
While both rf diode sputter and ion-beam deposition
methods have resulted in high-quality GMR films,11 they are
likely to be improved if the effects of the working inert gas
can be modeled so that better growth processes can be designed. Here, molecular dynamics has been used to investigate the effects of high-energy xenon ion or neutral impact
on the model 共111兲 Ni/Cu/Ni multilayer surfaces. This has
enabled a reduced order model to be developed that predicts
the composition profile through the thickness of repeated Ni/
Cu/Ni multilayers as a function of the impact energy and flux
of the inert gas ions or neutrals.
II. COMPUTATION METHOD
A. Interatomic potentials
FIG. 1. Schematic illustration of 共a兲 rf diode sputter and 共b兲 ion-beam deposition systems.
However, working gas ion and neutral atom fluxes are
also incident upon a growth surface during plasma sputter or
ion-beam deposition, and their effect upon interfacial structures has not been systematically investigated. To illustrate,
rf diode sputter and ion-beam deposition processes are schematically shown in Figs. 1共a兲 and 1共b兲, respectively. In the rf
diode sputter system, Fig. 1共a兲, an inert gas 共e.g., Xe兲 plasma
is initiated and maintained between a target and a substrate
by a rf power. A sheath voltage is set up near the target. Inert
gas ions created in the plasma are accelerated across the
sheath, striking the target surface at a near-normal-incident
angle. These ions can have energies between 100 and 1000
electron volts. Their impact with the target results in the
sputtering of metal atoms from the target. These sputtered
atoms are then transported to the substrate and are deposited
on the substrate surface. A second sheath is established near
the substrate, and it results in the acceleration of working gas
ions to energies of 50–200 eV.14 These ions then strike the
substrate with a near-normal-incident angle. In some conditions, the relative flux of these ions can be high 共ion-to-metal
flux ratio above 0.5兲.14 In the ion-beam deposition system,
Fig. 1共b兲, high-energy 共⬃600 eV兲 inert gas ions generated in
an ion-beam gun are used to sputter metal atoms off the
target. The sputtered metal atoms are then transported to the
substrate for the deposition. Under typical growth conditions
for GMR multilayers, the inert gas ions strike the target at an
incident angle near 45°. MD simulations indicated that at an
incident energy of 600 eV and an incident angle of 45°, a
High-quality molecular dynamics simulation of inert gas
ion or neutral impact with multilayer surface requires highfidelity interatomic potentials. An embedded atom method
共EAM兲 potential for the Cu–Ni alloy system developed by
Foiles17 was used to calculate the interaction forces between
Cu–Cu, Ni–Ni, and Cu–Ni atoms. This EAM potential was
well fitted to measured material properties such as the sublimation energies, the equilibrium lattice constants, the elastic
constants, the vacancy formation energies, and the heats of
solution. Because the EAM potential captures the local environment dependence of the atomic interaction, it realistically
describes the energetics near surfaces and interfaces. A pairwise universal potential18 was employed to calculate the interactions between a xenon ion or neutral and the metal atoms. In this potential, no distinction was made between a
xenon ion and a xenon neutral. This approximation was
adopted because the empirical universal potential was derived from experiments where both inert ions and neutrals
existed and their overall effects were measured. The universal potential was well fitted to a significant amount of experimental data, including those obtained directly from the
⬃50–200 eV inert gas bombardment of various solid
surfaces.18 It is, hence, expected to yield realistic results for
the analysis here.
B. Molecular dynamics simulation of inert ion
impacts
Interlayer mixing can only start when deposition of a
new metal begins on the surface of a previously deposited
layer. Two different interfaces, the Ni on Cu and the Cu on
Ni, exist in the Ni/Cu/Ni multilayers. The interlayer mixing
caused by the inert gas impacts with the multilayer surface
will depend upon both the type of interface and the depth of
the interface below the surface upon which inert gas ion/
atom bombardment is occurring. A molecular dynamics
model was set up to simulate inert gas ion or neutral impacts
Downloaded 11 Aug 2003 to 128.143.168.139. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/japo/japcr.jsp
J. Appl. Phys., Vol. 90, No. 7, 1 October 2001
X. W. Zhou and H. N. G. Wadley
3361
FIG. 2. Atomic configurations for a sample with one Ni monolayer above a
Cu crystal. 共a兲 before Xe impacts; 共b兲 after ten Xe impacts at 50 eV; 共c兲 after
ten Xe impacts at 100 eV; and 共d兲 after ten Xe impacts at 200 eV.
on both the Ni on Cu and the Cu on Ni surfaces, with the Ni
on Cu and the Cu on Ni interface boundaries set at different
depths below the surface. Cu and Ni underlayer crystals containing 84(112̄) planes in the x direction, 11共111兲 planes in
the y direction, and 48(11̄0) planes in the z direction were
created by assigning the atoms to the equilibrium lattice sites
共see, for example, the geometry in Fig. 2兲. Ni and Cu overlayer crystals of different thicknesses 关1–7 共111兲 planes兴
were then placed on top of the Cu and the Ni underlayers
using an epitaxial lattice extension. This procedure formed
the Ni on Cu and the Cu on Ni surfaces, with interface
boundaries located at systematically varied depths below the
surface. Inert gas atoms with an assigned kinetic energy were
injected from random locations above the samples to impact
the top 共111兲 surface at a normal-incident angle. Periodic
boundary conditions were applied in the x and z directions to
extend the lateral crystal dimensions. The bottom two atomic
planes were fixed at their equilibrium positions to prevent the
shift of the crystal during impacts. Under this condition, the
crystal can also be imagined as being extended into a bulk
共i.e., extended vertical thickness兲. A thermostat algorithm19
was applied to the two atomic planes above the fixed region
to maintain a fixed substrate temperature of 300 K. The effects of impact were then simulated by tracking the positions
of metal atoms and inert gas atoms using Newton’s equations
of motion.
III. IMPACT EFFECTS FOR THE NI ON CU SURFACE
Earlier studies have shown that when a hyperthermal
inert gas atom impacts a Cu or Ni surface, it can cause a
surface metal atom to penetrate the underlying lattice.20 This
may result in the exchange of a surface atom with an underlying atom. Because the cohesive energy of copper 共3.54兲 eV
is much lower than that of nickel 共4.45 eV兲,21 such atomic
penetration occurs more frequently in the copper lattice than
in nickel. As a result, the exchange probability of a Ni on Cu
interface has been found to exceed that of a Cu on Ni
interface.20 The exchange between overlying Ni atoms and
underlying Cu atoms is believed to be the main mechanism
FIG. 3. Atomic configurations for a sample with three Ni monolayers above
a Cu crystal. 共a兲 before Xe impacts and 共b兲 after ten Xe impacts at 200 eV.
for interlayer mixing. To explore this issue, we begin by
analyzing inert gas atom impacts with a Ni on Cu surface.
A. Atomic configurations
Molecular dynamics simulations were carried out for Xe
atom impacts with a crystal consisting of 1 ML of nickel
above a copper 共underlayer兲 crystal, Fig. 2共a兲. Ten Xe impacts were simulated, with a time interval of 10 ps between
each impact. Since the impact effect persisted a few picoseconds, the dynamics of each simulated impact was independent of the others. Representative atomic configurations after
ten Xe atom impacts for impact energies between 50 and 200
eV are shown in Figs. 2共b兲–2共d兲. In Fig. 2, dark and light
spheres represent Ni and Cu atoms, respectively. Figure 2共b兲
indicates that Xe impacts caused atomic exchanges between
Ni atoms in the Ni monolayer and underlying Cu atoms even
at the lowest energy of 50 eV. The extent of this impactinduced exchange increased with the atom impact energy.
Inert gas impacts in the hundreds of eV range with a
multilayered sample may even cause atomic exchange when
the interface boundary is several monolayers below the surface. Exemplary crystals with 3 Ni and 7 Ni monolayers
above Cu are shown in Figs. 3共a兲 and 4共a兲. Corresponding
surface configurations after ten Xe impacts at 200 eV are
shown in Figs. 3共b兲 and 4共b兲. Here, some surface monolayers
have been artificially displaced in the vertical 共y兲 direction
after the calculations to clearly expose the interface structures. It can be seen from Fig. 3共b兲 that 200 eV Xe impacts
can cause atomic exchange even when the Ni on Cu interface
boundary is 3 ML below the surface. However, by comparing with Fig. 2共d兲, the extent of mixing is seen to be greatly
reduced. Figure 4共b兲 indicates that when the number of
monolayers above the interface is increased to 7, no mixing
occurs on the Ni on Cu interface after 200 eV Xe atom
impacts.
Downloaded 11 Aug 2003 to 128.143.168.139. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/japo/japcr.jsp
3362
J. Appl. Phys., Vol. 90, No. 7, 1 October 2001
X. W. Zhou and H. N. G. Wadley
FIG. 6. Atomic configurations for a sample with one Cu monolayer above a
Ni crystal. 共a兲 before Xe impacts; 共b兲 after ten Xe impacts at 50 eV; 共c兲 after
ten Xe impacts at 100 eV; and 共d兲 after ten Xe impacts at 200 eV.
FIG. 4. Atomic configurations for a sample with seven Ni monolayers above
a Cu crystal. 共a兲 before Xe impacts and 共b兲 after ten Xe impacts at 200 eV.
where P, E, and n represent the mixing parameter, the energy
of a Xe atom, and the number of Ni monolayers above the
interface, respectively. The curves calculated using Eq. 共1兲
are also plotted in Fig. 5. They are seen to give a good fit to
the molecular dynamics simulation results.
B. Interlayer mixing
IV. IMPACT EFFECTS FOR THE CU ON NI SURFACE
To quantify the impact effects seen above, a mixing parameter was defined as the average number of exchanged Cu
atoms per Xe impact after ten Xe impacts. The mixing parameter obtained at different impact energies is plotted in
Fig. 5 as a function of the number of Ni monolayers above
the Cu underlayer. Clearly, the mixing parameter rapidly increased as the number of Ni monolayers decreased. It also
increased with the atom impact energy. It can be seen that 50
eV Xe impacts could only cause detectable mixing when the
Ni overlayer was 1 ML thick. For the 100 and 200 eV Xe
impacts, the numbers of Ni monolayers below which interfacial mixing could still occur were 3 and 6, respectively.
The mixing data can be fitted to an empirical relationship:
The results summarized in Fig. 5 indicate that inert gas
atom impacts with a Ni on Cu surface is a significant source
of interlayer mixing. For the 1 ML Ni on Cu surface, for
instance, approximately 0.7 Cu atoms is mixed in the Ni
layer during each Xe impact at 50 eV. The numbers of mixed
Cu atoms in the Ni layer per impact is increased to about 1.8
and 4.3 when the impact energy is increased to 100 and 200
eV, respectively. Clearly, the high probability of Cu mixing
in the Ni layer can cause a diffuse Cu composition profile
across the Ni on Cu interface. Because of differences in cohesive energies, the effects of Xe atom impacts with a Cu on
Ni sample are anticipated to differ from those of a Ni on Cu
sample.20 Knowledge of the impact behavior on the Cu on Ni
surface would help identify any asymmetric composition
profiles across the Ni on Cu and the Cu on Ni interfaces. The
Ni mixing in the Cu layer across the Cu on Ni interface is
also particularly important to the GMR application because
the mixed magnetic Ni atoms can destroy the magnetic coupling, resulting in ‘‘dead layers.’’ Here, separate simulations
were carried out for the Xe impact on the Cu on Ni surface.
冋冉
P⫽5.5⫻10⫺3 E 1.26 exp ⫺
冊
册
108
⫹0.122 共 n⫺1 兲 ,
E
共1兲
A. Atomic configurations
FIG. 5. Mixing probability as a function of the number of Ni monolayers
above a Cu crystal for different Xe impact energies.
The effects of ten Xe atom impacts on a sample with one
Cu monolayer above a Ni crystal are shown in Fig. 6. Like
the earlier Ni on Cu case, Fig. 6 indicates that Xe impacts
caused significant atomic exchange between Cu atoms in the
Cu monolayer and the underlying Ni atoms. This occurred
even at a Xe energy as low as 50 eV. The extent of the
exchange increased with the impact energy. However, by
comparing with Fig. 2, it can be seen that the probability of
exchange for the Cu on Ni samples is much lower than for
those where Ni was on a Cu crystal.
Downloaded 11 Aug 2003 to 128.143.168.139. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/japo/japcr.jsp
J. Appl. Phys., Vol. 90, No. 7, 1 October 2001
X. W. Zhou and H. N. G. Wadley
3363
FIG. 9. Mixing probability as a function of the number of Cu monolayers
above a Ni crystal for different Xe impact energies.
Cu on Ni interface, no mixing was observed during the simulation when impacts occurred at surfaces 6 ML above the
interface.
FIG. 7. Atomic configurations for a sample with three Cu monolayers above
a Ni crystal. 共a兲 before Xe impacts and 共b兲 after ten Xe impacts at 200 eV.
Xe atom impacts with Cu on Ni surfaces that contained
more than one Cu monolayer were also explored. The 200
eV Xe impacts on a surface containing three Cu monolayers
and a surface containing six Cu monolayers are presented in
Figs. 7 and 8. The crystals before impacts are illustrated in
Figs. 7共a兲 and 8共a兲, and the corresponding surface configurations after ten Xe impacts are shown in Figs. 7共b兲 and 8共b兲.
It can be seen from Fig. 7共b兲 that 200 eV Xe impacts also
cause atomic exchange even when the Cu on Ni interface is
buried 3 ML below the surface. However, the extent of the
exchange is noticeably lower than the corresponding Ni on
Cu case shown in Fig. 3共b兲. Figure 8共b兲 indicates that for the
B. Interlayer mixing
The mixing parameter for Cu on Ni samples was obtained from the MD simulations. Figure 9 shows this parameter as a function of the number of Cu monolayers above the
Ni crystal for the various Xe impact energies. The general
trend is similar to that of the Ni on Cu cases, Fig. 5. For 1
ML of Cu, each 50 eV Xe impact resulted in, on average,
about 0.1 exchanges. For the 100 and 200 eV Xe impacts, the
number of Cu monolayers below which significant interfacial
mixing occurs was about 2 and 4, respectively. The extent of
exchange for the Cu on Ni samples, is seen to be smaller
than that for the Ni on Cu samples. The mixing parameter
data are fitted to an empirical equation:
冋冉
Q⫽6.0⫻10⫺5 E 2.02 exp ⫺
冊
册
80
⫹0.566 共 n⫺1 兲 .
E
共2兲
The curves of Eq. 共2兲 are included in Fig. 9. They are seen to
well fit the molecular dynamics simulation results.
V. DISCUSSION
FIG. 8. Atomic configurations for a sample with six Cu monolayers above a
Ni crystal. 共a兲 before Xe impacts and 共b兲 after ten Xe impacts at 200 eV.
Molecular dynamics simulations above have indicated
that during the growth of metal multilayers, high-energy
共50–200 eV兲 Xe atom impacts with the growth surface are
likely to be a major source of interlayer mixing. To explore
ways of controlling this mixing, a method for predicting the
composition profile across an interface is highly desired. Extensive molecular dynamics calculations yielded general empirical relations of Eqs. 共1兲 and 共2兲. These are useful for
quantifying mixing parameters for Ni on Cu and Cu on Ni
interfaces as a function of both the impact energy and number of monolayers between surface and interface. A simple
reduced-order model can then be developed to calculate the
composition profile across the thickness of a multilayered
structure with an arbitrary number of Ni/Cu/Ni repeats.
The idea of the model is rather straightforward, Fig. 10.
Suppose that at a given time t during deposition, n monolayers (i⫽1,2,...,n) have been deposited on the (n⫹1)th substrate layer. The Cu and Ni compositions 共atomic fraction兲 in
Downloaded 11 Aug 2003 to 128.143.168.139. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/japo/japcr.jsp
3364
J. Appl. Phys., Vol. 90, No. 7, 1 October 2001
X. W. Zhou and H. N. G. Wadley
FIG. 11. Model predicted composition profiles across the thickness of the
Ni/Cu/Ni multilayers at different Xe impact energies and a fixed Xe:metal
ratio of 0.8.
⌬C 1 ⫽
N 1 C 2 P 1 ⫺C 1 N 2 Q 1
,
M
共4兲
and
FIG. 10. Schematic illustration of the reduced order model used to describe
atomic exchange between adjacent monolayers.
these n⫹1 monolayers can be represented as C i and N i (i
⫽1,i⫽2,...,i⫽n⫹1). To simplify the model, it is further assumed that no vacancies are created during deposition 共i.e.,
C i ⫹N i ⫽1兲. This is a reasonable assumption when the gradient of the vacancy concentration is small across the film
thickness. Under such a condition, the number of vacancies
is approximately the same in each monolayer. Because the
intent was to quantify the relative fraction of Cu and Ni in
each monolayer, we can deduct the vacancy. During each Xe
impact at the surface, there are certain probabilities at which
atomic exchanges occur between adjacent monolayers of all
n⫹1 layers shown in Fig. 10. These probabilities can be
written as P i and Q i , which represent, respectively, the number of exchanges between upper Ni and lower Cu atoms, and
between upper Cu and low Ni atoms. An index i is used here
to indicate that the exchanges occur between the ith layer
and the (i⫹1)th layer 共i⫽1, i⫽2,..., i⫽n兲. For the ith layer
where i⬎1 and i⬍n⫹1, the change of the layer Cu composition, ⌬C i , during a Xe impact, is caused by four exchanges: between Ni in the (i⫺1)th layer and Cu in the ith
layer, between Cu in the (i⫺1)th layer and Ni in the ith
layer, between Ni in the ith layer and Cu in the (i⫹1)th
layer, and between Cu in the ith layer and Ni in the (i⫹1)th
layer. Considering also the probabilities of finding the pair of
atoms to be exchanged, we can write
⌬C i ⫽
⫺N i⫺1C i P i⫺1⫹C i⫺1N i Q i⫺1 ⫹N i C i⫹1 P i ⫺C i N i⫹1 Q i
,
M
共3兲
where M is the total number of atoms per layer. For i⫽1 and
i⫽n⫹1, Eq. 共3兲 is reduced, respectively, to
⌬C n⫹1 ⫽
⫺N n C n⫹1 P n ⫹C n N n⫹1 Q n
.
M
共5兲
In Eqs. 共3兲–共5兲, P i and Q i can be obtained from Eqs. 共1兲 and
共2兲 for a given Xe energy. If the relative flux of Xe is represented by the ratio of Xe atoms to metal atoms, r, then the
growth of each metal monolayer is accompanied by M •r
times of Xe impacts. The following procedures were then
used to calculate the composition profile across a growth
multilayer thickness:
共1兲 1 ML of metal 共Cu or Ni兲 is added on the substrate.
Index is assigned to each layer and the layer compositions
are initialized 共C 1 ⫽1 for the Cu monolayer and C 1 ⫽0 for
the Ni monolayer兲.
共2兲 Eqs. 共3兲–共5兲 are used to update the layer compositions after one Xe impact.
共3兲 Step 2 is repeated a total of M •r times.
共4兲 A new metal monolayer is added on the surface. The
composition in the new layer is initialized to pure Cu for Cu
growth and to pure Ni for Ni growth. Alternative deposition
of 10 ML of Cu and 10 ML of Ni is simulated. The layer
index is reassigned after each new layer is added to ensure
that the index of the surface monolayer is 1.
共5兲 Steps 共2兲–共4兲 are repeated until the film has grown to
a desired thickness.
Using the model described above, the composition profile across the Ni/Cu/Ni multilayer thickness was explored at
different Xe energies and Xe:metal ratios. The Cu composition profiles at two Xe energies and a fixed Xe:metal ratio of
0.8 are shown in Fig. 11. It can be seen that even at the low
Xe impact energy of 50 eV, a diffused interface developed.
Because the exchange on the Ni on Cu surface is more significant than on the Cu on Ni surface, the Ni on Cu interface
is seen to be more diffuse. This phenomenon has been recently observed in three-dimensional atom probe
experiments.22 As the Xe energy was increased to 200 eV, the
interfaces were further diffused. At this energy, Cu is seen to
Downloaded 11 Aug 2003 to 128.143.168.139. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/japo/japcr.jsp
J. Appl. Phys., Vol. 90, No. 7, 1 October 2001
FIG. 12. Model predicted composition profiles across the thickness of the
Ni/Cu/Ni multilayers at different Xe:metal ratios and a fixed Xe impact
energy of 200 eV.
distribute into the entire Ni layers as no region with zero Cu
content existed. Contrarily, Ni did not distribute into the entire Cu layers as there are still regions with Cu content of 1.
This, again, was caused by significant exchange on the Ni on
Cu surface. During deposition of Ni, the underlying Cu was
continuously exchanged to the surface, which later became
the interior of the Ni layer. Another interesting finding in Fig.
11 is that at 200 eV Xe impacts, the probability for exchanging Cu onto the surface is so large that a Cu surface segregation can occur during the deposition of Ni.
The Cu composition profiles at two different Xe:metal
ratios and a fixed Xe energy of 200 eV are compared in Fig.
12. It can be seen that increasing the Xe:metal ratio has the
same effects as increasing the Xe impact energy. This is because increasing the Xe:metal ratio and Xe energy both increase the number of exchanges during deposition. Figure 12
indicates that at a small Xe:metal ratio of 0.1, regions of pure
Ni and pure Cu both existed. At the high Xe:metal ratio of 2,
the regions of pure Cu became narrower, while no regions of
pure Ni can be seen. A high exchange probability at the high
Xe:metal ratio also resulted in a Cu surface segregation during Ni deposition.
Due to the high computational cost, we could not do a
direct simulation of Xe impacts on a growth surface while
Cu or Ni were simultaneously deposited unless unrealistically high impact frequency and deposition rate are used. To
more efficiently study the effects, our simulations were carried out for atomically flat surfaces. We believe that the approximation used here may yield reasonably realistic results.
We know that the probability of atomic exchange between
adjacent planes is a function of the depth of the two planes
below the surface being impacted. If the impacts occurred on
a rough surface, then the depth of a given pair of adjacent
planes will be different at different 共lateral兲 spots on the surface. During deposition, however, this just means that the
same collision kinetics would occur at different lateral spots
at different times. A time integral would then minimize this
effect.
Hyperthermal energy sputtering deposition methods such
as rf diode or ion-beam sputter deposition techniques have
been successfully used to flatten the growth surface and,
X. W. Zhou and H. N. G. Wadley
3365
hence, to produce high-quality GMR multilayers. However,
the calculation here indicates that the 50–200 eV inert gas
atom impacts with a growth surface are likely to limit the
further improvement with these systems. Experiments have
shown that the insertion of 1–2 ML of Co between the
Ni81Fe19 /Cu interface can significantly improve the GMR
properties of the Ni81Fe19 /Cu/Ni81Fe19 multilayers.23 This
may be helpful to growth because mixing between the Co
and Ni81Fe19 alloy is not expected to affect the magnetic
properties of the layer, while the impact-induced mixing between the thermodynamically immiscible Co and Cu can be
subsequently reduced by annealing. Nevertheless, the recognition of this high-energy impact effect emphasizes the need
to reduce the high-energy atom flux. A number of ways can
be used to do this. For the rf diode sputter process, the optimization of rf power and chamber pressure14 or the application of substrate bias voltage can reduce both the flux and the
energy of inert ions impacting on the substrate. For the ionbeam sputter deposition process, a target bias technique24 can
be used to create a target sputtering at normal ion incidence
that results in much lower energy and flux of reflected neutrals than sputtering at near 45° of ion incidence. It might
also be necessary to refine the position of the substrate to
avoid the high flux zone of reflected neutrals, or to switch the
working gas from Ar to Xe because the flux and energy of
reflected Xe neutrals are both much lower than those of reflected Ar.15
VI. CONCLUSIONS
Molecular dynamics simulations of high-energy Xe impacts with Ni/Cu/Ni multilayer surfaces indicated that:
• 50–200 eV impacts caused significant exchanges between surface Ni and the underlying Cu, or surface Cu and
underlying Ni. The exchanges between surface Ni and underlying Cu were in general more significant than vice versa.
The probability of exchange increased with the impact energy.
• The probability of exchange rapidly decreased as the
number of monolayers on top of the interface increased. It is,
hence, important to reduce the probability of high-energy
impacts at the substrate surface during deposition of the first
few monolayers of a new layer.
• High-energy inert gas impacts with the growth surface
resulted in asymmetric composition profiles across the interfaces. The Ni on Cu interfaces were more diffuse than the Cu
on Ni interfaces.
• At a high inert gas impact energy or a high ratio of
inert gas atoms to metal atoms, significant exchanges between surface Ni atoms and underlying Cu atoms resulted in
a Cu surface segregation effect.
ACKNOWLEDGMENTS
The authors are grateful to the Defense Advanced Research Projects Agency 共A. Tsao and S. Wolf, Program Managers兲 and the National Aeronautics and Space Administration for support of this work through NASA Grant Nos.
NAGW 1692 and NAG-1-1964.
Downloaded 11 Aug 2003 to 128.143.168.139. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/japo/japcr.jsp
3366
1
J. Appl. Phys., Vol. 90, No. 7, 1 October 2001
S. Honda, S. Ohmoto, R. Imada, and M. Nawate, J. Magn. Magn. Mater.
126, 419 共1993兲.
2
P. M. Levy, J. Magn. Magn. Mater. 140-144, 485 共1995兲.
3
W. H. Butler, X. G. Zhang, D. M. C. Nicholson, and J. M. MacLaren,
Phys. Rev. B 52, 13399 共1995兲.
4
S. S. P. Parkin, Phys. Rev. Lett. 71, 1641 共1993兲.
5
G. A. Prinz, Science 282, 1660 共1998兲.
6
W. H. Butler 共private communication兲.
7
X. W. Zhou and H. N. G. Wadley, J. Appl. Phys. 88, 5737 共2000兲.
8
H. Kano, K. Kagawa, A. Suzuki, A. Okabe, K. Hayashi, and K. Aso, Appl.
Phys. Lett. 63, 2839 共1993兲.
9
X. W. Zhou and H. N. G. Wadley, J. Appl. Phys. 84, 2301 共1998兲.
10
X. W. Zhou and H. N. G. Wadley, J. Appl. Phys. 87, 553 共2000兲.
11
X. W. Zhou and H. N. G. Wadley 共unpublished兲.
12
J. C. S. Kools, J. Appl. Phys. 77, 2993 共1995兲.
13
S. Schmeusser, G. Rupp, and A. Hubert, J. Magn. Magn. Mater. 166, 267
共1997兲.
14
H. N. G. Wadley, W. Zou, X. W. Zhou, J. F. Groves, S. Desa, R. Kosut, E.
Abrahamson, S. Ghosal, A. Kozak, and D. X. Wang, Mater. Res. Soc.
X. W. Zhou and H. N. G. Wadley
Symp. Proc. 538, 323 共1999兲.
X. W. Zhou, S. Sainathan, and H. N. G. Wadley 共unpublished兲.
16
X. W. Zhou, W. Zou, and H. N. G. Wadley, Mater. Res. Soc. Symp. Proc.
616, 103 共2000兲.
17
S. M. Foiles, Phys. Rev. B 32, 7685 共1985兲.
18
R. E. Johnson, in Energetic Charged-Particle Interactions with Atmospheres and Surfaces, Physics and Chemistry in Space Planetology, edited
by L. J. Lanzerotti, M. Hill, and D. Stoffler 共Springer, Berlin, 1990兲, Vol.
19, p. 68.
19
X. W. Zhou, R. A. Johnson, and H. N. G. Wadley, Acta Mater. 45, 1513
共1997兲.
20
X. W. Zhou and H. N. G. Wadley, J. Appl. Phys. 87, 8487 共2000兲.
21
S. M. Foiles, M. I. Baskes, and M. S. Daw, Phys. Rev. B 33, 7983 共1986兲.
22
D. J. Larson, P. H. Clifton, N. Tabat, A. Cerezo, A. K. Petford-Long, R. L.
Martens, and T. F. Kelly, Appl. Phys. Lett. 77, 726 共2000兲.
23
S. S. P. Parkin, Appl. Phys. Lett. 61, 1358 共1992兲.
24
Internal Reports from Commonwealth Scientific Corporation, Alexandria,
Virginia 共1999兲.
15
Downloaded 11 Aug 2003 to 128.143.168.139. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/japo/japcr.jsp