the study of conductance in magnetic tunnel junctions with a thin
TRANSCRIPT
The study of conductance in magnetic tunnel junctions with a thin MgObarrier: The effect of Ar pressure on tunnel magnetoresistance andresistance area productA. Zaleski, J. Wrona, M. Czapkiewicz, W. Skowroński, J. Kanak et al. Citation: J. Appl. Phys. 111, 033903 (2012); doi: 10.1063/1.3679543 View online: http://dx.doi.org/10.1063/1.3679543 View Table of Contents: http://jap.aip.org/resource/1/JAPIAU/v111/i3 Published by the American Institute of Physics. Related ArticlesElectrical transport across metal/two-dimensional carbon junctions: Edge versus side contacts AIP Advances 2, 012132 (2012) Epitaxial Cr on n-SrTiO3(001)—An ideal Ohmic contact Appl. Phys. Lett. 100, 052106 (2012) Contact transport of focused ion beam-deposited Pt to Si nanowires: From measurement to understanding Appl. Phys. Lett. 100, 053503 (2012) Comment on “Simulation of Schottky and Ohmic contacts on CdTe” [J. Appl. Phys. 109, 014509 (2011)] J. Appl. Phys. 111, 026102 (2012) Preparation of Ohmic contacts to GaAs/AlGaAs-core/shell-nanowires Appl. Phys. Lett. 100, 042103 (2012) Additional information on J. Appl. Phys.Journal Homepage: http://jap.aip.org/ Journal Information: http://jap.aip.org/about/about_the_journal Top downloads: http://jap.aip.org/features/most_downloaded Information for Authors: http://jap.aip.org/authors
The study of conductance in magnetic tunnel junctions with a thinMgO barrier: The effect of Ar pressure on tunnel magnetoresistanceand resistance area product
A. Zaleski,1,2,a) J. Wrona,1 M. Czapkiewicz,1 W. Skowronski,1 J. Kanak,1 and T. Stobiecki11Department of Electronics, AGH University of Science and Technology, Al. Mickiewicza 30,Krakow 30-059, Poland2Faculty of Physics and Applied Computer Science, AGH University of Science and Technology,al. Mickiewicza 30, Krakow 30-059, Poland
(Received 28 September 2011; accepted 23 December 2011; published online 3 February 2012)
The thickness dependence of tunnel magnetoresistance and resistance area product in Co40Fe40B20/
MgO wedge/Co40Fe40B20 magnetic tunnel junctions (MTJs) has been studied for multiple Ar
partial pressure (PAr) values during MgO sputtering. The extension of the simple equivalent circuit
model [B. Oliver et al., J. Appl. Phys. 91, 4348 (2002)] has been suggested in order to include
different transport mechanism contributions to the overall conductance of the MTJ as a function of
the MgO barrier thickness. Parameters of the model, used for quantitative description of the
conductivity of unpatterned MTJ stacks, have been analyzed as a function of PAr. VC 2012 AmericanInstitute of Physics. [doi:10.1063/1.3679543]
INTRODUCTION
Magnetic tunnel junctions (MTJs) based on CoFeB/
MgO/CoFeB trilayers are excellent candidates for future
spintronic devices, because they have a high tunnel magneto-
resistance (TMR) ratio and low resistance area (RA) prod-
uct.1 The controlled deposition of the MgO barrier in such
structures is one of the key steps to achieving optimal MTJ
parameters. The barrier smoothness, thickness (dMgO), and
(001) CoFeB/MgO/CoFeB texture play an important role,
particularly for the low RA product region with
dMgO< 1 nm, due to the complex problem of the MgO crys-
tallization on the amorphous CoFeB layer.2,3 It is well
known that optimal MgO deposition is the crucial factor in
obtaining high TMR and low RA product values.3–6
The tunneling type of carrier transport is characterized by
an exponential increase in resistance with increasing barrier
thickness that corresponds to the linear dependence of the RA
product logarithm as a function of dMgO. Fe/MgO/Fe MTJ
samples prepared using molecular beam epitaxy showed devi-
ation from a linear dependence of the RA product (in the log
scale) below dMgO¼ 1.5 nm, as reported by Yuasa et al.7
Later, it was demonstrated that such deviation may start below
dMgO¼ 1.0 nm by slightly changing the stack structure and
keeping the same growth conditions.8 For the sputtered
CoFeB/Mg/MgO/CoFeB MTJs, pure tunneling behavior was
reported for dMgO higher than 0.70 nm (see Fig. 1(b) in
Ref. 9). In our earlier studies, the deviation from a linear
dependence of the log RA product versus dMgO was similarly
observed at dMgO< 0.76 nm for sputtered CoFeB/MgO/
CoFeB wedge samples.10 Generally, it is known that, between
3 and 5 monolayers (MLs) of MgO (1 ML¼ 0.21 nm), there
is a transition between the amorphous and (001)-crystallized
MgO barrier in post-annealed MTJs.2
The production of several-MLs-thick barriers using RF
sputtering appears to be a big technological challenge.
Therefore, the characterization of inhomogeneities (e.g., hot-
spots) and defects (e.g., pinholes) in the barrier on the wafer
level is of importance.
Current-in-plane tunneling (CIPT) is a widely used tech-
nique for measuring the TMR and RA product on blanket
MTJ films without patterning.11 This technique, in combina-
tion with an X-ray diffraction and atomic force microscopy,
was used for studying the influence of PAr during MgO depo-
sition on the barrier thickness dependence of the TMR and
RA product.12 However, further analysis of the obtained data
was necessary in order to explain changes of the TMR and
RA product in the ultra-low dMgO range (i.e., below
0.76 nm). Several authors used simple equivalent circuit
models to describe transport mechanisms in the MTJ nano-
pillars with a thin tunnel barrier.13–17 This model assumes
that measured resistance of the MTJ can be expressed as a
combination of one tunneling resistance and one resistance
of the metallic nanobridge connected in parallel.
Nanobridges (or nanoclusters as they are called in
Ref. 18) are widely used in TMR read-heads to improve reli-
ability and to reduce the resistance at the cost of TMR signal
reduction.14,19 Komagaki et al.16 estimated their density in
[parts/lm2] by fitting a nanopillar breakdown voltage de-
pendence using Poisson distribution. They observed an expo-
nential decrease of nanobridge density with increasing dMgO.
In this paper, we suggest a simple approach that allows
one to describe quantitatively the effect of dMgO and PAr on
nanobridge contribution and, thereby, transport properties of
MTJ stacks, using an extended equivalent circuit model for
interpretation of CIPT results.
EXPERIMENTAL
Multilayer structures used in this study were deposited
onto thermally oxidized Si (100) wafers in a Timaris PVD
a)Author to whom correspondence should be addressed. Electronic mail:
[email protected]. Fax: þ48 (0)12 633 2398.
0021-8979/2012/111(3)/033903/5/$30.00 VC 2012 American Institute of Physics111, 033903-1
JOURNAL OF APPLIED PHYSICS 111, 033903 (2012)
cluster tool system from Singulus Technologies. All the
metallic layers were deposited by magnetron-dc-sputtering,
whereas the insulating MgO layer was rf-sputtered directly
from a sintered MgO target. The structure of the investigated
MTJs was the following: substrate/seed layers/PtMn(16)/
Co70Fe30(2.0)/Ru(0.9)/Co40Fe40B20(2.3)/MgO wedge/Co40Fe40
B20(2.3)/capping layers (thickness in nanometers). The MgO
wedge layer was sputtered using linear dynamic deposition
wedge technology. The applied working rf power density to the
MgO target during deposition was fixed at 6.6 W/cm2, while
the Ar partial pressure was varied from 1 mTorr to 15 mTorr.
After deposition, the MTJ structure was annealed in a high-
vacuum furnace at 360 �C for 2 h in a magnetic field of 10 kOe
(796 kA/m). Detailed information on sample preparation can be
found in Ref. 12.
Unpatterned MTJ wafers were characterized by CIPT
technique at room temperature. CIPT is an efficient tool,
which allows one to get information about electrical conduc-
tivity of the MTJ stack without a patterning procedure.11,20 It
uses the resistance versus magnetic field (R-H) minor loops
measured using probes with different spacing and special fit-
ting procedures20 in order to obtain values for RA product,
TMR, as well as minor loop coercivity and shift field. It is
noteworthy that information obtained using CIPT gives aver-
age values over an approximate length of 10 lm reciprocal
to wedge gradient direction. Further, we will focus on the
thickness dependence of TMR and RA product (in the low
resistance state) for samples with the MgO barrier sputtered
at PAr¼ 1, 2, 3.8, 5.6, and 15 mTorr.
Model description
An extension to an equivalent circuit model is proposed
for interpreting the relationship between the RA product and
dMgO. We have analyzed the MgO tunnel barrier thickness
dependence of the inverse of the RA product (further 1/RA
is called conductance) of the MTJ wafers measured using
CIPT. We have considered two conduction channels that cor-
respond to two resistors connected in parallel, characterized
by the tunneling and metallic transport mechanisms, respec-
tively. A peculiarity of our extended model is that we have
introduced the weight function as(dMgO), which has been
used to evaluate the contribution of tunneling and metallic
channel conductance for varied dMgO.
We have developed the following procedure in order to
evaluate as(dMgO) from the CIPT data.
Firstly, the RA product dependence on TMR has been
linearly extrapolated to the zero TMR level (TMR¼ 0%).
This gives us the value of the nanobridge resistance-area
product RA0. The procedure was first suggested by Oliver
et al. and applied for MTJs with an oxidized Al barrier.15
Similarly, the linear extrapolation of dMgO dependence on
TMR gives us the minimal nanobridge length c0. Figure 1
shows such extrapolations for all sets of CIPT data, corre-
sponding to PAr¼ 1, 2, 3.8, 5.6, and 15 mTorr. Extrapolated
data of RA0 and c0 have been shown in Table I.
Secondly, the relative contribution of nanobridges as has
been derived. Assuming that the sum of the nanobridges and
tunnel barrier (at) contributions is expressed as as þ at ¼ 1,
the total RA product is equal to21
RA ¼ RAt=at � RAs=as
RAt=at þ RAs=as¼ RAt � RAs
RAtas þ RAsat:
The nanobridge resistance-area product is given by
RAs¼RA0c0/dMgO. Values of RAt can be obtained by the lin-
ear extrapolation of log RA versus dMgO data to the low
thickness range.
Following that, the relative contribution of nanobridge
as can be expressed as
as ¼RAsðRA� RAtÞRAðRAs � RAtÞ
: (1)
FIG. 1. (Color online) (a) TMR vs RA product for all studied PAr. Linear extrapolation shows a way of determining nanobridge resistance RA0. (b) TMR vs
MgO thickness. Linear extrapolation shows a way of determining nanobridge length c0.
TABLE I. Nanobridge length c0 and nanobridge resistance RA0 as a function
of Ar partial pressure during MgO sputtering (PAr).
PAr [mTorr] c0 [nm] RA0 [Xlm2]
1.0 0.66 6 0.04 0.74 6 0.13
2.0 0.65 6 0.03 0.49 6 0.08
3.8 0.63 6 0.02 0.57 6 0.06
5.6 0.74 6 0.04 0.51 6 0.04
15.0 0.78 6 0.05 0.36 6 0.15
033903-2 Zaleski et al. J. Appl. Phys. 111, 033903 (2012)
The set of as values for different dMgO can be calculated with
Eq. (1) using the total RA product value measured by CIPT.
Thirdly, the nanobridge contributions as for a set of
measured dMgO points has been fitted with the exponential
decay function as(dMgO) in Eq. (2) in order to obtain decay
parameter t0 and exponential prefactor n.
asðdMgOÞ ¼ n � exp
�� dMgO � c0
t0
�: (2)
Finally, the conductance as a function of the barrier thick-
ness has been fitted using expression (3), that is the essence
of our extended equivalent circuit model
1
RAðdMgOÞ ¼ Gtð1� asðdMgOÞÞ � expð�2k0ðdMgO � c0ÞÞ
þ asðdMgOÞ � c0
RA0 � dMgO: (3)
The first term corresponds to the tunneling mechanism via
the MgO barrier (possibly affected by hot spots). Its contri-
bution changes continuously as a function of dMgO via pre-
factor (1 – as(dMgO)). For the high dMgO range, this prefactor
is close to one and its physical meaning is that tunneling is
the dominant mechanism of MTJ conductance.
The second term is responsible for the metallic transport
via nanobridges with different lengths for dMgO normalized
by the c0 parameter. Weight function as(dMgO) controls the
contribution of these mechanisms.
The schematic drawing of the MgO wedge and equiva-
lent circuit model have been presented in Fig. 2. The wavy
lines are a schematic representation of the interface character
as shown on a high resolution transmission electron micro-
scope (HRTEM) cross-section image (Fig. 2(b)). The
HRTEM data shows that, for the nominal thickness of the
tunnel barrier dMgO> 0.76 nm, the MgO layer is crystalline
and rather wavy, but smooth on an atomic scale (detailed
results to be presented elsewhere).
Three regions of the MgO wedged shape barrier can be
distinguished in our model. For the ultrathin dMgO below the
nanobridge length c0, the transport is purely metallic and no
tunnel contribution is present (the left side in Fig. 2(a)).
Within the medium range of dMgO, above c0, some part of
the electrons tunnel via the barrier and the rest flow via nano-
bridges, which are schematically shown in Fig. 2(a) as
shaded regions, where the wavy profiles overlap. The contri-
bution of the nanobridges decreases with increasing barrier
thickness. The extended equivalent circuit has been used to
describe conductance as a function of dMgO in this region.
Finally, for the high dMgO range (corresponding to the
HRTEM cross-section shown in Fig. 2(b)), conductance is
determined only by tunneling and there is no metallic contact
between the electrodes (the wavy profiles do not overlap).
Fitting the thickness dependence of the conductance
using expression (3), we obtain an intrinsic tunneling con-
ductance Gt and tunneling electron wave vector in the barrier
k0. We keep earlier input parameters n, c0, t0, and RA0 fixed.
Discussion of the results has been presented further on.
Free electron approximation (considering only the com-
ponents of the wave vectors that are perpendicular to the
CoFeB-MgO interface) allows one to derive the values for
the barrier height above Fermi energy Ub applying the fol-
lowing relationship, similarly to Yuasa et al.:7
Ub ¼ðk0�hÞ2
2meff: (4)
We assume effective electron mass in the barrier to be
meff¼ 0.4�m0, with m0 being the electron rest mass.22
FIG. 2. (Color online) (a) The schematic drawing of the MgO wedge cross-section of the MTJ and (b) a cross-sectional HRTEM image of a real MTJ stack
with the nominal thickness dMgO¼ 1.02 nm. The conductance of the mixed transport region above minimal nanobridge length c0 has been modeled using an
extended equivalent circuit model. The conductance in the tunneling region decreases exponentially with the dMgO thickness and exponential prefactor Gt
being the intrinsic tunneling conductance.
FIG. 3. (Color online) The tunnel barrier thickness dependence of the RA
product, PAr¼ 3.8 mTorr. The dashed line shows the tunneling fit (valid for
thick MgO above 0.76 nm); the solid line shows an extended equivalent
circuit fit (valid in the whole range).
033903-3 Zaleski et al. J. Appl. Phys. 111, 033903 (2012)
RESULTS AND DISCUSSION
Figure 3 shows an example of the tunnel barrier thickness
dependence of the RA product for PAr¼ 3.8 mTorr. The tun-
neling fit represents the known exponential dependence of the
RA product for thick enough barriers, so that the effective tun-
neling via the MgO barrier is the dominant transport mecha-
nism.1,12 This tunneling fit is analogous to the one presented
in Fig. 4 of our earlier study.12 A deviation from the linear de-
pendence of log RA was observed for dMgO< 0.76 nm for
sputtered CoFeB/MgO/CoFeB wedge samples.10
Linear extrapolations of TMR dependence of dMgO and
the RA product (see Fig. 1) allow one to obtain values for c0
and RA0, respectively. This procedure was performed with
CIPT data for all studied PAr. Obtained values for c0 and RA0
as a function of PAr have been summarized in Table I. High
errors in the case of the 15 mTorr sample have been caused
by the low quantity of experimental points. Two tendencies
are worth pointing out. For the nanobridge length analysis,
there are clearly two groups of values corresponding to the
low (PAr� 3.8 mTorr) and high pressures (PAr� 5.6 mTorr).
The lowest nanobridge length was obtained for PAr¼ 3.8 mTorr.
This stack was also considered as optimal in the earlier study.12
Nanobridge resistance shows a tendency to decrease with Ar
pressure.
Further, as has been calculated for those dMgO that have
a RA product lower than values obtained by linear extrapola-
tion of logRA(dMgO) to the low dMgO range (see Fig. 3 in
Ref. 12) and for different PAr using Eq. (1). Then, the set of
nanobridge contributions as has been fitted as a function of
dMgO using Eq. (2). The fits have been presented in Fig. 4(a).
Obtained values for decay parameter t0 for varied PAr
have been presented in Fig. 4(b). A clear distinction between
a low and high PAr range can be seen.
Further, the fits of the conductance (i.e., the inverse
resistance-area product) as a function of the barrier thickness
within an extended equivalent circuit model using Eq. (3)
and c0, RA0, n, and t0 as input parameters have been pre-
sented in Fig. 5.
Excellent agreement between fits and experimental data
was achieved. Two sets of output parameters (Gt and k0)
were obtained from those fits. Gt is the parameter describing
the intrinsic conductance in the tunneling regime (for
as¼ 0).
Figure 6 shows the influence of PAr on parameters
derived from fitting.
Parameter k0 allows us to calculate the energy barrier
height in the quasi-classical Wentzel–Kramers–Brillouin
(WKB) approximation using Eq. (4). Although it has been
shown that WKB-based models may give too low values for
the barrier height,22 we have focused on the influence of PAr
on the barrier parameters. The highest value of k0 (and Ub,
correspondingly) has been achieved for 3.8 mTorr. Inverse
intrinsic tunneling conductance (1/Gt) has a tendency to
decrease with increasing PAr in the low pressure range. For
both parameters, there is a pronounced difference between
high and low PAr ranges, which correlates with the x ray dif-
fraction data reported earlier.12 There, the MTJ samples
showed lower values of full width at high maximum rocking
curve MgO (002) peaks (higher interplanar distances) for
low pressure (up to 3.8 mTorr) contrary to the high pressure
range. Correlation between the effect of the MgO structure
properties on the wave vector in the barrier requires detailed
calculation of the influence of defect types on the band struc-
ture of CoFeB/MgO/CoFeB and is beyond the scope of this
paper.
FIG. 4. (Color online) (a) Nanobridge contribution as and (b) the decay parameter for nanobridge resistance contribution as a function of Ar partial pressure
during MgO sputtering. The dashed line is a guide for the eye separating low and high PAr ranges.
FIG. 5. (Color online) The inverse resistance-area product as a function of
MgO thickness for different Ar partial pressures during MgO sputtering
fitted using an extended equivalent circuit model.
033903-4 Zaleski et al. J. Appl. Phys. 111, 033903 (2012)
Nanobridge length c0 and nanobridge decay parameter
t0 are related to the barrier roughness averaged over the
CIPT probing area. Higher values for c0 and t0 for high pres-
sure range (Table I and Fig. 4(b)) are in concordance with
the higher roughness (RMS or Rq) parameters calculated
from the AFM scans of incomplete MTJ stacks sputtered at
PAr¼ 3.8 and 5.6 mTorr (the growth stopped at the MgO
layer) reported in Ref. 12. Similar results were obtained by
Shen et al.,5 where they observed more than double the
increase of Rq from 0.14 to 0.40 nm for 3.8 and 5.8 mTorr,
respectively.
CONCLUSIONS
In summary, we have introduced a simple procedure to an-
alyze CIPT data for MTJ wafers using an extended equivalent
circuit model, which is valid for a wide range of MgO barrier
thicknesses. Extrapolation of TMR¼ f(dMgO, RA) data to the
zero TMR level has led to values for the minimal nanobridge
length c0 and resistance RA0. Further analysis has led to the
thickness dependence of nanobridge contribution as for differ-
ent PAr. Conductance fits give us values for the tunneling elec-
tron wave vector (in the barrier) and intrinsic tunneling
conductance Gt. The suggested model, supported by structural
measurements, has allowed us to divide investigated samples
into two distinct groups deposited in low and high PAr.
ACKNOWLEDGMENTS
We would like to thank Singulus AG for multilayers
deposition, Professor S. van Dijken and Dr. L. Yao for pro-
viding us with HRTEM images. The study was co-financed
through the Swiss Contribution project NANOSPIN-PSPB-
045=2010. A.Z. was supported under SPINSWITCH Project
MRTN-CT-2006-035327 and AGH. W.S. and T.S would
like to thank the Foundation for Polish Science MPD. Pro-
gramme co-financed by the EU European Regional Develop-
ment Fund and the Polish Ministry of Science and Higher
Education grants (IP2010037970 and NN 515544538).
1S. Yuasa and D. D. Djayaprawira, J. Phys. D: Appl. Phys. 40, R337
(2008).2S. Yuasa, Y. Suzuki, T. Katayama, and K. Ando, Appl. Phys. Lett. 87,
242503 (2005).3J. Cao, J. Kanak, T. Stobiecki, P. Wisniowski, and P. Freitas, IEEE Trans.
Magn. 45, 3464 (2009).4V. Harnchana, A. T. Hindmarch, A. P. Brown, R. M. Brydson, and C. H.
Marrows, J. Phys. Conf. Ser. 241, 012039 (2010); Y. Nagamine, H. Maehara,
K. Tsunekawa, D. Djayaprawira, N. Watanabe, S. Yuasa, and K. Ando, Appl.
Phys. Lett. 89, 162507 (2006).5W. Shen, D. Mazumdar, X. Zou, X. Liu, B. D. Schrag, and G. Xiao, Appl.
Phys. Lett. 88, 182508 (2006).6I.-J. Shin, B.-C. Min, J.-P. Hong, and K.-H. Shin, IEEE Trans. Magn. 45,
2393 (2009).7S. Yuasa, T. Nagahama, A. Fukushima, Y. Suzuki, and K. Ando, Nature
Mater. 3, 868 (2004).8R. Matsumoto, A. Fukushima, K. Yakushiji, S. Yakata, T. Nagahama,
H. Kubota, T. Katayama, Y. Suzuki, K. Ando, S. Yuasa, B. Georges,
V. Cros, J. Grollier, and A. Fert, Phys. Rev. B 80, 174405 (2009).9K. Tsunekawa, D. D. Djayaprawira, M. Nagai, H. Maehara, S. Yamagata,
N. Watanabe, S. Yuasa, Y. Suzuki, and K. Ando, Appl. Phys. Lett. 87,
072503 (2005).10W. Skowronski, T. Stobiecki, J. Wrona, K. Rott, A. Thomas, G. Reiss, and
S. van Dijken, J. Appl. Phys. 107, 093917 (2010).11D. C. Worledge and P. L. Trouilloud, Appl. Phys. Lett. 83, 84 (2003).12J. Wrona, J. Langer, B. Ocker, W. Maass, J. Kanak, T. Stobiecki, and
W. Powroznik, J. Phys.: Conf. Ser. 200, 052032 (2010).13T. Seki, H. Kubota, A. Fukushima, K. Yakushiji, S. Yuasa, K. Ando,
H. Maehara, S. Yamagata, H. Okuyama, and K. Tsunekawa, J. Appl. Phys.
108, 123915 (2010).14H. Maehara, K. Nishimura, Y. Nagamine, K. Tsunekawa, T. Seki,
H. Kubota, A. Fukushima, K. Yakushiji, K. Ando, and S. Yuasa, Appl.
Phys. Express 4, 033002 (2011).15B. Oliver, Q. He, X. Tang, and J. Nowak, J. Appl. Phys. 91, 4348 (2002).16K. Komagaki, M. Hattori, K. Noma, H. Kanai, K. Kobayashi, Y. Uehara,
M. Tsunoda, and M. Takahashi, IEEE Trans. Magn. 45, 3453 (2009).17P. K. George, Y. Wu, R. M. White, E. Murdock, and M. Tondra, Appl.
Phys. Lett. 80, 682 (2002).18J. J. I. Wong, L. Ramirez, A. G. Swartz, A. Hoff, W. Han, Y. Li, and R. K.
Kawakami, Phys. Rev. B 81, 094406 (2010).19M. Takagishi, K. Yamada, H. Iwasaki, H. N. Fuke, and S. Hashimoto,
IEEE Trans. Magn. 46, 2086 (2010).20http://www.capres.com for information about CIPT and its application
notes.21H. Xi, S. Franzen, J. I. Guzman, and S. Mao, J. Magn. Magn. Mater. 319,
60 (2007).22C. W. Miller, Z.-P. Li, I. K. Schuller, R. W. Dave, J. M. Slaughter, and
J. Akerman, Phys. Rev. B 74, 212404 (2006).
FIG. 6. (Color online) (a) Electron wave vector in the MgO barrier k0, (b) barrier height Ub, and (c) inverse intrinsic tunneling conductivity 1/Gt as a function of Ar
partial pressure during MgO sputtering. Ub was calculated from the values of k0 obtained using an equivalent circuit model using quasi-classical approximation.
033903-5 Zaleski et al. J. Appl. Phys. 111, 033903 (2012)