Charge-to-Spin Conversion and Spin Diffusion in Bi/Ag Bilayers Observed by Spin-Polarized Positron Beam

Abstract

Charge-to-spin conversion induced by the Rashba-Edelstein effect was directly observed for the first time in samples with no magnetic layer. A spin-polarized positron beam was used to probe the spin polarization of the outermost surface electrons of Bi/Ag/Al_{2}O_{3} and Ag/Bi/Al_{2}O_{3} when charge currents were only associated with the Ag layers. An opposite surface spin polarization was found between Bi/Ag/Al_{2}O_{3} and Ag/Bi/Al_{2}O_{3} samples with the application of a charge current in the same direction. The surface spin polarizations of both systems decreased exponentially with the outermost layer thickness, suggesting the occurrence of spin diffusion from the Bi/Ag interface to the outermost surfaces. This work provides a new technique to measure spin diffusion length.

Full text

Charge-to-Spin Conversion and Spin Diffusion in Bi/Ag Bilayers
Observed by Spin-Polarized Positron Beam
H. J. Zhang,1,* S. Yamamoto,2B. Gu,3H. Li,1M. Maekawa,1Y. Fukaya,1and A. Kawasuso1
1Advanced Science Research Center, Japan Atomic Energy Agency, 1233 Watanuki, Takasaki, Gunma 370-1292, Japan
2Quantum Beam Science Directorate, Japan Atomic Energy Agency, 1233 Watanuki, Takasaki, Gunma 370-1292, Japan
3Advanced Science Research Center, Japan Atomic Energy Agency, 2-4 Shirakata Shirane, Tokai-mura, Ibaraki 319-1195, Japan
(Received 22 January 2015; revised manuscript received 12 March 2015; published 22 April 2015)
Charge-to-spin conversion induced by the Rashba-Edelstein effect was directly observed for the first
time in samples with no magnetic layer. A spin-polarized positron beam was used to probe the spin
polarization of the outermost surface electrons of Bi=Ag=Al2O3and Ag=Bi=Al2O3when charge currents
were only associated with the Ag layers. An opposite surface spin polarization was found between
Bi=Ag=Al2O3and Ag=Bi=Al2O3samples with the application of a charge current in the same direction.
The surface spin polarizations of both systems decreased exponentially with the outermost layer thickness,
suggesting the occurrence of spin diffusion from the Bi/Ag interface to the outermost surfaces. This work
provides a new technique to measure spin diffusion length.
DOI: 10.1103/PhysRevLett.114.166602 PACS numbers: 72.25.Ba, 71.70.Ej, 73.20.At, 78.70.Bj
In the last few years, increased attention has been paid to
spintronics due to its potential industrial applications to
data processing and information storage. The charge-
to-spin conversion in nonmagnetic materials, a central
issue in spintronics, is usually realized via the spin Hall
effect (SHE), the Rashba-Edelstein effect (REE), and
topological insulators [1].
The REE is the energy splitting of spin bands induced by
spin-orbit coupling and broken spatial symmetry. In a
two-dimensional (2D) electron gas system, the REE
Hamiltonian is usually expressed as HR¼αRðk׈
zÞ·σ,
where αRis the Rashba parameter, kis the electron
momentum, ˆzis the unit vector of surface normal, and σ
is the vector of the Pauli matrix [2]. Giant REE has been
found in Bi/Ag, Pd/Ag, and Sb/Ag surface alloy systems by
using angle-resolved photoemission spectroscopy [3–5].
Recently, Rojas Sánchez et al. reported the spin-to-charge
conversion due to the giant REE at the Bi/Ag interface [6].
They used microwave spin pumping to inject a spin current
from a NiFe layer into a Bi/Ag bilayer and detected the
resulting charge current. They proposed that the spin-
to-charge conversion could be ascribed to the REE
coupling at the Bi/Ag interface but not the SHE. Their
findings imply that the REE is more efficient than the SHE
to produce spin-to-charge conversion in spintronics. It is
anticipated that the charge-to-spin conversion is also
possible due to the giant REE at the Bi/Ag interface.
Positronium (Ps), which is the bound state of a positron
and an electron, can only be formed at a local region where
the electron density is low enough (typically, less than
1013 cm−2in 2D density) [7]. Therefore, in a metal,
formation of Ps is only possible at the outermost surface
(vacuum side, a few Å away from the first surface layer
[8]). There are two types of Ps: ortho-Ps (spin triplet,
jS; mi¼j1;1i,j1;−1i, and j1;0i) and para-Ps (spin
singlet, jS; mi¼j0;0i). Para-Ps decays into two γrays
with energy of ∼511 keV, and is difficult to distinguish
from free positron-electron two-γannihilation. In contrast,
ortho-Ps, which decays mostly into three γrays (the decay
possibilities into other odd numbers of γrays are negligibly
small) with energy ranging between 0 and 511 keV, is
distinguishable from two-γevents. Inspired by the exciting
progress of spintronics in the last decade, a spin-polarized
positron beam was developed in order to detect the spin
polarization of the outermost surface electrons [9,10].
The change in the ortho-Ps annihilation intensity is
obtained by integrating the intensity over part of the energy
spectrum that is below 511 keV: R¼AL=AP, where ALis
the area under the energy curve in the low energy region
(from 383 to 468 keV), and APis the area under the 2γpeak
(from 494 to 528 keV). When the Ps formation probability
is low, the increment of Rfrom R0(subscript “0”means no
Ps formation) is proportional to the ortho-Ps intensity (F3γ
Ps),
ΔR¼R−R0∝F3γ
Ps:ð1Þ
In this study, Rand R0were measured using positron
implantation energies of Eþ¼50 eV and 12 keV,
respectively.
The asymmetry of ΔRthat is induced by the spin flip of
the outermost electrons (þP−↔−P−) can be written
as [11]
ΔRðþP−Þ−ΔRð−P−Þ
ΔRðþP−ÞþΔRð−P−Þ¼2ϵð1Þ−ϵð0Þ
2ϵð1Þþϵð0ÞPþPy
−;ð2Þ
where ϵð1Þand ϵð0Þare the detection efficiencies of
annihilation γrays from j1;1iplus j1;−1iand j1;0i,
PRL 114, 166602 (2015) PHYSICAL REVIEW LETTERS week ending
24 APRIL 2015
0031-9007=15=114(16)=166602(5) 166602-1 © 2015 American Physical Society

respectively. From the known values of Pþ,ϵ, and the
experimental asymmetry, the transverse spin polarization
(Py
−) can be determined. For our detector alignment
(perpendicular to the positron beam), the factor
½2ϵð1Þ−ϵð0Þ=½2ϵð1Þþϵð0Þ equates to a constant of 0.6.
A schematic of the spin-polarized Ps annihilation experi-
ment is shown in Fig. 1(a). A transversely spin-polarized
positron beam was generated by a 22Na source
(∼370 MBq) and an electrostatic beam apparatus. The
base pressure of the positron beam apparatus was
∼6×10−8Pa. The final beam diameter was 1 mm. The
spin polarization of the positron beam, Pþ, was measured
to be 0.3 [12]. The beam was guided to inject into the center
of a sample. A reversible direct current (jc), which was
perpendicular to Pþ, was applied to the two sample ends
through two electrodes. The beam energy Eþwas adjusted
to 50 eV by a deceleration tube from the initial value of
12 keV. The center of the sample was electrically grounded.
The Ps annihilation γrays were detected by using a high-
purity Ge detector.
The component of the surface spin polarization (P−)
along the yaxis was obtained from
Py
−¼P−cos ϕ¼ΔRþjc−ΔR−jc
0.18ðΔRþjcþΔR−jcÞ;ð3Þ
where ϕis the relative angle of P−to Pþ(yaxis), ΔRþjc
and ΔR−jcare the three-γannihilation intensities that
correspond to an input charge current density of þjc
and −jc, respectively. In this experiment, the charge current
was repeatedly reversed between þjcand −jc. To deter-
mine Py
−, the averages of all ΔRþjcand ΔR−jcwere
calculated. The positive (negative) sign of Py
−corresponds
to the direction of surface spin polarization in the yaxis (−y
axis) with an input charge current of þjcin the zdirection.
Two types of Bi/Ag bilayer structures, Bið0–5Þ=
Agð25Þ=Al2O3and Agð25–500Þ=Bið8Þ=Al2O3(numbers
in round parenthesis denote film thickness in nm), were
prepared on α-Al2O3½0001substrates. As shown in
Figs. 1(b) and 1(c), both samples have the Ag layer connected
to the two electrodes of the dc power supply. To determine
the resistivity of Bi films, three Bi films (100, 200, and
500 nm) were deposited on α-Al2O3½0001substrates [13].
Consequently, the resistivity of the present Bi films was
determined to be ∼300 μΩcm, which was approximately
60 times larger than that of Ag films (∼5μΩcm). Thus, the
charge currents mainly flow in the Ag layers.
The square-shaped substrates with length of 20 mm and
width of 5–7mm, were cut from α-Al2O3½0001wafers
(mean roughness <0.1nm). All film depositions were
carried out at a substrate temperature of 300 K. The
substrates were annealed at 873 K for 30 min in a vacuum
chamber (with a base pressure of 3×10−7Pa), which was
separated from the beam apparatus. The preparation of each
Ag=Bi=Al2O3sample was completed in this chamber. First,
the Bi layer was deposited onto the substrate by thermal
deposition with Bi granules (99.9999%). Subsequently,
using rf magnetron sputtering with an Ag target (99.99%),
the Ag layer was deposited onto the Bi layer in a pure Ar
(99.999%) ambient at a pressure of 0.3 Pa. The growth rates
of Bi and Ag in this chamber were 0.1 and 1.9nm=min,
respectively. During the Bi deposition process, the Bi
thickness was monitored by using a quartz crystal thickness
monitor (SQM-160, Sigma instruments, measurement error
of 0.1nm) that was positioned close to the substrate. The
Ag=Bi=Al2O3sample was then transferred to another
chamber that was connected to the beam apparatus. The
transfer took place through air and took approximately 20
minutes. To remove any oxide layer from the sample
surface, a 1 keV Arþsputtering was utilized.
Each Bi=Ag=Al2O3sample was prepared as follows:
The Ag film was deposited onto the substrate in the
above separated chamber. Subsequently, the Ag=Al2O3
sample was transferred to the chamber that was
connected to the beam apparatus through air within 20
minutes. After cleaning the Ag surface with a 1 keV Arþ
sputtering, the Bi film was deposited at a growth rate of
∼0.05 nm=min.
The crystallinity and surface roughness of samples were
characterized by XRD patterns (SmartLab, Rigaku) and
atomic force microscopy (AFM) observation (AFM5300E,
FIG. 1 (color online). The positron beam and the samples.
(a) Schematic of the spin-polarized positron beam. The sample
layer stack of (b) Bi=Agð25Þ=Al2O3and (c) Ag=Bið8Þ=Al2O3.
The yellow blocks represent the Mo electrodes.
PRL 114, 166602 (2015) PHYSICAL REVIEW LETTERS week ending
24 APRIL 2015
166602-2

Hitachi). Figure 2shows the XRD θ−2θcurves: The Ag
film in the Agð25Þ=Al2O3sample is a polycrystal with the
(111), (200), (220), and (311) planes. The Bi layer of the
Agð25Þ=Bið8Þ=Al2O3is also polycrystalline with the (012)
and (003) planes. The Ag layer of the Agð25Þ=Bið8Þ=
Al2O3exists mainly in the (220) orientation [14]. From
AFM images, the mean roughnesses of Agð25Þ=Al2O3and
Bið8Þ=Al2O3were determined to be approximately 1.1 and
2.5 nm, respectively. Additionally, the mean grain diam-
eters of Ag in Agð25Þ=Al2O3and Bi in Bið8Þ=Al2O3were
found to be ∼40 nm from AFM measurements.
Various thicknesses of Bi layers (dBi ¼0.1, 0.2, 0.3, 0.6,
1, 2, 3, and 5 nm) were deposited on Agð25Þ=Al2O3. The
same charge current of 0.1 A (corresponding to the 2D
current density of jc¼14–19 A=m) was applied. For the
Agð25Þ=Al2O3sample, the difference of ΔRjcwas rather
small at jc∼15 A=m and could only be observed at much
higher jc. Figure 3(a) shows ΔRjcfor Agð25Þ=Al2O3at
jc¼89.3A=m. The surface spin polarization of 3.2% that
is estimated from Fig. 3(a), which corresponds to 0.5% at
jc¼15 A=m, is probably induced by the spin Hall effect
in Ag film. Figure 3(b) shows that the difference of ΔRjc
of Bið0.3Þ=Agð25Þ=Al2O3is larger than that of the
Agð25Þ=Al2O3sample, even though jcis much lower.
Figure 4shows the Bi thickness dependence of the
surface spin polarizations that normalized to the values at
jc¼15 A=m. The surface spin polarization increases from
0.5 to 0.9% with increasing dBi from 0 to 0.2 nm, reaches
4.1% at dBi ¼0.3nm, and subsequently decreases gradu-
ally for dBi >0.3nm. Considering the Bi atomic radius
(0.15 nm), dBi ¼0.3nm is approximately one monolayer.
As shown by the solid line in Fig. 4, the above Bi thickness
dependence of the surface spin polarization can be fitted by
an exponential function:
FIG. 2 (color online). XRD patterns of (a) α-Al2O3substrate,
(b) Agð25Þ=Al2O3, (c) Bið8Þ=Al2O3, and (d) Agð25Þ=Bið8Þ=
Al2O3. The “filled diamond”marks represent the imperfections
in the α-Al2O3substrates.
FIG. 3. Variation of ΔRas a function of input charge current
of þjcand −jcfor (a) Agð25Þ=Al2O3at jc¼89.3A=m,
(b) Bið0.3Þ=Agð25Þ=Al2O3at jc¼18.9A=m, and (c) Agð25Þ=
Bið8Þ=Al2O3at jc¼17.5A=m.
FIG. 4. The surface spin polarization of Bi=Agð25Þ=Al2O3
samples as a function of Bi thickness. The six data points of
Bið≥0.3Þ=Agð25Þ=Al2O3were fitted to an exponential function
of Eq. (4).
PRL 114, 166602 (2015) PHYSICAL REVIEW LETTERS week ending
24 APRIL 2015
166602-3

Py
−ðdBiÞ¼Py
−ð0.3Þexp½−0.48ðdBi −0.3Þ:ð4Þ
Similarly to the Bi=Agð25Þ=Al2O3samples, Ag layers of
different thicknesses (dAg ¼25, 100, 200, 300, 400,
500 nm) were deposited on Bið8Þ=Al2O3. The input charge
current was also regulated to 0.1 A for each sample, and the
surface spin polarization was normalized to the value at
jc¼15 A=m. As shown in Fig. 3(c), the difference of
ΔRjcis observed at dAg ¼25 nm. More importantly, its
magnitude and sign are comparable and opposite, respec-
tively, to those of Agð25Þ=Al2O3and Bi=Agð25Þ=Al2O3.
The opposite sign indicates an opposite surface spin
polarization. Figure 5shows that the surface spin polari-
zation decreases with increasing dAg. Again, this can be
fitted by an exponential function:
Py
−ðdAgÞ¼Py
−ð25Þexp½−0.0028ðdAg −25Þ:ð5Þ
The observed opposite sign in the surface spin polari-
zation and the thickness dependencies for Bi=Agð25Þ=
Al2O3and Ag=Bið8Þ=Al2O3suggest that excess electron
spins generated at the Bi/Ag interface diffuse into both Bi
and Ag layers and eventually appear at the outermost
surfaces. Also, the current-induced spin polarization within
the Ag layers of Ag=Bið8Þ=Al2O3samples is overcompen-
sated by excess and opposite spins supplied from the Bi/Ag
interface.
We assume a simple exponential form of expð−d=λsd Þ
for spin diffusion, where λsd is the spin diffusion length and
the prefactors in the exponentials of Eqs. (4) and (5)
correspond to 1=λsd. Thus, we determine a spin diffusion
length of λsdðBiÞ¼1=0.48 ≅2.1nm for the Bi layer and
λsdðAgÞ¼1=0.0028 ≅357 nm for the Ag layer. The spin
diffusion length in Bi is comparable to a recent value of
λsdðBiÞ¼1.2nm that was obtained from the inverse SHE
of a Py/Bi bilayer [15]. Also, the above λsdðAgÞof 357 nm
does not conflict with the previous reports of 132, 152, 700,
and 300 nm [16–18].
The spin diffusion length is a critical parameter in
spintronics. The present study demonstrates that it has
the potential to quantitatively characterize the spin diffu-
sion length by detecting the surface spin polarization of
samples with different thicknesses of material upon the
same film with a known value of spin polarization [such
as Agð25Þ=Bið8Þ=Al2O3].
In the study of spin-to-charge conversion in Bi/Ag
bilayers, the authors attributed the spin-to-charge conver-
sion to the inverse REE but not the inverse SHE since the
spin Hall angle of a BiAg alloy (−2.3%) has the opposite
sign to their observation [19]. Considering the fact that the
above-obtained spin diffusion lengths agree with those
reported so far, the spin polarizations on the outermost
surfaces of the Bi/Ag system observed here may be a
consequence of the REE, which is the inverse mechanism
of the one observed by Rojas Sánchez et al. with the spin
pumping method.
The charge-to-spin conversion in a sample that contains
a magnetic layer has been observed before. In 2010, Miron
et al. detected a current-driven spin torque induced by the
REE in Ptð3Þ=Coð0.6Þ=Alð1.6Þ=SiO2[20]. In 2011, they
observed the perpendicular switching of a single Co
ferromagnetic layer in the same sample [21]. Our previous
report on current-induced spin polarization of six transition
metals (Pt, Pd, Au, Cu, Ta, and W) was tentatively
explained as the surface spin accumulation due to the
REE [10]. These samples were also associated with the
magnetic layer due to the ferromagnetic property of
the nanoscaled Pt and Pd. The validity of the explanation
still remains a problem that needs to be experimentally
addressed. In this sense, the present observation of opposite
spin polarizations at opposite surfaces in Bi/Ag bilayers, as
far as we know, is the first direct observation of the REE in
a sample with no magnetic layer inside.
The spin density hδsiresulting from the REE and a
charge current is given by [22]
hδsi¼m
eαR
eℏEF
jc;ð6Þ
where m
eis the effective electron mass, eis the elementary
charge, and EFis the Fermi energy. For the Bi/Ag[111]
system, m
e¼0.35 m0(m0is the electron rest mass),
αR¼3.05 ×10−10 eVm, and EF¼0.18 eV is calculated
from the Fermi wavelength kF¼0.13 Å−1and m
e[3].
Thus, at the Bi/Ag interface, hδsi≈5×1010 cm−2for
jc¼15 A=m. On a metal surface, Ps is formed at the
vacuum side where the electron density (n2D)islow
enough, typically, less than 1013 cm−2. For the Bi surface,
n2D¼ð0.5∼4Þ×1013 cm−2at the first surface layer
[23–25], which may nearly fulfill the above Ps formation
condition. For the Ag surface, such a low electron density is
available at a vacuum region, a few Å away from the first
surface layer [8]. Therefore, an observable spin polarization
FIG. 5. The surface spin polarization of Ag=Bið8Þ=Al2O3
samples as a function of Ag thickness. The data were fitted to
an exponential function of Eq. (5).
PRL 114, 166602 (2015) PHYSICAL REVIEW LETTERS week ending
24 APRIL 2015
166602-4

Py
−is estimated to be at least ð0.1–1Þ%. Thus, the order of
magnitude of the spin polarization observed here,
Py
−¼ð4–5Þ%, could be explained by the REE.
In conclusion, we demonstrate charge-to-spin conversion
in Bi/Ag bilayers by using spin-polarized Ps annihilation
spectroscopy. Direct evidence of spin diffusion is found by
analyzing the outermost layer thickness dependence of
surface spin polarization of Bi=Ag=Al2O3and Ag=Bi=
Al2O3samples.
We are grateful to J. Ieda and S. Maekawa of JAEA, T.
Seki, K. Takanashi, and E. Saitoh of Tohoku University for
their valuable suggestions and discussions. This work was
financially supported by JSPS KAKENHI under Grant
No. 24310072 and the NSFC under Grant No. 11475130.
*zhang.hongjun@jaea.go.jp
[1] S. Maekawa, S. O. Valenzuela, E. Saitoh, and T. Kimura,
Spin Current (Oxford University Press, Oxford, 2012).
[2] Y. A. Bychkov and E. I. Rashba, J. Phys. C 17, 6039 (1984).
[3] C. R. Ast, J. Henk, A. Ernst, L. Moreschini, M. C. Falub, D.
Pacilé, P. Bruno, K. Kern, and M. Grioni, Phys. Rev. Lett.
98, 186807 (2007).
[4] C. R. Ast, G. Wittich, P. Wahl, R. Vogelgesang, D. Pacilé,
M. C. Falub, L. Moreschini, M. Papagno, M. Grioni, and
K. Kern, Phys. Rev. B 75, 201401 (2007).
[5] L. Moreschini, A. Bendounan, I. Gierz, C. R. Ast,
H. Mirhosseini, H. Höchst, K. Kern, J. Henk, A. Ernst,
S. Ostanin et al.,Phys. Rev. B 79, 075424 (2009).
[6] J. C. Rojas Sánchez, L. Vila, G. Desfonds, S. Gambarelli,
J. P. Attané, J. M. De Teresa, C. Magén, and A. Fert, Nat.
Commun. 4, 2944 (2013).
[7] A. Held and S. Kahana, Can. J. Phys. 42, 1908 (1964).
[8] N. D. Lang and W. Kohn, Phys. Rev. B 1, 4555 (1970).
[9] A. Kawasuso, Y. Fukaya, M. Maekawa, H. J. Zhang,
T. Seki, T. Yoshino, E. Saitoh, and K. Takanashi, J. Magn.
Magn. Mater. 342, 139 (2013).
[10] H. J. Zhang, S. Yamamoto, Y. Fukaya, M. Maekawa, H. Li,
A. Kawasuso, T. Seki, E. Saitoh, and K. Takanashi, Sci.
Rep. 4, 4844 (2014).
[11] D. W. Gidley, A.R. Köymen, and T. W. Capehart, Phys.
Rev. Lett. 49, 1779 (1982).
[12] A. Kawasuso and M. Maekawa, Appl. Surf. Sci. 255, 108
(2008).
[13] C. Yang, Research Project (on the Internet) in Research
Experiences for Undergraduates, Materials Physics, Florida
University, 2008 (unpublished).
[14] M. A. Majeed Khan, S. Kumar, M. Ahamed, S. A.
Alrokayan, and M. S. AlSalhi, Nanoscale Res. Lett. 6,
434 (2011).
[15] D. Hou, Z. Qiu, K. Harii, Y. Kajiwara, K. Uchida,
Y. Fujikawa, H. Nakayama, T. Yoshino, T. An, K. Ando
et al.,Appl. Phys. Lett. 101, 042403 (2012).
[16] R. Godfrey and M. Johnson, Phys. Rev. Lett. 96, 136601
(2006).
[17] T. Kimura and Y. Otani, Phys. Rev. Lett. 99, 196604
(2007).
[18] Y. Fukuma, L. Wang, H. Idzuchi, S. Takahashi,
S. Maekawa, and Y. Otani, Nat. Mater. 10, 527 (2011).
[19] Y. Niimi, H. Suzuki, Y. Kawanishi, Y. Omori, T. Valet,
A. Fert, and Y. Otani, Phys. Rev. B 89, 054401 (2014).
[20] I. M. Miron, G. Gaudin, S. Auffret, B. Rodmacq, A. Schuhl,
S. Pizzini, J. Vogel, and P. Gambardella, Nat. Mater. 9, 230
(2010).
[21] I. M. Miron, K. Garello, G. Gaudin, P.-J. Zermatten, M. V.
Costache, S. Auffret, S. Bandiera, B. Rodmacq, A. Schuhl,
and P. Gambardella, Nature (London) 476, 189 (2011).
[22] P. Gambardella and I. M. Miron, Phil. Trans. R. Soc. A 369,
3175 (2011).
[23] C. R. Ast and H. Höchst, Phys. Rev. Lett. 87, 177602
(2001).
[24] T. Hirahara, I. Matsuda, S. Yamazaki, N. Miyata, S.
Hasegawa, and T. Nagao, Appl. Phys. Lett. 91, 202106
(2007).
[25] N. Marcano, S. Sangiao, C. Magen, L. Morellon, M. R.
Ibarra, M. Plaza, L. Perez, and J. M. De Teresa, Phys. Rev. B
82, 125326 (2010).
PRL 114, 166602 (2015) PHYSICAL REVIEW LETTERS week ending
24 APRIL 2015
166602-5