Highly spin-polarized room-temperature tunnel injector for semiconductor spintronics using MgO(100).

Abstract

The spin polarization of current injected into GaAs from a CoFe/MgO(100) tunnel injector is inferred from the electroluminescence polarization from GaAs/AlGaAs quantum well detectors. The polarization reaches 57% at 100 K and 47% at 290 K in a 5 T perpendicular magnetic field. Taking into account the field dependence of the luminescence polarization, the spin injection efficiency is at least 52% at 100 K, and 32% at 290 K. We find a nonmonotonic temperature dependence of the polarization which can be attributed to spin relaxation in the quantum well detectors.

Full text

Page 1

Highly Spin-Polarized Room-Temperature Tunnel Injector
for Semiconductor Spintronics using MgO(100)
X. Jiang,1,2R. Wang,1,2R.M. Shelby,1R.M. Macfarlane,1S.R. Bank,2J.S. Harris,2and S.S.P. Parkin1,*
1IBM Research Division, Almaden Research Center, San Jose, California 95120, USA
2Solid State and Photonics Laboratory, Stanford University, Stanford, California 94305, USA
(Received 24 January 2004; revised manuscript received 23 November 2004; published 11 February 2005)
The spin polarization of current injected into GaAs from a CoFe=MgO?100? tunnel injector is inferred
from the electroluminescence polarization from GaAs=AlGaAs quantum well detectors. The polarization
reaches 57% at 100 K and 47% at 290 K in a 5 T perpendicular magnetic field. Taking into account the
field dependence of the luminescence polarization, the spin injection efficiency is at least 52% at 100 K,
and 32% at 290 K. We find a nonmonotonic temperature dependence of the polarization which can be
attributed to spin relaxation in the quantum well detectors.
DOI: 10.1103/PhysRevLett.94.056601PACS numbers: 72.25.Hg, 72.25.Dc, 72.25.Mk, 72.25.Rb
Devices based on the manipulation of the spin state of
electrons and holes within semiconductors are of interest
today for possible sensor, memory, and logic applications
[1,2]. A prerequisite for the realization of many of these
devices is the development of solid state spin injectors at
room temperature. The first such injectors used dilute
magnetic semiconductors [3–5] but operated only at low
temperatures. These materials, however, overcame the
conductivity mismatch between the injecting and receiving
materials, which had been recognized as an impediment to
spin injection [6]. It is now appreciated that highly con-
ducting ferromagnetic (FM)metals can alsobeusedasspin
injectors by forming a resistive tunnel contact between the
FM metal and the semiconductor [7–14]. Traditional FM
metals, such as Fe, Co, and Ni, and their alloys, are thus
particularly attractive since they exhibit high Curie tem-
peratures, and their magnetic moments can readily be
directed by magnetic engineering concepts developed dur-
ing the past decade [15]. Using amorphous Al2O3tunnel
barriers and FM metals, spin injectors have shown electro-
luminescent polarization (ELP) of ?40% at 4.5 K [16] and
?20% at 80 K [17] from semiconductor optical detectors.
Spin injectors have also been formed from Schottky tunnel
barriers with a reported ELP of ?30%, but only at low
temperatures [12,14]. In these experiments, the ELP is
limited by the tunneling spin polarization (TSP) of the
injected electrons. It was predicted that much higher TSP
could be realized for certain crystalline ferromagnet/tunnel
barrier combinations due to strongly spin-polarized eva-
nescent decay of particular wave functions through the
tunnel barrier [18–21]. Recently, Parkin et al. have re-
ported TSP values of up to 85% and room-temperature
tunneling magnetoresistance values of ?220% in CoFe=
MgO?100? tunnel junctions [22], consistent with these
predictions. In this Letter we report a spin injector formed
from a crystalline MgO(100) tunnel barrier in conjunction
with a ferromagnetic CoFe layer. Optical measurements
using quantum well (QW) detectors reveal efficient spin
injection up to 290 K. The ELP is found to vary non-
monotonically with temperature, which is due to spin
relaxation inside the QW detectors.
The spin injection efficiency was measured optically
using GaAs=AlGaAs QW light emitting diodes (LEDs).
The LEDs were grown using molecular beam epitaxy
(MBE). Two LED devices with the following structures
are discussed: p-GaAs(100) substrate/570 nm p-AlGaAs
820 830 840 850 860 870 880
Wavelength (nm)
800 805810 815820825
(b)
(c)
H = -5 T
H = 0 T
H = 5 T
Sample II
T = 290 K
VT = 2.0 V
σ+
σ-
H = -5 T
EL
i s
n
e t n I
t
) . u . a ( y
Sample I
T = 100 K
VT= 1.8 V
H = 5 T
H = 0
2 nm
CoFe
Ta
MgO
(a)
FIG. 1.
electroluminescence (EL) spectrum of samples I (b) and II (c).
The thin and thick lines in (b) and (c) represent the left (? ? )
and right (? ? ) circular polarization components of EL,
respectively.
TEM image of the CoFe=MgO spin injector (a) and
PRL 94, 056601 (2005)
PHYSICAL REVIEW LETTERS
week ending
11 FEBRUARY 2005
0031-9007=05=94(5)=056601(4)$23.00056601-1
 2005 The American Physical Society

Page 2

buffer layer/75 nm undoped AlGaAs/10 nm undoped
GaAs/15 nm undoped AlGaAs/100 nm AlGaAs upper
layer/5nm undopedGaAs,
Al0:08Ga0:92As for sample I and Al0:16Ga0:84As for
sample II. The AlGaAs upper layer is doped n type (5 ?
1016cm?3) and p type (1 ? 1017cm?3) for samples I
and II, respectively. The LEDs were passivated with
arsenic in the MBE chamber, and then transferred in air
into a magnetron sputtering chamber to grow the spin
injector, where they were heated to 550?C to remove the
arsenic cap. After the samples cooled down to ambient
temperature, shadow masks were used to deposit the tunnel
barrier (?3 nm MgO) and the FM electrode (?5 nm
Co70Fe30capped with ?10 nm Ta to prevent oxidation)
which form the spin injector. The MgO barrier was depos-
ited by reactive magnetron sputtering in an argon-oxygen
gas mixture [22]. The CoFe and Ta layers were sputtered in
pure argon gas. The active area of the spin injector was
?100 ? 300 ?m2. Finally, the LEDs were annealed in
vacuum at 300?C for 1 h.
Figure 1(a) shows a high resolution transmission elec-
tron microscopy image of the CoFe=MgO spin injector.
Both the MgO and CoFe layers are very smooth and are
polycrystallinewitha strong(100)texture along thegrowth
direction. Such a crystallographic orientation is consistent
with the theoretically predicted orientation which gives
rise to a high tunneling spin polarization [19–21].
The electroluminescence (EL) was measured in a super-
conducting magnet cryostat. With a bias voltage (VT)
applied across the LED structure, spin-polarized EL was
collected from the front side of the sample, i.e., through the
CoFe and MgO films. The ELP was measured at various
temperatures and bias voltages in a perpendicular magnetic
field (H). In this measurement geometry the electron spin
polarization is simply related to the ELP by the optical
selection rules [23].
Figures 1(b) and 1(c) show the EL spectrum of sample I
at 100 K (VT? 1:8 V) and sample II at 290 K (VT?
2:0 V), respectively. The EL peaks at longer and shorter
wavelengths are due to the heavy hole (HH) and light hole
(LH) emissions [24], respectively. For both samples, the
EL intensities of the left (I?) and right (I?) circular
components are magnetic field dependent, giving rise to
a significant ELP, as the CoFe moment is rotated out of the
film plane. The sign of the ELP indicates majority spin
injection from CoFe. Henceforth, we focus only on the HH
emission and refer to its ELP as PEL, which is equal to the
electron spin polarization just prior to recombination in the
QW [23]. In this sense, PELsets a lower bound for the spin
injection efficiency since the electrons will very likely
undergo some spin relaxation before recombination. For
sample I, the HH emission is well resolved from the LH
emission due to its narrow linewidth (?1 nm). Therefore,
it is rather straightforward to determine PEL. In contrast,
the HH and LH peaks for sample II are broad at 290 K and
whereAlGaAsis
are thus less well resolved. In order to extract PELfor this
sample, we fit the EL spectrum with two Lorentzians and
calculate PELfrom the fit, using the integrated area under
the HH peak.
The magnetic field dependences of PELfor sample I
at 100 K and sample II at 290 K are depicted in
Figs. 2(a) and 2(b) (open circles), where PELis calculated
as PEL? ?I?? I??=?I?? I??. In each case the polariza-
tion increases rapidly with field up to ?2 T,when the CoFe
moment is rotated completely out of plane. Above 2 T, PEL
continues to increase with field approximately linearly, but
at a much lower rate, reaching 57% and 47% at 5 T for
samples I and II, respectively. A linear variation of PEL
with field above 2 T (referred to as background polariza-
tion hereafter) is observed for both samples over a wide
temperature range. The slope of the background usually
varies gradually from a negative value at low temperatures
to a positive value at high temperatures, crossing zero at
?40–50 K. Several factors may contribute to the back-
ground polarization. At low temperatures, thermalization
of electron spins in the QW due to Zeeman splitting could
give rise to a negative background since GaAs has a
negative g factor. At high temperatures, however, the
Zeeman energy is negligible compared to kT, and therefore
cannot explain the observed background polarization. It is
likely due to a field dependent spin relaxation rate and/or
electron-hole recombination time. It is well known that a
perpendicular magnetic field can suppress D’yakonov-
Perel’ (DP) spin relaxation in GaAs [23], which would
therefore give rise to a positive background. Moreover, we
-60
60
-40
-20
0
20
40
60
-5 -4 -3 -2 -1012345
-60
-40
-20
0
20
40
-5 -4 -3 -2 -1
H (T)
012345
Sample I
T = 100 K
VT = 1.8 V
P
L
E
)
%
(
PC
)
%
(
CoFe electrode
Pt electrode
(a)
(c)
EL polarization
SQUID data
(d)
(b) Sample II
T = 290 K
VT = 2.0 V
FIG. 2.
[(c) and (d)] for sample I at 100 K and sample II at 290 K (open
circles). The crosses in (a) are PELof a control sample with a
Pt electrode. The solid lines in (c) and (d) show the field
dependence of the CoFe moment measured with a SQUID
magnetometer at 20 K, which has been scaled to allow com-
parison with PC.
Magnetic field dependence of PEL[(a) and (b)] and PC
PRL 94, 056601 (2005)
PHYSICALREVIEWLETTERS
week ending
11 FEBRUARY 2005
056601-2

Page 3

found that the light intensity from the QW increased with
increasing field, implying a shorter recombination time at
higher fields which would also give rise to a positive
background.
The EL polarization after subtraction of the linear back-
ground (PC) is shown in Figs. 2(c) and 2(d) (open circles),
which is a measure of spin polarization when the magnetic
field influence on the ELP is excluded. PCvalues as high as
52% and 32% were obtained at 100 and 290 K, respec-
tively. The solid lines in Fig. 2(c) and 2(d) show the field
dependence of the CoFe moment measured in a perpen-
dicular magnetic field with a superconducting quantum
interference device (SQUID) magnetometer at 20 K. The
excellent agreement between the EL and SQUID data
further confirms that the large ELP originates from spin
injection.
To rule out possible artifacts of our measurement setup,
we measured PELof a control sample, which had the same
QW detector as sample I, but had a nonmagnetic Pt elec-
trode in place of CoFe [crosses in Fig. 2(a)]. The polariza-
tion at 100 K was small, ?1%, and had a very weak field
dependence. Photoluminescence experiments with linearly
polarized pump light also gave a small polarization (<2%)
and a weak field dependence. These results proved that the
effects of polarization-dependent light absorption or reflec-
tion by the metal and semiconductor layers were very
small.
The bias and temperature dependence of PCare shown
in Fig. 3 for the two samples. The relatively small con-
finement potential of the GaAs=Al0:08Ga0:92As QW re-
sulted in weak EL signals at high temperatures and con-
sequently limited the measurements on sample I to below
100 K. In contrast, measurements on sample II were pos-
sible up to room temperature due to the use of a deeper
GaAs=Al0:16Ga0:84As QW. For both samples, PCdecreased
with increasing bias at a given temperature. A similar bias
dependence was observed in optical experiments and was
attributed to spin relaxation through the DP mechanism
before photoexcited electrons reached the QW [25,26]. In
semiconductors lacking inversion symmetry, DP spin re-
laxation occurs due to spin precession about an effective
magnetic field whose orientation and magnitude depends
on the electron momentum. Larger electron momentum at
higher bias results in a bigger effective field and conse-
quently more rapid spin relaxation [23].
A nonmonotonic temperature dependence of the ELP
was found for both samples, which is illustrated most
clearly in Fig. 4. The bias voltage is VT? 1:8 and 2.0 V
for samples I and II, respectively. The ELP depends on the
1.71.8 1.92.0
0
10
20
30
40
50
1.8 1.9 2.0 2.12.2
0
10
20
30
40
50
20 K
60 K
100 K
41 K
80 K
PC
)
%
(
Sample I
(a)
Sample II
(b)
VT (V)
20 K
140 K
260 K
80 K
200 K
VT (V)
FIG. 3.
(a) and II (b). Note the different bias ranges for (a) and (b).
Bias and temperature dependence of PCof samples I
2040 60 80 100
0
10
20
30
40
50
60
0 50 100
Temperature (K)
150 200250300
0
10
20
30
40
50
Temperature (K)
L
E

o
P
la z i r a i t on %
(
)
(a)
Sample I
VT = 1.8 V
PEL at 5 T
PC
(b)
Sample II
VT = 2.0 V
0.0 0.5 1.0
VT (V)
1.5 2.0
0
50
100
150
200
250
300
100 K
81 K
60 K
41 K
( I µ )
A
18 K
Sample I
FIG. 4.
samples I (a) and II (b). The open and closed squares correspond
to values of PELat 5 Tand of PC, respectively. Note the different
temperature ranges for (a) and (b). The inset of (a) shows the I-V
curves of sample I at various temperatures.
Temperature dependence of the EL polarization of
PRL 94, 056601 (2005)
PHYSICALREVIEW LETTERS
week ending
11 FEBRUARY 2005
056601-3

Page 4

spin relaxation rate and the electron recombination time in
the QW detectors. The DP spin relaxation rate in a QW is
given by ??1
s
/ ?pT, where ?pis the momentum relaxation
time and T is the temperature [27]. At very low tempera-
tures, ?pis dominated byionized impurity scattering which
has a weak temperature dependence, so that ?pT and,
consequently, ??1
s
increase with temperature. At higher
temperatures, when polar optical phonon scattering domi-
nates the momentum scattering, ?pT and, therefore, ??1
decrease with increasing temperature [28]. As a result, the
ELP tends to increase with temperature. The electron
recombination time also varies with temperature [29,30]
and could contribute to the temperature dependence of the
ELP. Both the spin relaxation rate and the electron recom-
bination time are dependent on the details of the QW
detectors, which likely accounts for the quantitative differ-
ences between samples I and II.
A few subtle points require further discussion. First, the
threshold voltage of the LED device decreases with in-
creasing temperature [see the inset of Fig. 4(a)]. However,
the light emission efficiency drops rapidly at high tempera-
tures. As a result, larger currents are required to obtain
enough EL signal at high temperatures. Second, the ap-
plied bias VTis across the entire LED structure. As the
temperature changes, the total voltage drop across the
MgO barrier and the n- or p-AlGaAs depletion region
(V1) can vary slightly even if VTremains constant. How-
ever, changes in V1would give rise to a monotonic tem-
perature dependence of the ELP and thus cannot account
for the experimental results. In addition, current-voltage
measurements suggest that the change of V1with tempera-
ture at a given VTis small and, therefore, could not signi-
ficantly influence the temperature dependence of the ELP.
Spin relaxation mechanisms other than the DP mecha-
nism, such as the Elliot-Yafet (EY) and Bir-Aronov-
Pikus (BAP) mechanisms [2,23], cannot account for the
increase of the ELP with temperature at higher tempera-
tures. The EY spin relaxation rate is proportional to the
momentum scattering rate and would, therefore, give rise
to a decreased ELP with increasing temperature, while
BAP relaxation is weak in undoped QWs and cannot give
rise to the observed temperature dependence. Finally, we
note that DP spin relaxation in bulk semiconductors has a
rate proportional to T3[23] and so such relaxation in the
GaAs and AlGaAs layers between the injector and the QW
is unlikely to give rise to the pronounced nonmonotonic
temperature dependence which we found.
The observation of efficient spin injection up to 290 K
using a CoFe=MgO tunnel injector is consistent with the
high Curie temperature of CoFe and the weak temperature
dependence of spin-dependent tunneling. The actual spin
injection efficiency will be higher than that inferred from
the polarization of the QWelectroluminescence because of
spin relaxation in the QW detector. Moreover, the spin
relaxation is strongly temperature dependent, thus giving
s
rise to a nonmonotonic temperature dependence of the
ELP. The MgO based spin injector can readily be fabri-
cated by sputter deposition. Moreover, the MgO barrier
prevents intermixing of the FM metal and semiconductor,
leading to improved device thermal stability [31]. These
desirable features make MgO based tunnel spin injectors
attractive for future semiconductor spintronic applications.
We thank Phil Rice for XTEM analysis and acknowl-
edge Glenn Solomon for useful discussions. We thank
DARPA SPINS for support of this work.
*Electronic address: parkin@almaden.ibm.com
[1] D.D. Awschalom, D. Loss, and N. Samarth, in Nano-
Science and Technology, edited by P. Avouris, K. von
Klitzing, H. Sakaki, and R. Wiesendanger (Springer,
Berlin, 2002).
[2] I. Zutic, J. Fabian, and S. Das Sarma, Rev. Mod. Phys. 76,
323 (2004).
[3] R. Fiederling et al., Nature (London) 402, 787 (1999).
[4] Y. Ohno et al., Nature (London) 402, 790 (1999).
[5] B.T. Jonker et al., Phys. Rev. B 62, 8180 (2000).
[6] G. Schmidt et al., Phys. Rev. B 62, R4790 (2000).
[7] E.I. Rashba, Phys. Rev. B 62, R16267 (2000).
[8] H.J. Zhu et al., Phys. Rev. Lett. 87, 016601 (2001).
[9] A.T. Hanbicki et al., Appl. Phys. Lett. 80, 1240 (2002).
[10] T. Manago and H. Akinaga, Appl. Phys. Lett. 81, 694
(2002).
[11] V.F. Motsnyi et al., Appl. Phys. Lett. 81, 265 (2002).
[12] A.T. Hanbicki et al., Appl. Phys. Lett. 82, 4092 (2003).
[13] S. Takahashi and S. Maekawa, Phys. Rev. B 67, 052409
(2003).
[14] C. Adelmann et al., cond-mat/0409103.
[15] S.S.P. Parkin et al., Proc. IEEE 91, 661 (2003).
[16] O.M.J. van ’t Erve et al., Appl. Phys. Lett. 84, 4334
(2004).
[17] V.F. Motsnyi et al., Phys. Rev. B 68, 245319 (2003).
[18] P. Mavropoulos, N. Papanikolaou, and P.H. Dederichs,
Phys. Rev. Lett. 85, 1088 (2000).
[19] W.H. Butler et al., Phys. Rev. B 63, 054416 (2001).
[20] J. Mathon and A. Umerski, Phys. Rev. B 63, 220403
(2001).
[21] X.-G. Zhang and W.H. Butler, Phys. Rev. B 70, 172407
(2004).
[22] S.S.P. Parkin et al., Nat. Mater. 3, 862 (2004).
[23] F. Meier and B.P. Zakharchenya, Optical Orientation
(North-Holland, Amsterdam, 1984).
[24] Generally speaking, electron recombination with excited
heavy holes may also contribute to the shorter wavelength
emission (so-called ‘‘LH’’ emission).
[25] H. Sanada et al., Appl. Phys. Lett. 81, 2788 (2002).
[26] E.A. Barry, A.A. Kiselev, and K.W. Kim, Appl. Phys.
Lett. 82, 3686 (2003).
[27] M.I. D’yakonov and V.Yu. Kachorovskii, Sov. Phys.
Semicond. 20, 110 (1986).
[28] V.I. Puller et al., Phys. Rev. B 67, 155309 (2003).
[29] J. Feldmann et al., Phys. Rev. Lett. 59, 2337 (1987).
[30] M. Gurioli et al., Phys. Rev. B 44, 3115 (1991).
[31] R. Wang et al., Appl. Phys. Lett. 86, 052901 (2005).
PRL 94, 056601 (2005)
PHYSICAL REVIEW LETTERS
week ending
11 FEBRUARY 2005
056601-4