Vol 458 | 26 March 2009 | doi:10.

1038/nature07879

LETTERS
Electromotive force and huge magnetoresistance in
magnetic tunnel junctions
Pham Nam Hai1, Shinobu Ohya1,2, Masaaki Tanaka1,2, Stewart E. Barnes3,4 & Sadamichi Maekawa5,6

The electromotive force (e.m.f.) predicted by Faraday’s law reflects MnAs nanoparticles as a bottom electrode, a 2.1-nm-thick AlAs tunnel
the forces acting on the charge, –e, of an electron moving through a barrier with a 1-nm-thick GaAs spacer, and a top electrode comprising a
device or circuit, and is proportional to the time derivative of the 20-nm-thick hexagonal NiAs-structure MnAs thin film. Zinc-blende
magnetic field. This conventional e.m.f. is usually absent for MnAs thin films are difficult to fabricate because the bulk phase is not
stationary circuits and static magnetic fields. There are also forces stable4. However, using phase separation techniques5–8, a GaMnAs
that act on the spin of an electron; it has been recently predicted1,2 alloy ferromagnetic semiconductor layer, when annealed at ,480 uC,
that, for circuits that are in part composed of ferromagnetic materials, generates self-assembled zinc-blende MnAs nanoparticles of diameter
there arises an e.m.f. of spin origin even for a static magnetic field. 2–3 nm. Figure 1b shows a transmission electron microscopy lattice
This e.m.f. can be attributed to a time-varying magnetization of the image of our typical MTJ. Dark areas indicated by white circles
host material, such as the motion of magnetic domains in a static correspond to zinc-blende MnAs nanoparticles in which Mn atoms
magnetic field, and reflects the conversion of magnetic to electrical have precipitated in the zinc-blende crystal structure. Spin-dependent
energy. Here we show that such an e.m.f. can indeed be induced by a tunnelling processes occur between these MnAs nanoparticles and the
static magnetic field in magnetic tunnel junctions containing zinc- top NiAs-structured MnAs film through the AlAs tunnel barrier.
blende-structured MnAs quantum nanomagnets. The observed Figure 2a shows current–voltage (I–V) characteristics of a round,
e.m.f. operates on a timescale of approximately 102–103 seconds 200-mm-diameter MTJ, measured at 3 K; data are shown for zero
and results from the conversion of the magnetic energy of the super-
paramagnetic MnAs nanomagnets into electrical energy when these
magnets undergo magnetic quantum tunnelling. As a consequence, a a 100 μm
Au
huge magnetoresistance of up to 100,000 per cent is observed for
certain bias voltages. Our results strongly support the contention Hex. MnAs film (20 nm)
that, in magnetic nanostructures, Faraday’s law of induction must A GaAs spacer (1 nm)
be generalized to account for forces of purely spin origin. The huge
AlAs barrier (2.1 nm)
magnetoresistance and e.m.f. may find potential applications in high
ZB MnAs (φ ≈ 3 nm)
sensitivity magnetic sensors, as well as in new active devices such as
nanoparticle
‘spin batteries’. GaAs matrix (10 nm)
Three ingredients are important to the observation of a large spin-
GaAs:Be (20 nm)
derived e.m.f. The first is an ensemble of superparamagnetic nanometre-
sized magnets with a large spin S < 200. Owing to a very large magnetic p+GaAs(001) substrate Insulator
anisotropy, the magnetic moment m is aligned along the z direction with
a component Sz 5 6S of the spin in this direction. A static magnetic field b
H 5 Hz splits these two ground states (with Sz 5 6S) by an energy
2mH 5 2SgmBH (where g is the g-factor and mB is the Bohr magneton).
It is this appreciable energy difference that drives the e.m.f. Second, these Hexagonal MnAs
nanomagnets constitute an essential conductive path through our mag-
netic tunnel junctions (MTJs), but have such a small capacitance C that GaAs
the Coulomb energy U 5 e2/(2C) for adding or removing electrons AlAs
exceeds the thermal energy kBT, effectively blocking sequential electrical
conduction3. However, as is commonplace, there are spin-flip channels
of many-body origin that conduct under this ‘Coulomb blockade’.
Third, for a temperature T 5 3 K, an S < 200 nanomagnet would not 10 nm
Zinc-blende GaAs:MnAs
usually relax within our ten-minute timescale. However, the spin-flip
channels mix Sz 5 2S with 2S11 and ultimately the two ground states
Figure 1 | Device structure. a, Schematic structure of our magnetic tunnel
Sz 5 6S. With the conduction of a single electron, relaxation –S R S
junctions (MTJs), comprising hexagonal (Hex.) NiAs-structure MnAs
occurs, the electron gains an energy 2SgmBH, and for an ensemble this (20 nm)/GaAs (1 nm)/AlAs (2.1 nm)/GaAs:MnAs (10 nm) thin films grown
results in a steady current driven by an e.m.f. E 5 2SgmBH/e. on a p1GaAs(001) substrate. The GaAs:MnAs film contains zinc-blende-
Normally an MTJ consists of metallic thin-film ferromagnetic structure (ZB) MnAs nanoparticles. w, diameter. b, Transmission electron
electrodes and a thin tunnel barrier made of an insulator. The MTJs microscopy lattice image of an MTJ. White circles indicate the areas of zinc-
in this study are unique (Fig. 1a); they consist of zinc-blende-structured blende MnAs nanoparticles.
1
Department of Electrical Engineering and Information Systems, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8656, Japan. 2Japan Science and Technology Agency,
4-1-8 Honcho, Kawaguchi-shi 332-0012, Japan. 3Physics Department, University of Miami, Coral Gables, Florida 33124, USA. 4TCM, Cavendish Laboratory, University of Cambridge,
Cambridge CB3 0HE, UK. 5Institute for Materials Research, Tohoku University, Sendai 980-8577, Japan. 6CREST, Japan Science and Technology Agency, Tokyo 100-0075, Japan.

489
©2009 Macmillan Publishers Limited. All rights reserved
LETTERS NATURE | Vol 458 | 26 March 2009

a 0.6 Figure 2 | Transport characteristics of an MTJ. a, Current–voltage (I–V)

characteristics of a 200-mm-diameter MTJ, measured at 3 K without (black
T=3K
curve) and with (red curve) a magnetic field of 10 kG applied in plane along
H // GaAs[110]
0.4 the GaAs[110] azimuth. In a magnetic field of 10 kG, the origin of the I–V
H = 0 kG
curve was shifted towards positive bias by 21 mV, corresponding to an e.m.f.
H = 10 kG generated by the MTJ. Inset, voltage dependence of the MTJ resistance at
e.m.f. zero field. A Coulomb blockade of U < 50 meV was observed. b, Current
0.2
versus magnetic field of an MTJ measured at 10, 15 and 20 mV. Inset,
induced e.m.f. as a function of magnetic field H measured by the
compensation method (see text). Error bars reflect a systematic uncertainty
I (µA)

0.0
due to the time dependence of the e.m.f. c, Voltage across an external
3 resistance (RL) connected directly to the MTJ in a magnetic field without a
current source. Black and red curves are the output voltages of the MTJ when
–0.2 the external resistance is respectively infinitely large (that is, open circuit)
R (MΩ) 2
and 200 kV. The grey curve is the black curve but reflected about the vertical
1 axis. The inset shows that the voltage lasts for at least 10 min.
–0.4
0
–100 –50 0 50 100 MTJ. With a static magnetic field, this e.m.f. causes the top MnAs film
V (mV) and the bottom MnAs nanoparticles to have positive and negative
–0.6
–80 –60 –40 –20 0 20 40 60 80 potentials, respectively. Effects of the e.m.f., induced by a static mag-
V (mV)
netic field, can also be clearly seen in Fig. 2b, which shows the
b 0.10 magneto-current measured at positive bias voltages smaller than
T=3K 21 mV. For low fields, the current is positive and increases with
H // GaAs[110] increasing magnetic field, revealing an ordinary negative magnetore-
20 mV
0.05
sistance. However, for fields greater than about 5 kG, the current
15 mV
decreases with increasing magnetic field, and eventually becomes
negative. The reversed current at 10 kG indicates that the induced
e.m.f. is larger than the applied bias voltage. When the external voltage
0.00 compensates the internal e.m.f., no current flows in the circuit. The
25 10 mV
I (µA)

inset in Fig. 2b shows the magnetic-field dependence of the induced

20 e.m.f. measured by this compensation method.
e.m.f. (mV)

–0.05 15 Figure 2c shows the voltage drop across an external resistance (RL)
10 connected directly to the MTJ as a function of magnetic field (‘spin
battery’ configuration). The black curve is the output voltage of the
5 MTJ when the external resistance is infinitely large (that is, open
–0.10
0 circuit), and the red curve is the voltage drop across an external
0 2 4 6 8 10
H (kG) resistance of 200 kV. For finite fields, a voltage appears across the load
resistance, demonstrating that the MTJ indeed constitutes a ‘spin
–0.15
–8 –4 0 4 8 battery’. As seen in the inset of Fig. 2c, the voltage lasted for at least
H (kG) 10 min, implying a decay constant t < 102–103 s. When the magnetic
c 10
8
field was turned off, the voltage immediately disappeared, showing that
T=3K
6 Open circuit H // GaAs[110]
the intrinsic time constant RC of the MTJ is short. Indeed, using a
resistance R < 106 V and a capacitance C < NC0—where the number
V (mV)

Open circuit
4 H = 10 kG
8 1 2 3 4 of nanoparticles is N < 109 and the capacitance of a single nanoparticle
2 R = 200 kΩ
L
is C0 5 e2/2U < 10218 F (U 5 50 meV, as discussed later, is the
0 Coulomb energy of zinc-blende MnAs nanoparticles)—RC is expected
0 2 4 6 8 10 to be about 1023 s. The induced magnetic e.m.f. continually pumps
6 Time (min)
charges through the nanoparticles, thereby maintaining the output
V (mV)

voltage. Thermoelectric effects are not responsible for the observed

(+) e.m.f., as discussed in Supplementary Information.
4 V RL Figures 3a and 3b show the magnetoresistance of the MTJ measured
(–)
with bias voltages V of 220, 215, 210, 25 and 21 mV. For
RL =
200 kΩ V 5 220 mV, a large magnetoresistance ratio (defined as
2 (Rmax 2 R10kG)/R10kG) of 1,220% was observed. This ratio is equal to
(I10kG 2 Imin)/Imin, which is the appropriate figure of merit for this
device when used as a magnetic field sensor. With decreasing V, the
magnetoresistance ratio increased, reaching 100,000% for the smallest
0
–12 –8 –4 0 4 8 12 bias voltage (21 mV). This is three orders of magnitude greater than
H (kG) conventional giant magnetoresistance (GMR)9,10 or tunnelling magne-
toresistance (TMR)11–14, and comparable to the colossal magnetore-
sistance (CMR) observed in some manganese perovskites15,16. An even
magnetic field (black curve), and for a 10 kG magnetic field (red larger magnetoresistance ratio might be obtained if the junction was
curve) applied along the easy axis of magnetization of the top MnAs biased with a voltage smaller than –1 mV. This huge magnetoresistance
thin film (that is, along the [1120]MnAs and [110]GaAs axes in the is a consequence of the combined effect of the induced e.m.f. and
film plane). The polarity of the applied bias voltage is defined as the Coulomb blockade in our zinc-blende MnAs nanoparticles. The inset
voltage of the top MnAs film with respect to the substrate. The I–V in Fig. 2a shows that the resistance of the MTJ increased sharply when
curve, for 10 kG, is shifted towards the positive bias voltage region by jVj , 50 mV, implying a Coulomb energy U < 50 meV. The resistance
21 mV. This shift corresponds to an e.m.f. of 21 mV generated by the at small bias is therefore quite large, as electron tunnelling through the
490
©2009 Macmillan Publishers Limited. All rights reserved
NATURE | Vol 458 | 26 March 2009 LETTERS

b 107
a 4,000 MR ≈ 105%
H // GaAs[110]
T=3K –5 mV 106
3,500
–1 mV

R (Ω)
105
3,000

104
2,500

R (kΩ) 2,000
103
–10 –8 –6 –4 –2 0 2 4 6 8 10
H (kG)
1,500 c 3
–5 mV
–10 mV 2
1,000

(dR/dH)/R (kG–1)
–15 mV 1
500
–20 mV 0
0 –1
–10 –8 –6 –4 –2 0 2 4 6 8 10
H (kG) –2

–3
–10 –8 –6 –4 –2 0 2 4 6 8 10
H (kG)

Figure 3 | Huge magnetoresistance. Magnetoresistance of the MTJ effect. c, (dR/dH)/R at V 5 25 mV. There are evident peaks corresponding
measured at V 5 220, 215, 210, 25 mV (a) and –1 mV (b). A huge to the level crossings (see text) around 0, 1.5, 3.0 and 4.5 kG, marked by the
magnetoresistance (MR) as large as 100,000% is observed. This huge value is coloured arrows. The white arrows show the positions of steep change of
a consequence of the induced e.m.f. combined with the Coulomb blockade resistance.

nanoparticles is almost suppressed by the Coulomb blockade. The the number of channels and in mobility, which occur at each level
e.m.f. induced by a magnetic field causes the MTJ to be self-biased in crossing, are reflected in transport properties. When the field
the same direction as the applied bias voltage, with a concomitant direction is reversed, the dominant tunnelling direction of magneti-
dramatic reduction of the effective resistance. zation is also reversed. Therefore, an initial asymmetric population of
An e.m.f. induced by a static magnetic field was first theoretically zinc-blende MnAs nanomagnets, with opposite magnetization direc-
predicted for magnetic nanostructures, such as magnetic nanowires tions, will result in an asymmetric V(H) curves (Fig. 2c).
or spin-valves1,2. Viewed macroscopically, this motive force reflects We examine the e.m.f. data for branches with identical slopes,
the energy conservation requirements17,18. A domain of magnetic 2gmBS, but displaced by a constant amount 2gmBSHa, and we also
moment m has an energy 1mH when anti-parallel to the field but look for sudden changes in resistance with this same period. In
2mH when parallel. In an anti-parallel field, an electron induces the Fig. 3a we indeed see there are such changes in the resistance and
transition 1m R 2m. The magnetic energy change 2mH is transferred that they are separated by approximately 1.5 kG (corresponding to
to the electrical sector, generating an e.m.f. which drives a current in Ha). In Fig. 3c we show the resistance derivative with evident equally
an external load17. spaced peaks indicated by the coloured arrows. The e.m.f. data
The experimental Coulomb blockade energy of 50 meV is suffi- (Fig. 2c) show equally evident branches. For positive H there are
ciently large that it almost totally suppresses transport via the two e.m.f. open circuit (black curve) branches, which extrapolate
single-electron states. The result is a very large zero field resistance.
In addition to these electronic levels, there is a set of levels reflecting ---
the spin state jSzæ, where Sz is the angular momentum quantum num-
ber of the nanomagnet order parameter of magnitude S. Such nano- –(S – 2)
S–4
magnets have been much studied19. Figure 4 illustrates the relevant
energy scheme for a dominant uniaxial anisotropy, and the well-
–(S – 1)
known phenomenology is briefly described in the legend. Electrons S–3
AS2
can tunnel through the nanomagnet to the bottom electrode, despite
the blockade, provided that the magnetization (S state) of the nano- –S
magnet also tunnels between the magnetic levels of Fig. 4. If the elec- 2gµBS(H – 2Ha)
S–2
tron is added to an initial magnetic state j2Sæ, this can couple to an
entirely similar but empty state with the equivalent magnetic ground
2gµBS(H – Ha)
state jSæ and which differs in energy only by the relatively small S–1
Zeeman term 2gmBSH. The theory, outlined in Supplementary
Information, shows that this amount of energy is given to, or taken 2gµBSHa
2gµBSH
from, the tunnelling electron, resulting in an e.m.f. S
Equating our open circuit e.m.f. of 21 mV with 2gmBSH/e implies
that S < 200. With increasing field H, when the energy of the state Figure 4 | Magnetic energy. The magnetic levels have, approximately, an
j2Sæ is greater than that of jS21æ, it becomes possible to conduct energy 2ASz2, with A being the uniaxial magnetic anisotropy constant.
using these states, which involves tunnelling from –S to S21. Now There are two equivalent ground states | 6Sæ separated by a large energy
the magnetic matrix elements are larger, and the conductivity greater. barrier ,AS2. Magnetic tunnelling occurs between the states | 6Sæ. A
magnetic field adds gmBSzH to the energy and lifts the state | 2Sæ above | Sæ by
The e.m.f. now corresponds to the difference in energy of these states, 2gmBSH and effectively suppresses tunnelling. Level crossings, and
as in Fig. 4, and is reduced by 2gmBSHa, where Ha is the anisotropy tunnelling, occur for periodic values of field with period 2gmBSHa, where
field. Depending upon the real or effective load, with increasing H Ha 5 A/gmB is the anisotropy field. At such crossings the magnetic relaxation
many new potential e.m.f. branches open as a function of H, and each rate is enhanced19. In our experiments these same level crossings open new
is displaced relative to its neighbour by 2gmBSHa. Clearly, changes in transport channels and lead to new e.m.f. branches.
491
©2009 Macmillan Publishers Limited. All rights reserved
LETTERS NATURE | Vol 458 | 26 March 2009

back to approximately 3.0 kG and 4.5 kG. Only the 3.0 kG branch is 3. Fulton, T. A. & Dolan, G. J. Observation of single-electron charging effects in small
tunnel junctions. Phys. Rev. Lett. 59, 109–112 (1987).
evident for negative field. Because the external load is then zero, the 4. Sanvito, S. & Hill, A. N. Ground state of half-metallic zinc-blende MnAs. Phys.
first missing branch appears when the internal e.m.f. is compensated Rev. B 62, 15553–15560 (2000).
by an external bias. The data are shown in the inset of Fig. 2b and 5. Sato, K., Katayama–Yoshida, H. & Dederichs, P. H. High Curie temperature and
roughly agree with the extrapolation of the green dashed line nano-scale spinodal decomposition phase in dilute magnetic semiconductors.
Jpn. J. Appl. Phys. 44, L948–L951 (2005).
(labelled 1) in Fig. 2c. When a load of 200 kV is added, the e.m.f. is 6. Moreno, M., Trampert, A., Jenichen, B., Daweritz, L. & Ploog, K. H. Correlation of
reduced, but it remains the case that the similarly identified branches structure and magnetism in GaAs with embedded Mn(Ga)As magnetic
extrapolate back to the same fields, which in turn correspond to the nanoclusters. J. Appl. Phys. 92, 4672–4677 (2002).
coloured arrows in the transport data shown in Fig. 3; this lends 7. Yokoyama, M., Yamaguchi, H., Ogawa, T. & Tanaka, M. Zinc-blende-type MnAs
nanoclusters embedded in GaAs. J. Appl. Phys. 97, 10D317 (2005).
strong support to our model of magnetic co-tunnelling. The relaxa-
8. Kwiatkowski, A. et al. Structure and magnetism of MnAs nanocrystals embedded
tion times of superparamagnetic systems are notoriously long19 and in GaAs as a function of post-growth annealing temperature. J. Appl. Phys. 101,
quite compatible with the observed lifetime (,102–103 s) of our 113912 (2007).
e.m.f. The magnetization measurement results described in 9. Binasch, G., Grünberg, P., Saurenbach, F. & Zinn, W. Enhanced magnetoresistance
in layered magnetic structures with antiferromagnetic interlayer exchange. Phys.
Supplementary Information strongly support our interpretation of
Rev. B 39, 4828–4830 (1989).
the behaviour of zinc-blende MnAs nanomagnets. 10. Baibich, M. N. et al. Giant magnetoresistance of (001)Fe/(001)Cr magnetic
We have shown that a static magnetic field results in an e.m.f. superlattices. Phys. Rev. Lett. 61, 2472–2475 (1988).
compatible with predictions. This cannot reflect a time derivative 11. Moodera, J. S., Kinder, L. R., Wong, T. M. & Meservey, R. Large magnetoresistance
at room temperature in ferromagnetic thin film tunnel junctions. Phys. Rev. Lett.
of the magnetization (hM/ht) and the usual Faraday’s law of induc-
74, 3273–3276 (1995).
tion as this would change sign with the applied field. It lends strong 12. Miyazaki, T. & Tezuka, N. Giant magnetic tunnelling effect in Fe/Al2O3/Fe
support to the contention that, in magnetic nanostructures, junction. J. Magn. Magn. Mater. 139, L231–L234 (1995).
Faraday’s law must reflect forces due to both the electric and mag- 13. Parkin, S. S. P. Giant tunnelling magnetoresistance at room temperature with
MgO (100) tunnel barriers. Nature Mater. 3, 862–867 (2004).
netic dipole charges of the electrons. The resulting huge magnetore-
14. Yuasa, S., Nagahama, T., Fukushima, A., Suzuki, Y. & Ando, K. Giant room-
sistance together with the e.m.f. may have potential applications as a temperature magnetoresistance in single-crystal Fe/MgO/Fe magnetic tunnel
completely new kind of magnetic sensor with ultrahigh sensitivity, junctions. Nature Mater. 3, 868–871 (2004).
and as active devices such as ‘spin batteries’. 15. Jin, S. et al. Thousandfold change in resistivity in magnetoresistive La-Ca-Mn-O
films. Science 264, 413–415 (1994).
16. Tokura, Y. (ed.) Advances in Condensed Matter Science Vol. 2, Colossal
METHODS SUMMARY Magnetoresistance Oxides (Gordon & Breach, 2000).
Our MTJs consist of ferromagnetic hexagonal NiAs-structure MnAs (20 nm)/ 17. Barnes, S. E., Ieda, J. & Maekawa, S. Magnetic memory and current amplification
GaAs (1 nm)/AlAs (2.1 nm)/GaAs:MnAs (10 nm) thin films, grown by molecular devices using moving domain walls. Appl. Phys. Lett. 89, 122507 (2006).
beam epitaxy (MBE) on p1GaAs(001) substrates. To fabricate the zinc-blende 18. Barnes, S. E. & Maekawa, S. Currents induced by domain wall motion in thin
MnAs nanoparticles, we grew a 10-nm-thick Ga0.94Mn0.06As thin film at 240 uC ferromagnetic wires. Preprint at Æhttp://arxiv.org/abs/cond-mat/0410021v1æ
(2004).
and then annealed at 480 uC for 20 min in the MBE growth chamber, during which
19. Barbara, B. et al. Mesoscopic quantum tunnelling of magnetization. J. Magn. Magn.
phase separation occurred in the GaMnAs layer and the zinc-blende MnAs nano- Mater. 140–144, 1825–1828 (2002).
particles of diameter 2–3 nm were formed in the GaAs matrix. After the growth, we
fabricated round, 200-mm-diameter mesa diode structures for transport measure- Supplementary Information is linked to the online version of the paper at
ments. The MTJs were mounted and cooled to 3 K in a refrigerator unit (JANIS www.nature.com/nature.
Research) equipped with an external electromagnet (0–10 kG). Transport mea- Acknowledgements This work was partly supported by Grant-in-Aids for Scientific
surements were carried out with a conventional two-terminal direct-current (d.c.) Research No. 18106007, No. 19048018 and No. 20686002, the Special
method using an Advantest R6246 two-channel voltage/current source/monitor. Coordination Programs for Promoting Science and Technology, and R&D for
An Advantest R6871E multi-meter was used for monitoring the voltage drop Next-Generation Information Technology by MEXT, PRESTO of JST, and EPSRC
across the external resistance connected to the MTJ with a magnetic field (the ‘spin (UK). We thank B.-H. Yu for his help in the transport measurements. P.N.H.
battery’ configuration). The sweep rate of the magnetic field was 10 kG min21. acknowledges a JSPS Research Fellowship for Young Scientists and the Global COE
Program (CO4).
Full Methods and any associated references are available in the online version of Author Contributions P.N.H. designed the experiment, fabricated the samples,
the paper at www.nature.com/nature. collected most of data and performed analysis of data; S.O. set up measurement
apparatuses and gave experimental advice; M.T. managed and planned the
Received 5 August 2008; accepted 12 February 2009. research and supervised the experiment; and S.E.B. and S.M. developed the
Published online 8 March 2009. theoretical explanation of the experiment. All authors discussed the results and
commented on the manuscript.
1. Barnes, S. E. & Maekawa, S. Generalization of Faraday’s law to include
nonconservative spin forces. Phys. Rev. Lett. 98, 246601 (2007). Author Information Reprints and permissions information is available at
2. Barnes, S. E. Spin motive forces, ‘‘measurements’’, and spin-valves. J. Magn. Magn. www.nature.com/reprints. Correspondence and requests for materials should be
Mater. 310, 2035–2037 (2007). addressed to M.T. ([email protected]).

492
©2009 Macmillan Publishers Limited. All rights reserved
doi:10.1038/nature07879

METHODS
Our MTJs consist of ferromagnetic hexagonal NiAs-structure MnAs (20 nm)/
GaAs (1 nm)/AlAs (2.1 nm)/GaAs:MnAs (10 nm) thin films. The GaAs:MnAs
film contains zinc-blende-structure MnAs nanoparticles in a GaAs matrix
(Fig. 1a). These structures are grown by MBE on p1GaAs(001) substrates.
First, we grew a 20-nm-thick Be-doped GaAs buffer layer on a p1GaAs(001)
substrate at 580 uC. After cooling the substrate temperature to 240 uC, we grew a
10-nm-thick Ga0.94Mn0.06As film, a 2.1-nm-thick AlAs tunnel barrier and then a
1-nm-thick GaAs spacer. Then the structure was annealed at 480 uC for 20 min in
the MBE growth chamber, during which phase separation occurred in the
GaMnAs layer and the zinc-blende MnAs nanoparticles of diameter 2–3 nm were
formed in the GaAs matrix. When annealed at higher temperature (.550 uC),
the more stable hexagonal NiAs phase precipitates. By inserting a very thin GaAs
layer between AlAs and MnAs, the quality of the top MnAs film is improved.
Furthermore, by observing the reconstruction in the reflection high-energy
electron diffraction (RHEED) patterns of the GaAs layer, we can control the
annealing temperature more precisely. Finally, a 20-nm-thick type-A hexagonal
NiAs-structure MnAs thin film was grown at 240 uC as a top electrode. Here, the
epitaxial relationship of type-A MnAs on the semiconductor layer is
(1100)MnAs // (001)GaAs, [1120]MnAs // [110]GaAs, and [0001]MnAs //
[110]GaAs, and the easy magnetization axis is [1120]MnAs in the film plane20,21.
After completing the growth, post-growth annealing was carried out in the
growth chamber at 310 uC for 10 min to improve the crystal quality of the top
MnAs film.
After the initial growth, we fabricated, by standard photolithography and
chemical etching, round, 200-mm-diameter mesa diode structures. We spin-
coated an insulating negative resist on the sample, opened a contact hole with
180 mm diameter on top of the mesa, and fabricated a metal electrode by evaporat-
ing Au on this surface. The MTJs were mounted and cooled to 3 K in a refrigerator
unit (JANIS Research) equipped with an external electromagnet (0–10 kG).
Transport measurements were carried out with a conventional two-
terminal direct-current (d.c.) method using an Advantest R6246 two-channel
voltage/current source/monitor. An Advantest R6871E multi-meter was used for
monitoring the voltage drop across the external resistance connected to the MTJ
with a magnetic field (the ‘spin battery’ configuration). The sweep rate of the
magnetic field was 10 kG min21. The results presented here (e.m.f. and huge mag-
netoresistance) were observed in many MTJs fabricated by the same process.

20. Tanaka, M. et al. Epitaxial orientation and magnetic properties of MnAs thin films
grown on (001) GaAs: Template effects. Appl. Phys. Lett. 65, 1964–1966 (1994).
21. Tanaka, M. Ferromagnet (MnAs)/III–V semiconductor hybrid structures.
Semicond. Sci. Technol. 17, 327–341 (2002).

©2009 Macmillan Publishers Limited. All rights reserved