Appl. Phys. Lett., Vol. 85, No. 9, 30 August 2004

1. APPLIED PHYSICS LETTERS VOLUME 85, NUMBER 9 30 AUGUST 2004
Exchange biasing of the ferromagnetic semiconductor Ga1−xMnxAs
K. F. Eid, M. B. Stone, K. C. Ku, O. Maksimov, P. Schiffer, and N. Samartha)
Department of Physics and Materials Research Institute, Pennsylvania State University, University Park,
Pennsylvania 16802
T. C. Shih and C. J. Palmstrøm
Department of Chemical Engineering and Materials Science, University of Minnesota, Minneapolis,
Minnesota 55455
(Received 10 December 2003; accepted 13 July 2004)
We demonstrate the exchange coupling of a ferromagnetic semiconductor ͑Ga1−xMnxAs͒ with an
overgrown antiferromagnet ͑MnO͒. Unlike most conventional exchange biased systems, the
blocking temperature of the antiferromagnet ͑TB = 48± 2 K͒ and the Curie temperature of the
ferromagnet ͑TC = 55.1± 0.2 K͒ are comparable. The resulting exchange bias manifests itself as a
clear shift in the magnetization hysteresis loop when the bilayer is cooled in the presence of an
applied magnetic field and an enhancement of the coercive field. © 2004 American Institute of
Physics. [DOI: 10.1063/1.1787945]
Ferromagnetic semiconductors (FMSs) are the focus of Samples are grown by low temperature molecular beam
extensive research for applications in semiconductor epitaxy on (001) semi-insulating, epi-ready GaAs substrates
spintronics.1,2 These hybrid materials offer new functionality bonded to a mounting block by indium. After thermal de-
because their carrier-mediated ferromagnetism can be ma- sorption of the oxide layer, a high temperature GaAs layer is
nipulated by employing established techniques used in semi- grown under standard conditions, followed by a low tem-
conductor technology.3 Several device geometries have been perature GaAs layer grown at 250° C. The magnetically ac-
demonstrated using the “canonical” FMS Ga1−xMnxAs, in- tive region of the sample consists of a 10-nm-thick
cluding magnetic tunnel junctions,4–6 spin polarized light Ga1−xMnxAs layer ͑x = 0.08͒, capped with a thin Mn layer
emitting diodes,7 and spin-tunable resonant tunnel diodes.8 It whose thickness is either ϳ4 nm or ϳ8 nm. After the growth
is important in this context to develop a means of “exchange of the Ga1−xMnxAs layer, the sample is moved from the
biasing” ferromagnetic semiconductors for device applica- growth chamber into an adjoining high vacuum buffer cham-
tions. This entails exchange coupling the FMS to a proximal ber and the As cell is cooled to room temperature. Once the
antiferromagnet (AFM), and has been observed in ferromag- As pressure in the growth chamber decreases to an accept-
netic (FM) metals, intermetallic alloys, and oxides.9–12 able level, the substrate is reintroduced into the growth
The hallmark of an exchange biased system is a hyster- chamber, and the Mn capping layer is grown at room tem-
esis loop which is not centered about zero magnetic field, but perature. Alternatively, the sample is transferred in vacuum
is shifted either to positive or negative fields depending upon to an adjacent As-free growth chamber for the Mn capping.
the orientation of the applied field while cooling the sample Exchange-biased samples have been obtained using both
through the blocking temperature ͑TB͒ of the AFM layer. schemes. These precautions are to avoid the formation of FM
This is typically accompanied by an increase in the coercive MnAs during the growth of the capping layer.
field ͑HC͒ of the FM. In most examples of exchange bias, The epitaxial growth is monitored by in situ reflection
TB Ͻ TN Ӷ TC, where TN is the Néel temperature of the AFM high energy electron diffraction (RHEED) at 12 keV. The
Ga1−xMnxAs surface shows a clear reconstructed ͑1 ϫ 2͒
and TC is the Curie temperature of the FM. The FMS
Ga1−xMnxAs is of current interest because TC can be varied RHEED pattern; the Mn surface shows a weaker, yet streaky
and unreconstructed pattern whose symmetry is suggestive
from ϳ10 to ϳ150 K by varying parameters such as Mn
of the stabilization of a cubic phase of Mn. The thickness of
concentration and sample thickness;13 in principle, this al-
the Ga1−xMnxAs layer is calculated from RHEED oscilla-
lows studies of exchange bias in a FM where TB can be
tions, while the thickness of the Mn overlayer is estimated
above or below TC.
from RHEED oscillations of MnAs whose growth rate is
We are aware of only one reported attempt to exchange
mainly determined by the sticking coefficient of Mn. The Mn
bias a FMS, using MnTe/ Ga1−xMnxAs bilayers.14 Although
concentration in the ferromagnetic layer is estimated at x
the FMS showed an increased HC, there was no conclusive ϳ 0.08 based upon a calibration as a function of the Mn cell
evidence for any exchange coupling between the FMS and temperature ͑T = 785° C͒.16 Magnetization measurements are
the AFM. We demonstrate here unambiguous exchange bias performed using a superconducting quantum interference de-
in a Ga1−xMnxAs epilayer on which a thin layer of Mn has vice (SQUID) magnetometer, where samples are either field
been deposited. Rutherford backscattering scattering (RBS) cooled or zero field cooled from room temperature to the
measurements show that—upon removal from the vacuum measuring temperature in a finite or zero magnetic field, re-
chamber—the layer of Mn oxidizes to form MnO (an AFM spectively.
with TN = 118 K 15). The postgrowth removal of the samples from the mount-
ing block requires a modest annealing cycle for a few min-
a)
Electronic mail: nsamarth@phys.psu.edu utes at ϳ200° C during which the top Mn layer oxidizes to
0003-6951/2004/85(9)/1556/3/$22.00 1556 © 2004 American Institute of Physics
Downloaded 27 Aug 2004 to 129.169.10.56. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp

2. Appl. Phys. Lett., Vol. 85, No. 9, 30 August 2004 Eid et al. 1557
FIG. 1. Hysteresis loops of a Ga1−xMnxAs ͑t = 10 nm͒ / MnO͑t ϳ 4 nm͒ bi-
layer measured at T = 10 K, after cooling in the presence of different mag-
netic fields: (a) H = 2500 Oe, (b) H = −2500 Oe, and (c) H = 0. In (d), we
show the hysteresis loop measured at T = 10 K for an uncapped Ga1−xMnxAs
͑t = 15 nm͒ control sample, after field cooling in H = 1000 Oe. The diamag-
netic and/or paramagnetic background has been subtracted from these (and
FIG. 2. Panels (a)–(c) provide a summary of the temperature dependence of
all subsequently presented) hysteresis loops.
the hysteresis loops and the hysteresis loop parameters for the sample de-
scribed in Fig. 1. The hysteresis loop at any given temperature is measured
after cooling the sample from T = 100 K to the measurement temperature in
form MnO. The nature of the Mn layer is probed by RBS a field of H = 2500 Oe. Panels (a) and (b) show the field cooled hysteresis
analysis using 2.3 and 1.4 MeV He+ ions with both normal loop at T = 5 K and T = 30 K, respectively. Panel (c) shows the temperature
and glancing angle detector geometries, corresponding to dependence of the exchange field, HE = ͉͑HC− + HC+͒ / 2͉, and the coercive
field, HC = ͉͑HC− − HC+͒ / 2͉ for T Ͻ TN. The inset panel (d) shows the
scattering angles of 165° and 108°, respectively. Both ran- temperature-dependent magnetization at H = 20 Oe, after field cooling in H
dom and ͗001͘ channeling measurements are conducted. The = 10 kOe.
channeling measurements show that the Mn overlayer is
completely oxidized with stoichiometry close to MnOx (x
ϳ 1 to 1.5); the thickness of this MnO layer is consistent origin. As the temperature approaches TC, the hysteresis loop
with the estimate from RHEED oscillations. collapses into a single nonhysteretic curve of zero coercivity
Figure 1(a) and 1(b) depict two hysteresis loops for the (not shown), indicating that the Ga1−xMnxAs layer has be-
Ga1−xMnxAs ͑10 nm͒ / MnO ͑4 nm͒ bilayer, measured at T come paramagnetic. We do not observe any training effects
= 10 K after cooling the sample in a magnetic field of H for a hysteresis loop performed multiple times at a particular
= ± 2500 Oe. Out of a total of 12 samples, similar data are temperature after a single field cooling of the sample.
reproduced for three other samples, with MnO layer thick- Figure 2(c) summarizes the temperature dependence of
ness of both 4 and 8 nm. The “horizontal” offset of the hys- the hysteresis loop parameters HE and HC. From this figure,
teresis loops from the origin in a direction opposite to that of we estimate TB = 48± 2 K. The temperature dependence of
the cooling field is a clear signature of exchange coupling in the saturation magnetization [Fig. 2(d)] indicates that TC
the bilayer. Control measurements are used to rule out ex- = 55.1± 0.2 K; both the value of TC and the magnitude of the
trinsic effects: Figure 1(c) shows that cooling the sample in saturation magnetization are similar for the uncapped and
zero applied magnetic field results in a slightly biased hys- capped samples, indicating that the Mn overlayer does not
teresis loop. The presence of some bias suggests that a small affect the FM state of the Ga1−xMnxAs layer except through
remnant magnetic field was present in the SQUID while the exchange biasing. Fig. 2(c) shows that HE undergoes a
cooling the sample. Such a field partially magnetizes the FM monotonic decrease with temperature until vanishing at TB,
layer, which in turn sets the bias in the AFM layer. The while HC passes through a broad peak near TB before becom-
sheared appearance of the magnetization curve suggests that ing zero at TC. The absence of a second peak in HE at TC and
there are multiple domains in the FM and that the sample of any exchange bias in the paramagnetic state further sug-
was very close to being zero-field and zero-magnetization gest that TC Ͼ TB.18
cooled. Figure 1(d) shows that the exchange bias shift is Figure 3 shows the dependence of HE and HC on the
absent in a sample of similar thickness and grown under magnitude of the cooling field. Cooling fields as small as
identical conditions, but without the AFM overlayer. There is H = 25 Oe produce a significant exchange bias, and both HE
also a notable increase of the coercive field for the and HC vary only slightly with the magnitude of cooling field
Mn-capped sample compared to typical values for uncapped up to H ӷ HE, HC. These data indicate that once exchange
samples,17 which is consistent with the expected effects of bias has been established, it becomes insensitive to the mag-
exchange biasing.9 nitude of the cooling field. Establishing exchange bias re-
Figure 2(a) and 2(b) compare the field-cooled hysteresis quires only a field large enough to saturate the ferromagnetic
loops at 5 and 30 K, respectively. In these plots, the sample layer at temperatures above TB. Although the coupling en-
is cooled from T Ͼ TC to the measuring temperature in a field ergy in this system ͑ϳ3 ϫ 10−3 erg/ cm2͒ is small compared
of H = 2500 Oe. With increasing temperature, the hysteresis to that in typical exchange bias systems, we observe rela-
loops become narrower, and their centers move closer to the tively large exchange bias fields due to the relatively low
Downloaded 27 Aug 2004 to 129.169.10.56. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp

3. 1558 Appl. Phys. Lett., Vol. 85, No. 9, 30 August 2004 Eid et al.
Nonetheless, the demonstrated feasibility of exchange bias-
ing of this FMS and the identification of the accompanying
difficulties provides an important starting point for additional
experiments.
This research has been supported by the DARPA-SPINS
program under Grant Nos. N00014-99-1093, -99-1-1005,
-00-1-0951, and -01-1-0830, by ONR N00014-99-1-0071
and by NSF DMR 01-01318. We thank J. Bass for useful
discussions.
1
H. Ohno, in Semiconductor Spintronics and Quantum Computation, edited
by D. D. Awschalom, D. Loss, and N. Samarth, (Springer, Berlin, 2002),
p. 1.
2
FIG. 3. Hysteresis loop parameters at T = 10 K, as a function of the cooling S. A. Wolf, D. D. Awschalom, R. A. Buhrman, J. M. Daughton, S. von
field for the sample described in Fig. 1. Measurements are taken after cool- Molnar, M. L. Roukes, A. Y. Chtchelkanova, and D. M. Treger, Science
ing from T = 100 K to T = 10 K in the respective magnetic fields. Note that 294, 1488 (2001).
3
the field axis is split at H = 1.1 kOe and is depicted on different scales for H. Ohno, A. Shen, F. Matsukura, A. Oiwa, A. Endo, S. Katsumoto, and Y.
magnetic fields less than or greater than this value. Iye, Appl. Phys. Lett. 69, 363 (1996).
4
M. Tanaka and Y. Higo, Phys. Rev. Lett. 87, 026602 (2001).
5
S. H. Chun, S. J. Potashnik, K. C. Ku, P. Schiffer, and N. Samarth, Phys.
magnetization of the FM layer. Samples with an 8-nm-thick Rev. B 66, 100408(R)(2002).
6
MnO layer have a similar value of HE as the ones with a R. Mattana, J.-M. George, H. Jaffres, F. Nguyen Van Dau, A. Fert, B.
4 nm layer, suggesting that the critical thickness of MnO Lepine, A. Guivarc’h, and G. Jezequel, Phys. Rev. Lett. 90, 166601
required for exchange biasing is less than 4 nm.10 This small (2003).
7
Y. Ohno, D. K. Young, B. Beschoten, F. Matsukura, H. Ohno, and D. D.
value of the critical thickness is consistent with the large Awschalom, Nature (London) 402, 790 (1999).
anisotropy constant K reported for bulk MnO ͑K = 6 8
H. Ohno, N. Akiba, F. Matsukura, A. Shen, K. Ohtani, and Y. Ohno, Appl.
ϫ 103 ergs/ cm3͒.19 Phys. Lett. 73, 363 (1998).
9
Finally, we comment on the reproducibility of our re- W. H. Meiklejohn and C. P. Bean, Phys. Rev. 102, 1413 (1956).
10
J. Nogués and I. Schuller, J. Magn. Magn. Mater. 192, 203 (1999).
sults: we attribute the low yield of successfully exchange- 11
X. W. Wu and C. L. Chien, Phys. Rev. Lett. 81, 2795 (1998).
biased samples to the extreme reactivity of Mn with GaAs.20 12
P. J. van der Zaag, R. M. Wolf, A. R. Ball, C. Bordel, L. F. Feiner, and R.
This may create a detrimental interfacial layer when the Jungblut, J. Magn. Magn. Mater. 148, 346 (1995).
13
samples are annealed during removal from the mounting K. C. Ku, S. J. Potashnik, R. F. Wang, S. H. Chun, P. Schiffer, N. Samarth,
block. To test this hypothesis, we intentionally annealed a M. J. Seong, A. Mascarenhas, E. Johnston-Halperin, R. C. Myers, A. C.
Gossard, and D. D. Awschalom, Appl. Phys. Lett. 82, 2302 (2003).
sample that shows exchange bias for 10 min at 220° C; these 14
J. K. Furdyna, X. Liu, Y. Sasaki, S. J. Potashnik, and P. Schiffer, J. Appl.
conditions—only slightly more severe than those used for Phys. 91, 7490 (2002).
sample removal—completely eliminate the exchange bias. 15
M. S. Jagadeesh and M. S. Seehra, Phys. Rev. B 23, 1185 (1981).
16
RBS analysis on this annealed sample still shows a MnO S. J. Potashnik, K. C. Ku, R. Mahendiran, S. H. Chun, R. F. Wang, N.
overlayer with similar characteristics to the exchange biased Samarth, and P. Schiffer, Phys. Rev. B 66, 012408 (2002).
17
S. J. Potashnik, K. C. Ku, R. F. Wang, M. B. Stone, N. Samarth, P.
sample; however, we find a slight increase in the channeling Schiffer, and S. H. Chun, J. Appl. Phys. 93, 6784 (2003).
minimum yield and also a small enhancement of the Ga 18
The absence of any appreciable background magnetization beyond TC in
channeling interface peak by ϳ1 ϫ 1015 atoms/ cm2. This either of the samples measured demonstrates the absence of observable
suggests the formation of a Mn– GaAs interfacial layer, con- MnAs clusters or any ferromagnetism derived from the Mn capping lay-
sistent with studies of thermally induced reactions between er;this is verified by carrying out magnetization measurements above the
Curie temperature of MnAs ͑320 K͒.
Mn and GaAs.20 Such interfacial reactivity will likely play 19
F. Keffer and W. O’Sullivan, Phys. Rev. 108, 637 (1957).
an important role in attempted exchange biasing 20
J. L. Hilton, B. D. Schultz, S. McKernan, and C. J. Palmstrøm, Appl.
Ga1−xMnxAs using other AFMs such as MnTe and FeMn. Phys. Lett. 84, 3145 (2004).
Downloaded 27 Aug 2004 to 129.169.10.56. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp