Annealing Dependence of Exchange Bias in MnO/Ga1−xMnxAs Heterostructures
Journal of Superconductivity: Incorporating Novel Magnetism, Vol. 18, No. 3, June 2005 ( C 2005)DOI: 10.1007/s10948-005-0019-9Annealing Dependence of Exchange Biasin MnO/Ga1−x Mnx As HeterostructuresK. F. Eid,1 O. Maksimov,1 M. B. Stone,1 P. Schiffer,1 and N. Samarth1 We discuss the annealing-dependence of exchange bias in Ga1−x Mnx As epilayers that are overgrown with Mn. Although pure Mn is a known antiferromagnet, we ﬁnd that the as- grown Mn overlayer does not produce any exchange coupling with Ga1−x Mnx As. Rather, our studies indicate that annealing in air is essential for creating the standard signatures of exchange bias (a clear shift in the magnetization hysteresis loop and an increased coercive ﬁeld). We use X-ray photoelectron spectroscopy to characterize the compositional depth pro- ﬁle of both as-grown and rapid thermal annealed samples. These studies demonstrate that the cleanest exchange bias arises when the Mn overlayer is completely converted into MnO. KEY WORDS: exchange bias; ferromagnetic semiconductor; antiferromagnet.1. INTRODUCTION HE ). This occurs when the sample is cooled through the blocking temperature (TB ) of the AF in the pres- Ferromagnetic semiconductors derived from ence of either an external magnetic ﬁeld or a net mag-III–V semiconductors such as GaAs and InAs have netization in the ferromagnetic layer . This shiftattracted signiﬁcant attention recently within the is typically accompanied by an increase in the coer-context of semiconductor-based “spintronics” [1,2]. cive ﬁeld (HC ) of the ferromagnetic layer. This phe-These materials present a unique opportunity for nomenon is the subject of intense study in a varietycontrolling magnetism through electrical gates , of magnetic materials and is exploited in different ap-and have also been incorporated in a variety of proof- plications .of-principle device conﬁgurations [4–6]. The ability We have recently demonstrated the ﬁrst success-to exchange bias spintronic devices derived from ful exchange biasing of a ferromagnetic semiconduc-these materials will add to their viability for future tor (Ga1−x Mnx As) by an overgrown antiferromagnetapplications, as well as offer a new system for ba- (MnO) . In these experiments, metallic Mn issic scientiﬁc research. For example, it is known from ﬁrst epitaxially grown on top of Ga1−x Mnx As instudies of metallic ferromagnetic multilayer struc- an ultrahigh vacuum chamber. Upon removal fromtures that exchange biased read heads are more sta- the vacuum chamber, the Mn overlayer is uninten-ble than hybrid sensors . Exchange biased spin- tionally oxidized during removal of the wafer fromvalves are also preferred for devices because of the the indium-bonded mounting block. Rutherforddesirable shape of their magnetoresistance curves backscattering (RBS) measurements clearly indicate. the presence of MnO, a well-known antiferromagnet Exchange biasing of a ferromagnet by a vicinal that is presumably responsible for the exchangeantiferromagnet (AF) manifests itself as a shift of biasing effect. This conclusion is supported by thethe magnetization hysteresis loop, making it centered observation that Al-capped samples do not exhibitabout a nonzero magnetic ﬁeld (the exchange ﬁeld, any exchange bias since they are impervious to oxidation. Furthermore, RBS studies also suggest1 Department of Physics and Materials Research Institute, Penn- that the extreme reactivity between Mn and GaAs sylvania State University, University Park, Pennsylvania 16802. requires a delicate balance between the need for 421 0896-1107/05/0600-0421/0 C 2005 Springer Science+Business Media, Inc.
422 Eid, Maksimov, Stone, Schiffer, and Samarthoxidation and the preservation of a clean inter- ∼8 nm is deposited. Mn growth is performed at roomface. For instance, modest annealing at temperatures temperature in order to prevent inter diffusion andaround 230◦ C completely destroys the exchange bias. chemical reaction between Mn and Ga1−x Mnx As lay- Given these circumstances, it is important to de- ers [11,12].velop a better understanding of the oxidation of the The growth mode and surface reconstructionMn overlayer and to hence obtain systematic control are monitored in situ by reﬂection high-energy elec-over the creation of exchange biased Ga1−x Mnx As tron diffraction (RHEED) at 12 keV. The thick-devices. In this paper, we focus on the growth and ness of the Ga1−x Mnx As layer is calculated fromcharacterization of the antiferromagnetic MnO layer RHEED oscillations, while the thickness of theand discuss the dependence and sensitivity of ex- Mn layer is estimated from RHEED oscillations ofchange coupling on the annealing conditions. In par- MnAs whose growth rate is mainly determined byticular, we use a series of annealing experiments to the sticking coefﬁcient of Mn. The Mn concentrationdeﬁnitively demonstrate that metallic Mn grown on in Ga1−x Mnx As is x ∼ 0.06, estimated from the MnGa1−x Mnx As does not by itself create exchange bias, ﬂux.and that the oxidation of the Mn layer to MnO is es- Magnetization measurements are performed us-sential to produce an exchange coupling. ing a commercial superconducting quantum interfer- ence device (SQUID) magnetometer. Samples were measured with the magnetic ﬁeld in plane along the2. EXPERIMENTAL DETAILS (110) direction as a function of both temperature and applied magnetic ﬁeld. Surface morphology is inves- Samples are grown by molecular beam epitaxy tigated using an atomic force microscope running in(MBE) on “epi-ready” semi-insulating GaAs (100) a tapping mode. Si cantilevers with a spring constantsubstrates that are bonded with indium to molyb- of 13–100 N/m and a nominal resonance frequencydenum blocks. Removal of the samples from the of 240–420 kHz are used. Images are recorded withmounting block entails heating up the entire block to a resolution of 300 × 300 pixels at a scan rate ofmelt the bonding layer of indium. In contrast to the 1 micrometer/s. The surface and sub-surface com-usual mounting, when indium is uniformly dispersed position is examined by X-ray photoelectron spec-under the entire wafer, for samples studied in this troscopy (XPS) using a Kratos Analytical Axis Ul-communication only the edges of the sample wafer tra system. The Ga 2p3/2 , As 2p3/2 , Mn 2p3/2 , 2p1/2 ,are indium bonded to the block. This procedure pro- 3p, and O 1s photoelectrons are excited using avides sufﬁcient thermal contact during MBE growth monochromatic Al K X-rays (hv = 1486.6 eV). XPSand allows the removal of a signiﬁcant portion of the quantiﬁcation is performed by applying the appropri-wafer without any post-growth annealing. ate relative sensitivity factors (RSFs) for the Kratos The MBE growth is performed in an Applied instrument to the integrated peak areas. These RSFsEPI 930 system equipped with Ga, Mn, and As effu- take into consideration the X-ray cross-section andsion cells. The substrates are deoxidized using stan- the transmission function of the spectrometer. Lowdard protocol, by heating to ∼580◦ C with an As energy (<5 eV) electrons are used for charge neu-ﬂux impinging on the surface. A 100 nm thick GaAs tralization. The charge-induced shifts are correctedbuffer layer is grown after the deoxidization. Then, using the binding energy of C–C hydrocarbon linesamples are cooled to ∼250◦ C for the growth of a (285.0 eV) as an internal standard.5 nm thick low temperature GaAs layer, followedby a 10 nm thick Ga1−x Mnx As layer (x ∼ 0.06). Thegrowth is performed under the group V rich condi- 3. RESULTS AND DISCUSSIONtions with a As:Ga beam equivalent pressure ratioof ∼12:1. After the Ga1−x Mnx As growth, the sam- Figure 1a shows the (1×2) surface reconstruc-ples are transferred in situ to an adjoining ultra high tion in a RHEED pattern for a Ga1−x Mnx As epi-vacuum buffer chamber and the As cell is cooled layer prior to the Mn growth. Figure 1b and c showsto the rest temperature of 110◦ C to avoid formation RHEED patterns for Mn deposited on Ga1−x Mnx Asof MnAs clusters during the Mn growth. When the with the incident electron beam along the  andAs pressure in the growth chamber decreases to an ¯  directions. The RHEED pattern consists ofacceptable level, the wafers are reintroduced. Then, sharp, elongated streaks and its symmetry is sugges-the Mn capping layer with a thickness of ∼4 nm or tive of the stabilization of a cubic phase of Mn. This
Annealing Dependence of Exchange Bias in MnO/Ga1−x Mnx As 423 Fig. 1. (a) RHEED patterns for Ga1−x Mnx As showing the two-fold reconstruction prior to Mn growth; (b) and (c) RHEED patterns for Mn epilayers grown on Ga1−x Mnx As with different incident directions of the electron beam.indicates that, in contrast to previous reports , pieces from the same wafer: one piece is from theMn can be epitaxially grown at room temperature. A indium-free portion of the wafer and is not heatedrepresentative atomic force microscope scan of the after removal from the vacuum chamber, while theMn surface is shown in Fig. 2. The epilayer is atom- other is from the indium-bonded portion and henceically ﬂat with a root mean square (rms) roughness undergoes a rapid thermal anneal to ∼220◦ C duringof ∼0.2 nm, similar to the rms values reported for sample removal. The zero magnetization at tempera-Ga1−x Mnx As epilayers . tures above TC indicates that the sample is of good Figure 3a shows the temperature dependence of quality without MnAs clusters. Although we ob-the magnetization for a sample with a Curie tem- serve no difference in the TC of the indium-free andperature (TC ) of 85 K. The data is shown for two indium-mounted portions of the sample, we do note Fig. 2. Tapping-mode atomic force microscope scan of an as-grown Mn/Ga1−x Mnx As heterostructure with 8 nm Mn thickness.
424 Eid, Maksimov, Stone, Schiffer, and Samarth collected for the samples annealed to 200◦ C and 250◦ C, we only compare the 250◦ C annealed sam- ple to the as-grown sample. Figure 4 depicts high- resolution Mn 2p spectra for the as-grown and an- nealed samples. Both the positions and the shapes of the Mn 2p lines differ for the annealed and as grown samples. The Mn 2p3/2 line for the as grown part is centered at ∼641.0 eV. Its position is in agreement with the binding energy of MnO (Mn2+ ), indicating that the Mn at the surface is oxidized. (Metallic Mn0 has a 2p3/2 line at 639 eV, while lines from Mn2 O3 (Mn3+ ) and MnO2 (Mn4+ ) have binding energies of ∼641.7 eV and ∼642.5 eV, respectively .) The observation of two satellite lines spaced by 5.5 eV from 2p3/2 and 2p1/2 lines further supports this as- signment. These satellite excitations are typical forFig. 3. Magnetization as a function of temperature and applied MnO [15,16] and are not present in either Mn2 O3 ormagnetic ﬁeld for MnO/Ga1−x Mnx As heterostructures. Hystere- Mn2 O3 [17–20]. The Mn 2p3/2 line for the annealedsis loops are measured at T = 10 K, ﬁeld cooled from T = 200 K piece is visibly shifted to a slightly higher binding en-with H = 1 kOe. (a) Temperature-dependent magnetization mea- ergy of ∼641.3 eV. The two satellite lines disappearsured with H = 100 Oe for an indium-free portion of the as-grown while a new satellite line at ∼663.5 eV appears, indi-sample and for an indium-bonded portion. (b) Field-cooled hys-teresis loop for indium-free portion of sample. (c) Field-cooled cating that the as-grown Mn has oxidized to Mn2 O3hysteresis loop for indium-mounted portion of sample. (d) Field- .cooled hysteresis loop for indium-free portion of sample after in- We also measure the compositional depth de-tentional annealing at T = 200◦ C for 1 min. A diamagnetic back- pendence in ion milled samples using 4 keV Ar+ .ground has been subtracted from the data shown. A paramagnetic Figure 5a shows the relative percentage-depth proﬁlebackground has been subtracted from the data shown in panels(a)–(c). of Mn, O, Ga, and As for the as-grown part. Even prior to sputtering, As (Ga) 2p lines are detected. This does not necessarily indicate As (Ga) diffusionthat the former has a smaller low-temperature satu- and is, most probably, due to the fact that the Mn ﬁlmrated moment compared to the latter. Figure 3b and cshows the magnetization (M) of the bilayer structureas a function of the applied magnetic ﬁeld (H) afterthe samples are cooled to the measuring temperature(T = 10 K) in the presence of an external magneticﬁeld of 1 kOe. Figure 3b shows the magnetizationof an indium-free part of the wafer. The magnetiza-tion curve is symmetric about zero applied ﬁeld, in-dicating absence of exchange bias. Figure 3c showsa shifted hysteresis loop measured for an indium-mounted portion of the sample. Finally, Fig. 3d showsa hysteresis loop of an indium-free portion of thesample that was intentionally annealed at 200◦ C for1 min. The center of the hysteresis loop is also shiftedfrom zero. These results clearly demonstrate that an-nealing is necessary to create exchange bias in theMn/Ga1−x Mnx As bilayer structures. To further understand these results, we per-form detailed XPS studies on the indium-free por-tion of the wafer. We study the as-grown sample, as Fig. 4. The Mn 2p spectra acquired at the surface for thewell as pieces that are annealed for 1 min at 200◦ C (a) annealed and (b) as-grown In-free parts of the Mn/and 250◦ C. Since qualitatively similar spectra were Ga1−x Mnx As heterostructure.
Annealing Dependence of Exchange Bias in MnO/Ga1−x Mnx As 425 centration after 250 s. A larger O content is present with O being more uniformly distributed with an O/Mn ratio of ∼1. Thus, the Mn ﬁlm is more sig- niﬁcantly oxidized upon annealing. We furthermore conclude that oxidation results in the expansion of the crystal lattice (∼1.8 times assuming complete conversion of Mn0 to MnO). Since the MnO ﬁlm is thicker, we do not see As (Ga) lines when the top layers are studied. This interpretation is further supported if we consider Mn 2p lines as a function of depth (Fig. 6). The Mn 2p3/2 line from the annealed sample is characteristic for MnO. Its posi- tion and shape remains constant and only the inten- sity decreases due to a decrease in Mn content. Thus, the annealed ﬁlm is nearly uniformly oxidized withFig. 5. Relative concentration as a function of depth of Mn, O, As,and Ga in the (a) as-grown and (b) annealed indium-free parts ofa Mn/Ga1−x Mnx As heterostructure. The nominal thickness of Mndeposition is 8 nm.thickness is comparable to the probing depth. Thedata show that the As (Ga) content increases withdepth, while the Mn content decreases. The change islinear and the content becomes constant after 150 s,indicating the depth of the Ga1−x Mnx As interface. ASigniﬁcant amount of O is present close to the surfacewith an O/Mn ratio of approximately 1. However, Ocontent drops rapidly with depth, and only traces arepresent after 100 s with the O/Mn ratio consistentlyless than 1/3. This suggests that only the top portionof the ﬁlm is oxidized and metallic Mn0 is presentcloser to the Ga1−x Mnx As interface. Figure 5b illustrates the relative percentage-depth proﬁle for the annealed piece of the indium-free mounted sample. In contrast to the as-grownpart, we do not detect As (Ga) lines prior to sput-tering, and the As (Ga) traces become evident only Fig. 6. The Mn 2p spectra acquired at the different depthsafter sputtering for 100 s. This is followed by a linear for (a) annealed and (b) as-grown indium-free piece of achange in content, and ﬁnally a stabilization in con- Mn/Ga1−x Mnx As heterostructure.
426 Eid, Maksimov, Stone, Schiffer, and SamarthMnO being the dominant form of Mn throughout the REFERENCESwhole layer. In contrast, the Mn 2p3/2 line from theas-grown part exhibits a low energy shoulder after 1. H. Ohno, in Semiconductor Spintronics and Quantum Com-60 s and shifts to 639 eV after 90 s. Satellite lines putation, D. D. Awschalom, D. Loss, and N. Samarth, eds. (Springer-Verlag, Berlin, 2002), p. 1.also disappear at this point. This clearly indicates 2. S. A. Wolf, D. D. Awschalom, R. A. Buhrman, J. M.that while the surface layers of the as-grown part Daughton, S. von Molnar, M. L. Roukes, A. Y. Chtchelka-are oxidized, metallic Mn dominates in the bottom nova, and D. M. Treger, Science 294, 1488 (2001). 3. H. Ohno, D. Chiba, F. Matsukura, T. Omiya, E. Abe, T. Dietl,layers. Y. Ohno, and K. Ohtani, Nature 408, 944 (2000). 4. Y. Ohno, D. K. Young, B. Beschoten, F. Matsukura, H. Ohno, and D. D. Awschalom, Nature 402, 790 (1999). 5. M. Tanaka and Y. Higo, Phys. Rev. Lett. 87, art. no. (2001).4. CONCLUSIONS 6. S. H. Chun, S. J. Potashnik, K. C. Ku, P. Schiffer, and N. Samarth, Phys. Rev. B 66, 100408(R) (2002). 7. L. Thomas, J. Luning, A. Scholl, F. Nolting, S. Anders, J. In summary, we have demonstrated that epitaxi- Stohr, and S. Parkin, Phys. Rev. Lett. 84, 3462 (2000).ally grown Mn/Ga1−x Mnx As heterostructures do not 8. K. Nagasaka, Y. Seyama, L. Varga, Y. Shimizu, and A. Tanaka, J. Appl.Phys. 89, 6943 (2001).exhibit any exchange bias until the Mn layer has been ´ 9. J. Nogues and I. Schuller, J. Magn. Magn. Mater. 192, 203completely oxidized into MnO by thermal annealing (1999).in air. XPS measurements as a function of depth show 10. K. F. Eid, M. B. Stone, K. C. Ku, O. Maksimov, P. Schiffer, N. Samarth, T. C. Shih, and C. J. Palmstrom, Appl. Phys. Lett. 85,that the as-grown Mn layer oxidizes partially into 1556 (2004). Also K. F. Eid, M. B. Stone, O. Maksimov, T. C.MnO upon exposure to air, with some metallic Mn Shih, K. C. Ku, W. Fadgen, C. J. Palmstrom, P. Schiffer, andregions close to the Mn/Ga1−x Mnx As interface, as N. Samarth, J. Appl. Phys. 97, 10D304 (2005). 11. X. Jin, Y. Chen, X. Lin, D. Dong, Y. Chen, M. Xu, W. Zhu, X.well as some regions of Mn2 O3 at the surface. Rapid Wang, X. Shen, and L. Li, Appl. Phys. Lett. 70, 2445 (1997).thermal annealing to ∼200◦ C completely converts 12. J. L. Hilton, B. D. Schultz, S. McKernan, and C. J. Palmstrøm,the Mn layer into MnO, creating an exchange-biased Appl. Phys. Lett. 84, 3145 (2004). 13. T. Sadowski, J. Domagala, J. Bak-Misiuk, S. Kolesnik, K.system. Swiatek, J. Kanski, and L. Ilver, Thin Solid Films 367, 165 (2000). 14. R. J. Iwanowski, M. H. Heinonen, and E. Janik, Chem. Phys. Lett. 387, 110 (2004).ACKNOWLEDGMENTS 15. F. Muller, R. de Masi, D. Reinicke, P. Steiner, S. Hufner, and K. Stowe, Surface Sci. 520, 158 (2002). 16. S. P. Jeng, R. J. Lad, and V. E. Henrich, Phys. Rev. B 43, 11971 The research has been supported by DARPA/ (1991).ONR under grants N00014-99-1093, and -00-1-0951, 17. S. A. Chambers and Y. Liang, Surface Sci. 420, 123 (1999).by ONR N00014-99-1-0071 and by NSF DMR 01- 18. M. A. Stranick, Surface Sci. Spectra 6, 31 (1999). 19. M. A. Stranick, Surface Sci. Spectra 6, 39 (1999).01318. We thank Dr Shallenberger for useful discus- 20. M. C. Militello and S. W. Gaarenstroom, Surface Sci. Spectrasions and assistance with XPS measurements. 8, 200 (2001).