2706 IEEE TRANSACTIONS ON MAGNETICS, VOL. 41, NO. 10, OCTOBER 2005 Exchange Bias and Giant Magnetoresistance in Spin Valves With Angström-Scale Antiferromagnetic Layers at 5 K K. L. Perdue , M. J. Carey , P. D. Sparks , and J. C. Eckert , Member, IEEE Department of Physics, Harvey Mudd College, Claremont, CA 91711 USA Hitachi Global Storage Technologies, San Jose, CA 95120 USA We have studied the effects on the exchange bias of decreasing the antiferromagnetic layer to the Angström-scale regime in order to shed light on the minimum required thickness of the antiferromagnet. We have deposited IrMn layers between 0.2 and 2 nm on spin valves and measured the exchange bias by examining hysteresis loops at 5 K using the giant magnetoresistance of the spin valves. The exchange bias persists for IrMn thicknesses down to 0.4 nm and has a maximum at 1.6 nm. Because the ultra-thin layers create an exchange field, the origin of at least one component of exchange biasing must have a similarly short length scale. Index Terms—Antiferromagnetic materials, giant magnetoresistance, interface magnetism. I. INTRODUCTION T HE hallmark of exchange bias—a coupling between an an- tiferromagnet and a ferromagnet—is the shift in the hys- teresis loop, causing it to be asymmetrical about zero applied field. The magnitude of the shift is the exchange field . In an exchange biased system, there is usually an accompanying increase in the coercive field, . The exchange bias effect is present below the blocking temperature, , which is gen- erally lower than the Néel temperature of the antiferromagnet. Microscopic models of exchange biasing currently focus on the formation of domains in the antiferromagnet or the ferromagnet, or on a small induced moment in the antiferromagnet [[1] and references therein]. Insight into the requirements for a micro- scopic model can be obtained by investigating the dependence of the exchange biasing on the thickness of the antiferromag- netic layer in the ångström-scale regime. Recent work by Hoffmann, et al. [1] using polarized neutron scattering suggested the existence of a net moment in the antifer- romagnet near the interface, and argued against the formation of a domain wall in either the ferromagnet or the antiferromagnet. Recent work [2] using dynamic magnetic anisotropy mea- surements found no exchange bias in NiFe/IrMn bilayer sys- tems, presumably at room temperature, for IrMn thicknesses below about 1 nm. We have previously reported exchange bias for IrMn thicknesses near this thickness limit at lower temper- atures in spin valve systems [3]. Investigations of the IrMn/Co and FeMn/Co systems by Ali, et al. [4], [5] over the antiferro- magnet (AF) thickness range of 1.4 nm to 11.5 nm showed sub- stantial exchange and coercive fields for even the thinnest sam- ples at 2 K so presumably the blocking temperatures are above 2 K. Stern [6] also investigated the IrMn/Co system for the AF thickness range 1.4 nm to 2.6 nm and measured and blocking temperatures. For calibration, the measured blocking temperature for his 1.4 nm IrMn sample was 100 K. Our cur- Digital Object Identifier 10.1109/TMAG.2005.855224 rent work extends the thickness range of IrMn to 0.2 nm in an IrMn/Co Fe system. To ensure that the majority of sam- ples will be below the blocking temperature, we report measure- ments at 5 K. We use the hysteresis in the magnetoresistance as the probe of and and the GMR, defined as R/R. II. PROCEDURE Films were deposited by DC magnetron sputtering. The chamber has a base pressure of Torr and 5 N pure argon was further purified before being let into the chamber at 2 mTorr. A target-substrate distance of about 7 cm was used. Rates near 1 /s were used for all layers. The substrate was rotated to insure uniformity. The rotation speed was nearly 40 rpm, but was varied to insure that an integral number of rotations occurred during each layer deposition. The shutter open/close speed is significantly less than 1 second, so that the layer thicknesses are accurate. Initial calibrations were per- formed using x-ray reflectivity, and the layers were deposited at the calibrated powers. Layer thicknesses were determined by shutter open times. The samples had the structure: 50 Ta/30 NiFe/10 CoFe/30 Cu/30 CoFe/ /50 Ta, where ranges from 2 to 20 , and were deposited on a silicon substrate. These thicknesses span the range where the onset of exchange bias could be expected. The magnetoresistance along the easy axis was measured using a Quantum Design Physical Proper- ties Measurement System. The samples were pinned in a field of T at 350 K for one hour and cooled to 5 K in a field of Oe. Hysteresis loops were measured by sweeping the ap- plied field up to 3000 Oe, then back down to Oe, twice to get two generations of loops. These exchange bias systems typ- ically exhibit training effects, so we will compare results from first generation scans to be consistent. III. RESULTS In Fig. 1, we plot resistance versus magnetic field for a spin valve with a 16 IrMn antiferromagnetic layer. The first gen- 0018-9464/$20.00 © 2005 IEEE