Tuning exchange bias

The exchange bias shift of the hysteresis loop, H E , in antiferromagnetic/ferromagnetic layer systems can be easily controlled ~ within certain limits ! by cooling in zero ﬁeld from different magnetization states above the antiferromagnetic Ne´el temperature, T N . This indicates that for moderate cooling ﬁelds, H E is determined by the magnetization state of the ferromagnet at T N , and not by the strength of the cooling ﬁeld. © 1999 American Institute of Physics. @

2][3][4][5][6][7][8] Larger cooling fields ͑i.e., well above the saturation field of the FM layer͒ do not further affect the FM layer.However, if the cooling field is large enough, it can affect the AFM layer.Thus, new phenomena such as a reduction or increase of H E at large cooling fields [8][9][10][11] or ''positive'' exchange bias can be observed. 86][17] If the samples are cooled in H cool ϽH C in the initial virgin magnetization curve, then the loop shift is much smaller than the one obtained from H cool ϾH C . 3 In this letter we show that the exchange bias field, H E , can be tuned by cooling in zero field from different magnetization states, indicating that the role of the cooling field is not to induce H E but only to have a single FM domain state above T N and thus a maximum exchange bias effect.This is of obvious technological importance: It allows one to select the desired value of exchange bias via a simple postgrowth cooling procedure.For example, for the case of hybrid spin valve sensors the value of H E can be tuned such that the field induced resistance change occurs at very small applied fields.
We studied two different exchange bias systems, FeF 2 ͑AFM͒/Fe ͑FM͒ and CoO ͑AFM͒/Co ͑FM͒, both of which allow for easy cooling and warming the samples across the AFM Ne ´el temperature, T N (FeF 2 )ϭ78 K and T N (CoO) ϭ291 K.
The FeF 2 /Fe samples were prepared using electron beam deposition.90 nm of FeF 2 were deposited at a rate of 0.2 nm/s at a substrate temperature T S ϭ200 °C on MgO ͑100͒ substrates, followed by a 12 nm Fe layer, at a rate of 0.1 nm/s, at T S ϭ150 °C and a 3 nm capping layer of Al, at a rate 0.1 nm/s, at T S ϭ150 °C.The base pressure of the chamber was better than 1ϫ10 Ϫ7 mbar and the pressure during deposition was lower than 3ϫ10 Ϫ6 mbar.The thickness of the different layers was controlled by a calibrated quartz oscillator.The FeF 2 layer grows in the ͑110͒ orientation. 18he Co/CoO samples were grown in a molecular beam epitaxy chamber with a base pressure of 5ϫ10 Ϫ11 mbar.A Co layer of 8 nm was grown at room temperature at a rate of 0.3 nm/min on a hydrogen terminated Si͑111͒ substrate. 19,20he CoO layer ͑100 nm͒ was deposited at a rate of 1 nm/min by evaporating Co in an oxygen atmosphere of 2 ϫ10 Ϫ6 mbar using a substrate temperature of T S ϭ110 °C.The chemical composition of the CoO layer was controlled with quantitative Auger spectroscopy.Electron diffraction ͓reflection high-energy electron diffraction ͑RHEED͒, lowenergy electron diffraction ͑LEED͔͒ indicate a ͑111͒ orientation of the CoO layer.
For the FeF 2 /Fe samples the hysteresis loops were measured at Tϭ10 K using a superconducting quantum interference device ͑SQUID͒ magnetometer, applying fields up to Ϯ2000 Oe.If the sample was cooled through T N , as is customary, in a field large enough to saturate the FM layer, H cool ϭϩ2000 Oe, but small enough not to affect the AFM layer, the loop shift at Tϭ10 K was H E ϭϪ330 Oe.We denote this field by H E,max .For the remaining discussion, the samples were cooled through T N in zero field (H cool ϭ0).However, different magnetization states, m(Hϭ0), were set up at Tϭ85 K, i.e., TϾT N , before the zero field cooling procedure.To obtain the different magnetization states minor loops were carried out.As shown in the inset of Fig. 1, the sample was saturated in a negative field of, e.g., H ϭϪ4H C .Then the field was increased to 0ϽHϽϩH SAT , where H SAT is the saturation field ͑see inset Fig. 1͒ and finally the field was reduced back to Hϭ0.This procedure establishes different magnetizations above T N in the range Ϯm R , where m R is the remanent moment of the FM.
The magnetization was recorded during the cooling procedure, showing that the magnetization did not change more than 1.5% between Tϭ85 K and T N ϭ78 K and not more than 4% between Tϭ85 and 10 K.If the sample is cooled in H cool ϭ0 from remanence m R , the loop shift is similar to the one for field-cooled samples, 1 i.e., H E ϷϪ͉H E,max ͉ ͑see Figs. 1 and 2͒.Nevertheless, if the sample is cooled in H cool ϭ0 from negative remanence, i.e., m(Hϭ0)ϭϪm R , the loop shifts in the opposite direction, i.e., H E Ϸϩ͉H E,max ͉ ͑similar to what is observed in samples cooled in negative fields 18 ͒.6][17] For intermediate magnetizations ͉m(H ϭ0)͉Ͻm R , the loops shift continuously between ϩ͉H E,max ͉ and Ϫ͉H E,max ͉ ͑Figs. 1 and 2͒.Thus, we can tune H E by setting up the appropriate magnetization state, m(Hϭ0), above the Ne ´el temperature before the zero-field cooling procedure.
The same experimental procedure was carried out with a CoO/Co sample using a temperature of Tϭ300 KϾT N ϭ291 K to perform the minor loops in order to set up the different remanent magnetization states.The hysteresis loops were measured at a temperature of 100 K applying fields up to Ϯ1000 Oe.The results shown in Fig. 3 exhibit an almost identical dependence of the exchange bias field on the initial magnetization as found above for the FeF 2 /Fe system.
These results indicate that the shift H E is determined by the magnetization state at TϭT N and not by H cool ͑for moderate cooling fields͒, i.e., exchange bias can be obtained and H E be controlled without a cooling field.
These results are of particular interest for device fabrication where exchange bias is employed for pinning the magnetization direction of a ferromagnetic layer ͑e.g., spin valve sensors, 21 magnetoresistance devices 22 ͒.H E usually is obtained from cooling or film deposition in an external field and H E is determined by the FM and AFM layer thicknesses and their microstructures. 23,24Here we demonstrated that H E can be tuned after device fabrication.
As shown in Fig. 1, the hysteresis loops of the FeF 2 /Fe samples consist of two components, i.e., two loops shifted to positive or negative fields with different relative weights depending on the magnetization state before cooling.The same behavior is found for the CoO/Co samples ͑not shown͒.For the cases with mϽm R which exhibit two components of the hysteresis loops, we infer that the FM layer for TϾT N is divided into domains with orientations parallel and antiparallel to the field direction during the minor loop.During cooling each ferromagnetic domain determines locally the exchange bias in the area of the domain.At low temperatures, these areas with different exchange bias shift the loop in opposite directions.Thus, it is the relative amount of one or the other type of domain which controls the shift of the loop.This is confirmed by the behavior of the hysteresis loop when the system is cooled starting from m(Hϭ0)ϭ0 ͑see Fig. 4͒.If the sample is cooled in m(Hϭ0)ϭ0 as obtained from a minor loop procedure, an unshifted loop with two symmetrical components, one shifted to positive fields and one shifted to negative fields, is obtained ͑Fig.4͒.However, if the sample is cooled in m(Hϭ0)ϭ0 as obtained from a  demagnetization procedure ͑at Tϭ300 K͒, the two component feature of the loop is much less pronounced ͑Fig.4͒.The demagnetization process reduces the size of the domains and randomizes their directions, thus the effect of the domains is greatly reduced.These results indicate that the role of H cool is only to ensure that the FM layer is in a uniformly magnetized state, which will guarantee the homogeneity ͑i.e., only one component͒ of the shifted loops at low temperatures.It is noteworthy that these results also confirm that magnetometry measures the average coupling over the whole interface area.
In conclusion, we have shown that the exchange bias field H E can be tuned between Ϯ͉H E,max ͉ by cooling in zero field from different magnetization states set up by minor loops slightly above T N .This indicates that it is the magnetic state of the FM layer at T N and not the cooling field which controls the exchange bias below T N .This gives the opportunity to tune the exchange bias even after device fabrication.
Note added in proof: After submission of this letter it came to our attention that similar work on CoO/Permalloy has been carried out by Go ¨kemejier and Chien, J. Appl.Phys.85, 5516 ͑1999͒.

FIG. 1 .
FIG.1.Hysteresis loops for a FeF 2 /Fe bilayer at Tϭ10 K cooled in zero field (H cool ϭ0) from Tϭ85 K in different magnetization states, ͑a͒ m ϭϩ3.46ϫ10Ϫ4 , ͑b͒ mϭϩ1.89ϫ10Ϫ4 , ͑c͒ mϭϪ0.92ϫ10Ϫ4 , and ͑d͒ m ϭϪ3.29ϫ10Ϫ4 emu.Note that m R ϭ3.54ϫ10 Ϫ4 emu.The inset shows the procedure to set up different magnetization states at Tϭ85 K.The lines are guides to the eye.

FIG. 3 .
FIG.3.Dependence of the exchange bias field, H E , of CoO/Co at T ϭ100 K on the zero field magnetic moment, m(Hϭ0), at Tϭ300 K before the zero field cooling procedure.The solid line through the data points is a guide to the eye.