Enhancing the blocking temperature of perpendicular-exchange biased Cr2O3 thin films using spacer and buffer layers

In this study, we investigated the effect of spacer and buffer layers on the blocking temperature TB of the perpendicular exchange bias of thin Cr2O3 films, and revealed a high TB of 260 K for 20-nm-thick Cr2O3 thin films. By inserting a Ru spacer layer between the Cr2O3 and Co films and changing the spacer thickness, we controlled the magnitude of the exchange bias and TB. By comparing the TB values of the 20-nm-thick Cr2O3 films on Pt and alpha-Fe2O3 buffers, we investigated the lattice strain effect on the TB. We show that higher TB value can be obtained using an alpha-Fe2O3 buffer, which is likely because of the lattice-strain-induced increase of Cr2O3 magnetic anisotropy.


Introduction
Electric control of magnetization using the magnetoelectric (ME) effect has received considerable attention as a promising candidate for next generation low energy consumption magnetic recording devices 1) . The linear ME effect was theoretically predicted for Cr2O3 in 1960 2) and experimentally confirmed in 1961 3) . Few techniques at that time had utilized to deal antiferromagnets, so the practical potential of the discovery was not immediately realized. Since 2005, the ME effect in Cr2O3 has captured renewed attention because the electrical switching of perpendicular exchange bias has been demonstrated in bulk Cr2O3 single crystal/ferromagnet exchange coupling systems 4,5) . Here, antiferromagnetic domain reversal by applying both magnetic and electric fields 6) has been used to switch the perpendicular exchange bias of a Cr2O3/ferromagnet. This perpendicular exchange bias switching has recently been demonstrated in several-hundred-nanometer-thick Cr2O3 films deposited by the sputtering method, which yields more realistic device applications [7][8][9][10][11] . One major issue for device application is simultaneous realization of both further reduction of the Cr2O3 film thickness and increase in the operating temperature. Further reduction of the Cr2O3 film thickness is necessary to decrease the applied voltage and the aspect ratio of the recording bit. However, for thin Cr2O3 films, there is a rapid reduction in the blocking temperature TB of the perpendicular exchange bias with decreasing Cr2O3 thickness 12) . Here, TB represents the temperature at which the perpendicular exchange bias disappears and it is the upper limit of the operating temperature. Thus, in addition to the enhancement of the Néel temperature TN, for which several attempts have been reported recently [13][14][15][16] , to achieve high TB is also an important issue. Perpendicular exchange bias has been reported for Cr2O3 film thicknesses ≥ 30 nm 12) . However, the TB is as low as 80 K for 30-nm-thick Cr2O3 film when the unidirectional magnetic anisotropy energy JK (=HexMstFM) is 0.26 mJ/m 2 . Here, Hex, Ms, and tFM represent the exchange bias field, saturation magnetization and thickness of the ferromagnetic layer, respectively. If the Cr2O3 thickness is further decreased, the perpendicular exchange bias is not observed even at very low temperatures. For Cr2O3, the lower TB than TN (~307 K) can be qualitatively understood by using the Meiklejohn-Bean free-energy model (MB model) 17,18) . In the MB model, the TB is explained by the competition between the interface exchange coupling energy Jex and the product of the magnetic anisotropy energy KAF and thickness tAF of the antiferromagnet (KAFtAF). When KAFtAF > Jex, the exchange bias is observed (unidirectional anisotropy). When KAFtAF < Jex, the exchange bias disappears and only the enhancement of the coercivity Hc is observed (uniaxial anisotropy). Such appearance/disappearance of the exchange bias has also been observed for antiferromagnetic Mn-Ir alloy 19) . For the Mn-Ir alloy, the exchange bias at room temperature disappears and the Hc increases when the Mn-Ir thickness decrease below ~5 nm. That is, for a 5-nm-thick Mn-Ir film, the TB is approximately equal to the room temperature. For Cr2O3, because of its small KAF (~2.0 × 10 4 J/m 3 at low temperature 20) ) and large Jex, the TB becomes much lower than the TN even for a 500-nm-thick Cr2O3 film. The TB of the Cr2O3/Co exchange coupling system is easily controlled by changing the tAF or Jex 18) , which indicates that the TB of the Cr2O3/Co system can be well described by the MB model. According to the MB model, to realize a high TB in thin Cr2O3 layers, either the Jex needs to be decreased or the KAF needs to be increased. Decreasing the Jex can be achieved by inserting a thin metallic spacer between the Cr2O3 film and ferromagnet. Increasing TB by inserting a Pt spacer layer has been reported, where the reduction in the Jex was confirmed as a reduction in the Hex 18,21,22) . A higher KAF can be achieved by doping and inducing lattice strain.
The KAF consists of magnetic dipole anisotropy KMD and magnetocrystalline anisotropy (the fine structure anisotropy) KFS. Increasing the KAF by Al-doping has been reported, which is mainly mediated by increasing the KFS 23) . Artman et al. calculated the change in the KMD induced by the lattice parameter variation and found that the KMD increases with increasing c, decreasing a, or increasing ionic position w 20) . The KFS may also be affected by the lattice strain, but this effect has not been quantified. Because lattice strain can be easily induced for thin films by adjusting the lattice parameter of the buffer layer, a thin Cr2O3/Co system is favorable for investigating the effect of the lattice strain on the KAF. In this study, we attempted to achieve TB ≈ TN for a 20nm-thick Cr2O3 thin film through controlling the Jex by inserting a metallic spacer layer and enhancing the KAF by inducing lattice strain.

Experimental details
All the samples were fabricated by RF-DC magnetron sputtering with a base pressure below 5 × 10 −7 Pa. The sample structures were c-Al2O3 substrate/Pt 25 or α-Fe2O3 20/Cr2O3 20/Ru tRu/Co 1/Pt 5 (nm). In this study, we varied the Ru spacer layer thickness tRu and buffer layer material (Pt or α-Fe2O3). The oxygen reactive sputtering technique was used for deposition of Cr2O3 (-Fe2O3) by sputtering the Cr (Fe) metal target in an Ar/O2 mixed gas atmosphere. The substrate temperature during deposition was 873 K for the Pt buffer layer, 773 K for the Cr2O3 and -Fe2O3 oxide layers, and 406 K for the Ru spacer, Co ferromagnetic, and Pt capping layers. The magnetic properties were measured by superconducting quantum interference device (SQUID) magnetometry after cooling the sample in the presence of an applied magnetic field (+10 kOe) from 340 K, which is sufficiently above the TN of Cr2O3. During the measurements, the magnetic field was applied normal to the film surface.
Structural characterization was performed by X-ray diffraction (XRD) and cross-sectional transmission electron microscopy (TEM) measurements. Nanobeam electron diffraction was used for refining the lattice parameters of the thin Cr2O3 layer (20 nm thick) 24) .

Ru spacer layer effect
First, we investigated the effect of the Ru spacer layer on the Hex using Pt-buffered samples. We measured the dependence of the Hex on the Ru spacer layer thickness for the Ru spacer samples (Al2O3/Pt buffer 25/Cr2O3 250/Ru spacer tRu/Co 1/Pt 5), and compared the results with those of the Pt spacer samples (Al2O3/Pt buffer 25/Cr2O3 250/Pt spacer tPt/Co 1/Pt 5). In thin Cr2O3 films, because of the small KAFtAF, only the Hc enhancement will occur, making these films unsuitable for the Hex comparison. Thus, we used relatively thick Cr2O3 films (250-nm-thick). Figure 1 there is a rapid decrease in the JK with increasing tPt for tPt > 0.7 nm, while for the Ru spacer samples, the JK linearly decreases with increasing tRu. From these results, we assume the island growth of Pt for the Pt spacer samples when the Pt spacer layer is thin (tPt < 0.7 nm), which leads to a smaller change in the JK, reflecting the small change in the Cr2O3-Co distance. In contrast, the layer growth is expected for Ru of the Ru spacer sample, which leads to a gradual change in Cr2O3-Co distance and JK, reflecting the linear change in the Cr2O3-Co distance. It is noted that the thickness dependence is affected not only by the spacer material but also by the spacer layer deposition conditions (e.g., comparing the results of refs. 14, 18 and 25). These results show that the JK of Cr2O3/Co films can be controlled more easily by using Ru spacers than Pt spacers. On the other hand, the changes in the magnetic properties of Co largely depend on the spacer layer material (Pt or Ru), and they are independent of the spacer thickness (tPt or tRu). Here, we demonstrate the enhancement of the TB for a 20-nm-thick Cr2O3 film by inserting a Ru spacer layer.

Buffer layer effect
We investigated the effect of lattice strain induced by changing the buffer layer material on the KAF of Cr2O3.
For a Cr2O3 buffer layer, good lattice matching and oxidative resistance at 773 K (deposition temperature of Cr2O3) are required. In this study, we chose -Fe2O3 and Pt for the buffer layer materials. Both Cr2O3 and -Fe2O3 possess the corundum structure, so an epitaxial growth of -Fe2O3/Cr2O3 is expected. Pt is a face-centered cubic structure, but it preferentially orients along the [111] direction. Thus, the lattice matching is relatively good. In addition, both -Fe2O3 and Pt show good oxidative resistances. Figure 3 shows the schematics of the (0001) planes of Cr2O3 and -Fe2O3, and the (111) plane of Pt. Table 1 summarizes the lattice mismatch between Cr2O3 and the buffer layers. Because -Fe2O3 (Pt) has larger (smaller) lattice parameters than Cr2O3, we expected to obtain a-axis expanded (compressed) Cr2O3 using an (a) -Fe2O3 (Pt) buffer. We investigated the change in the KAF of these samples by measuring the TB, and compared the results with the theoretical predictions.

Structural characterization
The lattice parameters and morphology can be affected by changing the buffer layer. Thus, we first characterized the structural properties. Figure 4 shows the XRD patterns for 2 (out of plane, figure 4(a)) and 2 (in plane, figure 4(b)) scans of -Fe2O3-buffered and Pt-buffered samples. The 2 scan was carried out for the  Figure 5 shows the cross-sectional TEM images of the -Fe2O3-buffered ( figure 5(a)) and Pt-buffered ( figure 5(b)) samples. The TEM image confirmed the epitaxial growth of the -Fe2O3-buffered sample and existence of the two types of domains in the Pt-buffered sample (domain sizes of several tens of nanometers). The different domain sizes can also affect the TB. However, based on the MB model, the domain sizes perpendicular to the film surface are more important than the domain sizes in the film plane, because we discuss the relation between JK and KAFtAF. In this paper, we neglect the grain size effect, and additional investigations are underway. We then checked the lattice parameter a of Cr2O3 on -Fe2O3 and Pt. We have previously reported the a value of -Fe2O3and Pt-buffered 20-nm-thick Cr2O3 films from the XRD patterns 26) .
However, the a value of the -Fe2O3-buffered sample could not be accurately determined owing to the overlap between the Cr2O3 ( 0 3 30 ) and -Fe2O3 ( 0 3 30 ) Bragg peaks. In the present study, we performed nanobeam electron diffraction measurements 24) of the -Fe2O3-buffered sample and more precisely determined the a value of the Cr2O3 film. The a value of Cr2O3 estimated from the nanobeam electron diffraction was the same as that of the -Fe2O3 buffer estimated from in-plane XRD. Thus, we confirmed that the a value of Cr2O3 is locked by the -Fe2O3 buffer layer. From this result, we could identify that both the -Fe2O3 ( 0 3 30 ) and Cr2O3 ( 0 3 30 ) peaks are at 64.1° in figure 4(b). The refined lattice parameter values are listed in table 1. Both the present data and the data in 26) show that the a values for -Fe2O3-buffered Cr2O3 films are larger than that of the Pt-buffered Cr2O3 film, and both the samples have larger a values than bulk Cr2O3 (a = 4.95 Å). The unexpected a value of the Pt-buffered Cr2O3 could have resulted from the lattice relaxation due to misfit dislocations or dislocations at grain boundaries at Pt/Cr2O3 interface owing to a relatively bad lattice matching, while it is difficult to identify the dislocations from TEM images (Fig. 5 (b)).

Blocking temperature and magnetic anisotropy
Next, using 1.25-nm-thick Ru spacer samples, for which finite Hex and TB were obtained, we clarified the buffer layer effect on the TB using Pt-and -Fe2O3-buffered sample with 20-nm-thick Cr2O3. Figure 6 shows the temperature dependence of the JK for Pt-and -Fe2O3-buffered 20-nm-thick Cr2O3 (c-Al2O3/Pt 25 or -Fe2O3 20/Cr2O3 20/Ru 1.25/Co 1/Pt 5 (nm)). As shown in figure 2, TB ≈ 150 K for the Pt-buffered sample. A much higher TB ≈ 260 K was obtained for the -Fe2O3-buffered samples. Since the TN of the -Fe2O3-buffered 20nm-thick Cr2O3 film was lowered to 269 K due to the lattice strain 26) , we managed to obtain TB ≈ TN for a 20nm-thick Cr2O3 film. Because the JK values of these samples are almost equal, or slightly larger for the -Fe2O3buffered samples, the KAFtAF of the -Fe2O3-buffered samples must be much higher than that of the Pt-buffered samples. In other words, the KAF of the Cr2O3 films increased by using the -Fe2O3 buffer layer. If we assume the Mauri's domain wall model 27) , the higher KAF links to the higher Hex(JK). The slightly larger JK of -Fe2O3buffered samples can comes from the higher KAF, while more works are required for the confirmations. Based on the Mauri's model, the TN change also affect to the magnitude of the Hex. However the affection of Hex by the less than 10% difference in TN between -Fe2O3-buffered sample (TN ~ 269 K) and Pt-buffered sample (TN ~ 294 K) should be negligible small. In addition, we investigated the relations between the JK and TB for the Ptand -Fe2O3-buffered samples with different Ru spacer thicknesses. Figure 7 shows the TB values of the Pt-and -Fe2O3-buffered samples plotted against the JK at 50 K. Irrespective of the Ru spacer thickness, higher TB values were obtained for the -Fe2O3-buffered samples, while the TB decreased with increasing JK for both samples. These results clearly demonstrate that the KAF of the Cr2O3 layer is higher when using an -Fe2O3 buffer layer than a Pt buffer layer.
Moreover, we investigated the Cr2O3 layer thickness dependence of the JK and TB for the -Fe2O3-buffered sample. Figure 8 shows the temperature dependence of the JK for Al2O3/-Fe2O3 20/Cr2O3 tCr2O3/Ru 1.25/Co 1/Pt 5 (nm) with various Cr2O3 thicknesses (3 nm ≤ tCr2O3 ≤ 20 nm). We observed an exchange bias for a 5-nmthick Cr2O3 sample (TB ≈ 10 K), although thinner Cr2O3 samples exhibited no apparent exchange bias. As expected from the MB model, the TB decreases with decreasing Cr2O3 thickness, while the magnitude of the JK is almost unchanged in the 5 nm ≤ tCr2O3 ≤ 20 nm range. These results indicate that the MB model is qualitatively applicable for this system even for thin Cr2O3 regions (tCr2O3 ≤ 20 nm).
Because -Fe2O3 is an antiferromagnet, in addition to the lattice strain effect, an enhancement of KAF due to the interlayer coupling between the antiferromagnetic -Fe2O3 and Cr2O3 can be considered, as reported in the NiO/CoO system 28) . However, the observed dependence of the TB on the Cr2O3 thickness does not support this assumption. If the interlayer coupling effect is dominant, the TB will not decrease with decreasing Cr2O3 thickness because the -Fe2O3 thickness is constant. However, our data suggests that the TB decreases with decreasing Cr2O3 thickness. Thus, the interlayer coupling effect is negligibly small in this case, probably owing to the small KAF of -Fe2O3 (~2 × 10 4 J/m 3 at low temperature 20) ).
Using the MB model, we estimated the change in the KAF for 20-nm-thick Cr2O3 films due to the strain induced by the buffer layer. According to the MB model, at the critical point where the exchange bias abruptly disappears, the relationship KAF = Jex/tAF holds true. If we assume that the Jex is almost the same as the JK, we can estimate the KAF as KAF = JK/tAF. In fact, the KAF values of FeMn/NiFe 29) and IrMn/NiFe 19) at room temperature have been estimated from KAF = JK/tAF cr by determining the critical antiferromagnet thickness tAF cr .
In this study, we estimated KAF = JK cr /tAF. We determined the critical unidirectional magnetic anisotropy energy JK cr at 100 K by changing the Ru spacer layer thickness for Al2O3/Pt 25 or -Fe2O3 20/Cr2O3 20/Ru tRu/Co 1/Pt 5 (nm) structure sample. The critical Ru spacer layer thickness was 0.75 nm (1.25 nm) for the -Fe2O3-(Pt-) buffered sample, and JK cr = 0.37 mJ/m 2 (0.09 mJ/m 2 ) was obtained, as shown in Fig. 9. The estimated KAF values at 100 K were 1.9 × 10 4 J/m 3 for the -Fe2O3-buffered sample and 4.5 × 10 3 J/m 3 for the Pt-buffered sample.
Note that the TB of these samples are not exactly 100 K, but between 100 K and 150 K. Due to the rough determination of TB, the JK cr and the estimated KAF values are slightly underestimated. Because the calculation of the KAF using the MB model includes many assumptions, we could not obtain an exact absolute value of the KAF from these calculations. However, it can be used to compare the two samples with similar structures. In this study, only the buffer layers were different. The other film properties such as the Cr2O3 layer thickness, spacer layer, and Co layer thickness were maintained. Because the precise characterization of the KAF of antiferromagnetic thin films is considerably difficult, we believe that this is a good method to approximately estimate the KAF. Based on these concepts, it was found that the KAF of the -Fe2O3-buffered sample is nearly four times that of the Pt-buffered sample.
Here, we compare the experimental results and theoretical predictions assuming that the variation of the KAF

Lattice strain effect on blocking temperature and Néel temperature
Finally, we discuss the lattice strain effect on the TB and TN. The TN data for these samples are included in table 2. As previously reported by us 26) , the TN of an -Fe2O3-buffered sample is about 25 K less than that of a Ptbuffered sample, and the TN decreases with increasing a. There is a trade-off between the lattice strain effects on the TB and TN: the TB increases with increasing a, while the TN decreases. This dependence appears to be unfavorable. However, the lattice strain effect is stronger for the TB than the TN. Compared to the Pt-buffered sample, a 100 K higher TB was obtained for the -Fe2O3-buffered sample, while the TN reduction was about 25 K. Thus, it is possible to increase the TB with only a small reduction in the TN.

Conclusions
In this study, we discovered a high TB ≈ 260 K for a 20-nm-thick Cr2O3 thin film using a Ru spacer layer and an -Fe2O3 buffer layer. The Ru spacer enabled reproducible control of the magnitude of the Hex (JK), and by reducing the Hex (JK) we enhanced the TB. By changing the buffer layer material from Pt to -Fe2O3, a higher TB was attained. The enhancement of the TB may be due to the lattice strain induced KAF change, which we estimate four times higher KAF for -Fe2O3-buffered Cr2O3 film compared for Pt-buffered Cr2O3 film. We also clarified the trade-off between the TN and TB with respect to the lattice strain of Cr2O3, and demonstrated that the TB is more sensitive to lattice strain than the TN. Such a control of the TB is the first step towards utilizing the ME effect in Cr2O3 thin films. Combined with further improvement of material properties, these techniques for controlling the TB open up doors for the device application.