Magnetoelectric manipulation and enhanced operating temperature in antiferromagnetic Cr₂O₃ thin film

The first ME manipulation of exchange bias in a Cr₂O₃/Co all-thin-film system by an MEFC process was demonstrated by Ashida et al. in 2014,⁸⁰⁾ a decade after the manipulation in a Cr₂O₃ bulk single-crystal system demonstrated by Borisov et al.²³⁾ In the experiment conducted by Ashida et al., an Al₂O₃/Pt (25 nm)/Cr₂O₃ (250 nm)/Pt (1 nm)/Co (1 nm)/Pt (5 nm) structure with a high-quality Cr₂O₃ thin film fabricated by a reactive sputtering was used as the sample. Figure 2 shows the perpendicular exchange biased hysteresis loop of the sample at 200 K obtained by anomalous Hall effect (AHE) measurement; here, the switching is obtained by an MEFC process, in which antiparallel freezing magnetic H_fr and electric E_fr fields (blue solid circles) and parallel H_fr and E_fr (red open squares) are applied. Before this report, the magnetic properties of Cr₂O₃, especially those related to the exchange bias, had been extensively investigated. However, there had been few reports on the electric properties of Cr₂O₃ thin films.^43,88,89) The Cr₂O₃ film fabricated by reactive sputtering exhibited good magnetic and electric properties: it had a magnetic susceptibility almost equivalent to that of Cr₂O₃ single crystals⁷⁷⁾ and its leakage current was 4.0 × 10⁻⁵ A/cm² at 80 kV/cm, which is sufficiently small compared with that in a previous study.⁴³⁾ The realization of good magnetic and electric properties at the same time contributed to the first observation of the ME manipulation of exchange bias in a Cr₂O₃/Co all-thin-film system. Another key factor was the use of a Pt spacer layer inserted between Cr₂O₃ and Co as it reduced the exchange bias magnitude and prevented Co oxidation at the same time. Ironically, the huge perpendicular exchange bias obtained in the Cr₂O₃/Co all-thin-film system interrupted the ME manipulation, as we will discuss later. Thus, the reduction of the exchange bias magnitude by using a Pt spacer layer played an important role in the first observation. To date, various types of spacer layers have been developed, such as Pt,^{48,55,70,73,80)} Cr,^73,84) Ru,^60,66,67,73) and Au.⁷⁰⁾ These spacer layers are useful not only to control the exchange bias magnitude but also to adjust the magnetic properties of the ferromagnetic Co. The first observation also clarified the much larger manipulation energy, the EH product (the product of the applied electric and magnetic fields), in the all-thin-film system (EH ∼ 400 kOe kV/cm)⁸⁰⁾ than that in the bulk single-crystal system (EH < 30 kOe kV/cm).^24,38)

Zoom In Zoom Out Reset image size

After the first demonstration of ME manipulation, it was found to be essential to clarify the origin of the larger ME manipulation energy (EH product) in an all-thin-film system. One possibility is the suppression of the magnitude of the ME susceptibility α in the Cr₂O₃ thin film compared with that in Cr₂O₃ bulk single crystal. Since the difference in the ME energy ΔF = 2αEH becomes the driving force of the ME manipulation, if α is suppressed, the EH product required for the ME manipulation has to increase. To confirm this assumption, the measurement of α in a Cr₂O₃ thin film is required. However, the detection of the ME signal in Cr₂O₃ is not easy since the ME susceptibility α of Cr₂O₃ is relatively small (in this case, the ME signal is αE as the E field induces the magnetization). Such a difficulty is even more prominent for Cr₂O₃ thin films because the volume of a Cr₂O₃ thin film is more than three orders of magnitude smaller than that of a typical Cr₂O₃ bulk single crystal. Thus, several attempts to measure α values with higher sensitivity have been made so far. In the first measurements of the inverse ME effect by Astrov, the magnetization induced by an AC electric field was measured by surrounding pickup coils.^6,7) Later, Kita and co-workers used a superconducting quantum interference device (SQUID) magnetometer to detect the magnetization induced by a DC electric field.^10,11) After that, Borisov et al. developed a simple new approach using a conventional SQUID magnetometer; they applied an AC electric field to the sample and detected its magnetic response by using the standard AC susceptibility measurement procedure.¹³⁾ By using Borisov's approach, the first measurement of α in an Al₂O₃/Pt (25 nm)/Cr₂O₃ (500 nm)/Pt (25 nm) thin-film system was carried out.⁹⁰⁾ Soon after, Al-Mahdawi et al. developed a modified α-measurement setup using a conventional SQUID magnetometer and external lock-in amplifiers; they succeeded in measuring α for similar samples.⁹¹⁾ The scheme of the sample setup and the temperature dependence of α measured by Al-Mahdawi et al. are shown in Fig. 3. The temperature profiles of α in the Cr₂O₃ thin film measured by Borisov et al. and Al-Mahdawi et al.^90,91) are similar to that in the Cr₂O₃ bulk single crystal,^6–13) except for a small lowering of T_N and the relevant temperature in the Cr₂O₃ thin film. The most important finding is the almost identical peak value of α in both the Cr₂O₃ thin film (between 3.4 and 4.6 ps/m^90,91)) and the Cr₂O₃ bulk single crystal (between 3.8 and 5.1 ps/m^10–13)). These results clarified that, apart from the suppression of α, other dominant factors contribute to increasing the EH product in the Cr₂O₃ thin-film system.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Figure 4 summarizes the threshold ME manipulation energies (threshold EH products E_thH_fr) reported for both a Cr₂O₃ bulk single crystal and a Cr₂O₃ thin film in an MEFC process as a function of J_K/t_AF (noted as "Typical system").^23,64,80,84) Here, J_K = H_exM_st_FM is the unidirectional anisotropy energy induced by exchange bias; t_AF, H_ex, M_s, and t_FM are the Cr₂O₃ thickness, exchange bias magnitude, saturation magnetization of Co, and Co thickness, respectively. The J_K data in Fig. 4 were mainly measured at 250 K, whereas the data from Ref. 64 were measured at 265 K. As can be seen from Fig. 4, the E_thH_fr and J_K/t_AF values are linearly correlated, as indicated by the black dashed line that closely fits the data. These results allow us to estimate the following simplified free energy model for Cr₂O₃ under H_fr and E_fr in an MEFC process:^32,83)

$$\begin{equation} F \sim W_{\text{ME}} + W_{\text{EX}} = -\alpha E_{\text{fr}}H_{\text{fr}}\cos\theta + \frac{J_{\text{int}}}{t_{\text{AF}}}\cos\theta. \end{equation} \tag{ 2 }$$

Here, W_ME and W_EX denote the ME term and exchange bias term of the free energy, respectively. J_int is the interface exchange coupling energy, which is strongly correlated with J_K. Because it is a field cooling process, we neglect the antiferromagnetic anisotropy term K_AF sin² θ. The threshold condition for the switching at around T_N of Cr₂O₃ is

$$\begin{equation} \alpha E_{\text{th}}H_{\text{fr}} = \frac{J_{\text{int}}}{t_{\text{AF}}}. \end{equation} \tag{ 3 }$$

Since α is almost constant regardless of its form (bulk or thin film), if we assume a strong positive correlation between J_int and J_K, Eq. (3) indicates a linear relation between E_thH_fr and J_K/t_AF, which well agrees with the trend observed in Fig. 4. The relation in Eq. (3) can be qualitatively interpreted as follows: when we apply a positive H_fr, the neighbor ferromagnetic domains align along the magnetic field direction (positive direction) and the Cr₂O₃ antiferromagnetic domains receive induced unidirectional anisotropy from the ferromagnetic domains through the exchange bias. This is the same as the usual magnetic field cooling process in exchange bias systems. However, since Cr₂O₃ is an ME material, Cr₂O₃ antiferromagnetic domains can be controlled to the desired direction by applying both magnetic and electric fields. To switch the Cr₂O₃ antiferromagnetic domains against the unidirectional anisotropy induced by the exchange bias, an EH product greater than J_int/αt_AF is required. Given that the ME energy depends on the sample volume, whereas the exchange bias energy depends on the sample interface, a decrease in the Cr₂O₃ thickness will increase the threshold EH product. This should explain the larger ME manipulation energy for Cr₂O₃ all-thin-film systems. The threshold EH product is also proportional to J_int, i.e., the threshold EH product increases as the H_ex magnitude increases. Thus, referring to Fig. 2, we could easily realize the ME manipulation of exchange bias by reducing the H_ex magnitude by using a Pt spacer. However, the negative correlation between the ME manipulation energy and the Cr₂O₃ thickness is a critical problem, especially toward practical applications, where Cr₂O₃ films should be made as thin as possible. Nevertheless, the electric field required for ME switching increases inversely proportionally to the Cr₂O₃ thickness, and finally it exceeds the maximum breakdown voltage of the Cr₂O₃ thin film. That means we cannot achieve the ME manipulation of exchange bias for even thinner Cr₂O₃ films.

Zoom In Zoom Out Reset image size

This problem occurs as long as we use an exchange bias for readout, which imposes a limit on the Cr₂O₃ thickness. Moreover, this issue occurs not only for MEFC manipulation but also for isothermal manipulation. Recently, a possible way to overcome the problem was found; the unexpected finite magnetization in Cr₂O₃ plays an important role in this sense. Although Cr₂O₃ is an AFM, occasionally finite magnetization has been observed in Cr₂O₃ films.^{32,83,90,92–101)} The origin of such finite magnetization is still under debate; it has been attributed to piezomagnetism,⁹²⁾ boundary magnetization,^{90,93–97,102)} defect-mediated magnetization,^98–101) and so forth. In an exchange bias system, the magnetization from an antiferromagnetic layer sometimes evokes a curious positive exchange bias phenomenon,⁵⁹⁾ in which both the sign and magnitude of the exchange bias vary by changing the magnitude of the applied magnetic field H_fr during the magnetic field cooling process. Figure 5 shows an example of the positive exchange bias phenomenon observed in an Al₂O₃/Pt (25 nm)/Cr₂O₃ (250 nm)/Pt (t_Pt)/Co (1 nm)/Pt (5 nm) structure. Here, the exchange bias measured at 100 K is plotted as a function of H_fr; as can be seen, as H_fr increases, the exchange bias changes its sign from negative to positive. This phenomenon is explained by the competition between the exchange bias energy and the Zeeman energy of the Cr₂O₃ magnetization. With increasing H_fr, the magnitude of the Zeeman energy increases, whereas the exchange bias energy remains constant. As a consequence, at a certain H_fr, the Zeeman energy exceeds the exchange bias energy and the positive exchange bias state is stabilized. The results were well reproduced by an equation that takes into account both the free energy of the system and the thermal fluctuation effect, as shown in Fig. 5. A crucial finding is that the exchange bias energy and the Zeeman energy are fully compensated at the H_fr value at which the exchange bias becomes zero (x-intercept in Fig. 5). This idea can be utilized to reduce the ME manipulation energy in an MEFC process. As mentioned above, the ME manipulation energy strongly depends on the exchange bias energy; thus, if we can minimize the influence of the exchange bias energy by compensating the Zeeman energy, the ME manipulation will become significantly easier to achieve. Such an idea based on low-energy manipulation has been proposed by Al-Mahdawi et al.³²⁾ In a system with a finite Cr₂O₃ magnetization (M_Cr2O3), if we add the Zeeman term of M_Cr2O3 (W_ZM), the free-energy formula becomes

$$\begin{align} F &= W_{\text{ME}} + W_{\text{ZM}} + W_{\text{EX}}\\ &= -\alpha E_{\text{fr}}H_{\text{fr}}\cos\theta - M_{\text{Cr2O3}}H_{\text{fr}}\cos\theta + \frac{J_{\text{int}}}{t_{\text{AF}}}\cos\theta \end{align} \tag{ 4 }$$

and the threshold condition of the switching at around T_N of Cr₂O₃ is

$$\begin{equation*} \alpha E_{\text{th}}H_{\text{fr}} = -M_{\text{Cr2O3}}H_{\text{fr}} + \frac{J_{\text{int}}}{t_{\text{AF}}} \end{equation*}$$

or

$$\begin{equation} E_{\text{th}} = - \frac{M_{\text{Cr2O3}}}{\alpha} + \frac{J_{\text{int}}}{\alpha t_{\text{AF}}}\frac{1}{H_{\text{fr}}}. \end{equation} \tag{ 5 }$$

Equation (5) indicates that if a calibrated magnetic field H_fr = J_int/M_Cr2O3t_AF is applied, the exchange bias energy is fully compensated by the Zeeman energy of M_Cr2O3, and low-energy ME manipulation of the exchange bias becomes possible. This idea was experimentally demonstrated by using an Al₂O₃/Pt (25 nm)/Cr₂O₃ (1000 nm)/Pt (t_Pt)/Co (1 nm)/Pt (5 nm) system.⁸³⁾ For the demonstration, an accidentally obtained finite magnetization of Cr₂O₃ was used, the same as that also used for the observation of the positive exchange bias phenomenon.⁵⁹⁾ At that stage, we did not know the exact origin of M_Cr2O3 and we could not control its magnitude. However, we observed such magnetization in a reproducible fashion. Given the limitations to the applicable magnetic field, to obtain a fully compensated state (H_fr = J_int/M_Cr2O3t_AF) within an applicable magnetic field, we needed to design J_int and t_AF. In the demonstration, we used a 1000-nm-thick Cr₂O₃ film to obtain a fully compensated state with a relatively small magnetic field. The effective magnitude of J_int was controlled by changing the thickness of the Pt spacer layer. To study the relation of the compensation, an E_th vs 1/H_fr plot of the ME manipulation energy was used. In this plot, the y-intercept and slope are related to M_Cr2O3/α and J_int/αt_AF, respectively, as we can estimate from Eq. (5). Figure 6 shows E_th as a function of 1/H_fr in the context of MEFC manipulation.⁸³⁾ The x-intercept represents the fully compensated point (H_fr = J_int/M_Cr2O3t_AF). The fitting lines of all samples are characterized by almost the same negative y-intercepts, corresponding to the similar M_Cr2O3 values of the samples. The slopes of the fitting lines increase as the Pt thickness decreases, i.e., as the H_ex magnitude increases. Due to the negative y-intercepts owing to the existence of finite M_Cr2O3 values, all samples exhibit finite x-intercepts, thus indicating that the low-energy manipulation of the exchange bias is possible. The minimized threshold EH product obtained from Fig. 6 is also plotted in Fig. 4 by red solid diamonds (labeled as "System with M_Cr2O3"). While the "Typical system" data show a linear increase in the threshold EH product with increasing J_K/t_AF, the "System with M_Cr2O3" data exhibit an almost constant trend. Even for a 1000-nm-thick Cr₂O₃ system with an untunable small M_Cr2O3, we achieved a reduction of approximately two orders of magnitude for the threshold EH product. However, significant benefits will be obtained when the Cr₂O₃ sample becomes much thinner. Referring to Fig. 4, for J_K/t_AF ∼ 10⁵ emu/cm³ (corresponding to t_AF < 10 nm), the EH product required to overcome the exchange bias energy becomes ∼10⁵ kOe kV/cm. Thus, even if H_fr of 10 kOe is applied, E_th becomes as large as 10 MV/cm, which is higher than the applicable electric field, and ME manipulation can no longer be achieved. If a larger Cr₂O₃ magnetization is obtained, a reduction of several orders of magnitude for the threshold EH product will be possible, and the limitation of the Cr₂O₃ thickness can be overcome. Further elucidation of the origin of M_Cr2O3 and also ways to control it are issues under study.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

ME manipulation for Cr₂O₃/Co all-thin-film systems in an isothermal process was realized by two groups in 2015,^84,85) soon after the demonstration in an MEFC process. Since the necessity of a much higher voltage in an isothermal process than that in an MEFC process was assumed, significant effort was made to develop higher-breakdown-voltage samples. As a result, breakdown voltages of MV/cm order have been achieved, thus allowing the realization of isothermal ME manipulation in Cr₂O₃ all-thin-film systems. Figure 7 summarizes the isothermal manipulation of the exchange bias in an Al₂O₃/Pt (25 nm)/Cr₂O₃ (500 nm)/Cr (0.25 nm)/Co (1 nm)/Pt (5 nm) system.⁸⁴⁾ As can be seen, the switching of the exchange bias polarity by simply changing the isothermal electric field E_iso while maintaining the isothermal magnetic field H_iso was clearly demonstrated [Figs. 7(a) and 7(b)]. In these studies, a small coercive field of Co was realized by using a Cr spacer layer while maintaining its perpendicular magnetic anisotropy. Such a smaller coercive field (<50 Oe) than the exchange bias magnitude (∼100 Oe) enables Co magnetization switching at zero magnetic field, in addition to the exchange bias polarity reversal. In this study, repeated operations of exchange bias/Co magnetization switching were also demonstrated [Figs. 7(c) and 7(d)].

Zoom In Zoom Out Reset image size

ME manipulation experiments based on the MEFC process are crucial to understand the effect of the exchange bias energy on ME manipulation but are not suitable for practical applications. Although the possibility of heat-assisted operation using an MEFC process still remains, isothermal manipulation would be a more realistic candidate for spintronic applications. Thus, the understanding of isothermal ME manipulation is essential. However, it is not easy to investigate the manipulation properties in an isothermal process, since they require much higher EH product values for the switching than those in an MEFC process. Recently, some updates concerning the isothermal manipulation energy were reported. In order to understand the switching mechanism,^86,87) the magnetic field dependence of the threshold EH product was measured by using a large magnetic field (H_iso > 40 kOe). The results indicated that the nucleation and growth of the antiferromagnetic domains occur in the isothermal process. However, a full understanding of isothermal ME manipulation is yet to be achieved and further investigation is required. The investigation will be accelerated if the isothermal manipulation energy can be reduced.

Another important aspect of isothermal ME manipulation is the elucidation of the switching dynamics in the Cr₂O₃/Co exchange bias system. Since Cr₂O₃ is an AFM, ultimately, operations on the order of ps are expected,^103–106) whereas a slightly worse operation speed is expected to characterize in Cr₂O₃/FM exchange bias systems. Through pulse voltage application experiments, the dynamics of an Al₂O₃/Pt (20 nm)/Cr₂O₃ (150 nm)/Au (0.5 nm)/Co (0.4 nm)/Pt (1.5 nm) system was investigated by AHE measurement with a Hall bar of 2 × 2 µm² cross point.⁸⁶⁾ The switching time in isothermal switching depends on the applied EH product: when a sufficiently large EH product is applied, switching time periods shorter than 10 ns are obtained, whereas when the EH product is close to the threshold value, the switching time is on the order of 100 ns. These results suggest that the switching nature is governed by domain wall motion. The estimated minimal domain wall mobility was 2 × 10⁻³ m s⁻¹ Oe⁻¹ at 540 kV/cm. However, the domain wall mobility has room for further enhancement, as demonstrated by Belashchenko et al.¹⁰⁷⁾ in the prediction of a two-orders-of-magnitude higher maximum mobility for Cr₂O₃. In addition, the switching speed should increase with decreasing element size. In the future, operations with a time period below sub-ns order will possibly be realized, comparable with SOT-type¹⁰⁸⁾ or voltage-torque-type^36,37) ultrahigh-speed-operation MRAMs, or even faster.

Another notable topic is FM-free operation. As mentioned above, while the exchange bias has a strong merit in terms of readout, it interrupts the ME manipulation of Cr₂O₃. In addition, it can partially degrade the excellent properties of an antiferromagnetic material, such as the THz resonance frequency and no stray field. Although difficulties in the readout have to be overcome, FM-free structures offer the possibility of futuristic ultrahigh-performance spintronic device applications. As another kind of readout technique, the spin hall magnetoresistance of Cr₂O₃ has recently been measured.¹⁰⁹⁾ Moreover, the interface polarization of overlayer paramagnets, such as Pt and Pd on Cr₂O₃, was found, which was detected by X-ray magnetic circular dichroism (XMCD)¹¹⁰⁾ and AHE measurement.¹¹¹⁾ Kosub et al. demonstrated purely antiferromagnetic ME random access memory operation in a Pt (20 nm)/Cr₂O₃ (200 nm)/Pt (2.5 nm) structure by AHE detection of the induced interface polarization.¹⁰¹⁾ Of course, some issues still remain, such as a relatively large footprint owing to the need for AHE detection and a small AHE signal magnitude; however, this is an attractive technique for future applications.

Since the first demonstration of ME manipulation in a Cr₂O₃ all-thin-film system in 2014,⁸⁰⁾ significant progress has been made in the understanding of its mechanism and many beneficial properties have been elucidated. Recently, research studies motivated by device applications have also become active. For example, some studies related to the local writing of exchange bias have been exploited^65,71) for storage application;³²⁾ further applications of this field, including the use of the spin Seebeck effect,¹¹²⁾ spin transport,¹¹³⁾ and graphene transistors,^114–117) were also reported. Further progress is expected for electrically controllable AFM Cr₂O₃ thin films.

The Néel temperature of Cr₂O₃ (T_N ∼ 307 K), which is relatively high among the well-characterized ME materials, does not provide enough flexibility for practical applications. The insufficient T_N has been recognized as a paramount issue and several solutions have been proposed to overcome this limitation.

The first one concerns impurity substitution. Since the magnetic interaction in Cr₂O₃ is dominated by the Cr–Cr direct exchange interaction, any change in the Cr–Cr bond length strongly affects the effective exchange field. In addition, the impurity states introduced by doping can further affect the exchange coupling. In this regard, Mu et al., through electronic structure calculations, investigated the doping effect of several anions (V, Ti, Mn, Fe, Co, and Ni) and cations (N and B) on the effective exchange field of Cr₂O₃.¹¹⁸⁾ Although transition-metal impurities reduce T_N of Cr₂O₃, Mu et al. found that B doping can enhance T_N while retaining a large excitation gap, i.e., retaining its insulating properties. This is because the B doping introduces hybrid impurity states, which strongly enhance the exchange energies for the Cr neighbor ions. Mu et al. predicted an increase in the T_N of Cr₂O₃ by approximately 10% per 1% of O sites substituted by B. Later, Street et al.⁹⁴⁾ experimentally enhanced T_N of Cr₂O₃ up to 400 K by B doping. They estimated the change in T_N from the temperature dependence of the measured remnant boundary magnetization. Despite the experimental demonstration of T_N ∼ 400 K, several challenges still remain, including the difficulty of substituting an anion O site with B. Given that B easily becomes a cation, such as B₂O₃, a specific technique is required to fabricate B-doped Cr₂O₃. In their report, Street et al. adopted the pulsed laser deposition method, letting the Cr₂O₃ target be immersed in a B₁₀H₁₄ background gaseous atmosphere. In contrast, high-quality Cr₂O₃ films have recently been fabricated by a reactive sputtering method. The development of a fabrication technique for high-quality B-doped Cr₂O₃ by sputtering is one challenge. Another challenge is the stabilization of B into the Cr₂O₃ matrix. Since B has an exceptionally small atomic radius, it easily diffuses after repeated heat treatments. However, given the fact that B doping is the most promising solution for T_N enhancement, overcoming all these issues will be the key to practical applications.

The second approach is strain engineering. As previously specified, since the magnetic interaction in Cr₂O₃ is dominated by the Cr–Cr direct exchange interaction rather than the super-exchange interaction, changes in the bond length due to lattice strains strongly affect T_N. Strain engineering has been adopted for bulk Cr₂O₃ systems^119–123) but not for Cr₂O₃ thin films. Kota et al. investigated the effect of an isovolumetric elastic strain on T_N by first-principles calculations and Monte Carlo simulations.^123–125) The variation of the calculated T_N with the c/a ratio is shown in Fig. 8 (red solid circles). It was found that T_N is enhanced by a tensile strain and decreased by a compressive strain along the c-axis. The effect of the lattice strain is relatively large; in particular, more than 20% of the T_N enhancement was estimated to be due to 5% tensile strain in the crystal lattice. Such strain-induced T_N control has been partly demonstrated for Cr₂O₃ thin-film systems,^60,66) where several different buffer layers were used to control the lattice strain. T_N was identified in a low-magnetic-field M–T measurement using the Cr₂O₃/Ru/Co exchange bias system, where the magnetic anisotropy of Co was well tuned by using a Ru spacer layer for the detection of T_N.⁶⁶⁾ In Fig. 8, the experimentally obtained T_N of a buffered Cr₂O₃ film is plotted against the c/a ratio^60,66) in addition to the calculated T_N variation.¹²⁴⁾ As can be seen, the calculated T_N variation is well reproduced by experiments, thus confirming the correctness of the calculated data. Unfortunately, at present, experimental results have only been obtained for compressive strain regions, thus showing a decrease in T_N. To enhance T_N, it is necessary to develop suitable buffer layers.

Zoom In Zoom Out Reset image size

The third approach is to exploit the enhancement of spin correlations by using the AFM/AFM proximity effect. The idea is that when a thin Cr₂O₃ film is placed next to a high-T_N AFM, even at T > T_N of Cr₂O₃, retention of the antiferromagnetic correlation of Cr₂O₃ near the interface is expected because of the antiferromagnetic correlation with the neighbor high-T_N AFM. This behavior has been demonstrated for thick NiO (T_N ∼ 523 K) and thin CoO (T_N ∼ 293 K) bilayer systems.^126,127) Suggestive results have also been obtained for nanocomposite structures, including Cr₂O₃ and α-Fe₂O₃.¹²⁸⁾ Kota et al. calculated the spin correlation enhancement effect in the α-Fe₂O₃/Cr₂O₃ structure.¹²⁹⁾ In their study, α-Fe₂O₃ was selected for a high-T_N AFM as it has a high T_N (∼ 950 K) and the same corundum-type crystal structure as Cr₂O₃. The calculation results indicate that the spin correlation of Cr₂O₃ near the interface is sustained even at 1.5 times T_N of Cr₂O₃ (∼ 450 K) when an oxygen-divided interface is realized at the α-Fe₂O₃/Cr₂O₃ interface. On the basis of these calculation results, research on the α-Fe₂O₃/Cr₂O₃ structure has been going on. In order to obtain α-Fe₂O₃ with a perpendicular spin configuration, the enhancement of the Morin transition temperature of α-Fe₂O₃ by Ir doping was demonstrated.^130–133) The spin correlation length of Cr₂O₃ was experimentally estimated from a finite-size scaling effect of T_N.⁶⁶⁾ However, the T_N enhancement due to the spin correlation enhancement in an Ir-doped α-Fe₂O₃/Cr₂O₃ system has not yet been demonstrated, probably because of the interfacial mixing or metal-split-type interface at the Ir-doped α-Fe₂O₃/Cr₂O₃ interface. Thus, to achieve such a demonstration, an advanced interface control technique has to be developed.

To achieve a sufficiently high operating temperature, not only T_N but also the perpendicular magnetic anisotropy has to be enhanced. Essentially, the magnetic anisotropy has the role of maintaining the thermal stability. In addition, in exchange bias systems, the magnetic anisotropy is related to the blocking temperature of the exchange bias, T_B, as has been discussed on the basis of the MB model.^45,78,79) When the value of the K_AFt_AF product is not sufficient, T_B significantly lower than T_N has been observed in the Cr₂O₃/Co exchange bias system. The domain wall width is also related to the magnetic anisotropy.¹³⁴⁾ Thus, a high perpendicular magnetic anisotropy is required for devices characterized by domain-wall-related phenomena. For these reasons, effort has been made to enhance the perpendicular magnetic anisotropy. In this section, we describe the enhancement of the magnetic anisotropy by using lattice strain and doping. One of the most basic difficulties in this sense is the lack of an easy way to evaluate the magnetic anisotropy of an AFM. Typically, magnetic anisotropies of an AFM have been evaluated by antiferromagnetic resonance (AFMR) measurements.¹³⁵⁾ However, it is difficult to adopt such a technique of measurement for antiferromagnetic thin films. In this section, we will qualitatively discuss the magnetic anisotropy of Cr₂O₃ films based on the relation between the magnetic anisotropy and T_B, where T_B increases as the magnetic anisotropy increases.

As shown in the previous section, lattice strains affect T_N through a change in the exchange interactions. In addition, lattice strains distort CrO₆ octahedra, thus changing the single-ion anisotropy (or magnetocrystalline anisotropy) through the modulation of the crystal field. Kota and Imamura estimated the strain dependence of single-ion anisotropy by first-principles calculations and clarified that the single-ion anisotropy increases monotonically with a.¹³⁴⁾ Another notable finding is the extremely small single-ion anisotropy of Cr₂O₃ in the equilibrium state. A possible one-order-of-magnitude enhancement of single-ion anisotropy by a small lattice strain (∼1%) was indicated. Experimental results consistent with the calculated lattice strain dependence of the magnetic anisotropy trend were obtained for Cr₂O₃ (20 nm)/Ru (1.25 nm)/Co (1 nm) exchange bias systems with different buffer layers.⁶⁷⁾ The buffer layer was changed from Pt to α-Fe₂O₃, which increases a for the Cr₂O₃ layer, thus resulting in a large enhancement of T_B. A rough/semiquantitative estimation based on the MB model suggests an enhancement of the magnetic anisotropy of about fourfold by changing the buffer layer. These results confirm the effectiveness of the lattice strain for enhancing the magnetic anisotropy. It is important to note that, while increasing the lattice parameter a yields an increase in the single-ion anisotropy, it simultaneously causes a slight decrease in T_N. For example, when the magnetic anisotropy was enhanced fourfold, a decrease in T_N of 25 K (from 294 to 269 K) was observed. We have to take into account these contradictory properties when we utilize the lattice strain effect for operating-temperature engineering.

There have been some reports of the enhancement of the magnetic anisotropy of Cr₂O₃ by doping for a Cr₂O₃ bulk system^136,137) but not for a Cr₂O₃ thin-film system. This mainly reflects the difficulty in the evaluation of such enhancement in Cr₂O₃ thin-film systems. We fabricated Al-doped Cr₂O₃/Co exchange-biased thin films and evaluated the dopant concentration dependence of the magnetic anisotropy through the change in T_B.⁷²⁾ T_B increases as the Al content increases, thus indicating a monotonic increase in the magnetic anisotropy. The increment of T_B is much larger than that obtained by lattice strain effects, thus suggesting the effectiveness of Al doping in enhancing magnetic anisotropy, while it is difficult to evaluate the increment quantitatively. Although some theoretical considerations were presented, the mechanism has not yet been completely clarified. For example, the change in the magnetic anisotropy was discussed in terms of the change in lattice parameters; such changes, as specified above, can modulate the magnetic anisotropy by more than one order of magnitude. Figure 9 shows calculated single-ion anisotropy variations when both lattice parameters a and c are changed.⁷²⁾ The experimental values of the lattice parameters of Pt-buffered undoped Cr₂O₃, Pt-buffered Al-doped Cr₂O₃, and α-Fe₂O₃-buffered undoped Cr₂O₃ thin films are plotted in the same figure. As can be seen from Fig. 9, in Al-doped Cr₂O₃, both a and c decrease almost isotropically, reflecting the smaller ionic radius of Al³⁺ than that of Cr³⁺. As the Al content increases, the decrease becomes larger. However, Fig. 9 indicates that single-ion anisotropy is not significantly changed by isotropic deformation. This is in contrast to the expected large change in the single-ion anisotropy for α-Fe₂O₃-buffered undoped Cr₂O₃ (black open inverted triangle in Fig. 9), where c is increased and a is decreased. In addition, the experimentally observed enhancement of the magnetic anisotropy due to Al doping cannot be explained from changes in the electronic structure. These results indicate that, in addition to conventional lattice strains, there is a mechanism to enhance the magnetic anisotropy of Cr₂O₃, whose origin, however, is not yet clarified. To summarize, the parallel use of lattice strain and doping methods is a promising way to obtain a sufficiently high perpendicular magnetic anisotropy. Since, at present, the development of high-magnetic-anisotropy antiferromagnetic materials is mainly limited by the lack of evaluation techniques, progress in this area is crucial.

Zoom In Zoom Out Reset image size