Strong Impact of Underlayers on the Voltage‐Controlled Magnetic Anisotropy in Interface Engineered Co/MgO Junctions with Heavy Metals

The voltage‐controlled magnetic anisotropy (VCMA) effect in ferromagnets is of crucial interest for next‐generation non‐volatile magnetic memory technologies since it enables spin manipulation with low power consumption in addition to high‐speed operation and high writing endurance. Although intense efforts are made in the past decade to increase the efficiency of the purely electronic VCMA effect, the physical origin of this phenomenon is still elusive, particularly for interface‐engineered ferromagnetic materials. Here, a pathway is proposed to tune the VCMA effect by exploiting a combination of the insertion of heavy metals and underlayers. Focusing on Co/MgO junction deposited on Os underlayer, a large VCMA effect of −100 fJ V−1 m−1 is demonstrated with Pt insertion, which is ≈50% greater than the case deposited on Pt underlayer. On the other hand, smaller VCMA effect is observed with Ir insertion, and even a positive VCMA effect is observed with Os insertion, which are vastly different from the case deposited on Pt underlayer. The systematic variation of VCMA coefficient depending on the underlayer materials implies the effect of electron depletion from underlayers. The concepts demonstrated here can be applicable for other spintronic materials and may open a way to interfacial spin‐orbit engineering for VCMA effect.


Strong Impact of Underlayers on the Voltage-Controlled Magnetic Anisotropy in Interface Engineered Co/MgO Junctions with Heavy Metals
Hiroyasu Nakayama,* Takayuki Nozaki, Tomohiro Nozaki, and Shinji Yuasa DOI: 10.1002/admi.202300131 The VCMA effect is a phenomenon that induces a magnetic anisotropy change for the ferromagnet with applying an electric field through the insulator. Since application of an electric current is not necessary to manipulate spins, the VCMA effect has potential advantage toward the spintronic devices with ultra-low power consumption. In addition, the purely electronic VCMA effect enables spin manipulation free from chemical reactions or ionic displacements, which results in high-speed operation and high writing endurance. [3,4] As the origin of the purely electronic VCMA effect, electric-field modulation of the electronic occupation states, [5][6][7] the Rashba spin-orbit anisotropy, [8] and the magnetic dipole moment [9] have been proposed to describe phenomena at the junction between homogeneous ferromagnets and insulators.
Since the first report of VCMA in 2007, [10] the effect has been studied for a wide range of materials. In the early stages of the study, however, it was challenging to observe voltage-induced phenomena in ferromagnetic metals because of the extremely small Thomas-Fermi screening length of metals. [5,10,11] Therefore, a combination of ultrathin metallic films and liquid electrolyte were required for the first demonstration of the VCMA effect. [10] A year later, the VCMA effect in all-solid structure was first reported for a ferromagnetic semiconductor at low temperature. [12] Subsequently, the VCMA effect in all-solid structures at room temperature was achieved in ultrathin 3d transition metals. [13][14][15] To apply an extremely strong electric field on the surface, studies on the VCMA effect using an ionic liquid have also been conducted. [16][17][18] Although many experimental studies on the VCMA effect have been conducted, the sign of the VCMA effect in most materials is negative (see Figure 1), [2] and only a few materials exhibit a positive VCMA effect at room temperature; L1 0 -ordered FePd alloys [10,19] and Co/oxide structures such as Co/CoO x /HfO 2 , [20] Co/SrTiO 3 , [21,22] Co/Cr 2 O 3 , [23] and Co/CoO x /TiO x junctions. [24] A positive VCMA has also been reported for Co(0.4 nm)/MgO/ HfO 2 [25] and Ru/CoFeB/MgO structures with sputter-deposited MgO barrier. [26] For the case using sputter-deposited MgO barrier, however, it is also reported that the sign reversal of the VCMA effect occurs due to the subtle changes in the oxidation state at the junction interface. [27] The magnitude of the The voltage-controlled magnetic anisotropy (VCMA) effect in ferromagnets is of crucial interest for next-generation non-volatile magnetic memory technologies since it enables spin manipulation with low power consumption in addition to high-speed operation and high writing endurance. Although intense efforts are made in the past decade to increase the efficiency of the purely electronic VCMA effect, the physical origin of this phenomenon is still elusive, particularly for interface-engineered ferromagnetic materials. Here, a pathway is proposed to tune the VCMA effect by exploiting a combination of the insertion of heavy metals and underlayers. Focusing on Co/MgO junction deposited on Os underlayer, a large VCMA effect of −100 fJ V −1 m −1 is demonstrated with Pt insertion, which is ≈50% greater than the case deposited on Pt underlayer. On the other hand, smaller VCMA effect is observed with Ir insertion, and even a positive VCMA effect is observed with Os insertion, which are vastly different from the case deposited on Pt underlayer. The systematic variation of VCMA coefficient depending on the underlayer materials implies the effect of electron depletion from underlayers. The concepts demonstrated here can be applicable for other spintronic materials and may open a way to interfacial spin-orbit engineering for VCMA effect.

Introduction
With recent advances in the field of spin-orbitronics, the spin manipulation using interfaces has attracted increasing attention in terms of fundamental as well as the application point of view. One of the most promising approaches is the use of electric-field based spin manipulation via the voltage-controlled magnetic anisotropy (VCMA) effect in a junction between a ferromagnet and a non-magnetic insulator (see Figure 1). [1,2] www.advmatinterfaces.de VCMA effect is characterized by the VCMA coefficient given by ΔK PMA t/ΔE. Here, K PMA is the effective perpendicular magnetic anisotropy (PMA) energy, t is the thickness of the ferromagnet, and E = V bias /t barrier is the applied electric field, where V bias is the applied bias voltage and t barrier is the thickness of the tunneling barrier.
From an applications point of view, materials with a larger VCMA effect are desired to be achieved in all-solid structures. [1] One method is to utilize a heavy metal layer inserted at the ferromagnet/insulator junction interface. Nakamura et al., theoretically discussed the impact of inserting an atomic layer at an Fe/MgO junction and suggested that a greater VCMA effect might be obtained by using a 5d heavy metal as the insertion layer. [28] In their study, it was pointed out that the magnitude of the spin-orbit coupling of the inserted element and the position of the Fermi level with respect to the d-band level are crucial to obtain a large VCMA effect. Therefore, 5d transition metals such as Pt and Ir, which exhibit strong spinorbit interactions and magnetic proximity effects, are expected to be promising materials compared to 3d and 4d transition metals such as Cr [29] and Pd. [18] Another theoretical calculation predicts that Re and Os can be promising materials to exhibit larger VCMA effect by inserting into Fe/MgO junctions. [30] In experimental studies, the small improvement of the VCMA effect was reported for CoFeB/MgO junctions with Pt insertion, [31] and the importance of the heavy metals has been subsequently demonstrated in epitaxially grown Fe-based alloy thin films with Ir doping. [32,33] However, in their systems, the potential of interfacial effect may not be fully exploited since the diffusion of heavy metals into the ultrathin ferromagnetic films occurs during the annealing process at ≈300 °C. In our recent work, we have experimentally demonstrated an enhancement of the VCMA effect in a Co/MgO junction with Pt and Ir insertions, where the annealing process was not carried out. [34] Another method to optimize the VCMA effect is the utilization of an appropriate underlayer. Since the work function and the electronic states of the underlayer affect the adjacent ferromagnetic layer, the VCMA effect is also expected to exhibit a variation by introducing different underlayer materials. [35] While experiments with different underlayers have already been carried out in CoFeB/MgO junctions, [36] it has been difficult to systematically discuss the effect of the underlayer on the VCMA effect in these systems: Since the thermal annealing is necessary to obtain enough PMA for the evaluation of the VCMA effect, it is unavoidable during annealing that a large diffusion of underlayer heavy metals occurs into the ferromagnetic layer. [36,37] Therefore, the correlation between the effect of underlayer and the additional VCMA effect caused by the insertion of heavy metal at the junction interface is still elusive. In contrast to CoFeB/MgO systems, however, Co/MgO junctions have an advantage to obtain enough PMA without annealing since Co exhibits larger bulk PMA. Thus, by using Co/MgO junctions, it enables us not only to systematically study the correlation between the effect of insertion layers and the underlayers on the VCMA effect, but also to explore the optimum combination of the insertion layers and the underlayers.
In this study, we propose a method to tune the VCMA effect in Co/MgO junctions by optimizing the combination of the insertion materials at the junction interface and the underlayer materials. With focusing on the effect of underlayer materials, we have systematically investigated the tuning of the VCMA effect caused by the insertion of sub-atomic 5d transition metals (Pt, Ir, or Os). By introducing an Os underlayer, interestingly, a large VCMA coefficient ξ, as high as −100 fJ V −1 m −1 , is obtained for a Co/MgO junction with Pt insertion. The change in the VCMA coefficient due to the insertion of Pt layer Δξ is >2 times greater than the case deposited on Pt underlayer. On the other hand, we could not observe the enhancement of the VCMA effect with the insertion of Ir, which is clearly different from the case with Pt underlayer. Furthermore, even positive VCMA effect can be directly observed for the junction with the insertion of Os. These results imply the contribution of the electron depletion/doping from underlayer interfaces and suggest that the selection of the combination of insertion layers and underlayers is quite important to optimize the magnitude of the VCMA coefficient. These findings will provide a crucial piece of information for understanding VCMA physics as well as for further advances in the application studies of spintronic devices using the VCMA effect.

Magnetic Properties in Co/X/MgO Junctions Deposited on Os Underlayers
First, we conducted polar magneto-optical Kerr effect (polar-MOKE) measurements to characterize the magnetic properties of Co/X/MgO junctions deposited on Os underlayers (see Figure 2a and Experimental Section). To investigate the effect of insertion of the heavy metal layer on the magnetic properties of the Co, we measured the heavy metal layer X thickness dependence of the polar-MOKE hysteresis curves. In Figure 2b-d, the Kerr rotation angle for different metals, X, and thicknesses is plotted as a function of the perpendicular magnetic field H perp , where the thickness of the Co layer is fixed at t Co = 1.26 nm. Here, a gradual switching is obtained except for the case with an insertion of the Ir layer. The magnitude of the saturation magnetic field clearly decreases with increasing X thickness, reflecting an enhancement of the PMA at the Co/MgO interface. For the case with the insertion of Ir layer (see Figure 2c), a sharp switching due to the out-of-plane magnetic easy axis is obtained even when t Ir is 0.08 nm, indicating a greater enhancement of the PMA.
To evaluate the PMA, we next performed vibrating sample magnetometry (VSM) measurements. In Figure 3a, the saturation magnetic moment per unit area for Co/MgO junctions is plotted as a function of the nominal Co layer thickness t Co . The result clearly shows that the saturation magnetic moment per unit area increases linearly with increasing t Co . The linear fit to the data shown by the solid line gives the values of the saturation magnetization M S and the thickness of the magnetic dead layer t dead ; the slope and the intercept with the horizontal axis give M S = 1416 kA m −1 and t dead = 0.10 nm, respectively. Here, the magnitude of M S is close to the value obtained for the bulk cobalt [38] and the similar Co/MgO films with Pt underlayer, [34] guaranteeing the quality of the Co films. The obtained t dead is slightly greater than that of the Co films with Pt underlayer, which might be due to differences in the intermixing at the underlayer/Co interfaces and the magnetic proximity effect. Since it is known that almost no change was observed in the saturation magnetic moment per unit area in the multilayer films with ultrathin heavy metal layers X( = Pt, Ir, or Os) inserted, [34] hereafter, the values of M S and t dead for Co/MgO given above are used to determine the values of the PMA and the VCMA coefficient in the multilayer films with heavy metal layers inserted.

Perpendicular Magnetic Anisotropy
Here, the effect of inserting a heavy metal layer on the PMA is discussed. Since the Kerr rotation angle obtained by the polar-MOKE measurements is proportional to the perpendicular magnetization component of the ferromagnetic layer, M perp , the effective PMA energy K PMA of the ferromagnetic layer can be evaluated from the M perp (H perp ) area combined with the www.advmatinterfaces.de saturation magnetization value M S determined by the VSM measurements. K PMA is described as [13] where µ 0 is the magnetic permeability of vacuum. For the evaluation of K PMA , we took average of the data with different magnetic field sweep directions. Figure 3b-d shows the X thickness dependence of K PMA t Co * for multilayer films with insertion of Pt, Ir, and Os layers, where t Co *(= t Co −t dead ) is the effective thickness of the Co layer. The magnitude of the change in K PMA t Co * due to the insertion of the heavy metal layer is in following order: Ir > Os > Pt insertion, which is consistent with the case of Co/MgO junctions with Pt underlayer. [34] For Ir and Os, K PMA t Co * monotonically increases with the layer thickness, whilst Pt showed a slight decrease and a minimum value of K PMA t Co * . This decrease of K PMA t Co * can be due to the change of the interfacial PMA at Co/MgO interface with the insertion of Pt (also see Figure S1e in Section S1, Supporting Information).

Voltage-Controlled Magnetic Anisotropy Effect
Now, let us show the impact of inserting a sub-atomic heavy metal layer on the VCMA effect. For the VCMA measurements, we used multilayer devices (see Experimental Section). A schematic of the sample structure prepared for this study and optical microscope image of the sample are shown in Figure 4a,b. Typical polar-MOKE hysteresis curves of microfabricated multilayer devices without heavy metal layers inserted are shown in Figure 4c and the case with heavy metal layers inserted are shown in Figure 5a-c, where the thickness of Co is fixed at 1.0 nm. To show clear differences in the effect of the insertion layer, the thickness of the inserted Pt, Ir, and Os layers are set to 0.05, 0.05, and 0.09 nm, respectively. Here, gradual switching due to the in-plane magnetic easy axis was observed for all samples. For these multilayer devices, we applied a dc bias voltage. In Figures 4c and 5a-c, the red and blue data were measured under application of V bias = +0.8 V and −0.8 V, respectively. The insets show the enlarged figures of the main curves, and clear shifts in the saturation field are observed in all cases, indicating the occurrence of the VCMA effect. By using Equation (1), K PMA of the ferromagnetic layer can be determined in the same manner as in the case of Figure 3b-d. The magnitudes of K PMA t Co * obtained for each bias voltage condition are summarized in Figures 4d and 5d-f, respectively. Here, linear changes in K PMA t Co * are observed for all samples and the magnitude of the slope corresponds to the VCMA coefficient. The sign of the VCMA effect except for Co/Os/MgO is negative, which is consistent with previous results observed in Co/ MgO; [34,39] electric charge accumulation (depletion) results in a decrease (increase) in the PMA energy. For the Co/Os/MgO, interestingly, a positive VCMA effect is directly observed, which is the first demonstration of the positive VCMA effect in ferromagnetic metal/insulator junctions utilizing the insertion of a heavy metal layer. In Figure 5d-f, the magnitude of the VCMA coefficient increases with the insertion of the Pt layer, while it decreases with the insertion of the Os layer, and surprisingly,   respectively. These changes in the VCMA coefficient due to the insertion of the heavy metal layer Δξ are maintained over the thickness range of the inserted layers used in this study (see Figure 6). In Figure 6, the maximum values of the VCMA coefficient ξ in the Co/Pt/MgO, Co/Ir/MgO, Co/Os/MgO junctions reach −101, −44, and +12 fJ V −1 m −1 , respectively. Here, a VCMA coefficient as high as −100 fJ V −1 m −1 is observed, which is the highest level reported for the polycrystalline Co/MgO junctions with single dielectric material. [22] In the Co/MgO devices without insertion of X, the VCMA coefficient was evaluated as −40 ± 2 fJ V −1 m −1 , which is consistent with the value reported for Co/ MgO junction deposited on Pt underlayer (−41 ± 6 fJ V −1 m −1 ), [34] suggesting that the quality of the Co/MgO interface, which is important for the VCMA effect, is almost the same even though the material of the underlayer is different.
Here, we qualitatively discuss the mechanism of the tuning of the VCMA coefficient ξ by utilizing different X and underlayers. Figure 7 shows a summary of the VCMA coefficients obtained for Co/X/MgO with Pt [34] and Os underlayers, where the thickness of Co is 1.0 nm. Since the magnitude of the interfacial spin-orbit interaction in the metallic system can be tuned by the charge transfer, [40] the VCMA effect driven by interfacial spin-orbit interaction is also expected to be tuned by the electron doping/depletion from underlayers. One of the possible mechanisms is due to the difference of the work function among Pt, Ir, Os and Co. The work functions of Pt, Ir, Os, and Co are ≈5.6, [41] ≈5.3, [42] ≈4.5, [41] and ≈5.0 eV, [42] respectively. Since the work function of Os (Pt, Ir) is lower (higher) than Co, it is expected that Co is electron doped (depleted) at the interface with Os (Pt, Ir). [43] However, a significant difference in the VCMA effect was obtained in Os/Co/X/MgO junctions with Pt and Ir insertions (see Figure 7), even though Pt and Ir only show a small difference in their work functions. Therefore, a difference of work function alone cannot explain the experimental results. Another possibility is due to charge transfer between insertion and underlayers caused by alloying in the vicinity of interfaces. To explain the results in Figure 7, however, it might be necessary to consider a slight mixing of the underlayer materials into the Co layer, where it does not reach Co/MgO interface, since there is a certain distance between the underlayer and the junction interface. In fact, the values of dead layer and bulk magnetic anisotropy depending on the underlayer material may imply a slight mixing of the underlayer elements (discussed in Section 2.1 and Section S1, Supporting Information). In the case of alloying, Pt (Os) is expected to cause the electron doping (depletion) since Pt (Os) has more (fewer) valence electron than Ir and Co, and thus, the number of valence electron of inserted material can be tuned by appropriate selection of underlayer materials. For example, let us focus on the value of the VCMA coefficient obtained for Pt/Co/Ir/MgO in Figure 7. Since Pt tends to donate electrons as mentioned above, the VCMA coefficient of Pt/Co/Ir/MgO is expected to deviate from that of Os/Co/Ir/MgO and approach that of Os/Co/Pt/MgO. In fact, such results are obtained in Figure 7. A similar explanation can be given for Pt/Co/Os/MgO. Although further research is still necessary to understand the mechanism quantitatively, our experimental results suggest Figure 6. VCMA coefficient as a function of X thickness t X . a-c) t X dependence of the VCMA coefficient for Co/X/MgO multilayer films with Pt, Ir, and Os layers inserted. Δξ corresponds to the additional VCMA effect caused by the insertion of heavy metal X layers. www.advmatinterfaces.de that the number of valence electron of underlayer may be crucial to tune the additional VCMA effect due to the heavy metal insertion. Unlike previous studies that focused only on the insertion layer [18,[28][29][30][31] or the underlayer, [36] the above results enable us to tune the additional VCMA effect due to the heavy metal insertion by the appropriate selection of the underlayer materials, and are expected to provide a guideline for designing VCMA devices.
Finally, we discuss the future prospects of tuning the VCMA effect via control of the Co/MgO interface using heavy metals as insertion layers and underlayers. In our series of studies, we utilized single heavy metals, i.e., Pt, Ir, or Os as the insertion layer and Pt or Os as the underlayer. For more precise control of the VCMA effect and to obtain a larger VCMA coefficient, it may be necessary to introduce a wider range of elements as well as their alloys. Another way to increase the VCMA coefficient is to use high-k materials such as HfO 2 . A recent experimental study has demonstrated the correlation between the VCMA coefficient and the dielectric constant of the insulator. [44] In that study, it was also pointed out that the dielectric constant can be systematically tuned by changing the thickness of HfO 2 layer in the MgO/HfO 2 dielectric bilayers, and the obtained VCMA coefficient then reached ≈3 times greater than when using only MgO as insulator. Therefore, by introducing high-k materials for the polycrystalline Co/Pt/MgO junctions used in this study, a greater VCMA coefficient, -300 fJ V −1 m −1 class, may be expected. While we have focused on the VCMA effect at Co/X/MgO interfaces in the present work, studies on spinconversion phenomena such as the Edelstein effect and spinorbit torque driven at interfaces has attracted great deal of attention in the field of spin-orbitronics. [40,[45][46][47][48] By applying the heavy metal engineering technique used in the present study, the voltage control of such spin-orbitronic phenomena and a further improvement of spin manipulation efficiency using the interface may also be possible.

Conclusion
In summary, we have reported a pathway to tune the VCMA effect in Co/MgO junctions by optimizing the underlayer with the insertion of sub-atomic 5d transition metal layers (Pt, Ir, or Os) at the junction interface. By introducing Os underlayer, we obtained the VCMA coefficient up to −100 fJ V −1 m −1 for the Co/MgO junction with insertion of Pt, which is ≈50% greater than the case deposited on Pt underlayer and is the highest level reported for polycrystalline Co/MgO junctions using a single dielectric material so far. In contrast, the additional VCMA effect caused by the insertion of Ir was strongly suppressed, and even positive VCMA effect of +10 fJ V −1 m −1 was observed for the Co/MgO junction with the insertion of Os, which are significantly different from the case deposited on Pt underlayer. These results may be understood as due to the effect of electron depletion/doping from the underlayer interfaces, and thus, we anticipate this concept is applicable to other spintronic materials. These findings provide key information for VCMA physics and can be useful for the further development of the voltage-controlled spintronic device technologies.

Experimental Section
Sample Preparation: The sample system used for the VCMA measurements consists of Ta(5 nm)/Ru(10 nm)/Ta(5 nm)/Pt(2 nm)/ Os(4 nm)/Co(0.7-1.5 nm)/X (X = Pt, Ir, or Os)(0-0.28 nm)/ MgO(3 nm)/ITO(20 nm) multilayer film deposited on thermally oxidized silicon substrates, where the order of the layers is described from bottom to top. The deposition of the multilayer films was performed at room temperature by a combination of molecular beam epitaxy and sputtering without breaking the vacuum. The base pressure of molecular beam epitaxy and sputtering apparatuses were <5 × 10 −8 Pa and 5 × 10 −7 Pa, respectively. Most of the metal layers were deposited by magnetron sputtering, and only the Ir, Os, and MgO layers were formed by electron-beam evaporation. The thicknesses of Co, t Co , and heavy metal X, t X , were varied by using a linear shutter. The layer of heavy metal X was introduced to change the magnitude of the PMA and the VCMA effect. A transparent ITO (indium tin oxide) was used for the top contact, which enables to evaluate the magnetic properties of the ultrathin Co layer using polar-MOKE. To prevent heavy metals from diffusing into the Co layer, no annealing was done for any of the multilayer films. For the device fabrication, conventional optical lithography, ion-milling, and lift-off processes were used. The cross-sectional area of the rectangular VCMA element was 8 × 10 µm 2 .
Measurements of Magnetic Properties: To measure the magnetic properties of the multilayer films, Polar-MOKE and VSM are utilized. For the polar-MOKE measurements, a semiconductor laser with a spot size of 1.5 µm was used, which was much smaller than the cross-sectional area of the VCMA elements. The change in the Kerr rotation angle was measured with applying an external magnetic field perpendicular to the film plane. To measure the voltage controlled magnetic anisotropy, polar-MOKE apparatus equipped with a micro-probe system was used. For the VSM measurements, multilayer films of Ta(5 nm)/Ru(10 nm)/Ta(5 nm)/ Pt(2 nm)/Os(4 nm)/Co(0.94, 1.1, 1.26, 1.42, and 1.58 nm)/MgO(3 nm)/ ITO(20 nm) were used. All the measurements were conducted at room temperature.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.