Electric-field control of magnetic moment in Pd

Several magnetic properties have recently become tunable with an applied electric field. Particularly, electrically controlled magnetic phase transitions and/or magnetic moments have attracted attention because they are the most fundamental parameters in ferromagnetic materials. In this study, we showed that an electric field can be used to control the magnetic moment in films made of Pd, usually a non-magnetic element. Pd ultra-thin films were deposited on ferromagnetic Pt/Co layers. In the Pd layer, a ferromagnetically ordered magnetic moment was induced by the ferromagnetic proximity effect. By applying an electric field to the ferromagnetic surface of this Pd layer, a clear change was observed in the magnetic moment, which was measured directly using a superconducting quantum interference device magnetometer. The results indicate that magnetic moments extrinsically induced in non-magnetic elements by the proximity effect, as well as an intrinsically induced magnetic moments in ferromagnetic elements, as reported previously, are electrically tunable. The results of this study suggest a new avenue for answering the fundamental question of “can an electric field make naturally non-magnetic materials ferromagnetic?”

Scientific RepoRts | 5:14303 | DOi: 10.1038/srep14303 confirmed to have an fcc (111) texture using X-ray diffraction measurement. All of the Pt/Co/Pd and Pt/Co samples we used in this study were confirmed to have perpendicular magnetic anisotropy. Figure 1 shows the t Co dependence of the saturation magnetic moment per unit area (m s /S) for the both samples. As shown in the figure, the m s /S values of the Pt/Co/Pd samples were greater than those of the Pt/Co samples for all t Co values, indicating that a magnetic moment was induced in the Pd layer. Assuming that the magnetic moment is uniformly induced in the entire Pd layer, the induced magnetic moment per Pd atom is calculated to be ~0.1μ B (μ B is the Bohr magneton), the order of which is in good agreement with previous studies 32, 34,35 . The Pd layer thickness dependence indicates that the magnetic moment induced in the Pd layer saturates at t Pd ~ 2 nm (not shown), suggesting that ~2 nm is the distance limit of the proximity effect from the Co/Pd interface 35 . Thus, in the present sample (t Pd = 1.7 nm < 2 nm), a finite magnetic moment is expected to be induced in the uppermost Pd atomic layer, but the magnetic moment per Pd atom there is considered to be less than 0.1μ B because the magnetic moment induced by the ferromagnetic proximity effect is known to decrease with increasing distance from the interface 34,36 . Fabrication of the devices and experimental setup for magnetic moment measurement under applied electric field. An electric-double-layer (EDL) capacitor structure was used to modulate the electron density in the Pd surface 3,4,11,15,16,18,26 . The structure consisted of an Au gate electrode, a polymer film containing an ionic liquid (ionic liquid film) 18 , and a Pt/Co/Pd sample (Fig. 2). Two Pt/Co/Pd samples with t Co values of 0.10 and 0.19 nm (samples A and B, respectively) were used in the experiment. The ionic liquid we used was composed of an N,N,N-trimethyl-N-propylammonium (TMPA + ) cation and a bis(trifluoromethylsulfonyl)imide (TFSI -) anion. The EDL capacitor was fabricated by simply placing the ionic liquid film with the evaporated Au gate electrode (50 nm) on the Pt/Co/Pd sample. The area covered by the ionic liquid film (S ion-film ) was slightly less than the total area of the Pt/Co/Pd sample (S total ). Au wires were connected to the Au gate electrode and the Pt/Co/Pd metallic layers to apply a gate voltage V G between them. A positive V G corresponded to the direction of increase in the electron density at the Pd surface. The device was introduced into a superconducting quantum interference device magnetometer to measure the magnetic moment directly under the application of V G 6,18 .
Magnetic moment under electric field. Figure 3(a) shows the temperature T dependence of the perpendicular component of the magnetic moment m ┴ divided by S total for both samples A and B under V G = + 2.0 and − 2.0 V. After V G was changed at 300 K under a perpendicular magnetic field μ 0 H ┴ of ~20 mT, T was decreased to 10 K. Subsequently, μ 0 H ┴ was reduced to nearly zero (1.5 ± 0.1 mT), and the m ┴ shown in Fig. 3(a) was measured by increasing T. Below the Curie temperature T C , a clear difference in m ┴ /S total was observed with the change in V G for both samples: a positive (negative) V G , i.e., larger (smaller) electron density at the Pd surface, resulted in a larger (smaller) m ┴ /S total . The difference in m ┴ between V G = + 2.0 and -2.0 V (Δ m ┴ ( ± 2 V)) was found to increase linearly with decreasing temperature, as shown in Fig. 3(b), in which Δ m ┴ (± 2 V) was divided by S ion-film because the magnetic moment should be modulated in the area covered by the ionic liquid film. It should be noted that m ┴ was nearly equal to m s because the squareness ratio m ┴ /m s of the hysteresis loops was ~1 under V G = ± 2.0 V at a temperature of at least 10 K, as indicated in the inset of Fig. 3(a). The anomalous Hall effect was used to detect the hysteresis loop under the application of V G . Although the change in T C up to 100 K was reported for a Pt/Co sample with a similar device structure 18 , T C was not clearly dependent on V G in the present Pt/Co/Pd samples. Δ m ┴ ( ± 2 V) deviates significantly from the linear fitting immediately before it intercepts the horizontal axis, as indicated by the downward arrows in Fig. 3(b), which might imply the occurrence of a small change in T C .

Discussion
We first analysed the temperature dependence of m ┴ for the Pt/Co/Pd and Pt/Co samples. The reduction in the magnetic moment m near T C can be quantified using a critical exponent β as m ~ (1 − T/T C ) β . Figure 4(a) shows a double-logarithmic plot of the normalised m ┴ as a function of 1 − T/T C for several samples with T C in the range of 157-341 K, including sample B, under an applied electric field (see Table 1 for details of the samples). The value of β was determined from the slope of the linear fitting to the data 37,38 in the range of T L /T C (= 0.75) to T H /T C (= 0.87) (see Methods for details of the analysis). The value of m ┴ in the figure was normalised by its value at T H /T. The value of β was ~0.2 in the Pt/Co samples and ranged from 0.22 to 0.30 in the Pt/Co/Pd samples. From the magnified figure shown in Fig. 4(b), one can clearly see that as 1 − T/T C increased (in other words, as the temperature decreased), the difference in the normalised m ┴ values between the Pt/Co/Pd and Pt/Co samples increased, i.e., the magnetic moment in the Pt/Co/Pd samples increased more rapidly at lower temperatures. This behaviour occurred because the β value of the Pt/Co/Pd samples was higher than that of the Pt/Co samples and/or because the magnetic moment in the ferromagnetic Pd layer is greater at lower temperatures 35,37 . The latter factor is most likely dominant here because the linearity of the normalised m ┴ degrades as 1 − T/T C increases in the double-logarithmic scale. Other important points to consider in understanding the electric-field effect observed in this study are as follows. (i) A larger change in the magnetic moment was observed with the application of V G in the Pt/Co/Pd samples at lower temperatures, as shown in Figs 3(b) and 4(b). (ii) The data points for the Pt/Co/Pd samples without an applied electric field (indicated by the green symbols in Fig. 4) are almost entirely located between the ones obtained under positive and negative V G (indicated by the red and blue symbols) at any temperature T < T L . (iii) As shown in Fig. 3(b), the difference between m ┴ under positive and negative V G (Δ m ┴ (± 2 V)) increased linearly with decreasing temperature. This can probably be attributed to the linear temperature dependence of the magnetic susceptibility of Pd deposited on the layer consisting of the ferromagnetic element 35,37 . These experimental results, which show good reproducibility in similar structures (see Methods), demonstrate that the induced magnetic moment at the surface of the Pd layer was increased and decreased by the application of positive and negative V G , respectively.
We next analysed the change in m ┴ (Δ m ┴ ) for samples A and B upon application of an electric field. Figures 5(a,b) show Δ m ┴ /S ion-film at 10 K for both samples as a function of V G . Each data point was obtained from the temperature dependence of m ┴ at 10 K under each V G . V G was applied in the order indicated by the arrows in the figures. Although hysteresis behaviour was observed with round-trip V G application, the tendency of Δ m ┴ to increase (decrease) with positive (negative) V G application was reproduced in both samples. The change in the magnetic moment per Pd atom was determined from the linear fitting, as shown in the figures, based on the following assumptions. Pd has an fcc (111) structure, and only the magnetic moment in the uppermost atomic layer was changed because only the electron density there was changed attributed to Thomas-Fermi screening (see Methods for details of the capacitance measurement). The calculated value of the magnetic moment for sample A (B) was 0.050 ± 0.023μ B (0.078 ± 0.013μ B ) per V G of 1.0 V. The change in the electron number ΔN for sample A (B) per Pd atom and per V G of 1.0 V was calculated to be 0.049 (0.034). ΔN was determined from the capacitance C of the devices divided by S ion-film .
According to the ab initio calculation for Pd, the density-of-states peak appears at an energy level slightly lower than the Fermi level 29 . Thus, a decrease in the electron density should theoretically lead to an increase in the magnetic moment 5,12 . Our results, however, were the opposite. Support from theoretical calculations is clearly needed to understand our results. The state of the Pd in our sample structure was different from that in the bulk case in the following respects: (1) the magnetic moment was already induced in the Pd layer (thus, spin splitting was already induced); and (2) the Pd layer was very thin, and interfaces were formed between the Co and MgO layers. In addition, we note that in a Pt/Co system, in which the electric field induced changes in T C and a magnetic moment was reported 17,18 , ab initio calculations indicated that the number of 3d electrons can be decreased even when an electric field is applied in the direction of increase in the total electron number because the number of sp electrons increases 28 . This explains the discrepancy between the experimental results and the outcome suggested by the Slater-Pauling curve for the Co case. This scenario may be applicable even to the present 4d electron system (Pd). Furthermore, reversible chemical effects, such as the migration of oxygen atoms from the MgO 27 cap or of hydrogen absorption in the Pd layer 39 by V G applications might be related to a change in the electron state and thus to a change in the magnetic moment induced in Pd. From an experimental perspective, the next challenge is to induce ferromagnetism electrically in naturally non-magnetic materials beyond that attributable to the ferromagnetic proximity effect.

Methods
Determination of T C . Figures 6(a,b) show the Arrott plots 40 for two reference samples, a Pt/Co/Pd sample with t Co = 0.20 nm and t Pd = 1.7 nm and a Pt/Co sample with t Co = 0.32 nm, respectively. From the The magnetic moment in the Pt/Co/Pd samples increased more rapidly than that in the Pt/Co samples as 1 − T/T C increased (in other words, as the temperature was reduced). The data points for the Pt/Co/Pd samples to which an electric field were not applied (green symbols) were confirmed to lie almost entirely between the ones obtained under the application of positive and negative V G to sample B (red and blue symbols) at a temperature T < T L . Fig. 4. The thicknesses of the Pd, Co, and Pt layers (t Pd , t Co , and t Pt ); T C ; and the critical exponent β determined from the fitting (see Methods for details of the determination) are summarised.  plots, the T C of the Pt/Co/Pd (the Pt/Co) reference sample was determined to be 181 (329) K. The T/T C dependence of m ┴ at μ 0 H ┴ = 0.4 (± 0.1) mT for both reference samples is shown in Fig. 6(c), in which the vertical axis is normalised by an m ┴ value at T/T C = 0.87. In both samples, m ┴ decreased rapidly below T C . This was most likely because of the formation of a multi-domain state, as observed in similar Pt/Co samples 17,41 . The important point is that, as long as μ 0 H ┴ was the same (see Figs 6(c,d)), the two curves for Pt/Co/Pd and Pt/Co reference samples overlap very well in this temperature region, even though these samples have different sample structures and T C values. This suggests that as long as T C is known for one reference sample, one can determine the T C of another sample by comparing the m ┴ -T curves. Figure 7 presents a summary of the comparison of the results for the samples listed in Table 1 (coloured line) and the Pt/Co/Pd reference sample (black line). Adjusting the T C values of the samples resulted in the normalised curves overlapping well with the reference curve. The T C values of the samples listed in Table 1 were determined in this way. The difference in T C for these samples determined using the Pt/Co/ Pd and Pt/Co reference samples was at most 1-2%. The Arrott-Noakes (AN) plot 42 may provide more accurate T C for the present two-dimensional system. However, the difference in T C determined from the Arrott and AN plots is only 1-2% 17 . In comparing many samples without difficulty and confirming the most important finding of this study, i.e., the induced magnetic moment in the Pd layer being increased or decreased by applying an electric field, we believe that the Arrott plot and the above-mentioned way of determining T C were useful method. We note that using a similar Pt/Co/Pd sample with a T C of ~190 K (not shown in Table 1) confirms the reproducibility of the results, i.e., the data points for the sample without an applied electric field are almost entirely located between the points obtained for similar samples under positive and negative V G at any temperature T < T L .

Table 1. Properties of the Pt/Co/Pd and Pt/Co samples used in
Determination of the critical exponent β. The value of β was determined from the slope of the linear line fitted to the normalised m ┴ data in the range between T L /T C (= 0.75) and T H /T C (= 0.87) on a double-logarithmic scale (see Fig. 4a). The fitting range was determined from the linearity of the data. It was difficult to perform a reliable fitting to the data for sample A because there were not enough data points in the fitting range.
Capacitance measurement. The capacitance C of samples A and B was measured using a capacitance meter and applying an ac voltage with an amplitude of 0.1 V and frequency f of 100 Hz. In addition, the f dependence of C for a similar device showed that C increases slightly with decreasing f. Although further experiments are needed to determine an accurate value for Δ N under a dc gate voltage, the value of C estimated using the above experimental results for f = 0.01 Hz, which was the lowest value of f in the  Table 1. The black line indicates the results obtained for the Pt/Co/Pd reference sample. The results are indicated by coloured lines. All m ┴ was measured at 1.5 (± 0.1) mT and the vertical axes were normalised by the m ┴ value at T/T C = 0.87.