Spin–Orbit Torque and Current‐Driven Switching in Pt100‐yTby/Co/AlOx Trilayers

To decrease the energy consumption for the electrical manipulation of magnetization, the enhancement of the spin Hall effect through alloying is widely investigated, but the use of rare earth elements is rarely mentioned. This work reports the modification of the spin Hall effect on Pt by doping rare earth Tb atoms. The spin–orbit torque (SOT) performance is significantly enhanced in Pt100‐yTby alloyed heavy metal (HM) layer. Compared with the Tb‐free sample, the damping‐like effective field per unit current density increases to 1.9 times in the samples with Tb content between 5% and 10%. The critical current density for magnetization reversal is greatly reduced by 65% in a device with Pt87Tb13 HM layer and the in‐plane assistant field as small as ±20 Oe is sufficient for the deterministic switching in the same device. By magneto‐optical Kerr effect imaging, it is confirmed that the increased in‐plane field can effectively compensate the Dzyaloshinskii–Moriya interaction (DMI), which not only helps to reduce the critical current, but also facilitates the domain wall motion and is beneficial for the switching process. All results show that the Pt‐Tb alloy is a competitive candidate for low‐power spintronic devices.


Introduction
3][4][5] DOI: 10.1002/aelm.[11] Indeed, both the spin Hall effect and the interfacial Rashba-Edelstein effect are spinorbit coupling (SOC) correlated phenomena, which makes the SOT limited to materials with sizable SOC effects like Pt, W, Ta, etc. [12] Many efforts have been paid to improve the generation efficiency of spin current such as the exploration of topological insulators as the spin source, [13][14] modulation of the interface, [15][16] assembling different HMs on both sides of the FM layer, [17][18] alloying of the HM layer [19][20][21] and so on.[24] Ueda et al. [25] reported an enhancement of SOTs originating from the Co/Gd interface in Pt/Co/Gd heterostructure; an enhanced SOT was also reported in Pt/[Co/Ni] 2 /Co/Tb system generated by the Co/Tb interface; [26] L. Liu et al. [27] discussed the thickness dependence of Ho on the interfacial Dzyaloshinskii-Moriya interaction (DMI) in Pt/Co/Ho multilayers; Rare earth-transition metal (RE-TM) alloys [28][29] like TbCo provide an ideal platform to explore SOTs in ferrimagnets.Most previous studies of RE in spintronics focused on the RE/FM interface or RE-TM ferrimagnets, but the modulation of SHE on the heavy metal layer by doping RE is rarely mentioned.Due to the partially filled 4f band, Tb element exhibits large orbital and spin angular momentum, resulting in a large moment per atom and strong spin-orbit coupling.So Tb doping contributes to enhancing the bulk scattering caused by magnetic impurities. [30]Additionally, compared to the dopant of 3d, 4d, or 5d metal, the effect of the 4f electrons on the HM layer has not yet been clearly proposed.It is reported that the contribution of 4f electrons to the SHE is influenced by the proximity degree of the 4f level to the Fermi level  F and the method of alloying makes it possible to artificially modulate the proximity. [22]Therefore, it is In this work, we investigate the SOT performance in Pt-Tb alloy with different compositions.The current-induced effective fields and SOT efficiency have been quantified by the harmonic Hall voltage method.The deterministic current-induced switching is realized in all the samples and a small in-plane field of 20 Oe is sufficient for the deterministic switching.A significant reduction in critical current density is observed in the Pt 87 Tb 13 HM layer.The magnetization switching process is analyzed via the magneto-optical Kerr effect (MOKE) microscopy.Our work witnesses an effective improvement of SOT performance in Pt 100-y Tb y heavy metals.

Results and Discussion
Figure 1a gives the schematic diagrams of the electrical transport measurements and the film structure.Due to the wiring pat-tern, the anomalous Hall resistance (AHR) loops have a clockwise switching polarity as shown in Figure 1b (all following electrical tests follow the same polarity).The fairly square AHR loops indicate a desirable perpendicular magnetic anisotropy (PMA) in all the samples.The robust PMA is further verified by the effective anisotropic field H keff shown in Figure 1d, which is obtained by measuring the dependence of AHR on the in-plane field H x [Figure 1c] and fitting it according to the Stoner-Wohlfarth model: [31][32] The H keff value exhibits fluctuations with the Tb content, yet it remains consistently substantial overall, above 11 000 Oe.This strong PMA can be largely ascribed to the ultrathin Pt insertion of 0.55 nm.The coercivity field H c extracted from the AHR loops shows a similar trend to H keff .Figure 1e gives the resistivity  of Pt 100-y Tb y alloy layer.The  of the pure Pt layer is measured to be 33.9 μΩ cm, close to the previously reported values. [21]Along with the Tb content,  increases monotonically from 33.9 to 128.6 μΩ cm, implying a good dispersion of Tb atoms in the Pt matrix and no phase transition occurs. [33]igure 1f gives the saturation magnetization for different samples and a minimum M s of 877 kA m −1 is obtained at y = 10.The variation of M s may arise from the formation of a dead layer, which is relevant to the oxidation of FM surface during the deposition of AlO x layer or the diffusion of the FM and HM layers after annealing. [34]In consequence, the variation of M s together with the shunting effect of large resistance give rise to the gradually increased R H amplitude shown in Figure 1b, while the similar R H amplitudes in samples with y = 17 and 20 may originate from the interfacial spin-orbit coupling at HM/FM interface. [35]verall, all samples show suitable performance for further SOT measurements.
The harmonic Hall voltage measurement [36] was employed to measure the current-induced damping-like (H DL ) and field-like (H FL ) effective fields.With a sinusoidal current injected into the Hall bar along the x-axis, the first harmonic V 1f (in-phase) and second-harmonic V 2f (out-of-phase) Hall voltages are simultaneously detected as functions of an in-plane field.For the measurement of H DL , the in-plane field is swept along the x-axis (longitudinal) while it turns to the y-axis (transverse) for that of H FL .The values of H DL and H FL are derived from the following equations: [36] H In Equation ( 2), ± corresponds to the situation of M z > 0 or M z < 0, respectively.The notation r represents the ratio of the planar Hall resistance (PHR) to the AHR and its value for different samples varies from 0.07 to 0.114.Figure 2a,b gives the dependence of V 1f and V 2f on the in-plane field H x or H y in the y = 17 sample.As expected, [17,37] V 1f is quadratically related to H x or H y , while V 2f varies linearly with respect to them.The slopes of two V 2f curves for M z > 0 and M z < 0 are the same in the case of H x , whereas the signs of them become opposite when sweeping H y .Figure 2c,d plots the variations of H DL and H FL with the current density J of HM layer in the sample with y = 17.It is clear that both the H DL and H FL vary linearly with J, suggesting that the influence of Joule heating is negligible within this current range. [17,38]The signs of H DL and H FL indicate the directions of the effective fields, and it is found that the direction of H FL is independent of the magnetized state while that of H DL reverses with switching the magnetization.[41] H DL(FL) per unit current density is plotted as a function of Tb content in Figure 2e.For the pure-Pt sample, the value of H DL /J is about 2.85 Oe per 10 10 A m −2 , approximating the results previously reported. [17,42]After doping with Tb atoms, H DL /J rapidly increases to 5.4 Oe per 10 10 A m −2 , which is almost 1.9 times higher than that of the pure Pt, and remains at the same level for y = 5-10.Then a continual decline emerges with further increasing the Tb content.The variation of H DL /J with the Tb content is related to many factors and it can be explained by referring to the Equation ( 5): [19,21] where T int ,  SH ,  and  0 denote the interfacial spin transparency, spin Hall conductivity, resistivity of HM layer, and permeability of the vacuum, respectively.M s and t FM are the saturation magnetization and thickness of Co layer.Since the reduction of T int can be inhibited by inserting a thin Pt layer [21] due to the alloying of the HM layer, it can be assumed that T int is not the major factor causing changes in H DL /J.Due to the band structure origin of Berry curvature, which is crucial for SOT, Tb doping is thought to lead to a decrease in  SH . [43]The influences of  and M s on H DL /J as well as their variations with Tb content are clear from the Equation (5) and Figure 1.So at a low Tb content, the slightly reduced  SH is compensated by the combined effect of the increased  and the reduced M s , resulting in the increase of H DL /J.After y > 10, the decrease of  SH is further exacerbated, which, together with the rise of M s , makes the loss of H DL /J no longer be compensated, leading to the continual decline in H DL /J.In Figure 2e, the small and negative H FL /J is consistent with the previously reported H FL originating from the spin Hall effect (SHE) of Pt. [21,44] However, in the samples with Tb = 17 and 20, the signs of H FL become positive, while that of H DL are not affected.A reasonable explanation is that the origin of H DL and H FL are different in both two samples, with H DL stemming from the SHE of HM layer and H FL stemming from the interfacial Rashba effect.As for whether the Rashba effect occurs at the HM/FM interface or at the Co/AlO x interface, [45][46][47] this still needs to be verified in next step.In connection with the R H amplitude shown in Figure 1b, it seems to be more possible that the Rashba effect occurs at the HM/FM interface.The spin Hall angle (SHA) shown in Figure 2f is derived from the Equation ( 6): [48] For the pure Pt layer,  SH is determined to be ∼0.08,comparable to the previous reports. [17,21,41] SH shows a similar variation trend to H DL /J, but due to the influence of M s ,  SH drops faster after reaching the maximum value of 0.115 at y = 5.Since enhancement of SOT by doping relies on the intrinsic SOC properties of the matrix, [19,21] this result implies that Tb doping is more destructive to the original Pt band structure, compared to the doping elements which reach the optimal doping at a higher composition ratio such as Cu, [49] Au, [19] and Pd. [43]ompared to the SOT efficiency, the current-induced magnetization manipulation depends on more comprehensive factors.The influence of Pt 100-y Tb y composition on current-driven magnetization switching is investigated.Deterministic switching can be realized in all the samples with the assistance of a non-zero H x , as shown in Figure 3a.Under −600 Oe in-plane field, the required critical current density J c (J c is defined as the current density of Pt-Tb layer where R H changes its sign) exhibits a trend of falling first and then increasing with increasing the Tb content.From Figure 3b, a minimum J c about 1.4 × 10 11 A m −2 is acquired in the sample with y = 13, which is considerably reduced by about 65% compared to the Tb-free sample.Such a huge reduction in J c can be attributed to a combination of factors.First, J c is proportional to the saturation magnetization M s of the FM layer, [50][51] so samples with small M s (around y = 10) are conducive to attaining a low J c .The improvement of H DL /J (at y ≤ 10) significantly increases the efficiency of current-induced SOT generation, thereby further reducing J c .For sample with a low coercivity field H c (like Pt 87 Tb 13 ), the barrier to nucleation or domain wall motion (DMW) is correspondingly low.And under the assistance of the enhanced Joule heating effect that stems from the increased resistance of the HM layer, the barrier to switching is further reduced.Ultimately these factors above combine to create the minimum J c in sample with y = 13 rather than in the sample with higher  SH or H DL /J.Further investigation under different H x is introduced in the sample with y = 13.As can be seen from Figure 3c, deterministic switching by SOT is achieved over a wide range of fields, and the minimum H x can be as small as 20 Oe, which is beneficial for practical application.By flipping the H x from −20 to 20 Oe, the switching polarity of R H -J loop is also reversed, indicating that the opposite in-plane field can reverse the chirality of domain walls (DWs) as well as the associated current-driven DW motions, and ultimately results in the reversal of switching polarity.The critical current density J c extracted from the R H -J pulse loops is plotted as a function of H x in Figure 3d.Evidently, the increased in-plane field H x can greatly reduce the critical current density and the lowest J c is about 1 × 10 11 A m −2 .Overall, our results confirm that Pt 87 Tb 13 HM layer exhibits a decent SOT performance such as the low critical current density and the small assistant field, which is helpful in reducing the energy consumption of the device.
To understand the switching behavior at the level of magnetic domains, the polar MOKE imaging is employed.Figures 4 and 5 give the switching loops and the corresponding MOKE images for the switching achieved at H x = −600 and −50 Oe in the sample with y = 13.Prior to every test, the device is positively magnetized by a saturated field.The gray and dark areas in MOKE images denote the upward-and downward-magnetized domains, respectively.Under a field of −600 Oe, both the up-to-down and down-to-up switching processes are completed quickly.The reversed domains, as shown in Figure 4b,c, nucleate at defects or boundaries and then rapidly expand throughout the whole device.The DMI originating from the inversion symmetry breaking can stabilize the chiral Néel-type DW and has important effects on DW motion during the switching. [52]The value of the DMI effective field in this sample is determined to be about 400 Oe through the loop-shift method [53] (see S1, Supporting Information).Apparently, −600 Oe in-plane field is sufficient to overcome the DMI effective field and break the chirality of the DWs, thereby leading to a straightforward domain expansion, as depicted in Figure 4b,c.However, in the case of H x = −50 Oe, the DMI is not fully compensated, thus the switching behavior and DW motion may be significantly different.The nucleation of domains [square in Figure 5b] and the pinning of DWs [circle in Figure 5b,d] show higher sensibility to defects.Especially the rightward DW motion in Figure 5b is more likely to be pinned by defects or it needs to be driven by a higher pulse current during the up-to-down process.Tilted DWs marked by white arrows in Figure 5b,d are clearly visible, which can be taken as evidence of the competition between DMI and the external field. [54]To compensate the DMI effective field, the expansion of domains requires the formation of tilted DWs with large tilt angles. [55]In addition, more annihilations of residual domains can be observed near the end of switching.It can be seen from Figure 5c that after the up-to-down switching is completed, the downward domains (dark) at the right end of Hall bar partially flip back to the upward state (gray) as the current sweeps back from −38 mA.Note that this flip is not reflected in the R H -I pulse loop because of the far distance to the voltage channel.According to Ref. [55], this flip in Figure 5c can be explained in this way: when the dark domains expand to the right edge of Hall bar, the DW tilting almost vanishes.So under the combined actions of the uncompensated DMI effective field and the SOT of large pulse current, the edge DWs [56] are displaced to the left, resulting in the flipped domains.By comparing the switching processes at H x = −600 and −50 Oe, it is concluded that the increase of H x can effectively compensate the DMI effective field, which not only reduces the required current, but also makes the current-induced switching more deterministic and the DW motion more steady.

Conclusion
In summary, the modulation of spin Hall effect in Pt has been investigated by doping Tb atoms in the Ta (1 nm)/Pt 100-y Tb y (5 nm)/Pt (0.55 nm)/Co (0.7 nm)/AlO x (2 nm) multilayers.Through the harmonic Hall voltage measurement, a significantly enhanced SOT performance is observed at low Tb content.When the Tb content is in the range of 5-10%, the damping-like effective field per unit current density increases to 1.9 times in comparison with the Tb-free sample.The spin Hall angle reaches the maximum value of 0.115 at y = 5.For the current-induced magnetization reversal, the Tb doping makes a substantial reduction of 65% in critical current density J c and the minimum J c is acquired in sample with y = 13.In addition, an in-plane field as small as ±20 Oe is sufficient for the deterministic switching in the same sample with Pt 87 Tb 13 HM layer.All of these features are helpful to reduce the energy consumption of the device and are well suitable for industrial applications.Through the MOKE images, it is confirmed that increasing the in-plane assistant field can effectively compensate the DMI, which not only helps to reduce the current, but also facilitates the movement of domain walls and makes the current-induced switching more steady.Our study gives insight into the impact of the doping of RE atoms in heavy metal Pt and offers a new option for low-power spintronics devices.

Experimental Section
All samples with a structure of substrate/Ta (1 nm)/Pt 100-y Tb y (5 nm)/Pt (0.55 nm)/Co (0.7 nm)/AlO x (2 nm) were deposited on thermally oxidized Si substrate by DC/RF magnetron sputtering at room temperature.The base pressure was better than 4 × 10 −7 Torr and the working Ar pressure was kept at 2 mTorr during sputtering.The Pt 100-y Tb y alloy layer was deposited by a co-sputtering system utilizing Pt and Tb targets where the value of y denotes the atomic percentage (%) of Tb and it was changed by controlling the deposition rate of Pt while that of Tb was kept at 0.1717 Å s −1 .The deposition rates of Ta, Co, and AlO x were 0.3465, 0.223, and 0.2141 Å s −1 , respectively.After deposition, the entirety of the films were annealed under a high vacuum at 320 °C for 10 min to obtain a good perpendicular anisotropy.All devices used for electrical transport measurements were fabricated into Hall bars with current channels 20 μm wide and 200 μm long, by standard ultraviolet lithography and lift-off methods.Then Ti (10 nm)/Au (80 nm) electrodes were prepared in the same way to make better electrical contact.
All the transport measurements were carried out on a homemade platform where Keithley 6221 was used as the DC or AC current source and voltage signal was detected by Keithley 2812A nanovolt meter or Stanford SR 830 lock-in amplifiers.Spin-orbit toques were quantified by the harmonic Hall voltage method with a low-frequency sinusoidal current of 133.33 Hz passed into the Hall bars.Current-induced magnetization switching loops were obtained by sweeping the pulse current with a pulse width of 1 ms.At the same time, a MOKE microscope was used to characterize the magnetic domain distribution over this process.The resistance was measured by the 4-point method and the resistivity  of Pt 100-y Tb y alloy layer was calculated according to the shunt model of the parallel circuit.The magnetic properties were measured through a vibrating sample magnetometer (VSM).

Figure 1 .
Figure 1.a) Schematic diagrams of the electrical transport measurements and the structure of stacks.b) The anomalous Hall resistance (R H ) loops as a function of perpendicular magnetic field H z for all devices.c) Normalized R H versus in-plane field H x for both the "up" and "down" magnetized states in the sample with y = 17.d-f) The variation of (d) the effective anisotropic field H keff and the coercivity field H c , e) the resistivity of Pt 100-y Tb y layer, and f) the saturation magnetization with the Tb content of samples, respectively.

Figure 2 .
Figure 2. a,b) The harmonic Hall voltages V 1f(2f) versus (a) longitudinal (H x ) and (b) transverse (H y ) in-plane field for the sample with y = 17 under a sinusoidal current with an amplitude of 5 mA.The insets show the variation of V 2f which lags behind the current by 90°.c,d) The dependence of (c) damping-like (H DL ) and (d) field-like (H FL ) effective field on the current density J flowing through the HM layer in the sample with y = 17.e,f) The variations of (e) H DL(FL) /J and (f) the spin Hall angle along with the Tb content.

Figure 3 .
Figure 3. a) Current-induced switching loops at H x = -600 Oe and b) the corresponding critical current density J c for different samples.c) Current-induced switching loops under various H x in the y = 13 sample.d) The dependence of J c on the in-plane field H x in the y = 13 sample.

Figure 4 .
Figure 4. a) Current-induced magnetization switching loop at −600 Oe in-plane field in sample with y = 13.b,c) The corresponding MOKE images for (b) "up to down" and (c) "down to up" switching processes.

Figure 5 .
Figure 5. a) Current-induced magnetization switching loop at −50 Oe in-plane field in the sample with y = 13.b,d) The corresponding MOKE images for the (b) "up to down" and (d) "down to up" switching processes.c) The flipping process of domains at the right edge of Hall bar after the "up to down" switching is completed with current reverting from −38 mA.