Quantitative separation of planar Nernst effects in harmonic Hall measurements of spin–orbit torques in Pt/Co/HfO/Ta multilayers Authors: Changjin Yun, Yonghwan Jo, Mingu Kim, Seunghoon Yu, Jinhyeok Park, Seungmo Yang, Sungjoong Joo, Kungwon Rhie, Kyujoon Lee Published: 2025 ABSTRACT AbstractSpin–orbit torques (SOTs) provide a promising pathway for efficient magnetization switching in magnetic random-access memory (MRAM) devices, but their quantitative evaluation is complicated by thermal artifacts. While increasing the resistivity of heavy metals can enhance the spin Hall angle through scattering mechanisms, high-resistivity systems also exhibit sizable thermoelectric signals that obscure reliable SOT extraction. In this study, we investigate SOTs in Pt/Co/HfO/Ta multilayers with in-plane magnetic anisotropy, comparing samples with low and high-resistivity Pt layers prepared under different sputtering pressures. The high-resistivity device exhibits both enhanced damping-like torque and an additional unconventional second harmonic signal. By applying symmetry-based analysis to measurements under reversed magnetic field polarities, we separate and identify this component as originating from the planar Nernst effect (PNE), based on its sin2φ angular dependence, quadratic current scaling, and sign reversal across opposite Hall crosses. With thermal effects removed, we find a ~ 39% increase in the damping-like torque for the high-resistivity sample. Our separation technique provides a reliable route to disentangle SOT and thermal contributions in metallic heterostructures with in-plane anisotropy, aiding the precise evaluation and engineering of SOT-based spintronic devices.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-025-31904-3.Keywords: Spin–orbit torque, Spin Hall effect, Planar Nernst effect, Second-harmonic measurementSubject terms: Engineering, Materials science, Nanoscience and technology, Physics Introduction For magnetic random-access memory (MRAM) devices, efficient control of the magnetization in the free layer is essential for applications1–3. Achieving high-density integration in MRAM requires reliable current-induced magnetization switching, and accordingly two principal electrical writing schemes emerged: spin-transfer torque (STT) and spin–orbit torque (SOT). STT operates in a magnetic tunnel junction (MTJ) where a spin polarized current, generated by the reference layer, switches the free layer through a tunnel barrier4–6. For high-performance operation in STT-based devices, a high current density is needed to generate sufficient spin current to switch the magnetization. However, this inevitably increases stress on the insulating barrier, often leading to its degradation. To overcome these limitations, an alternative SOT has been proposed, in which an in-plane current in a heavy metal (HM) generates a transverse spin current via spin Hall effect in the bulk of the heavy metal and Rashba-Edelstein effect at the interface, enabling magnetization reversal without damaging the tunneling barrier7,8. However, the efficiency of SOT-based switching remains relatively low, and improving it is a major challenge. In order to achieve efficient magnetization reversal, maximizing SOT is a central objective. Reported strategies to increase SOT include: (i) controlling resistivity of HM and alloying HM which can modify electron scattering9–11; (ii) Interfacial engineering such as adding ultrathin “dusting” layers, to improve spin transparency and interfacial spin–orbit coupling12–17; (iii) using materials with strong spin–orbit interaction and intrinsically low symmetry (e.g., WTe2 …), which can produce large SOT magnitudes and even generate out-of-plane spin polarizations that enable efficient magnetization switching18,19. Among these methods, raising the resistivity of the heavy-metal is a practical route to achieving larger SOTs, reflecting both intrinsic scaling of the spin Hall angle and extrinsic contributions from grain-boundary scattering9,20–23. However, in such high-resistivity systems the measured SOT signals are likely to include substantial transverse voltages arising from Joule heating, complicating accurate extraction of the true SOT signals. These considerations motivate a rigorous separation of SOT and thermal contributions, especially in highly resistive material systems. SOTs are typically quantified by two complementary approaches: single-domain torque metrology and domain-wall-based methods. The former probes a uniformly magnetized state and directly extracts effective fields, whereas the latter infers torques from current-driven domain-wall (DW) motion, such as measuring the depinning current or field, analyzing the dependence of DW velocity versus longitudinal field, or current-induced loop shifts, observed via Kerr microscopy or anomalous Hall effect measurement24–27. Within the single-domain category, one common technique is spin-torque ferromagnetic resonance (ST-FMR), which uses microwave currents to drive ferromagnetic resonance and extract torque components from the rectified mixing voltage28,29. Alternatively, second-harmonic Hall measurements apply a low-frequency AC current and derive the damping-like and field-like effective fields from the second-harmonic transverse voltage signals where the thermal effects can deform the measurement results30,31. In in-plane magnetized HM/FM heterostructures, Joule heating can generate transverse voltages: an out-of-plane temperature gradient induces the anomalous Nernst effect (ANE), while an in-plane thermal gradient produces the PNE. These thermal contributions become more pronounced in materials with high-resistivity or in certain device geometries, and their azimuthal angular dependencies can resemble those of genuine SOT signals, complicating the separation and interpretation of measurements. Previous studies have demonstrated methods to eliminate the ANE artifact from harmonic Hall measurement30, and the PNE has been analyzed in single ferromagnetic films using second harmonic Hall measurement32. However, separating the PNE contribution in typical metallic HM/FM multilayers remains challenging, which motivates a more systematic analysis via second-harmonic Hall measurements. To address these challenges, we performed in-plane second harmonic Hall measurements on Pt/Co/HfO/Ta systems and introduced a procedure to separate thermal artifacts from SOT signals. We fabricated two sets of samples with different Pt growth conditions to obtain low-resistivity (LR) and high-resistivity (HR) Pt layers. Using harmonic Hall measurements, we compared the SOT efficiencies in the LR and HR devices. Under applied currents high enough to generate thermal gradients via Joule heating, both LR and HR devices exhibited an unconventional azimuthal angular dependence attributed to a thermal contribution. By applying a symmetry-based analysis, we separated this unconventional signal and precisely determined the pure damping-like and field-like SOT components. The unconventional signals followed a sin2φ azimuthal angular dependence and scaled quadratically with current, supporting their origin in Joule heating. While previous experiments and simulations have investigated thermoelectric effects, a quantitative experimental distinction between SOTs and longitudinal thermoelectric contributions has not been achieved in metallic multilayers. Our work bridges this gap by providing the first experimental framework that distinguishes and quantifies SOT and longitudinal thermal contributions in resistivity-tuned Pt/Co/HfO/Ta heterostructures. This analysis experimentally establishes a broadly applicable and experimentally simple method for accurate SOT quantification under realistic Joule heating conditions. Results and discussion Harmonic Hall geometry and separation of SOT and thermal effects Figure 1a schematically illustrates the harmonic Hall measurement geometry for the Pt/Co/HfO/Ta multilayer, where φ is the azimuthal angle of the magnetization measured from the current (x) axis. Figure 1b shows the azimuthal angle dependence of the first-harmonic Hall resistance ( of the LR device measured under various external magnetic fields (Hext = 100–400 mT). Here we observe no field dependence confirming the magnetization is completely aligned along the in-plane direction. Figure 1c presents the corresponding second-harmonic Hall resistance () measured under an in-plane external magnetic field of 200 mT, for applied current amplitudes ranging from 3 to 8 mA. At low currents (I < 6 mA), the angular dependence of follows the conventional behavior reported in previous SOT studies. However, at high currents (I ≥ 6 mA), an additional, unconventional angular dependence appears in the signal. According to Avci et al.30, the conventional SOT-induced components can be described analytically by, where Beff = Bext + Bani (where is the applied external magnetic field, and is the anisotropy field), is the second-harmonic resistance, RAHE is the anomalous Hall resistance, RPHE is the planar Hall resistance and BOe is Oersted field. BDL and BFL denotes the damping-like (DL) effective field and field-like (FL) effective field, respectively. The term I0α∇Tz represents the thermal contribution from the ANE. The ANE arises from an out-of-plane temperature gradient (∇Tz), as illustrated in Fig. 1a. Accordingly, the measured can be decomposed into damping-like, field-like/Oersted, and ANE components. Notably the thermal ANE contribution is independent of the magnitude of the external magnetic field. The conventional SOT-induced second- harmonic signal exhibits a cosφ angular dependence, meaning that measurements taken under + H and –H (reversed field polarity) are opposite in sign. Therefore, subtracting data taken at + H and –H isolates the conventional SOT contribution (hereafter denoted ) while adding those two data sets cancels out the conventional part and reveals the remaining unconventional component (). (a) Schematic illustration of the Hall bar used for measurements and a cross-sectional side view showing the temperature gradients within the Hall bar. (b) Azimuthal angular dependence of the first-harmonic Hall resistance for the LR device under various external magnetic fields. (c) Azimuthal angular dependence of the second- harmonic Hall resistance () for various applied current amplitudes with an external magnetic field (Hext) of 200 mT applied. To analyze the two contributions, we first isolated the conventional SOT component in the LR device using the method described above. Figure 2a displays a representative raw second-harmonic Hall resistance curve (black squares) measured under an external field of + 200 mT with an AC current of 8 mA. The red line represents the fitted (SOT + ANE), and the blue line represents the remaining (PNE-related). Figure 2b, c show the evolution of isolated and signals as a function of the azimuthal angle for various current amplitudes from 3 to 8 mA, respectively. Both components increase with current, implying that both signals originate from a current induced effect. A detailed analysis is provided below. (a) data measured at Hext = 200 mT and I = 8 mA (black squares). The red and blue lines indicate fitting curves for conventional and unconventional components, respectively. (b, c) Decomposed signals obtained by adding and subtracting the data under positive and negative external magnetic fields, corresponding to (b) and (c). The observed unconventional term is most plausibly due to the PNE caused by Joule heating. As demonstrated by Liu et al.32, an in-plane current produces Joule heating that generates a longitudinal temperature gradient along the x axis (∇Tx) because of the different thermal conductivities of the Au electrodes in oxide/ferromagnet heterostructures. This gradient gives rise to a PNE voltage (Fig. 1c) with a sin2φ angular dependence. The central region of the Hall bar becomes significantly hotter than the areas near the large, thick electrodes, which serve as efficient heat sinks. This nonuniform heating results in an in-plane thermal gradient (∇Tx​) pointing from the hotter center of the Hall bar toward the cooler electrode regions near the Hall crosses being cooled more effectively than the central section. Quantification of DL/FL effective fields Figure 3a, c show the separation of the DL and FL effective fields contributions from the for the LR and HR devices, respectively. In these plots, the second-harmonic data have been analyzed to extract the effective fields: solid lines correspond to the DL component , while dashed lines correspond to the combined FL and Oersted component . Using Eq. (1), the effective field was defined as Beff = Bext + Bani, where the anisotropy fields used in the analysis were Bani = 0.89 T for LR device and 0.88 T for the HR device. The slope of the linear fit for the plotted versus 1/Beff yields the effective DL field, and the slope of the FL component plotted versus 1/Bext yields the FL effective field. The y-intercept of the fit versus 1/Beff corresponds to the term. To verify the linearity of SOT with current, we measured the BDL and BFL+Oe at several currents and plotted them versus current density in Fig. 3b, d. Both BDL (black squares) and BFL+Oe (red circles) increase linearly with current density, confirming that once the thermal contributions (ANE and PNE) are removed from the second-harmonic signal, the remaining SOT-induced effective fields scale linearly with the applied current. and components extracted from data for (a) LR and (c) HR Pt. The linear fit of BDL and BFL indicated by the solid lines and dashed lines, respectively. The linear dependence of BDL (black squares) and BFL+Oe (red circles) with the current density are shown in (b) LR and (d) for HR Pt. Because the Oersted field shares the same symmetry as the FL effective field, we first evaluated its magnitude and subtracted it to obtain the exact BFL. Taking into account the different resistances of the Pt and Co layers, we estimated the fraction of current flowing in the Pt layer for each device using a parallel-resistance model, and calculated the corresponding Oersted field BOe following the method described in ref. [31]. The estimated BOe values were—0.131 mT and—0.109 mT for LR and HR devices, respectively. After subtracting BOe, we extracted BDL and BFL for both devices and plotted them as a function of current density in Fig. 3b, d. The extracted FL and DL effective field values were BFL = 0.197 ± 0.01 mT (LR) and BFL = 0.144 ± 0.03 mT (HR), and BDL = 1.37 ± 0.06 mT (LR) and BDL = 1.90 ± 0.05 mT (HR) per current density of A/. From the obtained BDL values, we calculated the effective spin Hall angle 25, giving values of 0.124 ± 0.005 (LR) and 0.163 ± 0.004 (HR). These values are in close agreement with those reported previously for Pt/Co system9,30. Notably, BDL increases while BFL decreases with increasing Pt resistivity. Both effective fields can be influenced by interfacial factors such as spin–orbit coupling, spin transparency, spin mixing conductance, spin-memory loss, interfacial intermixing/roughness as well as bulk factors such as bulk scattering, spin diffusion, and current distribution in the heavy metal33–37. By tuning the growth condition of Pt, we expect changes in crystallinity and surface morphology, as evidenced by the X-ray diffraction and atomic force microscopy results shown in Fig. S138. Particularly, the 39% enhancement of BDL or spin Hall angle in HR devices can be due to the change in both the intrinsic and extrinsic contributions of the spin Hall angle. In contrast, the 27% decrease in BFL may be attributed to bulk and interfacial modifications induced by growth conditions, with the combined effects of the factors discussed above contributing to this behavior. I2 scaling and sign reversal as evidence of PNE To verify the thermal origin of , we varied the amplitude of applied current. Joule heating generates thermal gradients in both the out-of-plane (z) and in-planes (x) directions, which induces ANE and PNE effects, respectively (see Fig. 1a). The magnitudes of the ANE and PNE contributions are expected to scale as I2. In Fig. 4a, b, the red circles and black squares represent the voltage expression of amplitude of the and , respectively. The PNE signals induced by ∇Tx increase approximately quadratically with current and also ANE signals increase quadratically with current. In both devices, the ANE signal appears at lower current levels than the PNE signals. This is because of the different heat-sinking pathways, in the z direction, the substrate acts as a large heat sink so even modest Joule heating produces a measurable ∇Tz which results in an appearance of ANE. In contrast, along the x direction, heat dissipation occurs mainly through the gold electrodes, which are smaller heat sinks compared with the substrate. As a result, a significant ∇Tx develops only at higher currents, leading to the later onset of the PNE. Furthermore, the onset of a detectable PNE contribution occurs at a much lower current in the HR device (~ 3 mA) than in the LR device (~ 6 mA), reflecting the stronger Joule-heating-induced thermal gradients in the HR system. A non-zero offset observed in the VPNE versus I curve originated from the heat generation and dissipation characteristics of the metallic multilayers. In the Pt/Co system, the high thermal conductivity of Pt enables efficient heat sinking to the substrate, so an observable in-plane temperature gradient across the Co layer appears only when the local Joule heating exceeds a certain threshold heating power. As a result, the VPNE signal does not start exactly from zero current but begins to increase above a certain threshold heating power. The VPNE versus I2 plot, considering only the current range above the onset of the PNE signal, is provided in Fig. S2. This graph clearly demonstrates the quadratic dependence on the PNE voltage on current, supporting its thermal origin. Current dependence of thermal effects for samples of (a) LR and (b) HR Pt. The black squares represent ANE component, while the red circles represent PNE component. (c) SOT + ANE and PNE measured at the left and right Hall crosses. The inset shows a schematic diagram illustrating the current-induced thermal gradient. To conclusively identify the unconventional component as PNE, we compared the second-harmonic signals at the left and right Hall crosses of our devices. Since the PNE arises from asymmetric heat flow to the gold contacts, the lateral temperature gradients at the left and right crosses point in opposite directions. Thus, the PNE contribution should reverse its sign between the two Hall cross locations, while the conventional SOT term should remain unchanged. Consistent with this expectation, Fig. 4c shows that the conventional SOT contribution is identical at both positions, whereas the unconventional term inverts its sign. This sign reversal, reflecting opposite directions of ∇Tx, confirms that the unconventional second-harmonic Hall voltages originate from PNE. Furthermore, to rule out any possible contribution from SOT, we validated our measurement method using a control sample with the Pt/HfO/Co/HfO/Ta structure, in which the insertion of HfO suppresses spin-current transfer from Pt to Co as shown in S3. Conclusions In summary, we have experimentally demonstrated the separation of SOT and longitudinal thermal effects in resistivity-tuned Pt multilayers. The Pt layers grown under different pressures exhibit a difference in resistivity and by measuring the SOTs in these two different systems, we observe an increase in the DL effective fields as the Pt resistivity increases. However, in the HR system we observed an unconventional component from low currents which had to be removed in order to precisely evaluate the SOT effective fields. By performing second-harmonic measurements under opposite magnetic field polarities and subtracting the data, we were able to quantitatively separate the unconventional contributions. The isolated unconventional component was found to scale with the square of the current and to reverse sign when the direction of the thermal gradient was reversed, unambiguously identifying it as originating from the PNE. These results demonstrate that accurate quantification of SOT induced effective fields is possible even in metallic multilayer structures with in-plane magnetization, where high resistivity can produce significant unwanted PNE signals. This approach will be valuable for the evaluation and optimization of SOT-based devices. Methods Thin films with the structure Pt (5 nm)/Co (2 nm)/HfO (5 nm)/Ta (2 nm) were deposited in-situ at room temperature on thermally oxidized Si (100) substrates with a SiO2 300 nm layer. The Pt, Co, and Ta layers were deposited by DC magnetron sputtering, while the HfO layer was deposited by RF magnetron sputtering. The base pressure of the deposition chamber was maintained at 6.0 × 10−8 Torr. An HfO/Ta capping layer was used to prevent oxidation and degradation of the active magnetic layer. The Pt layer was deposited under two different working pressures (5 mTorr and 11 mTorr) to tune its resistivity, while all other layers were deposited under the identical conditions, and all samples exhibited in-plane magnetic anisotropy. Hall-bar devices with a line width of 10 μm were fabricated using UV photolithography and Ar ion milling, as schematically illustrated in Fig. 1a. Following device fabrication, we conducted in-plane harmonic Hall measurements to quantify the damping-like (DL) and field-like (FL) components of the SOT under controlled angular and current conditions. During harmonic measurements, an external magnetic field was applied at a polar angle of θ = 90°, and the azimuthal angle (φ) was swept from 0° to 360°, while maintaining a constant magnetic field magnitude. An AC current with a frequency of 17.3 Hz was applied using a Keithley 6221 current source, and the first- and second-harmonic transverse voltages were simultaneously measured using SR–810 and SR–830 lock-in amplifiers. To confirm the linear dependence of the DL and FL contributions on the applied current, additional measurements were performed with varying current amplitudes. The resistivity of single Pt films deposited at 5 mTorr and 11 mTorr was measured using the standard four-probe method and was determined to be 52.96 ± 0.42 μΩ cm and 156.18 ± 1.85 μΩ cm, respectively. We refer to the 5 mTorr device as the low-resistivity (LR) device and the 11 mTorr film as the high-resistivity (HR) device. The anisotropy field (Bani) and saturation magnetization (Ms) were determined by anomalous Hall effect measurements and vibrating sample magnetometer (VSM), respectively. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Author contributions C. Y. planned the study and designed the experiments. K. L. and K. R. supervised the project. C. Y., Y. J., S. Y., J. P., and M. K. grew the thin films and fabricated the devices. C. Y. and Y. J. performed the harmonic measurements and characterized the magnetic thin films with assistance from S. Y. and S. J. C. Y. and Y. J. carried out the analysis of the electrical data. C. Y., Y. J., K. R., and K. L. wrote the manuscript with input from all co-authors. All authors discussed the results and contributed to the final manuscript. Funding The research was supported by the National Research Foundation of Korea (NRF) funded by Korea Government (Grant No. 2022R1C1C1008604, 2022R1A2C1005960) and the Korea government (MSIT) (No. RS-2024-00466232). Data availability The data supporting the findings in this work are available from the corresponding author with a reasonable request. Declarations Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. These authors contributed equally: Changjin Yun and Yonghwan Jo. Contributor Information Kungwon Rhie, Email: krhie@korea.ac.kr. Kyujoon Lee, Email: lkjoon@korea.ac.kr. References Associated Data Supplementary Materials Supplementary Material 1 Data Availability Statement The data supporting the findings in this work are available from the corresponding author with a reasonable request.