Interface effects of polycrystalline Fe₂O₃ thin films on Pt Open Access

Antiferromagnetic (AFM) spintronics have attracted much attention in recent years, particularly in the context of solid-state logic devices, which opens possibilities of new architectures with faster dynamics and more stable states, compared to ferromagnets (FMs) based devices.^1–3 The main challenges lie in manipulation of spin states in AFMs using electronic means, and in electronically reading the states of such devices. Some AFMs are conductive, allowing easy interfacing with electronics. These are limited, while there are many insulating AFMs of interest.^4,5

To study insulating ferromagnets (FMs), bilayers composed of an insulating material and a heavy metal (HM) with large spin orbit coupling (SOC) are commonly used. This was demonstrated for insulating ferrimagnetic materials, such as Y₃Fe₅O₁₂ (YIG)⁶ and Tm₃Fe₅O₁₂ (TmIG),^7,8 as well as more complex textures like spiral magnets; Cu₂OSeO₃^9,10 and CoCr₂O₄.¹¹ By utilizing the spin Hall effect (SHE), it is possible to gain insight on the magnetic ordering at or near the interface of the insulating layer.^6,8,12–14 Currents in the HM generate a spin current via the SHE that interacts with the FM layer. The scattered spin current generates a backflow, resulting in an increase of resistance through inverse-SHE (iSHE). This resistance is called spin Hall magnetoresistance (SMR). Studies on bilayers with Pt as the HM layer and an AFM layer, such as NiO,^7,8,15,16 Cr₂O₃,¹⁷ or α-Fe₂O₃,^18–20 showed that the effect is also applicable to AFM based bilayers, despite the negligible magnetization. This is owing to the second order nature of the effect—in spite of domains and sublattices magnetization summing up to 0—the spin conduction is achieved via spin transfer torque when the spins are non-collinear with the AFM.^21,22 Hence, the Néel vector and its relation to an external magnetic field is key to understanding the resistance changes in AFM based bilayers. This will make the MR of insulating AFM different from insulating FMs when measuring angle dependent magnetoresistance (ADMR) or transverse magnetoresistance (t-MR), e.g., in ADMR measurements, there could be a phase difference between the signal from an AFM and a FM.^21,23 Specifically in α-Fe₂O₃, the AFM we focus on, different electronic behaviors may exist between samples grown in different directions. For example, (0 0 0 1) and (1 0 1 2) oriented samples showed different anisotropies in ADMR related to the symmetries of the sample plane.^19,20 This raises the question whether bilayers including polycrystalline AFMs would show an averaged effect of all facets, or if there is a unique feature of polycrystalline AFMs that may be interesting and useful. It is important to explore polycrystalline materials from a pragmatic point, as they are easier to fabricate and do not require stringent deposition conditions as those needed for crystalline film growth.

In this paper, we investigate the bilayer devices of Pt and polycrystalline Fe₂O₃. We observe FM-like behavior in ADMR measurements, opposed to AFM-like behavior reported for single-crystal thin films. The out-of-plane transverse-MR (t-MR) measurements show a temperature dependent change of the sign, possibly indicating that the interface has a nontrivial complex mechanism at play, or simply due to exhibiting temperature dependent coupling, which could be related to Dzyaloshinskii–Moriya interactions in Fe₂O₃.^24,25 Additionally, we observe a non-trivial glass-like relaxation of the t-MR below 80 K. This effect may arise from some sort of proximity effect in the Pt layer coupled to the AFM, or possibly due to Pt and Fe₂O₃ creating a complex alloy at the interface which gives rise to this effect.

Physical vapor deposition of the samples was done on thermally oxidized Si $⟨ 1 0 0 ⟩$ substrates via magnetron sputtering at 21 °C, see supplementary material I for additional details. We focus on three sample compositions:

1. Pt(5 nm)/Fe₂O₃ (10 nm) referred to as Fox-10;

2. Pt(5 nm)/Fe₂O₃ (30 nm), Fox-30; and

3. Pt(5 nm)/Cu(2 nm)/Fe₂O₃ (10 nm), CuFox-10.

Patterned Hall bar devices were prepared via positive photolithography for purposes of electronic characterization. Additionally, we fabricated single layer Fe₂O₃ films on the same substrates and tested them for conductivity. The resistance of these Fe₂O₃ films was too high to measure, above the 200 giga-ohm input impedance of our equipment. Characterization using x-ray diffraction (XRD) and reflectivity (XRR), and magnetometry were performed for thickness, crystallinity, and bulk magnetic behavior. These measurements were performed on non-patterned samples deposited on Si $⟨ 1 11 ⟩$ substrates (with no oxide, since it complicates the XRD analysis). The samples were deposited simultaneously with the other Hall bar device samples, to have identical deposition conditions.

In Fig. 1, we present results of the magnetic and structural characterizations of the Fox-30 sample. Figure 1(a) shows temperature dependent magnetometry, which does not change much with temperature. There is no sign of the Morin transition, characteristic of hematite crystals, possibly due to the small grain and thin film sizes, as previously reported.^26,27 The value of volumetric magnetization at saturation is comparable with other studies of Fe₂O₃ thin films.²⁷ The inset shows a small widening of the hysteresis curve when the sample is cooled to 10 K. Figure 1(b) shows the XRR measurement as well as a fitting of the measurement using the software REFLEX.²⁸ The resulting film's thickness and roughness are presented in Table I, showing around 4.8 nm of Pt and 35 nm of Fe₂O₃, close to the planned thickness.

Figures 1(c) and 1(d) show XRD scans at a grazing angle and θ–2θ, respectively. The disappearance of the Si (1 1 1) peak at 28.5° in the grazing scan is expected. The α-Fe₂O₃ peaks (1 0 4) at 33.1° and (1 1 6) at 54° are present in both scans. This is indicative of the polycrystalline nature of the α-phase of iron oxide. The additional strong peak corresponds to the Pt (1 1 1) peak at 39.7°, where we see some preferential growth in this direction for platinum. We have also employed Raman spectroscopy to validate the stochiometric phase, presented in Fig. 2 (here, the resistance bridge samples were directly measured). Peaks near 229, 294, and 410 cm⁻¹ were found, typical of α-Fe₂O₃.²⁹ The largest peak is the SiO₂ substrate peak at 520 cm⁻¹. For measurements of Fox-10 and for resistivity vs temperature measurements of both films, please refer to supplementary material III and IV, accordingly.

Below, we first present all the magnetoresistance measurements, and then in the discussion, we consider their physical interpretation.

Magneto-transport properties of both Fox-10 and Fox-30 were measured and found to behave similarly. We present the results from Fox-10 measurements compared to CuFox-10 (i.e., to a sample where a thin 2 nm Cu layer separates between the Pt and Fe₂O₃ layers), see supplementary material II for measurement details and VIII for a comparison between the 10 and 30 nm films. Figures 3(a)–3(c) show the field dependent magnetoresistance (FDMR) ratio of the Fox-10 device as a function of the applied field for the n-axis—normal to the device plane, the j-axis—parallel to current flow, and the t-axis—perpendicular to current and in the device plane, see schematic representation in Fig. 3(a). The measurement is presented for three different temperatures in Figs. 3(a)–3(c) for 300, 150, and 10 K. In supplementary material V, we show the comparison of FDMR as a function of temperature for each field direction. At 300 K, there is a near-linear dependence on the absolute field magnitude that tapers off at high fields. There is nearly no difference between n-axis and j-axis dependence, while in the t-axis, there is a negative dependence. These results are consistent with SMR. Below 150 K, a near parabolic behavior develops at low fields, while in high fields, the dependence in n- and j-axes start to acquire a negative slope and deviate from one another. At 10 K, the trend of this deviation and the reversal of slope and sign continue. A small hysteresis can be observed in the t-axis dependence.

SMR arises from spin currents in the heavy metal (from the SHE) interacting with and absorbed by the insulating magnetic material. As a result, it is sensitive to the interface and does not differentiate between interfacial exchange interactions, such as the magnetic proximity effect in Pt, and the magnetic ordering of the magnetic material near the interface. We fabricated and measured a device with 2 nm Cu between Pt and Fe₂O₃ (CuFox-10) to minimize the contribution of the proximity effect. In essence, the proximity effect is the magnetization of the Pt layer close to the magnetic material, which will manifest as an anisotropic magnetoresistance (AMR) signal.^8,30,31

In this sample, as shown in Figs. 3(d) and 3(f), the dependence is near parabolic in low fields for all axes in 300 and 150 K. Additionally, we see no deviation between n- and j-axes dependence, and the scan in those two axes behave similar to a Pt film without a magnetic layer adjacent to it. At 2 K, we observe a dampening of n- and j-axes dependence, and a complete change of the t-axis field dependence to negative. The MR signal is smaller compared to the film without Cu, which is attributed to the spin-memory loss in the Cu/Pt interface.³² Nevertheless, the measurement does show attributes similar to proximitized Pt [compare to Fig. 3(c)], but only at these very low temperatures.

We measured the ADMR of both Fox-10 and CuFox-10 at 8 T, see supplementary material VII for additional fields. Figures 4(a)–4(c) show the ADMR of Fox-10 at a magnetic field of 8 T, for angle sweeps α, β, and γ at 300, 220, and 10 K, respectively. The angles are marked in the figures, where we use the standard definitions commonly used in ADMR reports.^21,33,34 In α and β sweeps, we observe that the ADMR is consistently following a high resistivity at 0° and 180°, and a low resistivity at 90° and 270°. This is typical for ferromagnetic materials, while some antiferromagnetic materials were shown to follow an opposite trend, specifically, single crystals of α-Fe₂O₃.^21,35 We also measured the ADMR in γ, which has the same amplitude as the difference of α and β (this is expected, and is indicative of the reliability of the measurement). There is a small signal that increases at low temperatures. The temperature dependent behavior of the ADMR magnitude is illustrated in Fig. 4(g), where Fox-10’s ADMR amplitudes (circles, extracted from fitting to a sinusoid) as a function of temperature are plotted as blue—α, red—β, and green—γ circles. For Fox-10 in α and β rotations, we see a downward trend in the ADMR amplitude as the temperature decreases, reaching a minimum around 150–200 K, and then, the ADMR amplitude increases with cooling below 150 K. For γ, the amplitude simply grows with cooling. In the CuFox-10 film, the SMR component of the ADMR behaves considerably different, as shown in Figs. 4(d)–4(f). Now, the magnitude of the β rotation is always increasing as the sample is cooled, in addition to having a lower amplitude. The lower amplitude can be partially attributed to spin-memory losses in the Cu–Pt interfaces.³² We also see a complete suppression of the ADMR in γ rotations down to 80 K, while at 2 K, a small signal is observed, with a sign opposite to that of Fox-10. Fitted amplitudes of the ADMR measurement of the CuFox-10 device are also presented in Fig. 4(g); orange triangles for β scans and dark green triangles for γ scans. Additional measurements which were used to generate Fig. 4(g) are presented in supplementary material VII.

We set the magnetic field to be parallel to the n-axis, i.e., out of the plane of the sample, and scan the field from positive to negative 8 T and back, while measuring the out of plane t-MR. This is, in essence, a Hall effect measurement. We subtracted the regular Hall effect contribution of Pt (or Pt/Cu), measured separately (see supplementary material II for details). The results of the transverse resistance under this setup are presented in Fig. 5 for both Fox-10 (a) and CuFox-10 (b). In Fox-10, the t-MR takes on a negative coefficient at 300 K. Interestingly, the effects at high field do not line up with the magnetometry characterization. As the temperature decreases so does the signal, until 220 K where it is nearly suppressed. There, at low fields, we observe a distinct signal lining up well with the magnetometry discussed previously, shown in the inset. Further cooling results in a reversal of the high field response. That is, the coefficient is now positive, continuing to grow as the device is cooled. Finally, at 10 K, we observed an opening of a hysteresis. A similar reversal of the t-MR signal was observed previously in the literature on a different system.^8,36

CuFox-10 shows a suppression of the t-MR signal, and the coefficient is positive and slightly increasing when the temperature decreases. Somewhat surprising, a reversal of the coefficient does appear at the lowest temperature measured of 10 K. Additionally, the low field effect seen in Fox-10 does not show up in this sample (around 220 K) nor does the opening of a hysteresis at low temperatures. This indicates that the effects observed are not only of possible uncompensated spins at the interface but are also due to interactions between the Pt and AFM layers, which we elaborate in the discussion.

We note in passing that at high fields there is still a downward slope in the Fox-10 film, which changes with temperature, while in the CuFox-10 film, it is flat. This can also be related to interface interactions between Pt and Fe₂O₃. Larger fields are required to study this effect further, which is beyond the scope of this paper.

We observed that the hysteresis which develops in the t-MR at low temperature changes with time. To investigate the nature of this hysteresis, we performed time dependent measurements. Since there is no hysteresis at 100 K, the first step of each measurement was to heat the device to 100 K to reset any magnetic memory of previous measurements. The device was then cooled at zero field to a set temperature and allowed to thermally stabilize. Set temperatures between 70 and 2 K in 10 K increments were used. We note that during this stage, the t-MR remained constant once the set temperature was reached, with no time dependence. We then ramp the magnetic field to 8 T at 0.96 T/min and ramp it down to zero. The device is then let to stay at the set temperature and 0 field for 30 min, during which we measure the change in the t-MR measurement.

We will call the point before the magnetic field is ramped as the “zero-magnetization” resistance, $ρ 0 t − M R$⁠. We measure the resistance change as a function of time, $ρ ( t ) t − M R$⁠, and compare it to the zero-magnetization resistance, $Δρ(t)= ρ 0 t − M R− ρ ( t ) t − M R$⁠. More detailed description of the measurement is presented in supplementary material IX.

Here, in the case of stretch exponents, $T m$ was found to be ∼150 s for all stretch parameters and temperatures. For logarithmic fits, $T m$ was found to be 2 s at 70 K, increasing with lower temperatures and reaching 14 s at 2 K. Although our system is not expected to behave as a logarithm due to the fact that at thermal equilibrium the magnetization should relax to zero, the logarithm might indicate that the effect is of much larger time scale than we can observe in the lab. We demonstrate the fits to different models for three temperatures in Figs. 7(a)–7(c) (more temperatures appear in supplementary material X). The parameters were calculated by a regression fit for the first 1000 s and the last 800 s were used as validation for the resulting fitting parameters. We see that the extended fits well describe the data. The simple exponent (in blue) cannot model the relaxation curves. Logarithmic models tend to have a relatively good fit for all temperatures. Hence, we cannot rule out completely that the scale of the system is larger than we can observe, but this is not common for spin relaxing systems and may be coincidental. Note that the parameter $T m$ was a free parameter for different temperatures in the logarithm fits. Both the stretch exponent models β = 0.35 and 0.6 have a relatively good fit. β = 0.35 is a better predictor at high temperatures (70 K), while at low temperatures (2 K), β = 0.6 is better.

The portion of the remanent and relaxing terms attained by the different models and for different temperatures is presented in Figs. 7(d) and 7(e), respectively. We see a growing remnant t-MR as the sample is cooled, while the relaxing t-MR first increases, then below 10 K, it decreases in amplitude (for all the models).

We start the discussion with the glass-like relaxation. It shows a transient change in the t-MR measurement (following a magnetic field sweep) that relaxes over time, combined with a constant (remanent) effect. Both the relaxation rate of the transient signal and the percentage of the constant signal increases when the temperature decreases, i.e., to achieve good fits, we found that assuming a remnant parameter was required. This indicates that there are two different contributing factors to the spin signal, where one time constant (of the remanent part) is on the adiabatic scale or, at least, much longer than the lab conditions allow to observe (since in the adiabatic limit, we assume the signal should relax back to the measurement before applying the field). It is, however, unusual for such behavior to arise in the system we measure, as Pt is not ferromagnetic on its own. We also measured the Fox-30 sample that showed a similar behavior with similar scales to the Fox-10 sample. However, CuFox-10 did not show any effect at all. This indicates that the behavior observed is not inherent to the Fe₂O₃ layer but has to do with the interactions of Fe₂O₃ when it is in direct proximity to the Pt layer. It could be related to uncompensated spins in Fe₂O₃, as our samples are polycrystalline with small grain sizes, known to give rise to disordered, uncompensated spins.^20,21 We would expect there to also be a relaxation signal in the longitudinal resistance as, typically, hall effects and magnetoresistance effects co-exist; however, we could not resolve the effect. Considering the noise (as we detail in the supplementary material), we conclude that if the longitudinal magnetoresistance exhibits relaxation, it is simply too weak to resolve out of the noise and in the order of 10⁻⁵ which would require a more sensitive measurement scheme. This does point to the effectiveness of the t-MR in measuring small signals, as the effect is clearly observed in this measurement.

Additional evidence that the effect is an interface effect comes from comparing the t-MR to magnetometry measurements. Magnetometry, which is related to the bulk of the magnetic material, shows magnetic saturation above 0.3 T. But, the t-MR show an increasing magnetic response up to high fields of a few tesla, before starting to saturate.^25,26

Considering that the Fe₂O₃ layer is thin and fairly granular, it is reasonable for uncompensated spins to appear between grains, leading to a collinear response in the magnetometry and lower field saturation. The t-MR is sensitive to the interaction of current with the uncompensated spins at the interface, showing a weaker response to magnetic fields at the interface with the Pt layer (requiring a larger magnetic field to align them). This agrees with the ADMR we measure, which also shows FM-like response that does not saturate at low fields, see supplementary material V–VII for FDMR in additional orientations. We emphasize that the Fe₂O₃ layer is insulating; so all measurements arise from the properties of Pt and the interface with the AFM layer, whether due to proximity effects, SMR, or other MR effects. Thus, a plausible explanation to the slow glass-like relaxation is that at high fields, there is a FM coupling between the (large) uncompensated spins in the AFM layer that result in a proximity induced FM in the Pt layer. The AFM spins start to relax when the magnetic field is reduced, but this is slowed down due to the FM behavior in the Pt layer, resulting in the slow relaxation and remanent effects observed.

The uncompensated spins discussed above, however, cannot directly explain the reversal of the t-MR signal as a function of temperature. It may also be related to the proximity magnetization of the Pt layer. This will occur if the anomalous Hall effect coefficient of the proximity magnetized Pt has an opposite sign to that of the interface effect. So, when the proximity effect increases as the temperature is lowered, the t-MR signal flips.⁸ This also explains why there is no reversal in CuFox-10 at the same temperature scale since here the proximity effect is largely suppressed. We note that at 10 K, we did see a sign flip emerge.

The proximity effect is inherently an exchange effect and has to be in direct contact with the heavy metal exhibiting such an effect. Naively, one would suppose that Cu either does not exhibit proximity effects or that it should all cancel out. It was reported to decouple heavy metals from ferromagnets while maintaining electrical contact.^8,12 Platinoids, however, are not as straight-forward, especially Pt and Pd. In bulk theoretical estimates, they are close to satisfying the Stoner criterion, which gives rise to the possibility to show some magnetism in the surface of these metals, especially under field effects,^42,43 nanoparticles, and thin films.⁴⁴ It is possible that the sign flip we see in CuFox-10 is a manifestation of Pt's thin film magnetism, but this requires further investigation to ascertain.

The γ rotation of the ADMR measurements (Fig. 4) also corroborate this picture, since at lower temperatures, the proximity effect is larger, so the AMR-like effect measured in the γ rotation increases. The other ADMR rotations are consistent with an SMR contribution, other than the difference between α and β arising from the abovementioned changes in the γ rotation signal. Note, however, that the AMR-like effect in Fox-10 is not the simply expected effect in a FM material (as this is not a FM material), i.e., the resistance is smaller for magnetic field parallel to current vs perpendicular to current,⁴⁵ while for CuFox-10, the standard response is observed. This type of response is also not expected from SMR theory, but was observed in other Pt bilayer systems,^8,46 while it was absent in Ru bilayers that lack the proximity effect. We do not have, at this point, a deeper understanding of the source for the behavior and hope our results will encourage further research of this phenomenon.

Here, $Δ ρ sat t − M R$ is the saturation resistance of t-MR at high fields as obtained in Fig. 5, $Δ ρ SMR$ is the amplitude of the ADMR in β sweep as obtained in Fig. 4, assuming only SMR is present in that sweep, and $ρ$ is the zero-field longitudinal resistance of the device. $θ S H$ is the spin Hall angle of Pt, $λ$ is the spin diffusion length of Pt, both obtained from the literature,^14,19 and d is the thickness, as obtained from XRR measurements. In terms of the orders of magnitude, we get comparable results to those found in the literature, although here we report the imaginary part with a similar order of magnitude as the real part.^12,16,19,35

It is shown that the real part does not change much at high temperatures, until below 100 K, where it increases for both samples (Fox-10 and CuFox-10) as the temperature is decreased. The real part relates to the dampening terms, such as spin transfer torques. The imaginary part relates to field-like interactions at the interface and is clearly changing sign in Fox-10 at 220 K, while in CuFox-10, it does so only at lower temperatures, between 80 and 10 K. While the Morin transition is absent from the magnetometry, Fe₂O₃ still should exhibit a transition in spin–spin interactions, as the Dzyaloshinski–Moriya interaction, inherent to Fe₂O₃, is dampened when the material is cooled.^24,25,47 The fact that Fox-10 and CuFox-10 exhibit a different temperature dependence of the spin-mixing sign change indicates that this effect cannot come from only the temperature dependence of the Fe₂O₃ properties or its surface properties. Under the assumption that the interface between Pt and Fe₂O₃ is a clean boundary, this conductivity behavior can be interpreted as a change in the type of coupling in the Pt/Fe₂O₃ interface.^23,24,42 With the temperature ranges in mind, the Dzyaloshinski–Moriya interaction in Fe₂O₃ could be playing a role here, becoming less dominant as the sample is cooled, influencing how the Fe⁺³ ions interact with the Pt atoms.^12,16,19,34

We attempt to consider other possible sources for MR signals reported in the literature, considering also the FDMR measurements. A Hanle magnetoresistance (HMR) in our sample is calculated to be of the scale of 5 × 10⁻⁵ and lower at 8 T, as we indeed measured from the bare Pt control, and to have a parabolic shape. Thus, it cannot explain what we see in the AMDR or FDMR of our Fox-10 and Fox-30 samples, although it may contribute to them.^48,49 Likewise, the ordinary magnetoresistance from the Pt layer, as we calculate the mobility to be between $0.562$ and $0.746 c m 2 Vs$ in our sample, is expected to be on the order of 10⁻⁶ or lower at 8 T and parabolic in shape.

Our measurements, taken as a whole, indicate that there is a different kind of interaction between the HM Pt and the interface with the polycrystalline AFM Fe₂O₃. One option we can consider plausible is a mixed layer at the interface consisting of some Pt–Fe₂O₃ alloy, that has some non-trivial temperature dependence similar to that reported in a MnPSe3/Pt bilayer.^46,50 We hypothesis that there may be a non-standard proximity effect between the Pt and Fe₂O₃ interfaces in this system, influenced by the phase changes of Fe₂O₃, which resulted in the MR properties we have observed. Thus, the overall picture of the results can be mostly explained by the two contributions: (1) proximity effect, as the Cu separation layer nearly entirely eliminates effects observed in the Fe₂O₃/Pt bilayer, and (2) the uncompensated spins appearing due to the small grains and interface of the polycrystalline Fe₂O₃ film, resulting in the magnetoresistance effects observed in high magnetic fields. Our results show that one needs to pay close attention to the interface in the case of the polycrystalline Fe₂O₃/Pt bilayer system, as the results are not explainable by SMR and proximity alone. Further measurements of the interface's structural and chemical properties (such as high-resolution TEM and XMCD) can help to resolve the exact origin; however, these were beyond the scope of this study.

In summary, we have successfully fabricated bilayers of HM Pt and insulating AFM polycrystalline Fe₂O₃ and measured their magneto-transport properties for various magnetic fields, angles, and temperatures. We compared the Pt/ Fe₂O₃ device where Pt is in direct contact to the AFM layer, to a sample with a thin Cu layer in between, decoupling Pt from the AFM. We find that there are many effects that have a considerable contribution originating from the proximity effects in Pt and its coupling to interface magnetization of the polycrystalline AFM layer, which are significantly different than those measured in crystalline AFM layers. However, the displayed proximity effect has different traits compared to the previous measurements of proximity effect with insulating FM films, indicating that there is a substantial contribution from the AFM nature of the film. A comparison between magnetometry and SMR shows that in this case, the polycrystalline thin film has a high field effect coming from the interaction at the Pt/Fe₂O₃ interface. SMR based devices are powerful for investigating interfacial spin response to magnetic fields, namely, through ADMR and t-MR setups. Due to this high sensitivity, we were able to observe the development of a glass-like relaxation at low temperatures. Such effects have been observed previously in FM/AFM bilayers, but not in proximity induced magnetism in the Pt layer of the Pt/AFM bilayers. Our results indicated that a glass-like relaxation can also appear if the Pt layer has FM-like properties induced by the proximity effect from uncompensated spins in the coupled AFM layer. An investigation of the interface itself, such as interactions between Fe³⁺ ions and Pt atoms, in experiments such as XMCD may help to further understand the nature of spin-mixing conductivity and proximity effects in heavy metal/ polycrystalline AFM bilayers.

The supplementary material includes a PDF file of additional graphs of measurements and comparisons of the devices, as well as the detailed protocols and analyses used for the relevant sections.

We would like to thank Jonathan Shvartzberg and Professor Yeshurun’s group at Bar-Ilan University for performing magnetometry characterization in Quantum Design’s MPMS; and Dr. Cohen-Taguri from Bar-Ilan University’s Institute of Nanotechnology and Advanced Materials for measurement and analysis of x-ray reflectivity and diffraction of the thin films. This research was supported by the Israel Science Foundation, Grant No. 1499/23.

The authors have no conflicts to disclose.

Vladimir Kostriukov: Conceptualization (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Software (equal); Validation (equal); Visualization (equal); Writing – original draft (equal); Writing – review & editing (equal). Lidor Geri: Investigation (equal); Software (equal); Validation (equal); Visualization (equal). Amos Sharoni: Conceptualization (equal); Funding acquisition (equal); Investigation (equal); Methodology (equal); Supervision (equal); Validation (equal); Visualization (equal); Writing – original draft (equal); Writing – review & editing (equal).

The data that support the findings of this study are available from the corresponding author upon reasonable request.