WorksPhysics and Astronomy
Interface Engineering Enabled Low Temperature Growth of Magnetic Insulator on Topological Insulator
Abstract
Combining topological insulators (TIs) and magnetic materials in heterostructures is crucial for advancing spin-based electronics. Magnetic insulators (MIs) can be deposited on TIs using the spin-spray process, which is a unique non-vacuum, low-temperature growth process. TIs have highly reactive surfaces that oxidize upon exposure to atmosphere, making it challenging to grow spin-spray ferrites on TIs. In this work, it is demonstrated that a thin titanium capping layer on TI, followed by oxidation in atmosphere to produce a thin TiOx interfacial layer, protects the TI surface, without significantly compromising spin transport from the magnetic material across the TiOx to the TI surface states. First, it was demonstrated that in Bi2Te3/TiOx/Ni80Fe20 heterostructures that TiOx provided an excellent barrier against diffusion of magnetic species, yet maintained a large spin-pumping effect. Second, the TiOx was also used as a protective capping layer on Bi2Te3, followed by the spin-spray growth of the MI, NixZnyFe2O4 (NZFO). For the thinnest TiOx barriers, Bi2Te3/TiOx/NZFO samples had AFM disordered interfacial layer because of diffusion. With increasing TiOx barrier thickness, the diffusion was reduced, but still maintained strong interfacial spin-pumping interaction. These experimental results demonstrate a novel method of low-temperature growth of magnetic insulators on TIs enabled by interface engineering.
Introduction
Realization of the quantum anomalous Hall (QAH) effect has been demonstrated in several magnetic topological insulators (MTIs). MTIs can be synthesized by doping magnetic elements into topological insulators (TIs) [5, 6, 7, 8, 9] [1, 2, 3, 4] or growth of intrinsic MTI compounds [10, 11, 19] [12, 13, 14, 15, 16, 17, 18]. However, these MTI materials require temperatures much lower than their magnetic to achieve quantization in the Hall effect. This deficiency has been attributed to inhomogeneity in magnetic doping which causes localized variation in surface state exchange gaps. Magnetism can also be induced in the surface of TI thin films via proximity exchange effect. This proximity induced magnetization (PIM) is envisioned by coupling TIs with a magnetic insulator (MI). PIM in TIs have been experimentally confirmed in heterostructures of TIs with magnetic insulators (MIs) such as rare earth garnets and EuS [20, 21, 22, 23, 24, 25, 26, 27, 28], MTIs [29, 30] and telluride van der Waals (vdW) magnets [31]. However, recent studies have pointed out the complexities in measuring proximity exchange in these systems, [31] which can be affected by a number of problems such as diffusion at the interface and band bending. Thus, providing a barrier for diffusion while permitting magnetic exchange in TI/MI heterostructures will be necessary for devices, and it is shown here that a interface barrier achieves this goal. Only a single study by Watanabe et al. [31] has shown PIM and QAH in a TI/MI/TI thin film sandwich structure of . However, even in this material system, the QAH effect was seen at extremely low temperature of 0.03 K. This low temperature is possibly due to: (1) High-quality MIs which were grown on TI thin films have very low Curie temperature, , such as with K [31] and EuS with of ~17 K [25]. (2) Even with 20-25% of the magnetic species, Cr in finite localized non-magnetic nano-regions may exist in.
MIs such as rare-earth garnets and ferrites are excellent materials that are magnetically ordered well above room-temperature, making them ideal candidates for QAH materials coupled with TIs. However, growth of MIs on top of TIs using a high vacuum process is challenging because of their high growth temperatures (> 600 °C) [32]. At these temperatures, TIs are thermally unstable [33, 34] and can melt or sublimate making them unusable for TI/MI QAH systems. The spin-spray process provides a solution for growth of MIs at much lower temperatures of ~100 °C [35, 36, 37, 38, 39]. Ferrites are MIs with a high (~750 °C), which makes them excellent candidates for realization of QAH and spintronic device applications, provided they can be heterostructured with TIs. However, an engineering challenge remains to be solved. TIs have a highly reactive surface, which undergoes reactions promoted by the topological surface states (TSS) electrons [40, 41, 42, 43]. Unfortunately, exposure to atmosphere leads to rapid oxidation of Bi, Sb, Se and Te, which is experimentally observed in this work, decouples the MI layer from the adjacent TI. Without a finite exchange interaction between the TSS and the magnetic material, none of the spintronic and quantum properties can be achieved.
In this work, interface engineering of TI/FM and TI/MI heterostructures using an ultra-thin interlayer will be presented. The interlayer has three important properties: (1) it protects the TI surface against oxidation when exposed to atmosphere after the TI growth; (2) it limits interdiffusion at the TI/FM and TI/MI interface; and (3) it maintains sufficient transparency for spin transport that is crucial for devices. The insertion layer acts as a barrier to interfacial diffusion of elements without significantly compromising interaction of magnetic moments with TSSs. First, the interfacial diffusion was examined with and without on samples of and spin-spray grown NZFO. Second, the spin transparency of these heterostructures was examined by measuring the spin-orbit torque (SOT) effects on the Gilbert damping in FMR experiments. Further information was obtained from magnetization measurements, TEM images and EDS scans. From these studies it is evident that low-temperature spin-spray growth of MIs on TIs can be achieved by interface engineering of TIs which is a significant achievement. This is expected to lead to further experimental explorations of interfacial topological properties on the surface of TI/FM, TI/MI, MI/TI and MI/TI/MI heterostructures and their eventual integration of these materials systems into practical quantum and spintronic devices.
Experimental Results and Discussion
The experimental section is organized as follows: (2.1) contains results on the effects of TI surface oxidation and interlayer on spin-pumping and interdiffusion; (2.2) describes the formation of an AFM interlayer via atomic diffusion; (2.3) contains results on TI/MI heterostructures grown by non-vacuum, low-temperature spin-spray deposition of NZFO on .
Effects of TI Surface Oxidation an Interlayer TiO on Spin-pumping and Interfacial Diffusion
Suppression of SOT on TI Surface due to Surface Oxidation
Kittel equation fitting for Py, , (oxidized)/Py and /Py samples. c) Illustration of suppressed spin pumping due to oxidation of on exposure to atmosphere, and d) large spin pumping in /Py with Ni- interfacial layer resulting from diffusion of Ni [42].
The spin transparency of TI/FM interfaces is conveniently determined in FMR experiments by measuring changes in the Gilbert damping. In high-quality TI/FM interfaces the precessing magnetization in the FM results in spin-pumping into the TI. This is because the strong SOC in TIs, resulting in ISHE and enhanced gilbert damping. On the other hand, no spin-pumping is seen when the spins are prohibited from transferring from the FM to the TI layer. On exposure to atmosphere, TI surfaces readily oxidize [40], which leads to topologically trivial insulators on the surface. Almost total loss of spin-pumping was observed in TI/FM heterostructures when the thin film was exposed to the atmosphere before deposition of the FM (20 nm thick Py).
To understand the effect of surface oxidation of TI on spin-pumping in TI/FM heterostructures, FMR experiments were performed at room temperature with the following samples: /Py, (oxidized in atmosphere)/Py, (3nm Al oxidized in atmosphere)/Py, and Al(10nm)/Py (control). For the sample (oxidized)/Py as the label suggests, was exposed to atmosphere before deposition of Py. The FMR linewidths were extracted by fitting a Lorentzian function to the FMR field scans (see Experimental Methods section and Supporting Information § 8). The Gilbert damping was extracted from a linear fit to the as a function of frequency using the relation, , as shown in Figure 1a. Here, is the inhomogeneous linewidth broadening and is the gyromagnetic ratio ( MHz/Oe for metallic films) [48]. Figure 1a shows that the /Py sample has a very large Gilbert damping, , compared to all the other samples, indicating that spin transport is nearly eliminated by adding interfacial layers. The sample with (oxidized)/Py had a much reduced , almost the same as a control sample of Al/Py. Note that when the metal-oxide is thick, as with the /AlO/Py sample, , which is higher than the control sample but still much lower than that of the /Py. However, note that in the /Py sample there is a naturally-diffused interlayer of antiferromagnetic (AFM) Ni- that remains effective for spin transfer and large damping [42].
Figure 1. Results of FMR measurements on TI/FM heterostructures with differing interfacial conditions. a) FMR linewidth as a function of FMR frequency for /Py, (oxidized in atmosphere)/Py and Py thin films, and b) FMR frequency as function of FMR resonant field and
Further, the effective magnetization, was extracted by fitting the FMR frequency, as function of FMR, field plot using the Kittel equation, as shown in Figure 1b. Here, is the FMR resonance field and is the uniaxial anisotropy field. A significant change in suggests change in magnetic anisotropy because of interfacial exchange interaction of the FM layer with TSS [46]. Further, exchange interaction of an FM with a large SOC material should also result in change in . However, due to the large thickness of the FM layer, this change is expected to be negligibly small [48]. Hence, a value of 2.8 MHz/Oe has been considered for the metallic FMs in this work. The fitting to Kittel equation reveals values of 1.22 kOe for Py, slightly reduced values of 1.15 kOe for (oxd)/Py, 1.14 kOe for /AlO/Py and a dramatically reduced value of 6.55 kOe for the /Py heterostructure. This suggests reduced interfacial magnetic interactions in in the (oxidized)/Py and /AlO/Py samples because of decoupling of the FM from the TSS of the TI represented by the schematics in Figure 1c,Figure 1d. In MBE grown TI thin film samples, surfaces are usually capped with Se or Te [53, 54, 55] [49, 50, 51, 52], which is evaporated by heating in a high vacuum before growing the FM layer in SOT devices. The presence of a highquality capping layer becomes especially significant for successfully growing MI ferrites using the low-temperature spin-spray process on TIs as substrates are exposed to the atmosphere during deposition. The choice of an effective capping/interface engineered layer is essential as interfacial magnetic interactions and spin transparency must not be significantly diminished through of the barrier for efficient spintronic device applications.
TiO Interlayer Effects on Diffusion in TI/FM Heterostructures
without breaking vacuum results in interfacial AFM interface which hypothetically can lead to magnetic gap opening in TSS. on exposure to atmosphere oxidizes the surface and pushes the TSS below the oxidized surface and the magnetic Py layer becomes decoupled from the surface. b) Reduction in magnetic moments arise from diffusion of Ni from Py and reacting with . Oxidized does not show diffusion and moment reduction effects. c) Large exchange bias of 82 Oe in /Py generated by AFM in the interface [42].
The atomic interdiffusion in TI/FM samples was investigated as a function of interlayers. Two cases are illustrated in Figure 2a, one where the Py is grown on the TI without breaking the chamber vacuum and the other where the TI was exposed to atmosphere before depositing the Py. Without breaking the vacuum, the /Py interface develops a topological nontrivial AFM interfacial layer because of diffusion and reaction of Ni with and forming [42, 43]. In contrast, oxidizing the surface of the TI prevents reaction and diffusion of Ni or Fe from Py and should not allow for forming a topological interfacial layer. To study the effect of surface oxidation of on the interface diffusion and reactions in /Py, room temperature measurements were performed on the samples of /Py, (oxidized)/Py and Py (control) samples. Because of interfacial diffusion of Ni, Figure 2b shows that the saturation moment of the Py layer decreases by . This happens because the diffused Ni reacts with to form compounds which are not ferromagnetically ordered at room temperature [42]. But as expected, no magnetic moment is lost when the is oxidized before deposition of the Py layer, as the oxide interlayer prevents diffusion of Ni out of the Py layer. Furthermore, as a result of inhibiting diffusion, the oxide interlayer prohibits Ni from reacting with the , which prevents formation of an AFM interlayer. The presence of an AFM interfacial layer next to the FM Py was found to induce a large spontaneous exchange bias field, Oe [42] at low temperatures (ZFC 6 K). However, as shown in Figure 2c, samples of (oxidized)/Py didn’t show any exchange-bias effects in contrast to the large exchange bias in /Py heterostructures. These results confirm the suppression of surface reactivity due to atmospheric oxidation of TIs. The oxides formed on the TI surface decouple the TSS chemically from the adjacent FM layer, although we will show that spin transfer can remain effective. These results further emphasize the need for interface engineering for integration of TIs with spin-spray grown MIs.
Figure 2. Results of TI/FM heterostructure with and without oxidized TI. a) Schematic representation of effect of oxidation of BiTe before deposition of FM Py. Deposition of Py on
Interface Engineering in TI/FM Heterostructures using
It is important to investigate how the thickness, of interlayers of Ti and its oxide in TI/FM structures affect the SOT and related magnetic properties. Ti possesses negligible SOC and is highly transparent to spin currents [45], which makes an excellent material for interface engineering of TIs with FM oxides for spintronic device applications. In order to understand the effects of the insertion layers in TI/FM heterostructures, crystalline -axis-oriented was grown using magnetron sputtering similar to the work in reference [42]. The growth of high-quality was followed by deposition of ultrathin Ti that was subsequently exposed to atmosphere. Py thin films of thickness 20 nm were then grown on the / bilayers to first study their effect on magnetic properties. Figure 3a shows the effects of interfacial diffusion and solid-state reactions by monitoring the reduction of Py moments. Ni and small amounts of Fe diffuse out from the Py and form interfacial compounds due to solid state reactions with [42, 43]. Looking at the saturation moment, it is clear that layers with thickness nm cannot prevent interfacial diffusion of the magnetic species, which is aided by surface reaction with . Note that the cited thickness, is that of the Ti thin film deposited on a Si/SiO substrate. The diffusion of primarily Ni and small amounts of Fe, leads to a moment loss of , which is comparable to the control /Py sample [42]. However, insertion layers of thickness nm act as excellent barriers to the diffusion of Ni from Py preventing any loss of magnetic moments.
Figure 3. Results of magnetization and FMR experiments on /Py structures with interlayers. a) loops measured at room temperature, showing no loss of moment for the two thicker layers. b) FMR linewidth as a function of FMR frequency, showing strong damping for all thickness of interlayers. Inset illustrates the spin current passing down thru the . c) FMR frequency as function of FMR field and Kittel equation fitting. d) Summary of
Gilbert damping for Py, and for , 1, 2, 3 nm thick interlayers, which far exceeds that of the Py control sample.
Next, spin-pumping effect was investigated as a function of thickness via the enhancement in Gilbert damping in FMR experiments. A dramatic improvement in spin pumping over insertion layers was observed, which can be seen by comparing the results of Figure 1a with those of Figure 3b–Figure 3d. Table 1 reports the values of various properties for the samples. Values of obtained for the samples were 0.074, 0.040 and 0.049 for the 1, 2 and 3 nm thick layers, respectively, and are comparable to that of 0.061 for the sample. All these samples hence show significantly enhanced compared to 0.011 for the control sample of Py, thus confirming a large spin-pumping effect. The spin-mixing conductance can be quantified using the spin-mixing conductance parameter given by , where is the thickness of the FM layer, is the enhancement in , is the reduced Plank’s constant and is the gyromagnetic ratio. The enhancement in for the and samples may also contain some contributions from the loss of magnetic moments due to the large interdiffusion. But, because of the complexity of the interfaces of these samples, the effect of spin-pumping may not be accurately isolated. The values reported for these samples were calculated based on the assumption that the enhancement in compared to control sample of Py is due to spin-pumping only, which may nevertheless be incorrect. However, as the samples with 2 nm and 3 nm barriers had no noticeable interfacial diffusion of Ni, Fe, they had large values as listed in Table 1. These large values indicate that TIs with interface-engineered heterostructures can be excellent candidates for efficient SOT devices. Further, Table 1 also lists saturation magnetization, determined from the loop measurements in Figure 3a. Using those values, the effective interfacial out-of-plane (OOP) anisotropy field values can be calculated using the relation, . The values listed in Table 1 show that (2 and 3 nm)/Py samples have and 2.15 kOe, respectively. The negative for the Py sample indicates that the magnetic anisotropy and easy axis of the sample is highly in-plane as expected. The /Py samples on the other hand show a large OOP anisotropy because of the exchange interaction of the interfacial magnetic moments of Py with the TSS of .
| Material | (nm) | (emu/cm) | (m) | (kOe) | (kOe) | |
| Py | - | 0.011 | - | 12.6 | -1.05 | |
| /Py | 0 | 0.061 | 6.32 | - | ||
| /Py | 1 | 0.074 | 6.56 | - | ||
| /Py | 2 | 0.040 | 9.53 | 2.02 | ||
| /Py | 3 | 0.049 | 9.40 | 2.15 |
Table 1. Properties for /Py heterostructures for different interlayer thickness. The parameters are described in the text.
Tuning of Topological AFM Interface with Barrier
Formation of an interfacial topological AFM layer generated by diffusion of Ni and solid-state reaction with was shown previously in ref [42, 43]. The solid-state reaction was found to result in loss of magnetic moment and emergence of a large spontaneous exchange bias in /Py heterostructures. The loop measurements at room temperature in Figure 4a show loss of magnetic moments in the /Py sample with the 1 nm thick , similar to the /Py sample deposited with the same growth conditions. The increase in barrier thickness was found to prevent the diffusion of Ni across the interface. In order to understand the effect of the interface in the /Py samples, loops were measured at 300 and 6 K in zero-field cooling (ZFC) conditions. The sample with nm clearly shows a large Oe and an enhanced Oe, as shown in Figure 4a. This arises from exchange interaction between the FM Py layer with the interfacial AFM layer [42]. This was however smaller than the Oe for BiTe/Py without a TiO barrier. This is due to the reduction in exchange interaction caused by the insertion of the ultra-thin TiO layer between the interfacial AFM Ni-BiTe and the Py layer. With an increase in thickness of the TiO barrier to nm, there is no diffusion of Ni and hence no is observed in the at low temperatures. This clearly shows the effect of interface engineering on controlling the interfacial properties in TI/FM heterostructures that will be essential for spintronic devices. These results also show the potential of exploring fascinating interfacial phases in TI/FM heterostructures.
Figure 4.
Figure 4 (PDF p. 15). Measurements of loop at 6 K (ZFC) and 300 K showing exchange bias of samples. a) Oe for nm at 6 K. b) No exchange bias for nm, and c) no exchange bias for nm. d) Comparison of exchange bias at 6 K for the samples. (Inset shows the diffusion-generated AFM layer below the for nm.
Heterostructures
Interfacial Morphology of Heterostructures
The following describes results of non-vacuum, spin-spray growth of the magnetic insulator on , which is facilitated using a interlayer. heterostructures were investigated for their morphology, diffusion and spin-pumping. The morphology of the interface was first studied using cross-section TEM imaging and EDS. As shown in Figure 6a, the is vdW-layered and highly -axis oriented as expected [42]. The TEM image also shows a disordered interfacial layer for nm that is created by diffusion and solid-state reactions. In contrast, as observed in the TEM images in Figure 6b-Figure 6c, with increase in thickness of the barrier layer to 2 nm and 3 nm the thickness of the disordered interface is significantly reduced progressively. The EDS line profiles in Figure 6d-Figure 6f show the intensity of the EDS signal. With the 1 nm barrier layer there is a clear interfacial layer primarily consisting of Bi, Te, Fe and O (Figure 5.6a) with O significantly diffusing nm into . This is possibly the primary factor for creation of the highly disordered interfacial layer observed in Figure 6a. For the 2 and 3 nm thick interlayer the thickness of the oxygen diffusion is reduced to nm and nm, respectively. Further, thickness of nm prevents any Fe diffusion as confirmed by TEM images as well as EDS line profiles. This demonstrates the effectiveness of the thicker barrier layers in the heterostructures.
Figure 6. Results of FMR experiments on . a) Linewidth as a function of FMR frequency, b) summary of Gilbert damping and c) FMR frequency as a function of field for NZFO and . d) Comparison of change in compared to the control sample of NZFO because of SOT in the layer.
Figure 5. TEM images for with thickness: a) 1 nm b) 2 nm, and c) 3 nm, showing c-axis-oriented vdW layered structure and interfacial layers. EDS line profiles of the TEM images for: d) 1 nm, e) 2 nm, and f) 3 nm samples. g) Comparison of EDS spectra above barrier. Concentration of diffused Bi and Te: h) above barrier and i) below barrier for comparison. j) Diffused Fe below barrier in the 1, 2, and 3 nm samples.
Diffusion of elements at the interface was analyzed with EDS by extracting the relative concentration of Bi and Te above the and Fe below the regions. The concentration of oxygen couldn’t be analyzed accurately as EDS results for light elements are inaccurate. For this analysis of interfacial diffusion only a small thickness of about 5 - 6 nm above and below the regions were considered (see Supporting Information § 11). As shown in Figure 6h, which is extracted from the Figure 6g, there is a small diffusion of Bi and Te across the interface to the NZFO side. Similarly, Figure 6j, which is extracted from the EDS spectra in Figure 6i, also shows diffusion of Fe from NZFO into . The diffused Bi relative atomic% values are 2.9%, 1.6% and 1.1% for the samples with 1, 2 and 3 nm , respectively. The values for diffused Te also show a similar trend of 2.6%, 1.4% and 1.0%, respectively. Interestingly, the diffusion of Fe shown in Figure 6j below the barrier layers from NZFO to the side was much higher at 11.6% for the sample with 1 nm and reduces significantly to 5.7% and 4.4% for the samples with 2 and 3 nm barriers, respectively. The concentrations of Ni and Zn are very small () throughout the cross-section of the NZFO layer and are within the noise range of EDS signal as shown in Figure 6g,Figure 6i. Hence, diffusion of these elements could not be determined accurately. Nevertheless, the significant diffusion of Fe in the 1 nm samples form a disordered interface that results in a fascinating AFM interfacial phase which will be described in § 3.2.1.
Room Temperature Magnetic Properties of / /NZFO Heterostructures
NZFO thin films of thickness nm were grown on / bilayers using the method described in refs. [38, 39] (see Experimental Methods). The deposition was made at a substrate temperature of , which is not expected to cause any phase change in the layer. The Gilbert damping extracted from the slope of the FMR linewidth versus frequency in Figure 6a clearly shows a smooth decreasing trend with increasing thickness, shown in Figure 6b and listed in Table 2. The NZFO control sample had an value similar to the values of the / /Py samples with 2 nm and 3 nm . The significantly enhanced for the 1 nm insertion layer indicates a large spin-pumping and possibly some spin memory-loss (SML) effects in the / /NZFO heterostructure [56] [46]. Because of the large thickness of the NZFO layer, the overall Gilbert damping enhancement effect is reduced in the / /NZFO with and 3 nm samples. This is due to the increase in distance between the MI and TI because of the presence of the TiOx insertion layer. The enhancement in Gilbert damping indicates that the diffused oxygen in the samples is extremely low and does not significantly affect the interfacial spin transparency, unlike TIs exposed to atmosphere. The decreasing trend in with the increase in is similar to the trend observed with Py as the FM layer. In contrast to thick metallic FM films, MIs show more pronounced changes in magnetism because of proximity to large SOC materials [46,48]. Fitting the Kittel equation to the versus relation, the values of and of the samples as shown in Figure 7c and Table 2. The values for and 3 nm were very close to the NZFO control sample. However, the nm sample showed a dramatic reduction to less than half the other values. This change in has possible contributions from: (1) the exchange interaction between the magnetic moments in the MI and TI, as well as (2) new materials phases formed from diffusion and solid-state reactions in the interface, and (3) PIM in the TI due to the adjacent MI. However, these effects could not be isolated in this study and will need further analysis in the future. Furthermore, the for the samples obtained from the Kittel equation fittings in Figure 6c and listed in Table 2 were significantly different larger compared to the NZFO control sample. The large changes in plotted in Figure 6d signify strong exchange interaction effect of the magnetic moments in the MI with the TSSs in the TI surface [46]. The HRTEM and EDS measurements in the earlier section clearly indicated lack of diffusion in the (and Py) when TiOx has a thickness larger than 2 nm. This confirms a finite exchange interaction between the TSS and MI even with a finite separation between them.
Figure 7. Magnetic properties of BiTe/TiO/NZFO for various TiO thicknesses. Measured loops at 300 K and 6 K for: a) NZFO, b) nm, c) nm, and d) nm samples. e) Only the nm sample shows an exchange bias, Oe arising from formation of AFM interfacial layer at 6 K.
| Material | (nm) | (kOe) | (MHz/Oe) | (Oe) | (Oe) | |
| NZFO | - | 6.31 | 0.0118 | 2.81 | 0 | 161.71 |
| BiTe/TiO/NZFO | 1 | 2.52 | 0.0313 | 3.05 | 4.92 | 171.89 |
| BiTe/TiO/NZFO | 2 | 5.67 | 0.0134 | 2.92 | 0 | 171.75 |
| BiTe/TiO/NZFO | 3 | 6.56 | 0.012 | 2.79 | 0 | 175.51 |
Table 2. Properties for BiTe/TiO/NZFO heterostructures for different TiO interlayer thickness. The parameters are described in the text.
Interfacial AFM Phase in Heterostructures
An AFM interface layer was formed in the sample with nm, similar to that discovered in /Py samples. [42] AFM layers form at the interface in /Py, and /Py samples as a result of interdiffusion and solid state reaction with the TI surface states [42, 43]. In order to understand the effect of interfacial magnetic properties of the heterostructure samples, loop measurements were performed at 300 K and 6 K similar those shown in § 2.1.2 and 2.2. The loop at 300 K for the nm sample was well centered. However, at 6 K the loop showed a noticeable off-center shift of Oe signifying exchange bias from the interaction at the AFM-FM interface. The samples with thicker ( nm) barrier and the control NZFO sample had a well-centered hysteresis loop at 6 K as well as 300 K, indicating no discernable exchange bias. Table 2 lists values for the exchange bias, and coercive field, . This exchange bias is a clear signature of AFM order in the interface emerging in the ferrite heterostructures, again possibly due to solid-state surface chemical reactions promoted by the TSSs of . The AFM order in the interface may also arise from PIM in the disordered interface composed of Fe, Bi and Te, similar to V-doped /EuS interfaces in ref [47]. These novel magnetic interfacial layers can have topological nontrivial properties as was shown earlier, but the mechanism is still controversial. It is noteworthy that an effective barrier, TI/FM as well as TI/MI heterostructures can be grown without changing the crystalline phase of the TI while still maintaining magnetic interactions between MI (or FM) and TSSs. These experimental results are expected to open further exploration of exotic quantum states in interface-engineered TI/MI and TI/FM heterostructures.
Conclusions
Exposure of TI surface to atmosphere was shown to result in formation of oxides on the surface that are impervious to spin currents and coherent electron hopping. The surface oxidation of TI was shown to magnetically decouple FM grown on them and showed suppression of spin-pumping, interfacial solid-state reactions, and exchange interactions. An interface engineering method with barrier was presented. An ultra-thin layer () of was shown to protect the TI surface from oxidation without changing the topological properties. This was verified by coupling the TI, with the FM Py, which showed (1) large Gilbert damping enhancement, (2) interface solid-state reaction and (3) exchange bias. With increase in thickness beyond 2 nm, the interfacial diffusion can be blocked, while preserving a large spin-pumping effect in /Py heterostructures. Further, low-temperature atmospheric growth of NiZn-ferrite was used as the magnetic insulator in heterostructures. The magnetic properties of NZFO MI on bilayers were studied Similar to the /Py samples with an ultrathin dusting layer of (nominally, ), significant surface reactions were observed for the /NZFO samples, which also showed enhanced Gilbert damping and a small exchange bias at low temperatures (6 K). However, with higher barrier thickness, the interface reactions between and NZFO was prevented. The results in this study are expected to open a path towards further exploration of interface engineered TI/FM and TI/MI heterostructures for energy efficient spintronic device applications.
Experimental Methods
Growth of TI/FM and TI/TiO/FM Heterostructures Using Magnetron Sputtering
The TI, BiTe and Py thin films were grown using the same process conditions as the previous work in ref [42]. For studying the effect of surface oxidation, BiTe was exposed to atmosphere before deposition of FM. Further, Ti thin films were grown using DC magnetron sputtering with 30 W power and 3 mT Ar. The growth rate of Ti thin films was 0.019 nm/s, which were oxidized upon exposure to atmosphere. All the samples were capped with 3 nm of Al on the FM, which was subsequently oxidized to AlO on exposure to atmosphere.
Low-Temperature Growth of MI, NZFO Using Spin-Spray Process
Polycrystalline NiZn-ferrite thin films with thickness of 250 nm were deposited on a Si/SiO substrate using a homemade 24” spin spray. The spin-spray growth process for NZFO includes a precursor solution with 1.54 g/L of FeCl, 0.04 g/L of ZnCl, 0.03 g/L of NiCl and an oxidizer solution with 0.14 g/L of NaNO and 11.48 g/L of CHCOONa. The precursor solution had a pH level of 3-4 and the oxidizer solution had a pH level of 10.8. The process was carried out at a temperature of with 500 mL volume of solutions. Detailed discussion of the growth of the NZFO films using the spin-spray process is available in refs [38, 39]. XPS measurements of the NZFO samples revealed 77.4 %, 19.3%, 2.0% and 1.2 % average atomic concentrations of O, Fe, Ni and Zn, respectively.
Ferromagnetic Resonance
FMR experiments were performed using field-sweeping at constant frequencies. The FMR signal was detected using a lock-in amplifier for enhanced sensitivity. The RF field for exciting magnetization in the samples was provided by a co-planer waveguide. A pair of Helmholtz coils which produced an AC magnetic field (500 Hz) was used for the driving field. Control of the experiment and data acquisition was done using NI LabVIEW.
The parameters, FMR linewidth and resonance field were extracted by fitting a Lorentzian function to the FMR spectra, given by,
Here, is the absorbed RF power, are constants, and is the applied DC magnetic field.
Measurement of Hysteresis Loops
4.4. Measurement of Hysteresis Loops. Magnetization measurements were obtained using a Quantum Design MPMS XL-7 superconducting quantum interference device (SQUID) magnetometer [42, 43]. Hysteresis loop measurements were carried out at various temperatures between 6 and 300 K. The ZFC measurements were obtained while increasing the temperature in an applied field of 50 Oe. Room temperature measurements were also taken using a vibrating sample magnetometer (VSM).
TEM and EDS Measurements
4.5. TEM and EDS Measurements. Samples for TEM investigations were prepared by focused ion beam milling (FIB) using a ion source [42, 43]. Prior to TEM observation an additional cleaning procedure was performed by Ar-ion milling to reduce a surface amorphous layer and residual Ga from the FIB process. The TEM observations were performed using a Talos 200-FX (ThermoFiszher Scientific Inc.) TEM operated at an acceleration voltage of 200 kV. EDS measurements were performed using a ChemiSTEM (ThermoFisher Scientific) and processing of the spectra was performed using Esprit 1.9 (Brucker Inc.) software.
Acknowledgement
This work is partially supported by the U.S Army under grant no. W911NF20P0009, the NIH Award UF1NS107694 and by the NSF TANMS ERC Award 1160504. The work of DH and AF was partially supported by the National Science Foundation grant DMR-1905662 and the Air Force Office of Scientific Research award FA9550-20-1-0247. The work of KM was supported by
Air Force Research Laboratory under AFRL/NEMO contract: FA8650-19-F-5403 TO3. Studies employing the Titan 60-300 TEM was performed at the Center for Electron Microscopy and Analysis (CEMAS) at The Ohio State University with support through Air Force contract FA8650-18-2-5295. Certain commercial equipments are identified in this paper to foster understanding. Such identification does not imply recommendation or endorsement by Northeastern University and AFRL.
Conflict of Interest
The authors declare no conflict of interest.
Figure 8 (PDF p. 34).
Supporting Information: Interface Engineering Enabled Low Temperature Growth of Magnetic Insulator on Topological Insulator
Corresponding author: Nian X. Sun
E-mail: [email protected] (Corresponding author)
FMR Spectra for Studying the effect of Surface oxidation of

Figure S1. FMR spectra for Py, (oxidized)/Py, /AlO/Py and /Py samples for extracting , , and .
FMR Spectra for /TiO/Py Samples

Figure S2. FMR spectra for Py, /Py, /TiO (1, 2, 3 nm)/Py samples for extracting , , and .
FMR Spectra for /TiO/NZFO Samples

Figure S3. FMR spectra for NZFO, BiTe/TiOx (1, 2, 3 nm)/NZFO samples for extracting , , and .
EDS Color Maps and for Studying Interfacial Diffusion

Figure S4. EDS color maps for studying interfacial diffusion in, a) (1 nm)/NZFO, b) (2 nm)/NZFO, and c) (3 nm)/NZFO samples. The boxes in the Ti color maps denote the above Ti (blue) and below Ti (yellow) regions used for calculation of relative atomic % on Fe, Bi and Te.
Repeatability of FMR Results for /Py Heterostructures
A different set of Py, /Py and /TiO/Py samples were grown using the same sputtering process conditions. Similar to the samples in the main text, the TiO barrier thickness was varied from 1-3 nm. The gilbert damping, extracted from FMR linewidth and frequency plots, shown in Figure S5 demonstrate a similar trend as the results in the Main Text. The gilbert damping for the control sample of Py was 0.015. However, because of presence of the TI layer and large spin-pumping (also possible SML), the /Py and /TiO/Py samples showed large enhancement in . The values were 0.0204, 0.0179, 0.0176 and 0.0172 for TiO barrier thickness of 0, 1, 2 and 3 nm respectively.

Figure S5. a) FMR linewidth vs. frequency of /Py sample, b) Comparison of Gilbert damping in Py, /Py and /TiO/Py samples for varying thickness of barrier TiO.
References
- [5]W. Wang, Y. Ou, C. Liu, Y. Wang, K. He, Q.K. Xue, W. Wu, Direct evidence of ferromagnetism in a quantum anomalous Hall system, Nat. Phys. 14, 791 (2018).DOI
- [6]J. Teng, N. Liu, and Y. Li, Mn-doped topological insulators: a review, J. Semicond. 40, 081507 (2019).DOI
- [7]A. Tcakaev, V. B. Zabolotnyy, R. J. Green, T. R. F. Peixoto, F. Stier, M. Dettbarn, S. Schreyeck, M. Winnerlein, R. Crespo Vidal, S. Schatz, H. B. Vasili, M. Valvidares, K. Brunner, C. Gould, H. Bentmann, F. Reinert, L. W. Molenkamp, and V. Hinkov, Comparing magnetic ground-state properties of the V- and Cr-doped topological insulator (Bi,Sb)2Te3, Phys. Rev. B 101, 045127 (2020).
- [8]A. Tcakaev, V. B. Zabolotnyy, C. I. Fornari, P. Rüßmann, T. R. F. Peixoto, F. Stier, M. Dettbarn, P. Kagerer, E. Weschke, E. Schierle, P. Bencok, P. H. O. Rappl, E. Abramof, H. Bentmann, E. Goering, F. Reinert, and V. Hinkov, Incipient antiferromagnetism in the Eu-doped topological insulator Bi2Te3, Phys. Rev. B 102, 184401 (2020).
- [9]Y. Ni, Z. Zhang, I. C. Nlebedim, R. L. Hadimani, G. Tuttle, D. C. Jiles, Ferromagnetism of magnetically doped topological insulators in CrxBi2-xTe3 thin films, J Appl. Phys. 117, 17C748 (2015).DOI
- [10]J. Li, Y. Li, S. Du, Z. Wang, B.L. Gu, S.C. Zhang, K. He, W. Duan, Y. Xu, Intrinsic magnetic topological insulators in van der Waals layered MnBi2Te4-family materials, Sci. Adv. 5, eaaw5685 (2019).DOI
- [11]Z. Li, J. Li, K. He, X. Wan, W. Duan, Y. Xu, Tunable interlayer magnetism and band topology in van derWaals heterostructures of MnBi2Te4-family materials, Phys. Rev. B 102, 081107(R) (2020).DOI
- [19]H. Deng, Z. Chen, A. WołoŻ, M. Konczykowski, K. Sobczak, J. Sitnicka, I. V. Fedorchenko, J. Borysiuk, T. Heider, ŕ. PluciÄ,ski, K. Park, A.B. Georgescu, J. Cano, L.K. Elbaum, High-temperature quantum anomalous Hall regime in a MnBi2Te4/Bi2Te3 superlattice, Nat. Phys. 17, 36 (2021).DOI
- [20]X. Che, K. Murata, L. Pan, Q.L He, G. Yu, Q. Shao, G. Yin, P. Deng, Y. Fan, B. Ma, X. Liang, B. Zhang, X. Han, L. Bi, Q.H. Yang, H. Zhang, K. L. Wang, Proximity-Induced Magnetic Order in a Transferred Topological Insulator Thin Film on a Magnetic Insulator, ACS Nano 12, 5042 (2018).DOI
- [21]C.C. Chen, K. H. M. Chen, Y. T. Fanchiang, C. C. Tseng, S. R. Yang, C. N. Wu, M. X. Guo, C. K. Cheng, S. W. Huang, K. Y. Lin, C. T. Wu, M. Hong and J. Kwo, Topological insulator Bi2Se3 films on rare earth iron garnets and their high-quality interfaces, Appl. Phys. Lett. 114, 031601 (2019).DOI
- [22]V. M. Pereira , S. G. Altendorf, C. E. Liu, S. C. Liao, A. C. Komarek, M. Guo, H.-J. Lin, C. T. Chen, M. Hong, J. Kwo, L. H. Tjeng, and C. N. Wu, Topological insulator interfaced with ferromagnetic insulators: Bi2Te3 thin films on magnetite and iron garnets, Phys. Rev. Mat. 4, 064202 (2020).
- [23]C. Tang, C.-Z. Chang, G. Zhao, Y. Liu, Z. Jiang, C.-X. Liu, M. R. McCartney, D. J. Smith, T. Chen, J. S. Moodera, and J.Shi, Above 400-K robust perpendicular ferromagnetic phase in a topological insulator, Sci. Adv. 3, e1700307 (2017).arxiv.org/abs/1706.07971
- [24]A.J. Grutter, and Q.L. He, Magnetic proximity effects in topological insulator heterostructures: Implementation and characterization, Phys. Rev. Mat. 5, 090301 (2021).DOI
- [25]F. Katmis, V. Lauter, F.S. Nogueira, B.A. Assaf, M.E. Jamer, P. Wei, B. Satpati, J.W. Freeland, I. Erem[in], D. Heiman, P. Jarillo-Herrero & J.S. Moodera, A high-temperature ferromagnetic topological insulating phase by proximity coupling, Nature volume 533, 513 (2016).DOI
- [26]M. Li, C.Z. Chang, B.J. Kirby, M.E. Jamer, W. Cui, L. Wu, P. Wei, Y. Zhu, D. Heiman, J. Li, and J.S. Moodera, Proximity-Driven Enhanced Magnetic Order at Ferromagnetic-Insulator-Magnetic-Topological-Insulator Interface, Phys. Rev. Lett. 115, 087201 (2015).DOI
- [27]P. Wei, F. Katmis, B.A. Assaf, P. Jarillo-Herrero, D. Heiman, and J.S. Moodera, Magnetic proximity-induced symmetry breaking in topological insulators, Phys. Rev. Lett. 110, 186807 (2013); http://dx.doi.org/10.1103/PhysRevLett.110.186807
- [28]N. Andrejevic, Z. Chen, T. Nguyen, L. Fan, H. Heiberger, L.J. Zhou, Y.F. Zhao, C.Z. Chang, A. Grutter, and Mingda Li, Elucidating proximity magnetism through polarized neutron reflectometry and machine learning, App. Phys. Revs. 9, 011421 (2022).DOI
- [29]Q. Li, C. Xuan Trang, W. Wu, J. Hwang, D. Cortie, N. Medhekar, S.K. Mo, S.A. Yang, and M.T. Edmonds, Large Magnetic Gap in a Designer Ferromagnet-Topological Insulator-Ferromagnet Heterostructure, Adv. Mat. 34, 2107520 (2022).DOI
- [30]Q.L. He, X. Kou, A.J. Grutter, G. Yin, L. Pan, X. Che, Y. Liu, T. Nie, B. Zhang, S.M. Disseler, B.J. Kirby, W. Ratcliff II, Q. Shao, K. Murata, X. Zhu, G. Yu, Y. Fan, M. Montazeri, X. Han, J.A. Borchers and K.L. Wang, Tailoring exchange couplings in magnetic topological-insulator/antiferromagnet heterostructures, Nature Mat. 16, 94 (2017).DOI
- [31]R. Watanabe, R. Yoshimi, M. Kawamura, M. Mogi, A. Tsukazaki, X. Z. Yu, K. Nakajima, K. S. Takahashi, M. Kawasaki, and Y. Tokura, Quantum anomalous Hall effect driven by magnetic proximity coupling in all-telluride based heterostructure, Appl. Phys. Lett. 115, 102403 (2019).DOI
- [32]J. Ding, T, Liu, H. Chang, and M. Wu, Sputtering Growth of Low-Damping Yttrium-Iron-Garnet Thin Films, IEEE Magnetics Letters, vol. 11, 1 (2020).DOI
- [33]J. Buha, R. Gaspari, A.E.D.R. Castillo, F. Bonaccorso, and L. Manna, Thermal Stability and Anisotropic Sublimation of Two-Dimensional Colloidal Bi2Te3 and Bi2Se3 Nanocrystals, Nano Lett. 16, 7, 4217 (2016).DOI
- [34]M.P. Siegal, A.L. Lima-Sharma, P.A. Sharma, and C. Rochford, Correlating thermoelectric properties with microstructure in Bi0.8Sb0.2 thin films, Appl. Phys. Lett. 110, 141905 (2017).DOI
- [35]M. Abe, and Y. Tamaura, Ferrite-Plating in Aqueous Solution: A New Method for Preparing Magnetic Thin Film, Jap. J. App. Phys., 22, L511, (1983).DOI
- [36]M. Abe, T. Itoh, Y. Tamaura, and M. Gomi, Ferrite organic multilayer film for microwave monolithic integrate circuits prepared by ferrite plating based on the spray-spin-coating method, J. App. Phys., 63, 3774 (1988).DOI
- [37]M. Abe, Ferrite plating: a chemical method preparing oxide magnetic films at 24-100°C, and its applications, Electrochiimica Acta 45, 3337 (2000).DOI
- [38]O. Obi, M. Liu, J. Lou, S. Stoute, X. Xing, N.X. Sun, J. Warzywoda, A. Sacco Jr, D.E. Oates, and G.F. Dionne, Spin-spray deposited NiZn-Ferrite films exhibiting μr'>50 at GHz range, J. of App. Phys., 109, 07E527 (2011).DOI
- [39]X. Wang, H. Chen, X. Shi, Y. Gao, H. Lin, J.G. Jones, B.M. Howe, G.J. Brown, and N.X. Sun, A novel NiZn ferrite integrated magnetic solenoid inductor with a high quality factor at 0.7-6 GHz, AIP Advances 7, 056606 (2017).DOI
- [40]D. Kong, J.J. Cha, K. Lai, H. Peng, J.G. Analytis, S. Meister, Y. Chen, H.J. Zhang, I.R. Fischer, Z.X Shen, and Y. Cui, Rapid Surface Oxidation as a Source of Surface Degradation Factor for Bi2Se3, ACS Nano, 5, 6, 4698 (2011).DOI
- [41]H. Bando, K. Koizumi, Y. Oikawa, K. Daikohara, V.A. Kulbachinskii, and H. Ozaki, The time-dependent process of oxidation of the surface of Bi2Te3 studied by x-ray photoelectron spectroscopy, J. Phys. Condens. Matter 12, 5607 (2000).DOI
- [42]N. Bhattacharjee, K. Mahalingam, A. Fedorko, V. Lauter, M. Matzelle, B. Singh, A. Grutter, A. Will-Cole, M. Page, M. McConney, R. Markiewicz, A. Bansil, D. Heiman, N.X. Sun, Topological Antiferromagnetic Van der Waals Phase in Topological Insulator/Ferromagnet Heterostructures Synthesized by a CMOS-Compatible Sputtering Technique, Adv. Mat. 34, 15, 2108790 (2022).DOI
- [43]N. Bhattacharjee, K. Mahalingam, A. Fedorko, A. Will-Cole, J. Ryu, M. Page, M. McConney, H. Fang, N.X. Sun, Effects of Crystalline Disorder on Interfacial and Magnetic Properties of Sputtered Topological Insulator/Ferromagnet Heterostructures, arXiv:2205.09913v1 [cond-mat.mtrl-sci] (2022).DOI
- [45]H.Y. Lee, S. Kim, J.Y. Park, Y.W. Oh, S.Y. Park, W. Ham, Y. Kotani, T. Nakamura, M. Suzuki, T. Ono, K.J. Lee, and B.G. Park, Enhanced spin-orbit torque via interface engineering in Pt/CoFeB/MgO heterostructures, APL Materials 7, 031110 (2019).arxiv.org/abs/1803.06961
- [53]Y. Fan, P. Upadhyaya, X. Kou, M. Lang, S. Takei, Z. Wang, J. Tang, L. He, L.T. Chang, M. Montazeri, G. Yu, W. Jiang, T. Nie, R.N. Schwartz, Y. Tserkovnyak & K.L. Wang, Magnetization switching through giant spin-orbit torque in a magnetically doped topological insulator heterostructure, Nat. Mater. 13, 699–704 (2014).DOI
- [54]C. He, A. Navabi, Q. Shao, G. Yu, D. Wu, W. Zhu, C. Zheng, X. Li, Q. L. He, S. A. Razavi, K. L. Wong, Z. Zhang, P. Khalili Amiri, and K. L. Wang, Spin-torque ferromagnetic resonance measurements utilizing spin Hall magnetoresistance in W/Co40Fe40B20/MgO structures, Appl. Phys. Lett. 109, 202404 (2016).DOI
- [55]M. DC, R. Grassi, J. Y. Chen, M. Jamali, D. R. Hickey, D. Zhang, Z. Zhao, H. Li, P. Quarterman, Y. Lv, M. Li, A. Manchon, K. A. Mkhoyan, T. Low & J. P. Wang, Room-temperature high spin–orbit torque due to quantum confinement in sputtered BixSe(1−x) films, Nature Materials volume 17, pages800–807 (2018).DOI
- [56]M. Lim, H.W. Lee, Spin-memory loss induced by bulk spin-orbit coupling at ferromagnet/heavy-metal interfaces, Appl. Phys. Lett. 118, 042408 (2021).DOI