Giant spin torque efficiency in naturally oxidized polycrystalline TaAs thin films

The electrical manipulation of the spin degree of freedom in a material is a fundamental scientific aim that has a direct impact on nonvolatile, low dissipation magnetic memory applications. In this context, heavy metals and topological insulators have proven to be attractive platforms to study charge-spin interconversion due to the presence of significant spin Hall and Rashba-Edelstein effects [1, 2, 3, 4, 5, 6, 7, 8, 9]. More recently, Dirac and Weyl semimetals have started attracting interest for spintronics because of the potential for combining topological aspects of the band structure with increased electrical conductivity [10, 11, 12, 13, 14, 15, 16]. In particular, TaAs is an archetypal topological Weyl semimetal of contemporary interest [17] that has spin polarized surface states [18]. Although it is theoretically predicted to have a giant intrinsic spin Hall conductivity [19], this has yet to be demonstrated experimentally. Additionally, the role of natural oxidation has usually been neglected in topological materials, even though it has been shown that it can enhance the charge-spin interconversion efficiency in heavy metals and Dirac semimetals [20, 21, 13]. Several mechanisms have been proposed to explain this phenomenon, such as the presence of Rashba spin textures at the interface [22, 23, 24] or a theoretically predicted giant orbital Hall effect that can be further enhanced in oxide interfaces [25, 26].

In this paper, we report measurements of the charge-spin interconversion phenomenon in polycrystalline thin films of TaAs interfaced with a metallic ferromagnet ($\mathrm{Ni_{0.8}Fe_{0.2}}$) with and without natural surface oxidation. We show that at room temperature, the spin torque efficiency in oxidized TaAs films can be as large as $\xi_{\mathrm{FMR}}=0.45\pm 0.25$. We also present measurements of a completely in vacuo grown TaAs/$\mathrm{Ni_{80}Fe_{20}}$ heterostructure, obtaining an ST-FMR efficiency $\xi_{\mathrm{FMR}}=-0.27\pm 0.14$ with a spin Hall conductivity that agrees quantitatively with prior ab initio calculations for the Weyl semimetal phase of TaAs [19]. This is surprising since the films measured here are polycrystalline. Notably, the sign of the symmetric component of the ST-FMR spectra changes with oxidation indicating that the sign of the spin torque efficiency is opposite for pristine versus oxidized samples. We then present a systematic study of the spin torque efficiency in oxidized TaAs samples as a function of the ferromagnet layer thickness and extract a damping-like torque efficiency as high as $\xi_{\mathrm{DL}}=1.36\pm 0.66$. We also observe the presence of a significant field-like torque. Finally, we show that the sign change of the the spin torque efficiency also occurs in simple metal Ta/$\mathrm{Ni_{0.8}Fe_{0.2}}$ heterostructures when comparing oxidized and pristine interfaces.

We first discuss the synthesis of polycrystalline TaAs thin films using molecular beam epitaxy (MBE). TaAs crystallizes in a body-centered tetragonal structure ($\mathrm{I4_{1}md}$ space group) with a lattice constant of $a=0.3437$ nm and $c=1.1656$ nm [17, 18]. This makes it challenging to find an appropriate substrate for its thin film epitaxial growth given that most commonly available semiconductors do not have lattice constants in this range [27]. We note that MBE growth of thin films of the related Weyl semimetals, NbP and TaP, has been reported using a thin Nb buffer layer on (001) MgO [28]. We tried different substrates and find that GaAs (001) has the correct surface chemistry to nucleate TaAs growth. We carried out the synthesis in a VEECO 930 MBE chamber, monitoring the sample with reflection high energy electron diffraction (RHEED) at 12 keV. After desorbing the native oxide on an epiready semi-insulating (001) GaAs substrate, we grew 30 nm of GaAs at a thermocouple temperature of 720 $\mathrm{{}^{\circ}C}$ using Ga (5N) and As (5N) evaporated from different uncracked effusion cells. The Ga (As) beam equivalent pressure was $5\times 10^{-8}~{}(7\times 10^{-7})$ torr. We then cooled down the substrate to a thermocouple temperature of 400 $\mathrm{{}^{\circ}C}$ in the presence of As flux. At this point, we observed a RHEED pattern showing a $2\times 4$ GaAs surface reconstruction (Fig. 1 (a)). We then increased the temperature to 700-800 $\mathrm{{}^{\circ}C}$ and conjointly deposited As and Ta (the latter from a 4 pocket e-beam evaporator), obtaining the RHEED pattern shown in Fig. 1 (b) which displays characteristics of polycrystalline nature of the TaAs thin film (see Appendix A for details). We then transferred the TaAs films to an external electron beam evaporator for the deposition of $\mathrm{Ni_{0.8}Fe_{0.2}}$. Most of the samples studied here were transferred with a brief exposure to ambient, leading to an oxidized interface between TaAs and $\mathrm{Ni_{0.8}Fe_{0.2}}$. A few samples were transferred using a vacuum suitcase enabling the study of TaAs/$\mathrm{Ni_{0.8}Fe_{0.2}}$ heterostructures with a pristine interface. The complexity of the transfer procedure currently prevents us from doing a systematic study of unoxidized TaAs/$\mathrm{Ni_{0.8}Fe_{0.2}}$ heterostructures as a function of $\mathrm{Ni_{0.8}Fe_{0.2}}$ or TaAs layer thickness. The thickness of the $\mathrm{Ni_{0.8}Fe_{0.2}}$ was controlled by a quartz crystal monitor during growth, and the deposition rate of each layer was then confirmed ex situ using X-ray reflectivity measurements. We also grew simple metal Ta/$\mathrm{Ni_{0.8}Fe_{0.2}}$ heterostructures with and without natural oxidation on $\mathrm{SiO_{2}}$ substrates using electron beam evaporation for further spin torque analysis. All samples used in ST-FMR experiments were capped with 3 nm of Al to prevent oxidation of the $\mathrm{Ni_{0.8}Fe_{0.2}}$ layer. Our past studies show self-limited oxidation of such Al layers [13].

We studied the structural quality and crystalline phase of the TaAs films using X-ray diffraction. Figure 1 (c) shows a $\mathrm{2\theta-\omega}$ scan of a TaAs film, revealing (004), (008) and (112) peaks corresponding to the body centered tetragonal phase of TaAs. Measurements of the rocking curve around the (004) peak show a full width half maximum angle of 0.73^∘. A reciprocal space map around the TaAs (004) peak displays isotropic broadening, consistent with the polycrystalline nature of the film (Fig. 1 (d)). Figure 1 (e) shows the surface morphology of the film measured using atomic force microscopy (AFM). We observe several grains with a size of up to 100 nm $\times$ 50 nm. Height profiles of the AFM data in the vertical direction (Fig. 1 (f)) show 1 nm height variations between different grains that match the lattice constant of TaAs along the c-axis [18, 17]. We also used electrical measurements in 10 nm and 30 nm thick TaAs films using a 1 mm x 0.5 mm Hall bar configuration and determined a resistivity in the range $\mathrm{124\ \mu\Omega\cdot cm}$ to $\mathrm{318\ \mu\Omega\cdot cm}$. These values are 2 to 5 times larger than the values reported in bulk crystals of TaAs [29]. Magnetoresistance measurements at low temperature ($T=2$ K) show signs of weak localization and Hall effect measurements yield carrier densities on the order of $10^{21}-10^{22}\ \mathrm{cm}^{-3}$ (see Appendix).

To obtain a deeper understanding of the microscopic structure of the TaAs thin films, we performed high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM). Figure 2(a) shows a lower magnification image of the film cross-section having neighboring crystalline regions of different orientations, corresponding to the grains of TaAs about 5-10 nm in size. This can be better appreciated by taking the Fourier transform of smaller sections of the film showing unique crystal orientations, in order to identify the grain boundary regions as shown in Fig. 2(b). The polycrystalline nature of the TaAs film grown can possibly result from the large lattice mismatch between TaAs and GaAs, $17.5\%$ in the $\langle 100\rangle$ direction and $16.5\%$ in the $\langle 110\rangle$ direction. To obtain the chemical composition information from the film, we used energy dispersive X-ray (EDX) spectroscopy in the TEM measurements (Fig. 2(c)). Elemental distribution across the heterostructure showed a 1:1 stoichiometry between Ta and As in the film region (Fig. 2(d)). Additionally, the 2 nm region of darker contrast seen in the HAADF images corresponds to the presence of an oxide layer of TaAsO_x, on the film surface also seen in the elemental maps. This is produced by the air exposure of the sample between the growth and its measurement (see Appendix F). This is further confirmed by X-ray spectroscopy from the surface of the film with take-off angles of $30^{\circ}$ and $80^{\circ}$ with respect to the sample surface plane. The spectra reveals the presence of Ta, As, Ta₂O₅ and AsO_x on its surface (see Appendix C).

We now discuss measurements of charge-spin conversion using ST-FMR, focusing on the naturally oxidized TaAs/$\mathrm{Ni_{0.8}Fe_{0.2}}$ heterostructures. We used standard lithography techniques to pattern the films into $\mathrm{50~{}\mu m\times 10~{}\mu m}$ bars and deposited Ti/Au contacts for electrical measurement (Fig. 3 (a)). We have measured a total of 16 devices including TaAs and Ta films with and without natural oxidation of its surface before the $\mathrm{Ni_{0.8}Fe_{0.2}}$ deposition. Amongst these, we include the data of one TaAs film with a pristine surface. To perform the ST-FMR experiments (Fig. 3 (b)), we apply a radio-frequency charge current ($J_{c}$) along the heterostructure, which produces a flow of angular momentum (spin current, $J_{s}$) in a direction perpendicular to the flow of charge via the spin Hall or Rashba-Edelstein effects. The spin accumulation in the TaAs/$\mathrm{Ni_{0.8}Fe_{0.2}}$ interface, after being absorbed into the ferromagnet, applies a torque on its magnetization causing it to precess. This phenomenon can then be measured as a rectified voltage across the device ($V_{\mathrm{mix}}$) due to the mixing of the applied current and the anisotropic magnetoresistance of the ferromagnet [2] (see Appendix D). This allows us to obtain the spectra shown in Figs. 3 (c) and 3 (d). In addition to the spin generated torques, the flow of charge also produces an Oersted field that interacts with the magnetization and produces an out-of-plane torque that can be used to quantify the spin torque efficiency [2, 3]. We fit the resonance spectrum with a symmetric and antisymmetric Lorentzian contribution and determine the spin torque efficiency, which is given by [2, 4, 13]:

Here, $\xi_{\mathrm{FMR}}$ is the spin torque efficiency, $S$ ($A$) is the amplitude of the symmetric (antisymmetric) Lorentzian fit, $e$ is the charge of the electron, $\hbar$ is the reduced Planck constant, $\mu_{0}$ is the permeability of free space, $M_{S}=560\ \mathrm{kA/m}$ is the saturation magnetization of $\mathrm{Ni_{0.8}Fe_{0.2}}$ measured in a superconducting quantum interference device magnetometer), $t_{\mathrm{{TaAs}}}$ is the thickness of the TaAs layer, $t_{\mathrm{{NiFe}}}$ is the thickness of the ferromagnet, $M_{\mathrm{Eff}}$ is the effective magnetization obtained from fitting the Kittel FMR equation to the spectra and $H_{\mathrm{{Res}}}$ is the magnitude of magnetic field at which the resonance happens.

If the symmetric signal is produced solely by spin torque (in-plane) while the antisymmetric signal is produced by a combination of spin torque, Oersted field and other current-induced effective field-like torques (out-of-plane), we can write Eq. 1 as [3]:

Here, $\xi_{\mathrm{DL}(\mathrm{FL})}=(2e/\hbar)(4\pi M_{S}t_{\mathrm{NiFe}})(H_{\mathrm{OP}(\mathrm{IP})}/J_{c})$ is the efficiency of the in-plane, damping-like (out-of-plane, field-like) torques, defined as the normalized ratio of the field $H_{\mathrm{OP}(\mathrm{IP})}$ generated in the out-of-plane (in-plane) directions to the charge current flowing in the device [3]. In systems where the current-induced effective field-like torque is negligible, the spin torque efficiency is equal to the damping-like torque efficiency ($\xi_{\mathrm{FMR}}=\xi_{\mathrm{DL}}$).

First, we focus on the naturally oxidized samples (abbreviated as $\mathrm{TaAsO_{x}}$ in the rest of this work). In Fig. 3 (c), we observe the spectra of a 10 nm oxidized $\mathrm{TaAs}$ film with an 8 nm $\mathrm{Ni_{0.8}Fe_{0.2}}$ layer. When we fit the data, we see a large symmetric component in the spectrum ($\mathrm{V_{DL}}$) that is larger than the antisymmetric one ($\mathrm{V_{FL}}$) and has the same sign as platinum control samples (not shown). We extract the spin torque efficiency of this heterostructure using Eq. 1 and obtain $\xi_{\mathrm{{FMR}}}=0.45\pm 0.25$. For further insight into the origin of this large spin torque efficiency, we grew a set of samples with fixed TaAs thickness (10 nm) and different $\mathrm{Ni_{0.8}Fe_{0.2}}$ thickness. We compute the spin torque efficiency of each sample and observe that it monotonically increases with the thickness of the ferromagnet (Fig. 3 (g)). This indicates the presence of a significant current-induced positive field-like torque [30] and a giant damping-like torque with an efficiency higher than our largest experimental value (see Appendix E). To quantify this phenomenon, we perform a linear fit of the inverse spin torque efficiency as a function of the inverse ferromagnet thickness considering a 2 nm dead layer in the ferromagnet and use Eq. 2 to determine the damping-like and field-like torque efficiencies. These values can be as large as $\xi_{\mathrm{{DL}}}=1.36\pm 0.66$ and $\xi_{\mathrm{{FL}}}=0.30\pm 0.15$ (see Appendix E). At present, we cannot determine the spin Hall conductivity of oxidized TaAs due to the difficulty of precisely determining the electrical conductivity of $\mathrm{TaAsO_{x}}$. Nevertheless, the value of the damping-like and field-like torque efficiency is comparable to that measured in topological insulators ($\mathrm{Bi_{2}Se_{3}}$) [31, 4, 5] and higher than heavy metals (Pt, Ta)[3, 32, 33], Weyl semimetals ($\mathrm{WTe_{2}}$) [34], Dirac semimetals ($\mathrm{Cd_{3}As_{2}}$)[13] and Dirac nodal line semimetals ($\mathrm{IrO_{2}}$) [32].

Next we focus on the pristine (unoxidized) TaAs/$\mathrm{Ni_{0.8}Fe_{0.2}}$ heterostructure. Figure 3 (f) shows that the symmetric component of the spectra changes sign with respect to the one taken in oxidized samples (Fig. 3 (e)). We follow the same procedure described above and obtain the spin torque efficiency of non oxidized TaAs $\xi_{\mathrm{{FMR}}}=-0.27\pm 0.07$. The sign of the spin torque efficiency is opposite to naturally oxidized TaAs and Pt control samples (see Appendix D). Assuming the effective field-like torque is negligible in the pristine TaAs sample, this sets a lower bound on the spin Hall angle ($\theta_{SH}$) due to the less than ideal spin transparency of the $\mathrm{Ni_{0.8}Fe_{0.2}}$ interface ($\theta_{SH}\geq\xi_{\mathrm{{FMR}}}$) [3]. The spin torque efficiency can then be combined with our electrical conductivity measurements to obtain the spin Hall conductivity of TaAs. We find that in TaAs the lowest value of electrical conductivity measured in a Hall bar configuration is ($\sigma_{xx}=3144\ S\cdot cm^{-1}$) which implies that $|\sigma_{SH}|=\frac{\hbar}{2e}|\theta_{SH}|\sigma_{xx}\gtrsim 424\pm 110\ \frac{\hbar}{e}\frac{\mathrm{S}}{\mathrm{cm}}$. This experimental value agrees with the theoretical prediction of a negative giant spin Hall conductivity in TaAs and quantitatively agrees with its value computed from first principles [19].

To study the effect of the oxide on the giant spin torque efficiency measured in TaAs films, we measured the ST-FMR spectra of simple metal Ta/$\mathrm{Ni_{0.8}Fe_{0.2}}$ heterostructures evaporated on $\mathrm{SiO_{2}}$ with and without natural oxidation. We kept the Ta layer fixed at 6 nm, varied the $\mathrm{Ni_{0.8}Fe_{0.2}}$ thickness, and computed the spin torque efficiency of each device (Fig. 3 (f)). Even though there was variation among samples due to the lack of precise control over the oxide layer, we find that oxidation st the interface changes the sign of the spin torque efficiency: it is positive in naturally oxidized Ta ($\mathrm{TaO_{x}/Ni_{0.8}Fe_{0.2}}$) ($\xi_{\mathrm{{FMR}}}=0.037\pm 0.030$) while it is negative in non-oxidized Ta/$\mathrm{Ni_{0.8}Fe_{0.2}}$ heterostrucures ($\xi_{\mathrm{{FMR}}}=-0.032\pm 0.009$). These values are one order of magnitude smaller than the ones measured in TaAs; thus, the oxide itself cannot alone explain the large spin torque efficiency measured in TaAs. Nevertheless, this comparison with control Ta layers show once again our qualitative finding that the presence of oxide can change the sign of the symmetric component in the ST-FMR spectrum.

In conclusion, we demonstrated that it is possible to deposit thin films of TaAs on GaAs in a way that is compatible with standard III-V semiconductor processing. This allows us to experimentally measure a giant spin torque efficiency in naturally oxidized and pristine polycrystalline TaAs/$\mathrm{Ni_{0.8}Fe_{0.2}}$ heterostructures. We observe the presence of large damping-like and field-like torques with efficiencies that are greater in magnitude than other materials of contemporary interest. [31, 4, 5, 3, 32, 33, 34, 13, 32]. The extracted spin Hall conductivity of pristine TaAs is in remarkable agreement with theoretical predictions [19]. We show that the presence of an oxide produces a positive damping-like torque efficiency that changes the sign of the symmetric component of the ST-FMR spectrum. Several mechanisms have been identified as possible ways to produce spin torques due to oxidation, including enhanced spin Hall and Rashba-Edelstein effects [20, 13], material-dependent electron distribution near interfaces [21], and an oxide-enhanced giant orbital Hall effect in transition metals [25, 26]. The polycrystalline nature of our films may possibly also enhance the spin torque efficiency due to quantum confinement effects, akin to observations in topological insulators [35], but this is only speculation at present. The topological Weyl semimetal nature of TaAs with spin polarized surface states also likely plays a key role in enhancing the intrinsic spin Hall conductivity [19]. Determining and separating each of these contributions requires a higher control of the heterostructure that goes beyond the scope of this work. Nevertheless, our observation of a sign change in the ST-FMR spectra in oxidized samples is an easy and immediate way to discriminate between intrinsic effects and effects due to oxidation. The large spin torque efficiencies demonstrated in TaAs that can be further enhanced in the presence of oxide has immediate impact for fundamental research on spin-related phenomena in topological materials and in technological spintronic applications.