Spin-current probe for phase transition in an insulator

Abstract

Spin fluctuation and transition have always been one of the central topics of magnetism and condensed matter science. Experimentally, the spin fluctuation is found transcribed onto scattering intensity in the neutron-scattering process, which is represented by dynamical magnetic susceptibility and maximized at phase transitions. Importantly, a neutron carries spin without electric charge, and therefore it can bring spin into a sample without being disturbed by electric energy. However, large facilities such as a nuclear reactor are necessary. Here we show that spin pumping, frequently used in nanoscale spintronic devices, provides a desktop microprobe for spin transition; spin current is a flux of spin without an electric charge and its transport reflects spin excitation. We demonstrate detection of antiferromagnetic transition in ultra-thin CoO films via frequency-dependent spin-current transmission measurements, which provides a versatile probe for phase transition in an electric manner in minute devices.

Whilst neutron scattering is a powerful tool for studying spin fluctuations in materials, its availability is limited to large-scale user facilities. Here, the authors demonstrate how the pumping of pure spin currents can be used as a desktop probe to detect an antiferromagnetic transition.

A spin current refers to a flow of spin angular momentum of electrons in condensed matter¹. There are types of spin currents, including a spin current carried by conduction electrons and one carried by spin waves¹. The former type of spin current, conduction-electron spin current, can be detected by using the inverse spin Hall effect (ISHE)²,³,⁴,⁵,⁶,⁷, the conversion of a spin current into electric voltage via the spin–orbit interaction in a conductor, typically in Pt. Although conduction-electron spin current can reside only in metals and semiconductors, the latter type of spin current, called spin-wave spin current, can exist even in insulators.⁴,⁸ In fact, spin-wave spin currents have been studied in magnetic insulators and, very recently, in antiferromagnetic alloys and insulators⁹,¹⁰,¹¹,¹²,¹³,¹⁴,¹⁵.

For spin-current generation, one of the most versatile and powerful methods is spin pumping, an induction of spin current from magnetization precession in a magnetic metal into an attached metal via the exchange interaction at the interface⁵. Spin pumping was found to drive spin current also from magnetic insulators, such as Y₃Fe₅O₁₂ (refs ^{4, 6, 16, 17, 18, 19, 20, 21, 22}).

Various powerful theories have been constructed to describe the spin-pumping phenomenon²³,²⁴,²⁵. They commonly predict that the efficiency of spin pumping is sensitive to dynamical magnetic susceptibility at the interface between a magnet and a metal in a spin-pumping system. Therefore, spin pumping is sensitive to interface magnetic susceptibility, while standard magnetometry probes bulk properties which often hides interface signals. This raises an interesting hypothesis: when a very thin sample film is inserted at the interface of a spin-pumping system, spin pumping may reflect the dynamical susceptibility, which is directly related to spin fluctuation according to the fluctuation-dissipation theory, of the inserted thin sample film. Here we show that this is the case by using an antiferromagnetic transition in an ultra-thin film of CoO, and that spin pumping becomes an in situ microprobe for magnetic phase transition.

Results

Sample description

Figure 1c is a schematic illustration of the sample system used in the present study; we inserted an antiferromagnetic CoO thin film at the interface between Y₃Fe₅O₁₂ and Pt layers in a typical spin-pumping system Y₃Fe₅O₁₂/Pt to form Y₃Fe₅O₁₂/CoO/Pt. At low temperatures, CoO exhibits antiferromagnetic order²⁶. Here, Y₃Fe₅O₁₂ is a typical spin-pumping material, by which spin current is emitted when magnetization precession is excited⁴,⁶. Pt is used as a spin-current detector based on ISHE, in which a spin current is converted into an electric voltage in Pt perpendicular to the spin-current spin polarization direction²,³,⁴,⁶. When magnetization precession in Y₃Fe₅O₁₂ is excited by a microwave application, spin pumping is driven and then a spin current is injected from the Y₃Fe₅O₁₂ layer into the Pt layer across the thin CoO layer²³,²⁴,²⁵.

Figure 1. Concept and sample set-up.

(a) Inelastic scattering of polarized neutrons through an antiferromagnetic system. (b) Spin-current transmission through an antiferromagnetic system. (c) Experimental set-up of the spin-pumping measurement for the Y₃Fe₅O₁₂/CoO/Pt trilayer device. J_s denotes spin current injected from the Y₃Fe₅O₁₂ layer into the Pt layer through the CoO layer by spin pumping, which is detected as a voltage signal via the inverse spin Hall effect in the Pt layer. (d) A cross-sectional TEM image of a Y₃Fe₅O₁₂/CoO/Pt trilayer device. Scale bar, 10 nm. (e). A Co 2p XPS spectrum and Gaussian fitting analysis for the CoO layer in the Y₃Fe₅O₁₂/CoO/Pt trilayer device. TEM, transmission electron microscopy; XPS, X-ray photoemission spectroscopy.

Spin-pumping signal of Y₃Fe₅O₁₂/CoO/Pt system

Figure 2a shows the magnetic field dependence of microwave absorption spectra of Y₃Fe₅O₁₂ at various temperatures measured when a 5 GHz microwave is applied. At T=300 K, absorption peaks appear around H_FMR=±1.2 KOe, which correspond to ferromagnetic resonance (FMR) in the Y₃Fe₅O₁₂. With decreasing temperature, H_FMR is observed to slightly decrease, which is due to the temperature dependence of magnetization in the Y₃Fe₅O₁₂. The microwave absorption power P_ab at H_FMR is almost constant with changing T.

Figure 2. Spin-pumping detection of antiferromagnetic transition.

(a) Magnetic field (H) dependence of microwave absorption power (P) for a Y₃Fe₅O₁₂ film (3 μm in thickness) at various temperatures. P_ab denotes absorption power at FMR field. (b) Magnetic field (H) dependence of electric voltage (V) generated in the Y₃Fe₅O₁₂ (3 μm)/Pt (10 nm) bilayer film at various temperatures. (c) Magnetic field (H) dependence of electric voltage (V) generated in the Y₃Fe₅O₁₂ (3 μm)/CoO (6 nm)/Pt (10 nm) trilayer film at various temperatures. V_ISHE denotes the voltage signal at the FMR field. (d) Temperature dependence of V_ISHE for the Y₃Fe₅O₁₂ (6 μm)/Pt (10 nm) bilayer film. (e) Temperature dependence of V_ISHE for the Y₃Fe₅O₁₂ (3 μm)/CoO (6 nm)/Pt (10 nm) trilayer film. The inset shows the theoretical prediction of the spin conductance versus temperature in an antiferromagnetic system with S=1/2 (ref. ³⁸). Coefficient D denotes the spin conductance at a given frequency scaled by its maximum value D_max. (f) Temperature dependence of the XMLD signal ΔR_L3 for the Y₃Fe₅O₁₂ (3 μm)/CoO (6 nm)/Pt (1 nm) trilayer film (details are shown in Supplementary Note 1). The error bars in d–f represent the s.d. of multiple measurements at the same condition. XMLD, X-ray magnetic linear dichroism.

In Fig. 2b, we show the voltage V generated in the Pt layer in a simple Y₃Fe₅O₁₂/Pt spin-pumping system without a CoO layer measured by applying a 5 GHz microwave. At the FMR field H_FMR, a clear voltage peak appears at all temperatures. The sign of the peak voltage V_ISHE is reversed by reversing the polarity of the applied magnetic field, showing that the voltage peak is due to ISHE induced by spin current pumped from the Y₃Fe₅O₁₂ layer²,³,⁴,²³,²⁴,²⁵.

Figure 2d shows the temperature dependence of the peak voltage V_ISHE for the Y₃Fe₅O₁₂/Pt film without a CoO layer. V_ISHE decreases monotonically with decreasing the temperature. This monotonic decrease is attributed to the decrease in the resistivity of Pt and the increase in the magnetization damping in Y₃Fe₅O₁₂.

On the contrary by inserting a CoO layer in the simple spin-pumping system, a clear unconventional peak structure appears in the temperature dependence of the ISHE signal (Fig. 2c). Figure 2e shows the temperature dependence of the voltage peak intensity V_ISHE for the Y₃Fe₅O₁₂/CoO/Pt trilayer film. The V_ISHE for the Y₃Fe₅O₁₂/CoO/Pt trilayer film exhibits a clear peak at T=200 K, which is quite different from that for the Y₃Fe₅O₁₂/Pt bilayer film (Fig. 2d). The V_ISHE peak temperature is comparable to the Néel temperature of the CoO layer determined by an X-ray magnetic linear dichroism measurement using a synchrotron facility as shown in Fig. 2f (for detailed data, please see Supplementary Note 1 and Supplementary Figs 1 and 2). Furthermore, the temperature dependence of V_ISHE is similar to that of the magnetic susceptibility in a bulk CoO, in which the susceptibility is maximized around the Néel temperature T_N (refs ^{27, 28}).

Spin pumping with different CoO layer thickness

Figure 3a shows the temperature dependence of V_ISHE measured for various thicknesses of the CoO layer in Y₃Fe₅O₁₂/CoO/Pt trilayer films. In CoO films, the Néel temperature is known to decrease with decreasing the thickness of the film due to the finite size effect²⁹,³⁰. The observed CoO-thickness dependence of the peak temperature is consistent with this feature: the peak temperature decreases with decreasing the CoO layer thickness. All the results show that the V_ISHE peak position indicates the Néel temperature of the CoO layer, and that the V_ISHE enhancement around the antiferromagnetic transition can be related to the CoO-film susceptibility enhancement, which is a good measure for spin fluctuations.

Figure 3. Temperature dependence of spin-pumping signals in different systems.

(a) Temperature dependence of V_ISHE for Y₃Fe₅O₁₂/CoO/Pt trilayer films with different CoO-layer thicknesses (d_CoO=3, 6 and 10 nm). (b) Temperature dependence of V_ISHE for a Y₃Fe₅O₁₂/NiO(1.5 nm)/Pt film. The dash lines in a,b denote the peak positions.

Spin-pumping signal of Y₃Fe₅O₁₂/NiO/Pt system

To check the universality of the phenomenon, we measured another antiferromagnet: a NiO film (1.5 nm) in a Y₃Fe₅O₁₂/NiO/Pt system and found a similar peak structure and temperature dependence of V_ISHE (Fig. 3b). The peak position is consistent with the previous study on the Néel temperature in ultra-thin NiO films³¹. Furthermore, when a Cu layer is inserted between the Y₃Fe₅O₁₂ and the CoO (NiO) layers, we observed similar peak structures in the T-dependent ISHE signal (Supplementary Note 2 and Supplementary Fig. 3). This result suggests that the direct exchange coupling between Y₃Fe₅O₁₂ and an antiferromagnet is not necessary for the V_ISHE enhancement around Néel temperatures. Although observing such a phase transition in a single ultra-thin film was impossible without using large synchrotron facilities and a special X-ray magnetic linear dichroism spectrometer³²,³³,³⁴, our present method provides a way to probe it by a table-top experiment.

Discussion

Our present study stimulates further investigations of not only the spin transport near magnetic phase transitions but also microscopic spin-transport properties in antiferromagnetic systems. In antiferromagnetic insulators, incoherent thermal magnons and coherent Néel-order parameter dynamics¹² are considered to be responsible for transporting spins. Our experimental results in Fig. 3, however, show that V_ISHE is strongly suppressed towards lower temperatures in both cases of Y₃Fe₅O₁₂/CoO/Pt and Y₃Fe₅O₁₂/NiO/Pt systems; V_ISHE at 10 K is much less than that at T_N. Also, we notice that this feature was confirmed by some recent studies³⁵,³⁶. These results indicate that the spins are transported dominantly by incoherent thermal magnons rather than coherent Néel dynamics. At high temperatures, thermal magnons continuously evolve into thermal spin fluctuations, which would transport spin current above Néel temperature.

Such thermal spin dynamics both below and above T_N are well described by a bosonic auxiliary particle method³⁷. Using this method, the spin conductivity in an antiferromagnetic insulator was shown to be maximized near its Neel temperature³⁸, exactly like our V_ISHE (Fig. 2e inset). Since V_ISHE measures spin moments transferred across magnetic insulators, its enhancement directly reflects that of the spin conductivity. The spin conductivity and the magnetic susceptibility are in principle different quantities. However, their temperature dependences are rather similar because both are dominated by spin excitations with zero momentum transfer. Therefore, V_ISHE in our experimental set-up is a good measure for the spin dynamics and transition.

Moreover, another significant remark on our result which requires further theoretical understanding is the frequency dependence of the spin-pumping behaviour. Figure 4a shows the microwave frequency f as well as the temperature T dependence of V_ISHE/P_ab for the Y₃Fe₅O₁₂/CoO/Pt trilayer film, where V_ISHE is normalized by the FMR microwave absorption P_ab. For all the frequencies, V_ISHE/P_ab shows peaks around the Néel temperature T_N∼130 K. As seen in Fig. 4c, V_ISHE/P_ab exhibits different f dependence at different temperatures, namely V_ISHE/P_ab strongly depends on f only near T_N∼130 K, but it weakly depends on f at temperatures far from T_N. Such a strong frequency dependence implies that the observed phenomena reflect dynamical properties. Similar frequency dependence was also observed in the Y₃Fe₅O₁₂/NiO/Pt trilayer film, but it was absent in the Y₃Fe₅O₁₂/Pt bilayer film without antiferromagnetic layers (Supplementary Note 3 and Supplementary Fig. 4), showing that the observed prominent frequency dependence is characteristic to the antiferromagnetic layers near T_N. This type of strong frequency dependence cannot be explained by the coherent Néel dynamics or simple thermal magnons alone. First, coherent precession of Y₃Fe₅O₁₂ can pump antiferromagnetic magnons at the Y₃Fe₅O₁₂/antiferromagnetic interface, in analogy to the conventional spin pumping at the ferromagnet/metal interfaces¹². Second, the FMR dynamics inside of Y₃Fe₅O₁₂ itself should pump thermal magnons, which can subsequently diffuse across the structure, inducing the ISHE signal. This latter process relies on the breaking of the SU(2) symmetry of magnetic dynamics³⁹, which would be progressively enhanced with the increased ellipticity of the coherent dynamics in Y₃Fe₅O₁₂. Increased ellipticity of the Kittel mode in Y₃Fe₅O₁₂ at lower frequencies can thus be responsible for the enhancement of the conversion of the coherent precession into thermal magnons, leading to the observed increase in the signal. Alternatively, the finite lifetime of thermal magnons could be a source of the observed frequency dependence as in the case of the frequency-dependent spin conductance³⁸. In principle, the magnon lifetime depends on temperature and other intrinsic and extrinsic effects. A detailed theoretical analysis of the interplay of the coherent and incoherent magnetic dynamics in our heterostructure is, however, beyond the scope of this work. Constructing a comprehensive theory for spin-current transport in magnetic heterostructures is an important outstanding task for the development of novel spintronics based on quantum magnets.

Figure 4. Frequency dependence of spin-pumping signals in Y₃Fe₅O₁₂/CoO/Pt.

(a) A pseudo-colour plot of V_ISHE/P_ab as a function of the temperature T and the microwave frequency f for the Y₃Fe₅O₁₂ (3 μm)/CoO (3 nm)/Pt (10 nm) trilayer film. (b) Temperature (T) dependence of V_ISHE/P_ab at various exciting microwave frequencies. (c) Exciting microwave frequency (f) dependence of V_ISHE/P_ab at various temperatures. We did not show the data for frequencies lower than 4 GHz because the magnetization precession can be modulated due to the three-magnon interaction when f<4 GHz (ref. ⁴⁰).

Methods

Preparation of Y₃Fe₅O₁₂/CoO/Pt samples

A 3 μm-thick single-crystalline Y₃Fe₅O₁₂ film was grown on a (111) Gd₃Ga₅O₁₂ wafer by a liquid phase epitaxy method at 1,203 K in a PbO-B₂O₃ based flux. All the samples were cut from a same wafer into 1.5 × 3 mm² in size. CoO films with differencit thicknesses were coated on the Y₃Fe₅O₁₂ film by an radio frequency magnetron sputtering method. All the CoO films were prepared at 1,073 K to restrain the formation of cobalt oxide with other valence states and to improve the crystallinity. Then, 10-nm-thick Pt films were put on the top of the CoO films with a Hall-bar structure by an radio frequency magnetron sputtering method.

Sample characterization

Crystallographic characterization for samples was carried out by a X-ray diffractometry and transmission electron microscopy. A transmission electron microscopy image shows that the Y₃Fe₅O₁₂ film is of a single-crystal structure, and CoO and Pt layers are nearly epitaxially grown on the Y₃Fe₅O₁₂ film (Fig. 1d). An X-ray photoelectron spectroscopy method was used to confirm the chemical valence of the CoO layer (Fig. 1e).

Spin-pump experimental set-up

The spin-pumping measurement was performed in a physical property measurement system (PPMS), Quantum Design, Inc. To excite FMR in the Y₃Fe₅O₁₂ layer, microwave was applied by using a coplanar waveguide. The voltage signal between the ends of the Pt layer was measured by using a lock-in amplifier. Temperature dependence measurement was carried out from 10 to 300 K, after cooling samples from room temperature to 10 K in a 5,000 Oe magnetic field.

Data availability

All relevant data are available from the corresponding author on request.

Additional information

How to cite this article: Qiu, Z. et al. Spin-current probe for phase transition in an insulator. Nat. Commun. 7:12670 doi: 10.1038/ncomms12670 (2016).

Supplementary Material

Supplementary Information

Supplementary Figures 1-4, Supplementary Notes 1-3 and Supplementary References