Introduction

The exploration of spintronics has been advanced towards the manipulation of a pure spin current without a charge current. Due to the attributes of the larger angular momentum and zero charge current, pure spin current with much reduced Joule heating and power consumption is beneficial for information transport and processing1,2,3. Spin current is typically generated at a nonmagnetic heavy metal (HM) with strong spin-orbit coupling and a magnetic materials bilayered structure by physical phenomena such as the spin Hall effect (SHE)4,5, spin pum** effect (SPE)6,7, spin Seebeck effect (SSE)8,9, and spin photovoltaic effect (SPVE)1,10.

SPVE has attracted attention due to its ability to be excited by localized light and its faster response speed. In previous research, spin-polarized carriers were generated in the GaAs layer through the irradiation of circularly polarized light, and pure spin current was generated at the Pt/GaAs interface11. Recently, relevant works show that pure spin current can also be achieved at the Pt/Y3Fe5O12 (YIG) interface using unpolarized light1. This SPVE can be generated by photo-excitation of carriers near the interface and exists for light in the visible range1,10,12. Another work indicates that light-thermally induced temperature gradients (i.e., SSE) can excite magnons in YIG. Furthermore, the long-wavelength light causes the Fe3+ ions in the octahedra FeO6 from a high-spin state to a low-spin state, which increases the magnetization of YIG and subsequently enhances the magnon diffusion length13. In addition, a solution that separates the interface temperature gradient and bulk temperature gradients has been provided to analyze their contributions to spin currents14. For the light-induced spin current, most previous studies have focused on the generation and detection of magnons in YIG. The origin of light-induced spin current in the YIG system is still controversial. Therefore, it is worth exploring whether this light-induced spin current comes from SPVE or SSE.

Additionally, the mechanism of SPVE, i.e., the generation and transport of photo-excited spin current, has not been fully elucidated. Theoretically, SPVE is regarded as the bulk photovoltaic effect and its physical mechanism is similar to the “shift current” in nonlinear optical effects where the broken inversion symmetry (P) of the lattice is required3,20. Figure 1b shows the XRD θ−2θ pattern of the YIG/Gd3Ga5O12 (GGG) bilayers. The peak of YIG indicates a [111] growth orientation on the GGG (111) substrates. Due to the similar crystal structures and close lattice constants between GGG substrate and YIG film, the lattice mismatch between GGG (a = 12.383 Å) and YIG (a = 12.376 Å) is calculated as only 0.057%20. The YIG layer can be nicely epitaxially grown on the GGG surface according to the cube-on-cube relationship21. According to the 2dsinθ = nλ, the interplane spacing d111 of YIG is obtained as 1.801 Å, which is larger than the value (1.786 Å) in the standard PDF card (No. 98-001-1579). Anomalously large film d111 interplane spacing clearly indicates rhombohedral distortion of the YIG crystalline lattice (afilm = bfilm = cfilm and rhombohedral angle α < 90°)22. The surface topography of GGG substrate and YIG film is obtained by the atomic force microscopy (AFM), which shown in Fig. 1c. The root-mean-square surface (RMS) roughness is 0.08 and 0.27 nm, respectively. It indicates that the pretreated substrate has an atomically stepped GGG surface and facilitates the growth of atomically flat YIG film. Figure 1d shows the normalized static magnetic characteristics for the typical YIG (180 nm) film with the magnetic field applied perpendicular and parallel to the film plane at 300 K. The in-plane magnetic hysteresis loop is relatively square with a coercive field of about 5 Oe. The slanted out-of-plane magnetic hysteresis loop exhibits a coercive field of about 50 Oe and a saturation magnetic field of around 2 kOe. These properties are consistent with previous works and indicate the excellent quality of our epitaxial YIG film2,21.

Fig. 1: Properties of YIG/GGG films.
figure 1

a Crystal structure of YIG. The blue, green, purple, and red spheres represent the Tet.-Fe3+, Oct.-Fe3+, Dod.-Y3+ and O2- ions, respectively. b Representative θ−2θ XRD pattern for a YIG (180 nm)/GGG film. c AFM images of the processed GGG substrate and a representative YIG (180 nm)/GGG heterostructure. d Normalized hysteresis loops measured with the magnetic field along the in-plane and out-of-plane at 300 K, respectively. Ms is the saturation magnetization.

Full size image

Measurement of Light-induced spin currents

Figure 2a shows a schematic illustration of the sample structure and experimental configuration to measure light-induced spin current. A Xenon lamp was used to illuminate the W/YIG structure from either the top (as shown in Fig. 2a) or the bottom. The magnetic field H = 1 kOe is applied along the y axis to magnetize the YIG film to saturation, and the VISHE is measured along the x axis in the top W layer. The detailed experimental setup for light-induced spin current measurement can be seen Fig. S1 in Supplementary Information. The spin transport process in the W/YIG is briefly as follows: The spin-wave spin current is generated by the light in the YIG via the SPVE. At the W/YIG interface, a spin accumulates and the precessional spin angular momentum in YIG is transferred to the W layer, generating a conduction-electron spin current (JS). Finally, JS in the top W is detected by the voltage via the inverse spin Hall effect (ISHE). The more details will be discussed in theoretical analysis section.

Fig. 2: Measurements of the SPVE for different thickness/magnetization configurations by illuminating the top for the W/YIG (180, 350, 880 nm) samples [see the inset of Fig. 2b].
figure 2

a Schematic of the spin photovoltage measurement as Xenon lamp is introduced from the top of the sample, the GGG substrate is not shown here for simplification. b shows the magnetic field dependence of VISHE detected. c The voltage VTotal as a function of time was measured at different magnetic field directions along the y axis, in response to light that is turned on at 60 s and then turned off at 90 s. d The time dependence of VISHE was extracted in Fig. 2c by subtracting thermoelectric signal. e Schematic diagram of data processing of SPVE. SE is the abbreviation of the Seebeck Effect, which is a magnetic-field independent thermoelectric signal. The blue and red lines represent the signals at - H and + H, respectively.

Full size image

As an electrical method, the ISHE produces an electrical voltage from a spin current via spin-orbit coupling, which has been widely used for detecting spin currents9,14,23. We measured the optically generated VISHE for different illumination/magnetic field configurations in W/YIG with the thicknesses for W fixed at 5 nm and for YIG (180, 350, 880 nm), where all samples are of the same dimension and distance between electrical contacts. Figure 2b shows the detected field-dependent VISHE in response to light illuminating the W side of the structure (top-light configuration). The sign of VISHE are reversed by reversing the direction of the magnetic field, confirming the magnetic nature of the SPVE. Figure 2c shows the time dependence of the VTotal for different directions of magnetic fields in response to light that is first turned on at 60 s and then turned off at 90 s. Under light illumination, the temperature of sample increases due to thermal radiation, leading to an increase in the voltage at both ends of the W electrode, which mainly originates from the thermoelectric effect1,24. For experiments on the origin of this thermoelectric signal, please see the section D in the Supplementary Information. The magnitude of the measured voltage varies with the reversal of the direction of the magnetic field, indicating that the signal contains the field-independent voltage produced by the thermoelectric effect and the field-dependent voltage produced by the SPVE.

To minimize the effect of the thermal signals by the temperature gradient, before measuring the variation of VISHE with the magnetic field (Fig. 2b), first wait for 5 min under illumination. Additionally, the thermoelectric signals generated during the collection of VISHE data over time can be removed by processing. Figure 2e illustrates the schematic diagram of data processing for SPVE. The heat signal processing method can be effectively removed by a symmetrization procedure, VISHE can be expressed:

$${V}_{{{{{{\rm{ISHE}}}}}}\pm }={V}_{{{{{{\rm{Total}}}}}}\pm }-\left({V}_{{{{{{\rm{Total}}}}}}+}+{V}_{{{{{{\rm{Total}}}}}}-}\right)/2,$$
(1)

where VISHE± stands for SPVE signal, VTotal represents the original data measured by a nanovoltmeter, and the ± symbol represents the direction of the magnetic field. The processing method can effectively subtract the thermoelectric signal (i.e., SE in Fig. 2e and its value equals (VTotal+ + VTotal-)/2. Figure 2d depicts the time dependence of VISHE subtracted from the thermoelectric signal for W (5 nm)/YIG (180, 350, 880 nm) heterojunctions. The VISHE quickly reaches the maximum at 60 s when the light is turned on, and the VISHE detected on YIG with different thicknesses (1.18, 1.54, and 2.00 μV) are close to the saturated values of 0.95, 1.40, and 2.01 μV obtained from the field-swee** measurements, respectively.

Figure 3 shows the optically generated VISHE in response to illuminating the YIG side for the same sample as described in Fig. 2. The magnetic field dependence of VISHE is displayed in Fig. 3a. The VISHE saturation value of bottom illumination (VbsISHE) is smaller than that of top illumination (VtsISHE) for different thicknesses of YIG, and the ratio VbsISHE/VtsISHE is about 0.51 ± 0.04. This is because the light-receiving area of the bottom-light configuration required to fix the sample is about 0.6 times that of the top-light configuration. In addition, a portion of the light energy is absorbed by the GGG substrate, causing a further attenuation of the energy density transmitted to the YIG/W interface. Figure 3b shows the dependence of VTotal on time at different magnetic field directions. The data are presented in the same behavior as in Fig. 2c. The time dependence of VISHE with the subtracted thermoelectric signal is shown in Fig. 3c. When the magnitude and direction of the magnetic field are fixed, the reversal of the direction of light illumination does not change the sign of the VISHE. Conversely, the SSE leads to a change in the sign of VISHE due to the reversal of the direction of the temperature gradient by changing the direction of light illumination. This indicates that the VISHE signal originates from the SPVE rather than the SSE.

Fig. 3: Measurements of the SPVE for different thickness/magnetization configurations by illuminating the bottom for the W/YIG (180, 350, 880 nm) samples [see the inset of Fig. 3a].
figure 3

a The magnetic field dependence of VISHE. b The voltage VTotal as a function of time were measured at different magnetic fields directions along the y axis, in response to light that was turned on at 60 s and then turned off at 90 s. c The time dependence of VISHE were extracted in Fig. 3b by subtracting the thermoelectric signal.

Full size image

Heat transfer simulation

To further eliminate possible SSE in this experiment, the temperature gradient (ΔzT) along z direction will be provided through heat transfer simulation. For the case of the shorter light penetration depth and the thicker sample, it is easier to generate light-induced temperature gradient13,14. Hence, the larger-spot light with short-wavelength (450 nm), which is the case of low light intensity, is used for heat transfer simulation on a W (5 nm)/YIG (880 nm) model. The simulation results are shown in Fig. 4. Figure 4a, b show the distribution of energy density when light is shed on the top and bottom of the W (5 nm)/YIG (880 nm) model, respectively. The results verify that the thicker YIG (880 nm) cannot be penetrated by the short-wavelength light and the energy will be distributed on the surface, when the top or bottom of model is irradiated by a large-spot light. However, there is no obvious difference in the temperature distribution under different directions of light illumination (see Fig. 4c, d). Hence, the ΔzT is uniform, for the top-light and bottom-light configuration (as shown in Fig. 4e, f). Figure 4g, h show the change of the ΔzT with the distance from the interface for top-light and bottom-light configuration, respectively. Here, the ΔzT is only 10−4 K/μm, and such a small magnitude order of ΔzT is insufficient to induce the significant SSE.

Fig. 4: Heat transfer simulation for W (5 nm)/YIG (880 nm) model.
figure 4

a, c, and e are the three-dimensional color maps of the energy density distribution, temperature distribution and the ΔzT, for the top-light configuration with a large-spot size, respectively. b, d, and f show the corresponding bottom-light simulation results. The ΔzT are plotted as a function of distance from the interface for (g) top-light and (h) bottom-light configurations, respectively.

Full size image

In our experiment, the large-spot Xenon lamp with low energy (i.e., low light intensity) is used, so the ΔzT inside the sample will be very small. After reversing the direction of illumination, no reversal VISHE is observed in different thicknesses of YIG samples. Thus, the light-induced spin current we observed in the experiment belongs to the photo-excited spin current (i.e., SPVE). This is consistent with these previous photo-excited spin current reports where a large-spot light is also employed1,10,12.

Note that light-induced spin current is also considered as a light-thermally excited spin current (i.e., SSE), and the smaller-spot laser with higher energy (i.e., high light intensity) is used, and the thickness of samples is generally in the range of micrometer13,14,25. Consequently, a large ΔzT and reversal of the ΔzT sign with a reversed illumination direction led to the flipped sign of VISHE. Similarly, if a high light intensity is used to illuminate our sample, the SSE rather than SPVE should be observed. To validate this hypothesis, we choose the 405 nm laser converged by a convex lens, which is a smaller-spot laser with higher energy, and utilize it to illuminate the W (5 nm)/YIG (880 nm) sample. As we anticipated, the reversal of illumination direction causes a reversal sign of the VISHE. This experimental observation confirms that the spin current observed under this condition is mainly related to SSE [the detailed results shown in Fig. S3a, b]. Moreover, the corresponding simulation for the smaller-spot laser with higher energy is also conducted, and the simulated results are shown in Fig. S3c–f. This simulated results also show that after reversing the illumination direction, the magnitude of ΔzT can reach up to 0.94 K/μm (≫10−4 K/μm for large-spot light with low-energy) and the sign of ΔzT also is reversed, further proving the above analysis.

Therefore, Both SPVE and SSE contribute to the generation of spin current under light illumination. However, SPVE dominates under low light intensity, whereas SSE plays the dominant role under high light intensity. The debate on SPVE and SSE in YIG system is resolved.

Theoretical analysis

Next, we will discuss the SPVE mechanism in YIG system. Currently, a shift current model is proposed to explain the physical mechanism of photo-excited spin currents. A brief introduction to this model can be found in the Supplementary Information. According to the shift current model, the shift vector is odd under P symmetry, that is, \({R}_{{nm}}^{a}\left(k\right)=-{R}_{{nm}}^{a}\left(-k\right)\). Therefore, a material with P symmetry, the integration in Eq. (S3) is zero, and thereby the shift current conductivity is equal to zero (\({\sigma }_{{bb}}^{a}=0\))3,13,18,26. As shown in Figs. 2 and 3, the SPVE significantly enhances with an increase in the thickness of the YIG film, regardless of illumination direction. These results indicate that the photo-excited spin current is not only generated at the W/YIG interface. As shown in Fig. 5c, the propagation process of the spin current excited by photons in W/YIG is divided into three parts: within the YIG layer (A), at the W/YIG interface (B), and within the W layer (C).

Firstly, the process of A shows the propagation of the spin waves in YIG. The study of the ground state of YIG can only provide insights into the static properties of the system, i.e., properties measured at 0 K. A magnetic field is applied along the y direction of YIG, thus the magnetic moments from uncompensated spins of the Tet. Fe3+ ions are aligned along the y direction. When light is applied in this scenario, an individual spin is perturbed by photon, deviating from its equilibrium state. Due to the interactions between this spin and its neighboring spins, the perturbation induced by photon propagates throughout the entire system and forms the spin waves in YIG.

Secondly, there is a conversion from spin waves to spin current in the process of B. The spin angular momentum of the precession in YIG is transferred to the W layer by the s-d interaction, which conserves total angular momentum17,27,28. The boundary condition at the W/YIG interface can be expressed as17:

$${j}_{m}\left({d}^{-}\right)=\left(-{\mu }_{{{{{{\rm{B}}}}}}}/e\right)\left[ \, {j}_{s}^{\uparrow }\left({d}^{+}\right)-{j}_{s}^{\downarrow }\left({d}^{+}\right)\right],$$
(2)

Where e is electron charge, μB is the Bohr magneton, jm and js denote the magnon current density of YIG and spin current density of W, respectively. The interface YIG/GGG is assigned at z = 0, and d corresponds to the thickness of YIG. The symbol - (+) represents the decrease (increase) in spin angular momentum on the YIG (W) side at the interface, and ↑ (↓) corresponds to the spin-up (down) component.

The physics of the Eq. (2) is fairly transparent: the left term describes the flow of magnons generated under the action of photons, and the right term describes the spin current generated by the accumulation of magnons. Thus, the magnon current at the interface will be fully converted into the conduction-electron spin current in the W layer. In fact, spin-flop scattering caused by impurities and defects present at the W/YIG interface will reduce the outgoing spin angular momentum4.

Finally, the spin angular momentum is transferred from spin precession to conduction-electron spins, forming a pure spin current JS in the W layer (the process of C). This spin current in the W is detected by the voltage via the ISHE. The process that converts JS into a charge current JC is described. Assuming the spin-polarization vector direction as σ, two electrons moving in the opposite direction of JS have opposite spins: one parallel to σ and the other anti-parallel to σ in the pure spin current in W layer. These two electrons are bent to the same direction and induce Jc perpendicular to Js via spin-orbit coupling. The relationship between Js and Jc is thus given by the following equation9,24,29:

$${J}_{{{{{{\rm{C}}}}}}}={\theta }_{{{{{{\rm{SH}}}}}}}\frac{2e}{{{\hslash }}}{J}_{{{{{{\rm{S}}}}}}}\times \sigma,$$
(3)

where θSH represents the spin Hall angle (which is used to characterize the efficiency of the ISHE, and θSH of W is negative) and ℏ is in a material reduced Planck’s constant.

VISHE obtained from the data detected by the nanovoltmeter is given by30:

$${V}_{{{{{{\rm{ISHE}}}}}}}=\frac{A}{{\sigma }_{{{{{{\rm{C}}}}}}}}{J}_{C}=\frac{A}{{\sigma }_{{{{{{\rm{C}}}}}}}}{\theta }_{{{{{{\rm{SH}}}}}}}\frac{2e}{{{\hslash }}}{J}_{{{{{{\rm{S}}}}}}}\times \sigma,$$
(4)

where A is a coefficient related to the thickness of the film and σC is the electrical conductivity.

Discussion

In summary, we have prepared W/YIG (111) bilayers heterostructure by PLD method. SPVE and optically induced SSE are experimentally observed in this heterostructure via different light sources. Combined with heat transfer simulation in W/YIG model, it is found that large-spot light with low-energy produces photo-excited spin currents (i.e., SPVE), whereas small-spot light with high-energy generates light-thermally excited spin currents (i.e., SSE). Under large-spot light with low-energy illumination, SPVE is predominant and SSE is omitted thanks to the negligible temperature gradient. By means of symmetry analysis and experimental results, the model of shift current for SPVE is excluded. Based on the conservation of total angular momentum and the ISHE, a more reasonable model for SPVE in this system is presented. Our finding deepens the understanding for the origin of light-induced spin current in heavy metal/magnetic insulator bilayers, and SPVE with ultrafast time response may enable the design for high-speed spin photovoltaic devices.

Methods

Sample preparation

The high-quality epitaxial W/YIG (111) bilayers were fabricated on (111)-oriented GGG substrate (5 × 3 mm2). The GGG (111) substrate was ultrasonically cleaned with acetone, isopropanol and deionized water, and then annealed for 6 h in air at 900 °C. Epitaxial YIG films with different thickness were deposited on (111)-oriented GGG substrates by pulsed-laser deposition (PLD) with a KrF excimer laser (λ = 248 nm). Under an oxygen pressure of 83 mTorr, films were grown at 750 °C with a laser energy density of 1-2 J/cm2 and a repetition rate of 6 Hz. Subsequently, the films were annealed at 375 Torr of oxygen pressure and slowly cooled down. Then the samples were capped by 5-nm-W spin current detectors using a bar-shaped mask (0.4 × 4.4 mm2) by magnetron sputtering with Ar pressure at 2 mTorr.

Sample characterization

The layer thicknesses were calibrated by x-ray reflectivity (XRR) and controlled by the deposition time. The crystal structure of YIG film was characterized by X-ray diffraction (XRD, Bruker D8 Discover, Cu-Kα radiation, λ = 1.54056 Å). The surface characteristics of YIG and GGG substrate were detected by the atomic force microscopy (AFM, Cypher, Asylum Research), and the RMS is used to describe the surface roughness. The magnetic hysteresis loops of the YIG thin film at 300 K was recorded by magnetic property measurement system (MPMS3, Quantum Design). The Seebeck coefficients were measured by a Hall-effect measurement system (Linseis, HCS 1) with the Seebeck coefficient characterization.

SPVE measurements

The SPVE measurements were carried out in a home-made measurement system. The sample was placed in a vacuum chamber to reduce the noise signal in the process of measurement, and a Xenon lamp (~300 mW/cm2) or laser was irradiated on the top or bottom of the sample. The magnetic field was provided by an electromagnet, and a nanovoltmeter (Keithley, 2182 A) was used to measure VISHE across the W length. All measurements were carried at room temperature, and the schematic diagram of the experimental setup and sample structure is shown in Fig. S1.

Heat transfer simulation model

By the finite element method in COMSOL multiphysics modeling, local and non-local heating processes in transparent media with non-uniform temperature gradients is simulated31. When the size of the heated object and the size of the laser spot are both much larger than the wavelength, the Beer-Lambert law can be used to simulate the light-induced thermal conduction process in transparent materials well.

According to the energy conservation, the temperature field can be expressed as:

$$\rho {C}_{{{{{{\rm{P}}}}}}}\frac{\partial T}{\partial t}-\nabla \cdot \left(k\nabla T\right)={Q}_{{{{{{\rm{r}}}}}}},$$
(5)

where Qr is heat source term, ρ is the mass density, CP is the heat capacity at constant pressure, and k is the thermal conductivity32,33.

The simplified Beer-Lambert Law can be expressed as:

$$\frac{{e}_{i}}{{||}{e}_{i}{||}}\cdot \nabla {I}_{i}=-\kappa {I}_{i}$$
(6)

where ei represents the direction of the light beam and I is radiative intensity. When a light beam passes through a medium, energy is absorbed, and the radiative heat source term can be expressed as:

$${Q}_{{{{{{\rm{r}}}}}}}={\sum}_{i}\kappa {I}_{i}$$
(7)

During the heat transfer processes, temperature T affects radiative intensity I through the absorption coefficient κ, and radiative intensity I affects temperature T through the radiative heat source Q. The more details were discussed in the section B of the Supplementary Information.

Reporting Summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.