Introduction

An exciting prospect stemming from the combination of superconductivity and spintronics is the appearance of a spin-polarized supercurrent, capable of transporting not only charge but also a net spin component without any dissipation. However, superconductivity and ferromagnetism are two antagonistic macroscopic orderings1,2, in which the superconductivity can be destroyed by a strong exchange effect unless the Cooper pairing is of the triplet nature3,4,5,6. To this end, a robust synergy between superconductivity and spintronics is enabled by inducing spin-triplet pair correlations at superconductor/ferromagnet interfaces7,8,9,10,11,12. One compelling experimental evidence of spin-triplet supercurrents was reported in a half-metallic ferromagnet of CrO2 between two spin-singlet superconductors of NbTiN13. Subsequently, a series of experiments reported more evidence of spin-triplet pairing in Josephson junctions with inhomogeneous metallic ferromagnets14,15,16 or metallic magnets17,18,19 as the barrier materials.

To date, the earlier efforts have been primarily focused on materials of bulk nature. More recently, the rising stars of layered or van der Waals (vdW) materials, such as the two-dimensional (2D) ferromagnets CrI320,21, Cr2Ge2Te622,23, Fe3GeTe2 (F)24,25, and superconductor NbSe2 (S)26,27, provide unprecedented opportunities to explore spin-triplet supercurrents. In particular, by engineering vdW S/F interfaces, 2D superconducting spintronics can be developed without the constraint of lattice matching28,29,30.

In this paper, we report the first construction of a lateral Josephson junction composed of a vdW metallic ferromagnet (Fe3GeTe2) laterally connected between two layered spin-singlet superconductors (NbSe2), and use the S/F/S architecture to sustain skin Josephson supercurrents with long-range nature. The observed supercurrent across the ferromagnet can extend over 300 nm, and exhibits double-slit quantum interference patterns in an external magnetic field. Strikingly, the supercurrent displays pronounced skin characteristics, as signified by its peaked densities at the surfaces or edges of the ferromagnet. The present study provides a new platform for facile generation of spin-triplet supercurrents in superconducting spintronics based on 2D materials.

Results

S/F/S lateral Josephson junction

Fe3GeTe2 is a van der Waals ferromagnetic metal with a Curie temperature of 200 K as measured in our experiment, and it has a strong perpendicular magnetic anisotropy with an easy magnetization direction along the c-axis. Its hexagonal crystal structure with a space group of P63/mmc is composed of a layered Fe3Ge substructure that is sandwiched between two layers of Te atoms (Fig. 1b). The Fe3Ge substructure is comprised of two types of Fe atoms: Fe I and Fe II. Fe I sites are fully occupied while Fe II sites are only partially occupied. A frustrated triangular structure is created by the Fe I atoms, resulting in spins of the Fe atoms forming a noncoplanar structure (Fig. 1b). This noncoplanar structure acts as a fictitious magnetic field due to the presence of a nontrivial Berry phase in real space31,32,33,34. NbSe2 is known to be a van der Waals superconductor with spin-singlet pairing correlation (Tc = 7.0 K).

Fig. 1: S/F/S device architecture and temperature-dependent resistivity.
figure 1

a Schematic illustration of a S/F/S lateral Josephson junction. b Schematic diagram of the atomic structure of Fe3GeTe2, where showing noncoplanar spin textures with a fictitious field of h in a frustrated triangular lattices. c Optical image of the S/F/S device. d Temperature dependence of the S/F/S resistance by a four-terminal measurement and the applied current (Iapp) is 10 μA. Three transitions are identified, with the first two Tc1 = 6.9 K from superconducting NbSe2, and Tc2~5.4 K from the proximity-induced superconducting transition of the S/F bilayers at each end of the S/F/S. The red solid line represents the BKT transition using the Halperin–Nelson equation for fitting, which gives the third transition with a BKT temperature TBKT = 3.7 K. Source data are provided as a Source Data file.

Full size image

Figure 1c shows an optical image of a lateral vdW Josephson junction of S/F/S prepared by a dry-transfer method (details presented in the Methods section). The width w and length Lj (edge-to-edge separation between NbSe2) of the junction channel are ~2 μm and 300 nm, respectively. The thicknesses of the NbSe2 and Fe3GeTe2 are 25 and 22 nm, respectively. The whole device is shielded from air contamination by a top layer of hBN. To study the electrical properties of the S/F/S, we first measure the temperature dependence of the four-probe resistance (R–T) at zero magnetic field (R = V/Iapp, Iapp = 10 μA). With decreasing temperature, the R–T result (Fig. 1d) shows an abrupt resistance drop towards zero with three transition temperatures, with the first two corresponding to the superconducting transition of NbSe2 (6.9 K) and the proximity-induced superconducting transition of the S/F bilayers at each end of the S/F/S (5.4 K). When the temperature is further lowered, the observed tail of the S/F/S resistance drop is well fitted by the Berezinskii–Kosterlitz–Thouless (BKT) model with a BKT temperature of 3.7 K, strongly indicating that the zero-resistance state is approached as vortex-antivortex pairs bind together35,36.

Long-range supercurrent with spin-triplet nature

The measured current–voltage curve of the S/F/S (Fig. 2a) shows a clear superconducting feature of a sizable Josephson critical current Ic as large as 100  μA at zero voltage, which nevertheless is much smaller than the critical current of pristine NbSe2 (Supplementary Fig. 1). The S/F/S’s I–V curve does not exhibit a strong hysteretic behavior, which indicates that the junction is in an overdamped regime with a low resistance-capacitance product17,37. When the applied current exceeds Ic, the I–V curve exhibits a linear behavior crossing the origin and displays a normal state resistance Rn of 16 Ω.

Fig. 2: Transport characteristics of the S/F/S devices.
figure 2

a Current–voltage (I–V) curves for S/F/S Josephson junction at temperatures of 3 and 8 K under zero magnetic field, showing the Ic~100 μA at 3 K, and the inset shows the magnified plot of I–V curve around zero voltage. b Critical current Ic as a function of temperature with different Lj of 260, 280, 300 nm. The solid lines are theoretical fitting curves. c Current–voltage (I–V) curves for Josephson junctions with different channel lengths at temperatures of 3 K under zero magnetic field. Lj-dependent characteristic voltage IcRn product of the junctions in the inset. d Critical current Ic as a function of the external in-plane magnetic field with forward and reverse directions of field sweep. The arrows represent the direction of the scanning magnetic field. Source data are provided as a Source Data file.

Full size image

To further study the Josephson supercurrent, we measure the critical current as a function of temperature in the S/F/S with different channel lengths Lj. The T-dependent critical current (Fig. 2b) can be fitted by \({I}_{{{{{{\rm{c}}}}}}}(T) \, \approx \, {I}_{{{{{{\rm{c}}}}}}}(0){(1-\frac{T}{{T}_{{{{{{\rm{c}}}}}}}})}^{\alpha }\), where \({I}_{{{{{{\rm{c}}}}}}}(0)\) is the critical current at zero-temperature, and Tc is the S/F/S superconducting transition temperature37. We can derive the zero-temperature critical currents\(\,{I}_{{{{{{\rm{c}}}}}}}(0)\) to be 126.03, 90.37, 64.73 μA for Lj = 260, 280, 300 nm by applying different values of α, respectively. Again, these critical currents are one or two orders of magnitude smaller than those in pristine NbSe2 (Supplementary Fig. 1). As the channel length Lj increases, the zero-temperature critical current \({I}_{{{{{{\rm{c}}}}}}}(0)\) shows a tendency to decay. This is attributed to the reduction of proximity-induced Cooper pairs in the longer Fe3GeTe2 barrier. When the channel length increased to 450 nm, the I–V curve of the device exhibits a linear behavior, indicating that the superconducting current disappears completely (Fig. 2c). Lj-dependent characteristic voltage IcRn product of the junctions is shown in the inset of Fig. 2c. The fitting of an exponential decay function \({{\exp }}(-\frac{{L}_{{{{{{\rm{j}}}}}}}}{\xi })\) shows an estimated coherence length of ξ = 227 ± 18 nm. However, by using \({{{{{{\rm{\xi }}}}}}}_{{{{{{\rm{singlet}}}}}}}^{{{{{{\rm{FM}}}}}}}\approx \sqrt{\frac{\hslash D}{2{E}_{{{{{{\rm{ex}}}}}}}}}\)8,9,38,39, where D and \({{E}}_{{{{{{\rm{ex}}}}}}}\) are, respectively, the Fe3GeTe2 electron diffusion coefficient and exchange energy, we obtain the coherence length of the spin-singlet supercurrent in the Fe3GeTe2 layer to be ~3.5 nm, much shorter than the Lj. Therefore, the long-range nature of the supercurrents across the ferromagnetic Fe3GeTe2 layer observed in the S/F/S might be related to the proximity-induced spin-triplet superconducting correlations. To further prove that the long-range Josephson supercurrents is directly related to the nature of Fe3GeTe2, we replace the Fe3GeTe2 layer with a ferromagnet metal of Fe0.25TaS2 as a barrier to contract NbSe2/Fe0.25TaS2/NbSe2 lateral junction with the channel length of 200 nm. There is no Josephson supercurrent observed in such NbSe2/Fe0.25TaS2/NbSe2 junction (Supplementary Fig. 2).

Skin characteristics in spin-triplet supercurrent

To further investigate the quantum interference of S/F/S supercurrent, we measure the response of Ic to an external in-plane magnetic field that is perpendicular to the supercurrent channel. Figure 2d shows a double-slit quantum interference behavior in the amplitude of Ic as the magnetic field is varied, which is different from the phenomenon in a self-stacking NbSe2 junction (Supplementary Fig. 3). When the magnetic field is along the z-axis (Fig. 3a), the magnetic-field dependence of the critical current Ic(B) exhibits a pure sinusoidal double-slit pattern (Fig. 3b), indicating that the supercurrent density along the y-axis is characterized by only two conductive edge-channels, which exhibit significant depletion of bulk carriers. The full widths at half maxima of these two edge-channels are ~1.80 nm and 1.33 nm, as determined by Gaussian fitting (Fig. 3c). When the magnetic field is along the y-axis direction (Fig. 3d), the Ic(B) exhibits a mixed single-slit Fraunhofer and double-slit interference pattern (Fig. 3e). The supercurrent density along the z-axis (extracted from Fig. 3e) is relatively uniform, but with two pronounced peaks (Fig. 3f), indicating surface-dominated transport along the two edges of the Fe3GeTe2 strip.

Fig. 3: Skin effects in the supercurrent density of S/F/S Josephson junctions.
figure 3

a Schematic of the S/F/S with the magnetic field along the z-axis direction. b Differential resistance map across the junction at 3 K, showing a double-slit interference pattern. c Distribution of supercurrent density along the y-axis obtained from the inverse Fourier transform of the data in b. The Fe3GeTe2 layer thickness is 22 nm. d Schematic of the S/F/S junction with the magnetic field along the y-axis direction. e Differential resistance map across the junction at 3 K, showing a mixed single-slit and double-slit interference pattern. f Distribution of supercurrent density along the z-axis obtained from the inverse Fourier transform of the data in e, showing that the supercurrent can flow relatively more uniformly across the junction along the x-axis, but still peaked at the surfaces (now the two edges). The channel width w is 2 μm. Source data are provided as a Source Data file.

Full size image

In addition, the quantum oscillation of Ic(B) at both the out-of-plane and in-plane magnetic fields shows a clear hysteresis feature with an offset at the first maximum of the supercurrent from zero magnetic field, caused by the spontaneous magnetic flux from Fe3GeTe2 finite magnetization13,40. Typically, the total enclosed magnetic flux within a Josephson junction is the result of both the external flux generated by the magnetic field and an intrinsic flux induced by the magnetization of the barrier13,40,41, which means that the offset is dependent on the barrier moment. Therefore, when we measure the Ic(B) with forward and reverse directions of field scanning, the Fe3GeTe2 magnetization loop induces this offset from zero magnetic field. The Ic(B) at the in-plane magnetic field has a larger offset than at the out-of-plane magnetic field.

Discussion

Before closing, we conjecture on the possible mechanism for the observed long-range spin-triplet supercurrent. The notable aspect of the present work is that this long-range Josephson supercurrent exhibits a striking skin feature that is distinct from conventional bulk channels. We propose two possible physical mechanisms in the present system that might be responsible for this skin feature. First, we argue that both top and bottom surface of the Fe3GeTe2 layer has a mirror symmetry breaking, which can induce a Rashba spin-orbit coupling. In principle, the interplay of this Rashba spin-orbit coupling, the ferromagnetism of the bulk Fe3GeTe2, and the s-wave superconductivity of NbSe2 may induce a formation of 2D topological superconductivity on the surface of the Fe3GeTe2. In this case, the top and bottom surfaces of the Fe3GeTe2 could support an effective chiral p-wave topological superconductivity that was first proposed by Fu and Kane42. Of course this topological state is only a theoretical speculation that is consistent with interference patterns and symmetry restrictions. The experimental identification of topological superconductivity in the present system is certainly an exciting direction for further studies.

The second possible mechanism involves the noncoplanar structure (Fig. 1b) of the Fe atoms, which was predicted to produce a fictitious magnetic field. When a conventional s-wave superconductor is in contact with magnetic inhomogeneity, inhomogeneous magnetizations facilitates the conversion of spin-singlet Cooper pairs into spin-triplet pairs via spin-mixing and spin-rotation processes9,11,43. In the microscopic processes of entering the interface region to Fe3GeTe2, the electrons of singlet Cooper pairs of NbSe2 experience spin-dependent interface phase shifts44,45, then only a mixture of spin-singlet and spin-triplet (Sz = 0) Cooper pairs can be present46 (named spin-mixing). Thus, the noncoplanar spin textures of Fe3GeTe2 within frustrated triangular lattices31 have the ability to orient the spin-quantization axis to align with the local magnetization direction. Then, this can flip one of the spins in the triplet component (Sz = 0)45. As a result, the spin-triplet pairs (Sz = ±1), consisting of the same spin pairs with equal amplitude and opposite sign, are produced with a long decay length46, which is insensitive to the pair-breaking effect from the Zeeman field, but limited by a thermal coherence length23. However, noncoplanar spins formed in a system of triangular iron atoms have only been proposed in theory47, while no direct experimental evidence has yet been reported about the observation of this noncoplanar structure in the bulk Fe3GeTe2. Then experimental verification of the noncoplanar structure in current Fe3GeTe2 system also requires extensive further study. In conclusion, whether the long-range superconducting current has a spin-triplet state and the physical mechanism of the skin behavior of the superconducting current require more in-depth theory and experiments in the future.

In summary, we have experimentally demonstrated that a long-range supercurrent is generated at van der Waals interfaces in S/F/S lateral Josephson junctions. The decay length of the proximity-induced Cooper pairs with the same spin across the ferromagnet Fe3GeTe2 can be as long as 300 nm. More strikingly, the supercurrent density distribution profile provides the first evidence of skin feature of long-range spin-triplet superconductivity. Our work provides a new platform with superior control of interface properties for future exploration of novel physical properties and potential device applications of 2D superconducting spintronics.

Methods

Single-crystal growth

The Fe3GeTe2 single crystals were grown by a chemical vapor transport (CVT) method. The details of the growth are specified in our previous work48. The NbSe2 single crystals were also obtained by the CVT method. The synthesis process for the single crystal is shown in Supplementary Fig. 4. High-purity Niobium and Selenium (molar ratio Nb:Se = 1:2) in stoichiometric amounts were mixed with iodine (4 mg/cm2) as a transport agent before being sealed into an evacuated quartz tube. We raise the temperatures of the source and growth zones in the two-zone tubular furnace to 850 °C and 750 °C, respectively, at a rate of 5 °C/min and maintained at 850/750 °C for 10 days. Then the furnace was left to cool naturally to room temperature after the growth process, and the resulting shiny crystals were collected.

Characterization techniques

The crystal structures of Fe3GeTe2 and NbSe2 were characterized by X-ray diffraction. The magnetic properties of the single crystals were measured by a SQUID (MPMS3), and the thickness of the exfoliated samples was measured by an atomic force microscope.

Electrode processing

Electrodes with different geometries were patterned on a 300-nm-thick SiO2/Si substrate by ultraviolet lithography with Ti/Au (8/25 nm) metal deposited on the surface of substrate through magnetron sputtering. After the lift-off process, electrodes with different geometries were obtained for heterostructure transfer process. In this work, parallel electrodes with a spacing of 30 μm were used.

Heterostructure assembly by a dry-transfer technique

Mechanical exfoliation and dry-transfer processes were carried out within a glove box that maintained highly controlled levels of O2 and H2O (< 0.01 ppm) in order to minimize their presence. The fabrication process is shown in Supplementary Fig. 5. Firstly, we transferred the desired NbSe2 nanoflake onto the left side of the parallel electrodes. Then we stacked a layer of NbSe2 onto the right side of the parallel electrodes without touching the left NbSe2. A layer of Fe3GeTe2 was placed on the gap position bridging the two layers of NbSe2. Finally, we used a layer of hBN to encapsulate heterostructure for protection (Supplementary Fig. 6).

Electric transport measurements

The electric transport measurements were performed using a variable temperature cryostat provided by Oxford Instruments, which allowed for accurate temperature control from 300 K down to 2.4 K, and the magnetic field strength ranges from −1.4 T to 1.4 T. The resistance vs. temperature (R–T) curves and differential conductance resistance (dV/dI) were measured using a Keithley 6221 and 2182 A as the current source and voltmeter, respectively. Longitudinal electrical resistance and Hall resistance were acquired using a standard five-point measurement technique with the current set to be 10 µA.

Data processing

Due to the asymmetry in our nanoflake samples, the measured Hall resistance was mixed with the longitudinal magnetoresistance. Typically, We processed the Hall resistance data Rxy by using (Rxy (+B) – Rxy (–B))/2 to remove the contribution from the longitudinal magnetoresistance, and utilized (Rxx (+B) + Rxx (–B))/2 to eliminate the contribution from the Hall magnetoresistance to obtain longitudinal magnetoresistance Rxx. Besides, we define the temperature at which the resistance reaches half of the normal resistance as the transition temperature (Tc). We differentiate the I–V curve to obtain the I–dV/dI curve. The position at half of the difference between the maximum value and the minimum value in the dV/dI curve is defined as the Josephson critical current (Supplementary Fig. 7).