Interface-enhanced superconductivity in monolayer 1T′-MoTe₂ on SrTiO₃(001)

Abstract

Introducing superconductivity into two-dimensional (2D) films with nontrivial topology has been intensively pursued as one of the feasible scenarios to realize 1D topological superconductor. Prevailing endeavors mostly exploit the external gating or proximity effect of a traditional superconductor, by which the critical temperatures ($T_{\mathrm{c}}$) are limited to several Kelvin range. Here, we report on the discovery of interface-enhanced superconductivity in monolayer 1T′-MoTe₂ film. A thermally driven phase transition from Mo₆Te₆ nanowires to 1T′-MoTe₂ films, grown on SrTiO₃(001) surface by the molecular beam epitaxial methods, is demonstrated. A combined study of scanning tunneling microscopy/spectroscopy, electrical transport and magnetization measurements indicates the $T_{\mathrm{c}}$ of MoTe₂ film is around 30 K, two orders of magnitude larger than its 3D counterpart crystal. This study shows that interfacial engineering is an efficient way to tune monolayer 1T′-MoTe₂ film into superconducting states, and thus may pave the way toward higher-$T_{\mathrm{c}}$ 1D intrinsic topological superconductivity.

1 Introduction

Monolayer transition-metal dichalcogenides (TMDs) in 1T′ phase have been intensively studied as promising candidates for quantum spin Hall (QSH) insulators [1]. The QSH gap and the conducting edge channels have been detected in WSe₂, WTe₂ and MoTe₂ by scanning tunneling microscopy/spectroscopy (STM/STS) or transport measurements [2–4]. Ensuring research based on the nontrivial topological properties includes the strain-induced phase transition from semimetal to topological insulator in WTe₂ and symmetry-enforced topological metallic states along a grain boundary in MoTe₂ [5, 6]. Moreover, the coexistence of superconductivity and topologically nontrivial edge states has been long thought to realize one-dimensional (1D) topological superconductors (TSC) [7–12]. Sundry methods have been utilized to introduce cooper pairing into these topological 2D films. For example, voltage gating has been used to drive monolayer 1T′-WTe₂ into the superconducting phase with the critical temperature $(T_{\mathrm{c}}) \sim 1~\text{K}$ [13]. The proximity-induced superconductivity has been observed in WTe₂/NbSe₂ heterostructure with the $T_{\mathrm{c}} \sim 6~\text{K}$ [14]. Very recently, the enhancement of superconductivity in monolayer 1T′-MoTe₂ was reported, where the $T_{\mathrm{c}}$ increased up to ∼8 K in the hole-dominated region [15]. A change in the Coulomb interaction with dimensionality has been proposed to account for the enhancement in $T_{\mathrm{c}}$ via spin fluctuations. However, the transition temperatures achieved by the above endeavors are still relatively low. Alternative approaches are highly craved to realize higher-$T_{\mathrm{c}}$ superconductivity in these 2D topological films.

Interfacial engineering is an efficient way to tune the electronic properties of 2D materials [16]. In particular, interface-enhanced superconductivity has stimulated tremendous interest due to its great potential for raising $T_{\mathrm{c}}$ [17–22], for the most prominent example, monolayer FeSe grown on SrTiO₃ (STO) [18]. Interface-related processes such as interfacial charge transfer and electron-phonon coupling (EPC) are perceived as two critical mechanisms accounting for the enhancement effect [23–29]. Up to date, interface-enhanced superconductivity has not yet been achieved in monolayer 1T′-TMDs [30, 31]. In this work, we report on the observation of interface-enhanced superconductivity in monolayer 1T′-MoTe₂ film grown on the STO(001) substrate by molecular beam epitaxy (MBE). The ex situ electrical transport and magnetization measurements confirm the superconductivity and give the $T_{\mathrm{c}}$ of 30 K, two orders of magnitude increase compared with the 3D counterpart crystal ($T_{\mathrm{c}} \sim 0.25~\text{K}$) [32]. The U-shape superconducting gaps show the attenuation as a response to the temperature rise or external magnetic field.

2 Results and discussions

The wide-area STM image in Fig. S1 shows a step-and-terrace structure of the STO surface with a constant height of 0.4 nm. The corresponding magnified STM image in Fig. 1(a) shows a double-layer TiO₂-terminated $2 \times 2$ surface reconstruction [33–35]. This $2 \times 2$ reconstruction can be identified in the fast Fourier transformed (FFT) images [see details in Fig. S2]. In addition to the ordered structures, the irregular dark and bright features on the surface possibly associated with crystal defects such as O vacancies or Ti clusters are visible. Figure 1(b)–1(d) show a series of typical STM topographic images of $\mathrm{Mo}_{\mathrm{x}}\mathrm{Te}_{\mathrm{y}}$ on STO substrate at different growth temperatures. After growth, the samples were annealed at the same temperatures for 30 minutes under Te atmosphere. Figure 1(b) gives the STM image of 1D Mo₆Te₆ nanowires formed at 350 °C due to a Te-deficient environment, which has been well studied in the previous report [36]. With decreasing growth temperature down to 300 °C, the mixed phase of the Mo₆Te₆ nanowires and MoTe₂ islands arises [see Fig. 1(c)]. At optimal growth temperatures ranging from 240 °C to 260 °C, the Mo₆Te₆ nanowires will disappear and monolayer MoTe₂ film is synthesized accordingly. An overall large-scale STM image shows that the coverage of MoTe₂ film is ∼93% [see Fig. 1(d)]. The RHEED patterns before and after the MoTe₂ film growth are shown in Fig. 1(e). Using the lattice constant of STO ∼ 3.9 Å as a reference, one can get the lattice constant a of 1T′-MoTe₂ on STO ∼ 6.5 Å. 1T′ phase has a rectangle lattice, which can be taken as a distorted structure compared with 1T phase as illustrated in Fig. 1(f).

Figure 1

Epitaxy growth of Mo_xTe_y on STO(001) substrates. (a) STM image ($60 \times 60~\text{nm}^{2}$, $U = 1.1~\text{V}$, $I_{\mathrm{t}} = 50~\text{pA}$. U, sample bias voltage, $I_{\mathrm{t}}$, tunneling current.) of the STO(001) surface. The inset of (a) is the corresponding FFT image with the orange circles marking the $2 \times 2$ reconstruction. (b) STM image ($100 \times 100~\text{nm}^{2}$, $U = 2~\text{V}$, $I_{\mathrm{t}} = 100~\text{pA}$) of the Mo₆Te₆ nanowires on STO. (c) STM image ($100 \times 100~\text{nm}^{2}$, $U = 2~\text{V}$, $I_{\mathrm{t}} = 100~\text{pA}$) of the mixed phase of Mo₆Te₆ nanowires and 1T′-MoTe₂ islands on STO. (d) STM image ($200 \times 200~\text{nm}^{2}$, $U = 2~\text{V}$, $I_{\mathrm{t}} = 100~\text{pA}$) of monolayer 1T′-MoTe₂ film on STO. (e) RHEED patterns of STO substrate (top) and after monolayer 1T′-MoTe₂ film growth (bottom). (f) Schematic diagram of monolayer 1T′-MoTe₂ film on STO

Figure 2(a) shows the height of the MoTe₂ film in different regions. The height of the MoTe₂ film generally varies from 0.7 to 0.8 nm, which is lower than that grown on graphene (from 0.9 to 1 nm) [37]. This reduced height reflects the enhancement of the interlayer coupling between MoTe₂ film and STO substrate. In addition, the spatial-dependent height provides clues to different coupling strengths, resulting from varied stoichiometry and contrasting density of states between the MoTe₂ film and the STO surface. The epitaxial film consists of massive domains, which are generally dozens of nm² in size. Figure 2(b) shows the tunneling current image of the striped 1T′ phase with different orientations. Statistics of the orientations show that domains with an angle of 120° dominate [see Fig. 2(c)]. The atomically resolved image in Fig. 2(d) shows that the lattice constants a and b are $6.44\pm 0.05$ Å and $3.52 \pm 0.05$ Å, respectively. The enlarged lattice than bulk values ($a = 0.633\text{ nm}$ and $b = 0.347\text{ nm}$) suggests tensile strains, which are induced by $2 \times 2$ surface reconstruction of the STO substrate. The $dI/dV$ spectra shown in Fig. 2(e) reveal the electronic properties of the MoTe₂ film. In a smaller bias range (−0.2 to +0.2 V), we found that an energy gap attenuates the intensity of the density of state (DOS) near the Fermi level.

Figure 2

STM/STS characterization of monolayer 1T′-MoTe₂ films. (a) STM image ($80 \times 80~\text{nm}^{2}$, $U = 2~\text{V}$, $I_{\mathrm{t}} = 100~\text{pA}$) of the MoTe₂ film, with different step height shown in the right panel. (b) The tunneling current image ($50 \times 50~\text{nm}^{2}$, $U = 3~\text{V}$, $I_{t} = 30~\text{pA}$), showing different orientation of MoTe₂ islands. (c) Histogram of orientations of (b). (d) STM image ($20 \times 20~\text{nm}^{2}$, $U= 2~\text{V}$, $I_{\mathrm{t}} = 100~\text{pA}$) of the MoTe₂ film, showing 120° angle between the domains. Inset, atomically resolved STM image ($3 \times 3~\text{nm}^{2}$, $U = 50~\text{mV}$, $I_{\mathrm{t}} = 500~\text{pA}$) of the MoTe₂ film. (e) Tunneling spectra ($I_{\mathrm{t}} = 100~\text{pA}$) taken on the MoTe₂ film. The inset shows the spectra in a smaller bias range. The horizontal bars indicate the zero-conductance position of each curve

Strikingly, Fig. 3(a) exhibits a high-resolution $dI/dV$ spectrum as shown by the black circles within a narrower energy window (−20 to +20 mV). This U-shape gap is symmetric around the Fermi level, suggesting the occurrence of superconductivity. The influences of STO substrate are excluded in Fig. S3. We attempted to fit the STS results with Dynes model [38],

$$ N ( E,\Gamma ) =\operatorname{Re} \bigl[ ( E-i\Gamma ) / \sqrt{ ( E-i\Gamma )^{2} - \Delta ^{2}} \bigr], $$

where Δ represents the superconducting gap and Γ denotes the effective energy broadening. A nice fitting to the gap with a fully gapped s-wave function gives a Δ of 5.8 meV and a Γ value of 0.45 meV, as shown by the red curve. Compared with the previously reported superconducting energy gaps in MoTe₂ [39–41], this surprisingly large gap potentially indicates a higher $T_{\mathrm{c}}$ in MoTe₂/STO system. In general, the gap size varies from 5 meV to 6 meV in different regions. To distinguish the possible superconducting gap with a band gap or coulomb gap, we performed ex situ electrical transport measurements. Before being transferred out of the MBE high vacuum growth chamber, an amorphous Te capping layer with a thickness of 20 nm was deposited on MoTe₂ films at room temperature to provide protection from degradation in the air. Figure 3(b) exhibits the temperature dependence of the sheet resistance of sample No. 1. With decreasing temperature from 50 K to 2 K, the sheet resistance of the sample gradually increases and saturates, then begins to decrease at 30 K and drops to zero at 5 K, which is distinct with the behavior when the sample has a band gap or coulomb gap. The transport technique is a macroscopic probe compared with STS measurement. Therefore, we attributed this broad downturn of the resistance during the cooling to these possible reasons such as pronounced domain boundaries or the related defects in MoTe₂ film, and interstitial Ti clusters at the interface. The Te/STO heterostructure under the preparation conditions identical to Te/MoTe₂/STO heterostructure does not show superconducting behavior [see Fig. S4]. This indicates that the superconductivity comes from the monolayer MoTe₂.

Figure 3

Interface-enhanced superconductivity in MoTe₂ on STO. (a) A tunneling spectrum ($I_{\mathrm{t}} = 100~\text{pA}$) with a full gap (black circles) measured at 4.2 K and the Dynes model fitting (red solid line). Pronounced coherence peaks appear at ±5.8 mV. (b) Temperature dependence of the sheet resistance of sample No. 1 (Te/MoTe₂/STO), showing the superconducting transition occurs at ∼30 K. The inset depicts the schematic diagram of the four-probe measurement configuration. (c) M-T curves of sample No. 2 (Te/MoTe₂/STO) measured under a 30 Oe perpendicular magnetic field, showing the $T_{\mathrm{c}} \sim 30~\text{K}$. (d) Temperature dependence of the $dI/dV$ spectra ($I_{\mathrm{t}} = 100~\text{pA}$) and the corresponding extracted normalized ZBC. The spectra are spatially averaged due to the great thermal drift. (e) Magnetic field dependence of the $dI/dV$ spectra ($I_{\mathrm{t}} = 100~\text{pA}$) and the corresponding extracted normalized ZBC. The horizontal bars indicate the zero-conductance position of each curve

To further confirm the superconductivity, we carried out ex situ magnetization measurements for sample No. 2. The temperature dependence of VSM magnetization of the sample is shown in Fig. 3(c), measured in both zero-field cooling (ZFC) and field cooling (FC) modes. The magnetic field is 30 Oe, applied perpendicular to the STO(001) plane. Clearly, the M-T curves exhibit a drop at $T_{\mathrm{c}} \sim 30~\text{K}$, indicating the diamagnetic response. Please notice that the data include the contributions from MoTe₂ film, Te protection layer and STO substrate. Therefore, the background signal from the protection layer and substrate leads to the rise in the M-T curves at low temperatures [see Fig. S5]. Considering the transition temperature 30 K as the $T_{\mathrm{c}}$ and the largest gap size 6 meV in the tunneling spectra, $\Delta /k_{B} T_{c}$ of the sample is ∼2.32 ($k_{B}$ is the Boltzmann constant), which is larger than the weak coupling BCS condition with a value of 1.76. We further checked the temperature and magnetic field dependence of the $dI/dV$ spectra. With increasing temperature, the coherence peaks are gradually weakened and the gaps become shallower [see Fig. 3(d)]. The extracted normalized zero-bias conductance (ZBC) remains to be zero until the temperature is beyond 13.8 K. Remarkably, when applying the external magnetic field along the Z direction, the superconducting gaps are rather robust against the magnetic field as shown in Fig. 3(e). The extracted normalized ZBC remains to be zero until the magnetic field is beyond 9 T.

To further disclose the possible mechanism of the enhancement of the $T_{\mathrm{c}}$, we discussed the interface between MoTe₂ and STO. The work functions of STO and MoTe₂ are 4.5 eV and 4.9 eV [42, 43], respectively. Electrons would transfer from the STO substrate to the MoTe₂ films. Oxygen vacancies on the STO surface also induce charge transfer and enhance binding energy at the interface. Moreover, it’s proposed that the smaller interlayer distance between the superconducting layer and oxide substrate corresponds to the stronger interfacial bonding and then contributes to the optimization of superconductivity [44, 45]. As shown in Fig. 2(a), the lower height reflects the smaller interlayer distance between the MoTe₂ film and the STO substrate. Therefore, our results reveal that strong interface coupling at the MoTe₂/STO heterostructure plays an important role.

In summary, monolayer 1T′-MoTe₂ film on the STO(001) substrate has been successfully synthesized. The interface-enhanced superconductivity in MoTe₂/STO heterostructure with the $T_{\mathrm{c}}$ of 30 K has been demonstrated by ex situ electrical transport and magnetization measurements. Our work will stimulate more investigations on interfacial superconductivity in monolayer 1T′-TMDs on oxide substrates. More importantly, our findings establish the groundwork for a possible path toward the realization of topological superconductivity with high $T_{\mathrm{c}}$ by interfacial engineering.

3 Methods

The epitaxy growth and characterization of MoTe₂ films were performed in an ultrahigh vacuum MBE-STM system (Unisoku 1300) with a base pressure of $1 \times 10^{-10}$ Torr. The 0.5% wt Nb-doped STO(001) substrate was prepared by direct heating up to 950 °C for 1 h. The MoTe₂ films were then grown on STO(001) surface by coevaporating high-purity Mo (99.95%) and Te (99.999%) from electron-beam evaporators and standard Knudsen cells, respectively, with the flux ratio of Mo:Te ∼1:20. The growth process was monitored by in-situ reflection high energy electron diffraction (RHEED) and the growth rate was ∼50 minutes per monolayer. The STM topographic images were acquired under a constant current mode and processed with WSXM software. The tunneling spectra were acquired at ∼4.2 K unless otherwise stated. For wide (narrow)-energy-scale tunneling spectra, the magnitude of the bias modulation for the lock-in technique was 5 (0.5) mV at a frequency of 951 Hz. The electrical transport measurements were performed in a physical property measurement system (PPMS-9T) with the standard four-probe method. The vibrating sample magnetometer (VSM) magnetization measurements were performed in a magnetic property measurement system (MPMS3).