With the tremendous success of graphene in studying two-dimensional (2D) physics, layered transition metal dichalcogenides (TMDs) have rapidly risen and provided a new platform for investigating new physics and material properties in 2D limit.^[1] Unlike graphene, TMDs crystals have a variety of structures with 2H-, 1T-, 1T′-, and T_d-phases.^[2] It is well known that most of 2H- and 1T-compounds possess semiconducting behavior, whereas the structure with 1T′ and T_d phases usually exhibits a semimetal structure.^[3–6] In addition, TMDs have also been intensively explored for exotic phenomena such as non-saturating magneto-resistance (MR) in WTe₂,^[7,8] pressure-induced superconductivity in MoTe₂,^[9] ionic gated superconductivity in MoS₂,^[10] intrinsic Ising superconductivity in NbSe₂,^[11] etc.

Among those TMDs, 1T′ MoTe₂ has attracted considerable interest in recent years as a Weyl semimetal candidate.^[12–15] Recently, unconventional superconductivity in bulk and few-layer flakes, as well as topological phase transition induced by temperature and strain were reported.^[9,16,17] In particular, it was discovered that the superconductivity in MoTe₂ flakes is very different from that in the 2H-structure, such as ionic gated MoS₂^[10] and thin-layer NbSe₂.^[11] This is because the 1T′ structure is expected to show anisotropic spin-orbit coupling due to inversion symmetry broken along both in-plane and out-of-plane direction.^[17] Moreover, the thin flakes are expected to host the topologically protected edge states, which may be combined with their superconductivity, and thus resulting in nontrivial topological properties. Therefore, thin-flake MoTe₂ is a potential candidate for searching topological superconductivity, which has potential applications in quantum computation.^[18] As a result, the nature of superconductivity in MoTe₂, which remains unclear, needs further studying.

In this article, we investigate the superconducting properties of the mechanically exfoliated few-layer MoTe₂ flakes by using transport measurements. As reported in Ref. [19], the surface of MoTe₂ is unstable in the ambient condition because of the weak Mo–Te bond energy, which can induce unpredictable surface oxidation and artifacts in film samples. To ensure the intrinsic properties of MoTe₂ flakes, we encapsulate MoTe₂ by using hexagonal boron nitride (h-BN) film to protect it from oxidizing. As a result, we have successfully fabricated cleaner samples with much fewer defects than the CVD flakes.^[17] We find that our MoTe₂ flakes are superconducting and have an onset superconducting transition temperature T_c as high as 5.3 K in the thinnest sample (4 nm), measured in this article, which significantly exceeds the reported transition temperature T_c of the bulk counterpart (∼ 0.1 K).^[9] Besides, the T_c of the thin flakes can be tuned by as large as 320 mK through applying a gate voltage. Hence, our study may provide an attractive material platform for studying new type of superconductivity in the topological environment.

Bulk single crystals of 1T′-MoTe₂ were synthesized via a flux method with NaCl. The h-BN/MoTe₂ heterostructure devices were assembled by using a dry transfer method. First, MoTe₂ and h-BN flakes were separately exfoliated onto different SiO₂ (300 nm)/Si substrates by using Scotch tape. Subsequently, we applied a transparent glass slide with polydimethylsiloxane/polypropylene carbonate (PDMS/PPC) films to pick up a large-sized h-BN flake, and then used it to pick up a small MoTe₂ flake. Aligned with a micro-manipulator stage and through optical microscopy, the chosen MoTe₂ flake is transferred to the prefabricated Hall-bar Cr (5 nm)/Au (50 nm) electrodes on an h-BN/SiO₂ (300 nm)/Si substrate. All the mechanic exfoliation and transfer processes were performed in argon gas glove box. Example of an encapsulated h-BN/MoTe₂ heterostructure device is shown in Fig. 1(b).

Fig. 1.

	Figure Option ViewDownloadNew Window
	Fig. 1. (a) Measured Raman spectra for 4.0-nm and 7.6-nm MoTe2 sample at 300 K, where red solid curve is for 4-layers MoTe2, which is the same as literature data[30] and inset shows optical microscope image of device MoTe2. (b) Schematics of sandwiched structure of few-layer MoTe2 device.

Raman spectrum was measured with a He–Ne 632.8-nm laser at room temperature. The beam was focused into a diameter of 1 μm, and irradiates the sample through a 25 × magnification objective. Magneto-transport measurements are conducted in an Oxford Variable Temperature Insert Measurement System with magnetic field up to 13 T and temperature down to 1.5 K. The resistance is measured by using low-frequency lock-in technique with excitation currents ranging from 0.01 μA to 0.1 μA. The thickness of each sample is determined by the AFM after the transport measurement.

Figure 2(a) shows the normalized resistance (R/R_{250 K}) as a function of temperature for the exfoliated MoTe₂ flakes with thickness ranging from 4.0 nm to 9.3 nm. The sample structure is further characterized by Raman spectra measured at 300 K where we do not find any sign of T_d-phase in thin flakes (i.e., the thickness of 4 nm and 7 nm, see Fig. 1(a)). All samples exhibit a metallic behavior in the whole temperature range before entering into the superconducting state. The well-known first-order structural transition in a MoTe₂ single crystal is not observed in our thin-flakes,^[17,21] suggesting that thin samples may remain in the 1T′ phase in the low-temperature range.

Fig. 2.

Figure Option ViewDownloadNew Window

Fig. 2. (a) Temperature dependence of the normalized resistance (R/R250 K) for MoTe2 crystals with thickness ranging from 4.0 nm to 9.3 nm, with inset showing low-temperature zoom-in of the normalized resistance (R/R10 K) as a function of temperature. (b) Superconducting phase transition temperature Tc as a function of sample thickness with Tc being defined as 90% and 60% of normal state resistance at 10 K, and the blue dotted line referring to bulk Tc,0 cited from Ref. [8]. (c) Electrical Hall resistivity data as a function of a magnetic field for various sample thickness values. (d) Carrier concentration and mobility (inset) at different sample thicknesses at 5 K, extracted from two-band model fit.

The residual resistance ratio defined as RRR = R_{250 K}/R_N, where R_N is the normal state resistance above the superconducting transition temperature (here we take N = 10 K), varies from 3.2 (t = 4.0 nm) to 7.5 (t = 9.2 nm), which is approximately twice higher than that of CVD film.^[19] From the electrical Hall resistivity extracted through the two-band model (Fig. 2(d)),^[22] we can estimate the dominant charge (electron) mobility, μ_e, of 4-nm-thick MoTe₂ flake at 634 cm²/V∼s, i.e., μ_e = 634 cm²/V∼s, and charge (electron) carrier density n_e at 6.4 × 10¹⁸ cm⁻³, i.e., n_e = 6.4 × 10¹⁸ cm⁻³. These characteristics compare favorably with what was reported in CVD samples, where μ = 40 cm²/V∼s on the 3.8-nm-thick layer which is approximately a factor of 10 lower than our results.^[19] In general, the value of RRR and μ both suggest a high-quality sample in the exfoliated MoTe₂. Next, we turn to the superconducting properties. A zoom-in of the normalized resistance (R/R_{10 K}) is shown in the inset of Fig. 2(a) and the layer thickness-dependent critical superconducting transition temperature T_c is presented in Fig. 2(b), in which the T_c is defined as 90% (black curve) or 60% (red curve) of the normalized R_{10 K}, respectively. For comparison, we also note the T_c of bulk 1T′-MoTe₂ crystal used in Ref. [9]. First, all samples show superconductivity below 6 K. However, we do not observe the complete zero resistance due to the base-temperature limitation of our cryostat. An important observation here is that the value of T_c significantly rises up to 5.3 K (with criterion of 0.9R_{10 K}) for the 4-nm-thick sample, which is substantially higher than that of bulk MoTe₂ (T_c = 0.1 K)^[9] and T_c = 3.91 K of 30-nm-thick T_d-MoTe₂.^[17] Generally, as the thickness decreases, disorder and interaction effect will reduce the density of mobile charge carriers, and thus causing its superconductivity to be suppressed and eventually entering into an insulating state in the highly disordered regime,^[21,23–25] which is in good agreement with our Hall measurement results as shown in Fig. 2(d). A similar characteristic has been observed in many superconducting materials except for some special cases such as single crystal nanowire arrays.^[26,27] Moreover, Cui et al. reported the similar thickness dependence of T_c in T_d-MoTe₂ very recently.^[17,28]

To explain such an enhancement of superconductivity, we try to analyze the electrical Hall resistivity measured at the normal state of 5 K, which are shown in Figs. 2(c) and 2(d). Owing to the fact that the Hall resistivity, ρ_xy, exhibits non-linear magnetic field dependence, we calculate the carrier density and mobility through the two-band model.^[22] The dominant carrier (electron) density is nearly constant in a range from 7.6 nm to 4.5 nm and suddenly drops in the thinnest sample (4.0-nm thick). For the case of MoTe₂ single crystal, Mandal et al. reported an enhancement of T_c after doping it with Re atom, where the maximum T_c reached 4.1 K in Mo_0.7Re_0.3Te₂ composition.^[29] They argued that the increase of the electron concentration might contribute to the enhancement of the electron-phonon coupling and density of states at the Fermi energy. However, this is not consistent with our relationship between carrier concentration and T_c.

Another possible scenario is electrochemical doping during the fabrication of thin-flakes: there can appear some doping effects induced by the surface contamination or originating from the silicon oxide substrate. However, we rule out such a scenario based on the comparably higher mobility and RRR, which implies that the degradation of our samples is smallest. Moreover, the thin-flakes are weakly coupled to the SiO₂/Si substrate, we believe it is unlikely to be influenced by the oxide substrate.

Meanwhile, a similar large enhancement of superconductivity was observed in another TMD material TaS₂ in 2016, indicating that the enhancement of the effective electron–phonon coupling plays a crucial role.^[30] There it was claimed that the inversed T_c dependence versus the number of layers is due to a relatively strong Coulomb repulsion.^[30] Our exfoliated samples are much cleaner with fewer defects than CVD samples. Thus, it may lead the electron-phonon interaction to increase due to a strong screening effect of Coulomb interaction originating from the higher quality samples.^[31] Although we discussed several possible scenarios, we are still not clear about the mechanism and origin of T_c enhancement when thickness decreases; these issues remain to be further investigated theoretically and experimentally.

In order to gain a better understanding of the superconductivity properties for exfoliated-flake samples, we map the magnetic field and angular dependence of magneto-resistance (MR) and reduced H–T phase diagram as shown in Fig. 3. Figures 3(a) and 3(b) clearly show that the dependence of MR on magnetic field in a 4-nm-thick MoTe₂ crystal at various temperatures (1.5, 2, 3, 5, and 10 K) ((Fig. 3(a)) and field angles, θ (Fig. 3(b)) at a fixed temperature of 1.5 K, where θ denotes the tilt angle between the normal direction of the crystal plane and the magnetic field as depicted in the inset of Fig. 3(c). When the magnetic field is close to the crystal basal plane (θ approaches to 90°), the superconductivity transition shows a higher upper critical field because of its 2D superconductivity nature, which is well matched with 2D Tinkham model fit (Fig. 3(c)). The estimated crystal-plane coherence length ξ_|| is ∼ 5.5 nm, which is longer than sample thickness. This also supports 2D superconductivity scenario.

Fig. 3.

Figure Option ViewDownloadNew Window

Fig. 3. Plots of magneto-resistance versus magnetic field at (a) various temperatures and (b) 1.5 K for different field orientations. All experimental data are taken from 4-nm-thick MoTe2. (c) Angle-dependent upper critical field Hc2, with solid red line denoting fitting to the data following the Tinkham formula for a 2D superconductor, and empty circles referring to our data, solid circles representing mirror-symmetrized data for clarity, and inset showing experimental configuration and magnetic field direction. (d) Normalized H–T phase diagram, with symbol red star denoting our data.

In Fig. 3(d), the upper critical field H_c2 in a parallel field configuration is significantly enhanced due to its strong spin–orbit coupling resulting from inversion symmetry breaking. Although we cannot reach the onset point of H_c2 at θ = 90° (H||In-plane) in this study because of the limitation of our superconducting magnet, we still obtain a strikingly strong magnetic anisotropy result (H_c2||/H_c2⊥ ∼ 22) in comparison with the magnetic anisotropy of T_d-MoTe₂.^[17]

Finally, we also examine the electric field gating effect with the help of a back-gate sweeping. Figure 4(a) shows that plots of MR versus magnetic field for a 4-nm-thick MoTe₂ layer with three different gate voltages (+60 V, 0 V, and −80 V). The H_c2 is determined by a deviation point from the normal state, which is marked with symbol star on the graph. The applied gate voltage is found to be able to cause T_c and H_c2 substantially change as shown in Fig. 4(b), with T_c modulation of 320 mK and H_c2 modulation of 400 mT, corresponding to ΔT_c/T_c of 6.7% and ΔH_c2/H_c2 of 16.2%. It is worthy to note that the modulation of T_c and H_c2 induced by gating effect may be explained from the variation of carrier concentration since other TMD material NbSe₂ is reported to have less than 10% modulation by applying gate voltage,^[32] which value is more or less consistent with ours. Nevertheless, the large enhancement of T_c as thinned down would be caused by a deeper mechanism, which needs further studying.

Fig. 4.

	Figure Option ViewDownloadNew Window
	Fig. 4. (a) Variations of magnetoresistance with magnetic fields at 1.5 K for three different back-gate voltages (+60, 0, −80 V). Each dotted line is linear fitting to find onset of corresponding upper critical field. (b) Plot of Tc and Hc2 versus gate voltage, with dotted lines guiding to the eyes.

In this work, we develop a method of fabricating the atomically thin MoTe₂ flakes, which minimizes the degradation from exposure to the ambient environment and thus maintains the intrinsic properties of the material. The thin MoTe₂ flakes prepared by exfoliation method are characterized to be cleaner with much fewer defects than the CVD-grown samples reported previously. We find that the exfoliated MoTe₂ flakes are superconducting and possess an onset superconducting transition temperature T_c of up to 5.3 K, which significantly exceeds the T_c of the bulk counterpart (∼ 0.1 K). The superconductivity in few-layer MoTe₂ is consistent with 2D Ising spin crossover as demonstrated by the in-plane magneto-resistance measurement. Moreover, the unconventional superconductivity of the thin film can be modulated by an applied gate voltage, which tunes the T_c and H_c2 by as large as 320 mK and 400 mT, respectively. These results need further studying in the aspect of TMDs superconductivity in the topology ground.