In situ Raman spectroscopy across superconducting transition of liquid-gated MoS₂

Molybdenum disulfide (MoS₂) and other group-VI-B transition metal dichalcogenides (TMDs) are members of the family of layered materials. They possess a wide variety of crystal structures and phases and a correspondingly rich set of electronic properties.¹ The crystal structure of TMDs can be classified into three groups: triangular prismatic phases, octahedral phases, and mixed phases. In the triangular prismatic phases, six chalcogen atoms form a triangular prism [Fig. 1(a)]. Different stacking of the monolayers produces either the 2Ha, 2Hb, or 3R bulk single crystal. In octahedral phases, six chalcogen atoms form an octahedron [Fig. 1(a)]. This octahedron can be easily distorted depending on the constituents and carrier density resulting in a number of derivative bulk phases referred to as the 1T, 1T', 1T'', 1T''', and Td phase. In the 4Hb and 6R mixed phases, both the triangular prism and octahedron atomic arrangements co-exist.

The most stable phase of MoS₂ is the triangular prismatic 2Hb phase. According to the ligand field theory,² intrinsic 2Hb-TMDs (and other triangular phases) are insulating or semiconducting, while octahedral phases are metallic. The latter can also become superconducting.^3–5 Semiconducting TMDs have been widely studied using the electric double layer transistor (EDLT) device concept. In this device, the TMD channel is covered with an ionic liquid or electrolyte that is contacted by a counter electrode.^6–17 Application of a voltage between the channel and the counter electrode induces charge carriers in the channel either electrostatically or through intercalation of ions into the van der Waals gaps. Ionic liquid based devices rely exclusively on field effect doping. In contrast, electrolyte based devices have two modes of operation. At small applied voltages, charge carriers are induced by the field effect only. At larger voltages, however, intercalation takes place and increases the doping further.⁹ The EDLT geometry offers the possibility to explore the physical properties over a very wide range of carrier densities due to the superior density tunability compared to conventional field-effect transistors (FETs). In particular, two phenomena can only be induced in EDLTs and not in FETs: superconductivity and structural phase transitions. Superconductivity is observed in many members of the group-VI-B TMDs, such as MoS₂, WS₂, and MoSe₂.^7–10 It has unleashed a flood of interest and research due to the Ising nature of the superconductivity with Cooper pairs possessing a strong resilience against an in-plane magnetic field.^11,12 A structural phase transition from the triangular prismatic to the octahedral phase has been observed in MoS₂ (Refs. 18 and 19) upon ion intercalation. In MoTe₂ (Ref. 13), the same phase transition was observed by strong field effect doping. A theoretical study has also predicted that such a structural phase transition may occur at large carrier densities.²⁰ 

The robustness of the field-induced superconducting state against an in-plane magnetic field has been accounted for by the coupling between the spin and valley degrees of freedom in the band dispersion of the triangular prismatic phase where inversion symmetry in the monolayer is absent.^11,12,21 This is a unique feature of the triangular prismatic TMD phases.²² However, we are unaware of any investigations of the crystal structure in the presence of the superconducting state, even though heavy electron doping can induce a structural phase transition.^{3,4,13,18–20} Recent reports have demonstrated that also in other scenarios without a triangular prismatic structure materials can host Ising superconductivity. It appeared for instance in few-layer stanine²³ as well as the Td phase of MoTe₂.⁵ In particular, the observations on Td-MoTe₂ call for a revisit of the origin of Ising superconductivity in MoS₂ and an explicit verification that the heavy electron doping does not induce a structural phase transition to the octahedral phase in the MoS₂-EDLT as such a phase transition would also account for the appearance of Ising superconductivity. This possibility has motivated the present work in which we attempt to confirm in situ the crystal structure of MoS₂ when the superconducting state is induced in the EDLT device structure as confirmed in low temperature transport measurements. Changes in the crystal structure can most easily be identified and assessed from optical signatures.^13,18,19 For instance, Raman spectroscopy detects crystal structure specific vibrational modes. Photoluminescence provides information about bandgap changes, and the output from the second harmonic generation reflects changes in the crystal symmetry.

Although gate-dependent confocal optical spectroscopy measurements on TMDs have been reported by many groups, such experiments are less common for the EDLT device configuration.^{18,19,24–26} It must in particular be emphasized that previous spectroscopic work did not allow to verify explicitly the carrier density and whether the investigated carrier density was large enough to reach the superconducting transition, because the deployed measurement did not offer simultaneous in situ confocal spectroscopy and Hall effect measurements at low temperature. Here, we provide direct evidence of the unaltered triangular prismatic crystal structure of MoS₂ by simultaneously performing electrical transport, Hall effect, and micro-Raman measurements in the low temperature regime where the superconducting ground state emerges.

MoS₂ flakes were mechanically exfoliated from a bulk single crystal using adhesive tape. They were then transferred onto a SiO₂/Si⁺⁺ substrate. The triangular phase of the exfoliated flake was confirmed before device fabrication at room temperature using a confocal Raman spectroscopy system (NT-MDT Spectrum Instruments). As demonstrated in research on TaS₂, thin exfoliated flakes may contain smaller areas of other minor phases different from the macroscopic bulk phase.²⁷ The Raman spectrum of our MoS₂ thin flake is shown in Fig. 1(b). Two Raman peaks near 380 and 410 cm⁻¹ are observed and associated with the E¹_2g and A_1g phonon modes of the 2Hb MoS₂ phase, respectively. The third peak around 520 cm⁻¹ originates from the silicon substrate itself. Schematics of the two MoS₂ lattice vibration modes are depicted as insets. The A_1g mode can also be observed in the Raman spectrum of octahedral MoS₂ and has similar frequency. However, the E¹_2g mode is prohibited. Another Raman mode, the E_1g mode around 290 cm⁻¹, appears instead for octahedral MoS₂ as reported both theoretically²⁸ and experimentally.²⁹ In addition, Raman spectra of octahedral MoS₂ are often accompanied by three broad peaks near 160, 230, and 330 cm⁻¹. These modes are referred to as the J₁, J₂, and J₃ modes. These modes originate from the formation of a superstructure when the octahedrons are distorted.²⁸ The absence of both the E_1g and J modes is a strong indication that the exfoliated thin MoS₂ flake remains in the triangular phase. The separation between the E¹_2g and A_1g Raman lines also provides information about the thickness of the flake. Here, the difference equals 24.45 cm⁻¹, suggesting a thickness of 4 or 5 monolayers.³⁰ 

After transfer onto the substrate, the flake is shaped into a rectangle by a dry etching processing step. Subsequently, metal electrodes consisting of a 10 nm Ti adhesion layer and 50 nm Au are patterned using electron-beam lithography, thermal evaporation, and resist liftoff. A big metallic pad is fabricated simultaneously next to the device. It acts as the side gate electrode [Fig. 1(c)]. We used the same ionic liquid [N,N-diethyl-N-methyl-N-(2-methoxyethyl) ammonium bis (trifluoromethyl-sulfonyl) imide (DEME-TFSI)] as prior works reporting Ising superconductivity.^18,19 Hence, the carrier density is induced electrostatically by the field effect. A small piece of a glass slide is placed on top to flatten the surface of the ionic liquid drop for optical accessibility. The device was then mounted into a home-built sample rod and loaded into a cryostat equipped with a superconducting magnet [Fig. 2(a)]. Our sample rod is equipped with a low-temperature compatible microscope objective and a piezo driven XYZ stage (Attocube systems AG) to focus the laser light and move the illumination spot. Other optical components are placed on a vibration isolated optical table to which the cryostat with superconducting magnet coil is rigidly attached. The light is guided to the sample with a free-space wave guide. The arrangement allows direct imaging of the device in order to locate the incident laser spot. The sample temperature can be controlled continuously from 300 K down to 2 K, and the sample temperature (T) is monitored with a Cernox resistance thermometer located below the chip carrier.

The transport characteristics of the device are summarized in Figs. 2(b)–2(e). The circuits to acquire the measured quantities are illustrated in the insets. The included optical image was taken at 220 K on the sample covered by the ionic liquid and the glass slide after it was mounted inside the cryostat. The image was recorded with back illumination and a CCD camera located on the optical table at a distance of about 3 m. A typical example of ambipolar behavior¹⁷ appears in the transfer curve [Fig. 2(b)], where the channel current (I_DS) increases under both positive and negative gate voltage (V_G) corresponding to electron and hole doping, respectively. At high electron doping (V_G = 4.3 V), the channel resistance (R_xx) decreases monotonically with descending temperature and saturates below 30 K, followed by a sharp drop below 10 K implying a superconducting transition [Fig. 2(c)]. The superconducting state can be easily suppressed by applying a perpendicular magnetic field [Fig. 2(d)]. Figure 2(e) plots the magnetic field dependence of the Hall resistance. It is recorded at 20 K to compare with a previous study.⁷ From the slope, a carrier density of 1.2 × 10¹⁴/cm² is deduced. This falls in the density range where superconductivity on MoS₂ has been reported previously.⁷ 

Finally, we address the crystal structure of the MoS₂ thin flake in the density regime where the sample becomes superconducting by performing in situ micro-Raman spectroscopy. A He-Ne laser with a wavelength of 632.8 nm is used for excitation. The R_xx–T curves under laser irradiation are plotted in Fig. 3(a). The investigated sample spots are marked with colored symbols in the optical image. The temperature (T) is monitored on the back side of the chip carrier. Hence, the actual temperature of the MoS₂ thin flake exposed to laser irradiation is likely somewhat higher. The resistance drop is less pronounced when the device is illuminated, suggesting heating by the incident laser irradiation. However, this heating effect is limited to the vicinity of the laser spot as can be seen from the green colored trace (indicated by the green square symbol) in Fig. 3(a) that was recorded when the laser spot is located outside of the sample region where the transport is measured. This green trace overlaps well with the R_xx–T curve recorded in the dark (black line), indicating that the laser irradiation hardly heats up this sample area. The stepwise drop of R_xx and non-zero resistance after the transition is likely due to an inhomogeneous carrier density distribution and/or a local detachment of the frozen ionic liquid.^8,10

Figure 3(b) shows Raman spectra of the device recorded at 20 K for four different positions along the channel. The E¹_2g and A_1g modes characteristic for the 2Hb phase of MoS₂ are clearly observed in all cases. In addition, the two peaks labeled as 2LA(M) and A_2u appear. These peaks correspond to Raman modes of 2Hb-MoS₂ as well, but are only observable under resonant excitation conditions.²⁵ In this study, resonant conditions are incidentally satisfied because the photon energy of the He-Ne laser (1.96 eV) almost matches the A-exciton energy at the K and K' points of MoS₂. Raman modes below 350 cm⁻¹ that can be attributed to either the E_1g mode or the J modes of octahedral MoS₂ are absent and, hence, a structural phase transition to octahedral phases can be unequivocally excluded.^28,29 The intensity drop below 230 cm⁻¹ is due to the notch filter, inserted in the optical path to block the reflection of the excitation laser beam. It is important that all four Raman spectra are similar to each other and do not show any signal suggesting a structural phase transition from 2Hb to octahedral phases. Finally, the temperature dependence of the Raman spectrum was recorded as the superconducting transition is crossed [Fig. 3(c)]. Apart from line sharpening with descending temperature, the Raman spectrum remains identical with no additional features emerging. This too confirms the assertion that the crystal structure remains in the triangular prismatic phase when the channel undergoes the superconducting transition. We note for the sake of completeness that our measurements are not able to identify whether a transition among triangular phases (2Ha, 2Hb, and 3R) takes place, but we can exclude a transition to any of the octahedral phases.

In summary, we have investigated the crystal structure of MoS₂ in the high density regime where superconductivity is induced using an EDLT device geometry. The home-built low temperature in situ confocal Raman spectroscopy system makes it possible to confirm that the superconducting state appears within the triangular prismatic crystal structure. The technique deployed here can be extended to a polarization-resolved low frequency measurement method to further investigate the gap anisotropy of the superconducting state.³¹ 

The authors appreciate technical support from S. Wahl and helpful comments from T. Machida. They gratefully acknowledge financial support from the graphene flagship core 3 program and the DFG priority program SPP2244 (J.H.S.), the Moritani Scholarship Foundation, the Murata Science Foundation, and PRESTO, Japan Science and Technology Agency (JST) (Grant No. JPMJPR20L5) (Y.J.Z.).