Electrically tunable metal–semiconductor behavior in 2H-MoTe₂

Among layered transition metal dichalcogenides (TMDCs), molybdenum di-telluride (MoTe₂) stands out as an intriguing material system due to its ability to undergo a structural phase transition from semiconducting to metallic under relatively moderate conditions.¹ This transition is significantly influenced by two critical parameters: entropy and ground state energy.² For MoTe₂, the ground state energy difference between the semiconducting (2H) and metallic (1 T) phases is only 30 meV/fu, compared to 800 meV/fu for molybdenum di-sulfide (MoS₂) and 250 meV/fu for molybdenum di-selenide (MoSe₂).³ This smaller energy gap makes structural phase changes in MoTe₂ considerably easier than in its TMDC counterparts.⁴ Various methods, including thermal and laser annealing (photo-bolometric effect), doping, plasma treatment, electric fields, and strain, have been employed to induce the 2H-to-1T phase transition in MoTe₂, with the goal of exploring its potential for ultra-fast phase-change memory applications.^5–9 However, the performance and endurance varied significantly based on the technique used to induce the phase change, owing to differences in the underlying physics. For example, temperature- and electric field-driven transitions, associated with short-range interaction effects within the MoTe₂ lattice, offer low operating voltages but pose a risk of significant chip-to-chip variation.^10,11 To date, the most promising and consistent phase-change memories are those based on the creation of tellurium vacancies, driven by electro-bolometric (current-induced) or photo-bolometric (laser-induced) effects.^12,13

One of the major challenges with vacancy-induced phase transitions in MoTe₂ is the switching time scale, a key parameter for memory performance.¹⁴ This time scale is constrained by the rate of Te vacancy generation and regeneration, which is highly dependent on device parameters.¹⁵ Efforts to improve performance have focused on contact engineering to speed up the phase transition via the same vacancy-mediated mechanism.¹⁶ However, this localized phase transition often results in a lattice with Mo-rich islands interspersed with defect-free regions, producing a mixed-phase material rather than a purely metallic or semiconducting material.^6,17,18 As a result, device performance becomes highly sensitive to process parameters, leading to potential reliability issues.¹⁸ These limitations raise fundamental questions about the feasibility of using reversible phase transitions in MoTe₂ for memory applications and their electrical tunability.^19,20 In our study, we address these concerns and demonstrate that a reversible semiconducting-to-metal-like transition can be induced through controlled Joule heating without the need for doping, high-temperature annealing, defect engineering, or mechanical strain. Notably, this transition occurs only on insulating substrates, such as glass, and is entirely suppressed on substrates with high thermal conductivity, like graphene and hexagonal boron nitride (h-BN).^21,22 Electrical and optical studies further reveal that the metallic-like behavior in 2H-MoTe₂ is predominantly driven by electron–phonon coupling.

Single-crystalline flakes of 2H-MoTe₂ were mechanically exfoliated from a bulk crystal (purchased from HQ Graphene) using scotch tape. The exfoliated flakes were first transferred onto a glass substrate, and their thickness was measured using an atomic force microscope (AFM) [see Figs. 1(f) and 1(g)]. The flakes were then picked-up with a PDMS stamp and transferred onto pre-patterned silicon and indium tin oxide (ITO) substrates to fabricate three-terminal and two-terminal devices for field-effect and conductivity measurements, respectively (see Fig. 1).

In this study, the flake thickness is 38 nm, with a lateral size of approximately 50 μm, and the channel length between the source and the drain is 30 μm. For field-effect studies, multilayer graphene (MLG) is first transferred over the gold electrodes to act as a global back gate (BG), while h-BN functions as a dielectric [see Figs. 1(a) and 1(b)]. The voltage to the MLG is supplied via the gold electrodes (BG). Both h-BN and MLG known for their high thermal conductivity also act as heat sinks, facilitating efficient heat dissipation and minimizing the impact of Joule heating on the lattice temperature. A 2H-MoTe₂ flake is then transferred onto the h-BN/MLG scaffold such that it connects the source and drain electrodes [see Fig. 1(c)]. The device schematic is shown in Fig. 1(d). Transconductance and transfer characteristics are examined by connecting the source and back gate to independent voltage sources while grounding the drain terminal. Electrical measurements were conducted using a Keithley Semiconductor Characterization System (SCS 4200). For control studies without a heat sink, pre-patterned ITO-coated glass substrates were used. Conductivity, gas, and light response studies were conducted in a two-terminal configuration, with the 2H-MoTe₂ flake bridging two ITO electrodes on a quartz glass substrate [see Fig. 1(e)]. Voltage- and temperature-dependent Raman spectra were obtained using a HORIBA LabRAM HR spectrometer, equipped with a 532 nm laser operating at 3 mW. Spectra were recorded over a temperature range of 303–473 K. The system's high spectral resolution down to 5 cm⁻¹ allows for detailed analysis of vibrational modes, while the laser spot size of 1 μm ensures spatial precision. Calibration and data acquisition were managed through the LabSpec 6 software. For hydrogen (H₂) response studies, the sample was placed in a vacuum-sealed gas chamber, with H₂ introduced through a mass flow controller to regulate gas flow precisely. The chamber was maintained under vacuum conditions to minimize interference from ambient gases, ensuring accurate detection of H₂-induced changes in conductivity.

First, two-terminal current–voltage (I–V) measurements were performed on 2H-MoTe₂ placed on a quartz glass substrate with indium tin oxide (ITO) electrodes. The aim was to study the material on an insulating substrate, which limits heat dissipation from the lattice and thus provides an optimal platform for investigating the electro-bolometric effect (Joule heating). Controlled heating was provided to the sample by placing it on a hot plate and the temperature was varied from 303 to 363 K in 5 K increments. Voltage (V_sd) was applied to one of the ITO electrodes while the other electrode was grounded. Due to the substrate's poor thermal conductivity, sufficient time was allowed between measurements to ensure that any temperature gradient between the sample and substrate was minimized. Figure 2(a) shows the temperature-dependent I–V characteristics of the device recorded over a range of temperatures. Below 1.3 V, the current increases with temperature indicating semiconducting behavior, while above 1.3 V, the current decreases with temperature suggesting a transition to metallic-like behavior. This behavior is further analyzed in Fig. 2(b), where the I–V characteristics at 303, 363, and 473 K clearly reveal two distinct regimes of transport. Below 1.3 V, the device exhibits semiconducting behavior, and above 1.3 V, it transitions to metallic-like behavior. In the semiconducting regime (below 1.3 V), as shown in Fig. 2(c), the I–V behavior is non-linear due to the presence of a Schottky barrier at the interface. However, this non-linearity diminishes with increasing temperature, leading to a more linear I–V response. This behavior arises because thermally activated carrier generation reduces the influence of the Schottky barrier, enhancing carrier transport. In the metallic-like regime (above 1.3 V), as shown in the inset of Fig. 2(c), the I–V response displays a linear relationship. In this regime, the applied voltage is sufficient to overcome localized states or barriers within the material, allowing free electrons in the conduction band to move freely. Consequently, the current increases proportionally with the applied voltage, confirming the linear I–V behavior expected for metallic conduction. Additionally, the results from the ln(I) vs 1/T plot [Fig. 2(d)] provide further evidence for the transition in conduction behavior. The analysis reveals a positive activation energy below 1.3 V and a negative activation energy above 1.3 V, as summarized in Table S2 in the supplementary material.²³ Below 1.3 V, the positive activation energy indicates thermally activated carrier generation, where increasing temperature facilitates the overcoming of the Schottky barrier and enhances carrier concentration in the conduction band. In contrast, above 1.3 V, the negative activation energy suggests that as temperature increases, the current decreases. This behavior can be attributed to dominant electron–phonon scattering effects, which impede electron transport in the metallic-like regime.

To determine the nature of this transition, we performed temperature- and voltage-dependent Raman spectroscopy on the sample. Initially, voltage-dependent Raman spectra were recorded at room temperature. For all values of V_sd, the E_2g peak (231 cm⁻¹) and the A_1g peak (170 cm⁻¹), corresponding to the 2H phase, were found to be more dominant [Fig. 3(a)]. These Raman peaks are characteristic of the semiconducting 2H phase, which exhibits out-of-plane and in-plane vibrations at 170 and 231 cm⁻¹, respectively.²⁴ In contrast, the metallic 1T phase, typically induced by structural transformation, exhibits distinct Raman peaks at 160 and 265 cm⁻¹, signifying changes in lattice symmetry.²⁵ Structural phase transitions in MoTe₂, such as the 2H to 1T transition, are well-documented in the literature and typically occur under extreme conditions. Thermal annealing studies report the transition at temperatures exceeding 873 K (600 °C), while laser heating requires localized power levels of approximately 10 mW.^13,26 In our study, the experimental conditions, with a maximum temperature of 473 K and a laser power of 3 mW, fall significantly below these thresholds for inducing a structural phase transition. A similar trend was observed in the temperature-dependent study [Fig. 3(b)], where no peak shifts were detected even at the highest temperatures tested. The absence of Raman peak shifts confirms that the observed transition from semiconducting to metallic-like is not associated with a structural transformation of the lattice. The reduction in intensity at higher temperatures and voltages can be attributed to increased electron–phonon interaction, which results in energy transfer from lattice to electrons.²⁷ Instead, the semiconducting-to-metallic transition in our study can be attributed to enhanced electron–phonon interactions driven by Joule heating. At higher voltages and temperatures, Joule heating increases lattice vibrations (phonons), leading to more scattering of conduction electrons, which results in metallic-like behavior. This interpretation aligns with the reduction in Raman peak intensity at higher temperatures and voltages, which reflects energy transfer from the lattice to the electrons rather than any changes in lattice structure.

For V_sd < 1.3 V, Joule heating is minimal, making the sample's resistance likely to be dominated by impurity scattering. However, when V_sd exceeds 1.3 V, the current in 2H-MoTe₂ becomes substantial enough to induce significant Joule heating, with electron transport becoming dominated by phonon scattering.

To validate this hypothesis, the sample was exposed to H₂ gas, and changes in conductivity were observed in both the semiconducting (V_sd = 1 V) and metallic (V_sd = 5 V) regimes. The study was carried out in vacuum and at 373 K to ensure that the sample's surface was clean for the H₂ gas to interact. H₂ gas, known for its negative Joule–Thomson effect (i.e., it heats up when expanding in a vacuum), is expected to transfer its heat of expansion to the 2H-MoTe₂ lattice, increasing conductivity in the semiconducting phase and decreasing it in the metallic phase. As expected, 2H-MoTe₂ exhibited a positive response in the semiconducting regime and a negative response in the metallic regime (Fig. 4). In the semiconducting regime, where impurity effects dominate, the heat generated by the expanding H₂ gas is gradually absorbed by the lattice, leading to an increase in carrier concentration through thermal excitation and hence the conductivity increases. In the metallic regime, dominated by phonon scattering, the interaction with H₂ gas increases the phonon density, resulting in decreased conductivity due to the increase in electron–phonon interaction. Both the positive and negative H₂ responses are slow, likely due to the sample's very small cross-sectional area, which limits the number of H₂ molecules that can interact with it within a given time frame.

To further support our observations, the sample is heated using a 532 nm laser and its photoresponse characteristics were investigated at low and high voltages. At low voltage (−1 V), a positive photoresponse was observed, with the channel current increasing upon exposure to the laser light [Fig. 5(a)]. This is because of the lattice getting heated-up due to laser light. Because the laser delivers significantly more energy than H₂ gas, the lattice temperature increased sharply, intensifying phonon scattering and resulting in sharp transitions. As a result, the channel resistance decreased during laser exposure and quickly recovered once the laser was turned off. The variations in current under laser illumination are due to thermal effects caused by laser-induced heating. To further investigate this behavior, we conducted additional photoresponse studies at a low temperature of 12 K. At this temperature, the laser-induced heat dissipates more effectively to surroundings. As a result, the device exhibited a consistent current response under laser illumination [Fig. 5(c)], indicating that the observed variations at 303 K are influenced by thermal factors caused by laser heating. When the measurement was repeated at a higher voltage (−3 V), no change in the channel current was observed [Fig. 5(b)]. This lack of response is likely due to the fact that in the Joule heating-dominated regime, the additional heating effect of the laser is less significant, leading to little or no change in phonon density upon laser exposure and, therefore, no observable photoresponse in the sample. However, 2H-MoTe₂, with its ability to transition between semiconducting and metallic-like states under Joule heating, shows promising potential for infrared sensing and thermal imaging, where temperature-induced changes in material properties play a key role in device performance. Its bolometric properties enable precise temperature detection, making it ideal for applications in security, medical diagnostics, and environmental monitoring. Additionally, its tunability through temperature changes allows effective integration into thermal imaging systems for industrial and biomedical applications. Beyond sensing, the ability of 2H-MoTe₂ to modulate resistance through thermal excitation makes it a promising material for neuromorphic computing, where resistance switching mimics synaptic behavior for brain-like data processing.

Having established that Joule heating is responsible for the metallic behavior beyond the threshold voltage, exploring the material's response as a function of electron density becomes intriguing. To modulate electron density in 2H-MoTe₂, a gate voltage (V_g) was applied via the MLG back gate, with h-BN serving as the dielectric [see Fig. 1(d)]. As V_g was varied from −5 to +5 V, a significant increase in channel current was observed [see Fig. 6(a)], confirming that electrons are the primary carriers contributing to the current. The experiment was repeated at an elevated temperature of 373 K to investigate any signs of a positive temperature coefficient of resistance. It was observed that the channel current at 373 K was higher across all ranges of V_ds and V_g compared to the current observed at room temperature [see Fig. 6(b)]. It indicates that there is no Joule heating-induced crossover from semiconducting to metallic-like regimes. For enhanced clarity, Fig. 7 presents a comparison of channel current in 2H-MoTe₂ at two different temperatures across three distinct gate voltages (V_g). Adjusting V_g indirectly modulates electron–phonon coupling by altering electron density. At positive V_g, the increased electron density enhances the screening effect, which in turn reduces the coupling strength. Conversely, at negative V_g, the lower electron density decreases the electron density leading to stronger electron–phonon coupling. As shown in the figure, the sample consistently exhibited semiconducting behavior for all V_ds values, regardless of changes in electron density. The absence of Joule heating-induced semiconducting-to-metal-like behavior can be attributed to the presence of h-BN and MLG layers beneath the 2H-MoTe₂ channel. Due to their high thermal conductivity, these materials efficiently dissipate heat from the channel, minimizing the impact of Joule heating even at higher V_ds values, thus preventing any crossover behavior. To further investigate, we conducted I–V measurements on the h-BN substrate to explore whether transitions could occur under higher Joule heating. Our results show that on the glass substrate, with its low thermal conductivity (∼1 W/m K), the transition occurs at 1.3 V and 12 μA shown in Fig. S1(a) in the supplementary material.²⁸ In contrast, on h-BN, with much higher thermal conductivity (∼300 W/m K), significantly higher voltage and current were required to induce the transition, at 2.8 V and 20 μA shown in Fig S1(b) in the supplementary material.²⁹ These findings offer valuable insights for device design, highlighting the importance of selecting substrates with tailored thermal properties. For low-power applications, substrates like glass enable transitions at lower voltages, while high-thermal-conductivity substrates like h-BN provide stability under high current and thermal loads.

In summary, we have demonstrated that the transition from semiconducting to metallic-like behavior in 2H-MoTe₂ can be induced by electrically tuning the channel current through Joule heating. At low voltages, the material exhibits semiconducting behavior, where thermal excitation increases the carrier concentration. Exposure to H₂ gas further enhances this effect, leading to increased conductivity. However, light exposure at low voltages results in a positive photoresponse, as thermal excitation from laser heating increases the carrier concentration. At higher voltages, significant Joule heating raises the lattice temperature, driving the material into a metallic-like state, where electron–phonon scattering increases, reducing conductivity. Consequently, H₂ gas exposure at high voltages leads to a negative photoresponse, as the gas adds extra heat and intensifies phonon scattering. Additionally, at higher voltages, the lattice temperature is already elevated, so the additional laser-induced heating has minimal effect, resulting in no noticeable photoresponse. The h-BN substrate, with its higher thermal conductivity, requires more Joule heating to induce the transition compared to the glass substrate. Our findings provide valuable insights into the lattice dynamics of 2H-MoTe₂, which could inform the optimization of optoelectronic device performance.

See the supplementary material for additional data supporting the findings presented in this study. Figure S1 illustrates the temperature-dependent I–V characteristics of 2H-MoTe2 on (a) glass and (b) h-BN substrates. Table S2 presents the activation energy values for different voltage regimes.

The work was supported by the Cross-Disciplinary Research Framework (CDRF) grant (Grant Reference No. C1/23/200) instituted by BITS-Pilani University. C. Malavika acknowledges the financial support from the TIFR Hyderabad. The authors thank the Central Sophisticated Instrumental Facility (CSIF) at BITS-Pilani K. K. Birla Goa Campus for providing access to experimental facility.

The authors have no conflicts to disclose.

B. Manoj Kumar: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Writing – original draft (equal). Shreya Gaonkar: Data curation (equal); Investigation (equal); Methodology (equal). Rashed H. Lone: Data curation (equal); Investigation (equal); Methodology (equal). C. Malavika: Investigation (equal); Methodology (equal). E. S. Kannan: Funding acquisition (lead); Supervision (lead); Writing – review & editing (equal).

The data that support the findings of this study are available from the corresponding author upon reasonable request.