Hexagonal MoTe2 with Amorphous BN Passivation Layer for Improved Oxidation Resistance and Endurance of 2D Field Effect Transistors

Environmental and thermal stability of two-dimensional (2D) transition metal dichalcogenides (TMDs) remains a fundamental challenge towards enabling robust electronic devices. Few-layer 2H-MoTe2 with an amorphous boron nitride (a-BN) covering layer was synthesized as a channel for back-gated field effect transistors (FET) and compared to uncovered MoTe2. A systematic approach was taken to understand the effects of heat treatment in air on the performance of FET devices. Atmospheric oxygen was shown to negatively affect uncoated MoTe2 devices while BN-covered FETs showed considerably enhanced chemical and electronic characteristic stability. Uncapped MoTe2 FET devices, which were heated in air for one minute, showed a polarity switch from n- to p-type at 150 °C, while BN-MoTe2 devices switched only after 200 °C of heat treatment. Time-dependent experiments at 100 °C showed that uncapped MoTe2 samples exhibited the polarity switch after 15 min of heat treatment while the BN-capped device maintained its n-type conductivity for the maximum 60 min duration of the experiment. X-ray photoelectron spectroscopy (XPS) analysis suggests that oxygen incorporation into MoTe2 was the primary doping mechanism for the polarity switch. This work demonstrates the effectiveness of an a-BN capping layer in preserving few-layer MoTe2 material quality and controlling its conductivity type at elevated temperatures in an atmospheric environment.

TMDs layers within electrically insulating and dielectric (for FET applications) thin layers, which can also prevent oxidation and deterioration of 2D TMD characteristics, is of a significant practical interest to advance 2D TMDs for device applications. The challenge of such passivation layers is in avoiding interaction of the 2D TMD layers with the adjusted electrically insulating materials, where typical thin dielectrics are mostly based on oxides (e. g., SiO 2 , Al 2 O 3 , HfO 2 ), which intrinsically bring oxygen atoms in contact with TMDs and can contribute to TMD deterioration. This becomes especially prevalent if elevated temperatures are encountered during dielectric layer deposition or device operation. Thus, an exploration of a non-oxide based and environmentally stable dielectric material to be used in conjunction with TMD FET devices, or other 2D materials such as graphene and black phosphorus, and would benefit the wider usage of such materials.
Boron nitride has emerged as a dielectric counterpart to TMDs due to its chemical stability, lack of dangling bonds, and wide bandgap (5.97 eV) characteristics 17 . Most notably, hexagonal boron nitride (h-BN), which has a similar crystal structure to graphene 18 , can be exfoliated onto various 2D materials to form heterostructures such as h-BN-MoS 2 [19][20][21] , h-BN-WS 2 22 , h-BN-MoSe 2 23 , and BN-WSe 2 -BN 24 with remarkable electrical properties 25 . While these studies clearly demonstrate the benefit of h-BN packaging with TMD for improved mobility, on/off ratio, and thermal management, the challenge of exfoliation routes for such 2D heterostructure device assemblies is prohibitive for the scalable and cost-effective device production, leading to the search of alternative routes for h-BN synthesis and integration with TMD structures 18 . Deposition of h-BN layers via chemical vapor deposition routes have been extensively explored but require high temperatures during synthesis and annealing [26][27][28][29] , reducing applicability for temperature sensitive TMDs.
Most recently, research involved in physical vapor deposition (PVD) of BN thin films has demonstrated the advantage of room temperature growth of amorphous BN (a-BN) layers on the order of 2 nm to 3 nm thickness over large areas, which show a bandgap of 4.5 eV, dielectric constant of about 6 and breakdown voltage on the order of 10 MV/cm 30,31 . This makes them comparable to exfoliated h-BN for dielectric applications. Passivation layers of PVD-grown a-BN in combination with graphene 32 , 2D WS 2 and MoS 2 33 , have shown improved field effect mobility and reduced background charge carrier concentration. Herein, we investigate devices composed of exfoliated 2 H MoTe 2 with a PVD-grown a-BN passivation layer to study chemical and structural stability of 2D MoTe 2 at elevated temperatures. The studies also include an analysis of FET device performance (including n-to p-type behavior transition, on/off ratio, and carrier mobility) after heating in air, covering a possible range of temperatures of such 2H-MoTe 2 /a-BN device applications in electronic devices. Formation of Te vacancy sites and the potential for atmospheric oxygen to form substitutional dopants were of critical attention in the performed study. We find this BN capping layer to be an effective barrier layer, which protects the 2D TMD material from detrimental chemical oxidation, and significantly reduces the rate of field effect mobility and on/off ratio degradation of FET devices as compared to the uncapped sample.

Results and Discussion
Material exfoliated and characterization. Figure 1a depicts an optical image of an exfoliated 2H-MoTe 2 flake. The optical contrast of few layers and bulk flakes on SiO 2 /Si substrates provides a good guide to the flake thickness, since it is known that 2D flakes that contain one to few layers show a darker optical color than bulk material due to altering optical contrast with the SiO 2 /Si substrate 34 . AFM from a non-contact scan is shown in Fig. 1b, demonstrating good thickness uniformity of the dark-green area from Fig. 1a. A line scan measures a height difference of 2.30 nm, which corresponds to tri-layer MoTe 2 35 . The Raman spectrum in Fig. 1c, clearly shows three distinct peaks typical for MoTe 2 at 170 cm −1 , 235 cm −1 and 289 cm −1 , corresponding to the A 1g , E 1 2g and B 2g phonon modes, respectively 36 . The A 1g mode is due to out-of-plane phonon vibrations and is typically weak for 532 nm laser excitations 37 . The E 1 2g peak is the dominant mode and corresponds to in-plane atomic oscillations, and it has been shown to exhibit a strong signal for MoTe 2 in the 2D form and is relatively weak for bulk material 36 . The B 2g mode corresponds to the out-of-plane interaction of adjacent layers and has an interesting property in which it is clearly present for 2 to 5 of MoTe 2 layers but absent for either monolayer or the bulk thickness. This peak was reported to have its strongest value for 2 layers and rapidly decreasing as the thickness increases until it disappears after 5 layers 34 . Therefore, the relative intensities of the E 1 2g and B 2g peaks can be used to determine layer thickness and verify AFM measurements as presented in Fig. 1b. For all MoTe 2 flakes selected for this study, the Raman spectra displayed a strong E 1 2g peak and less intense A 1g and B 2g peaks, which is in agreement with 3-layer thickness measurements with AFM (Fig. 1c).
In order to analyze the 2D MoTe 2 thermal degradation rate over time, Raman spectra comparing the 2H-MoTe 2 E 1 2g mode for samples heated at 100 °C for up to 60 min with 10 min increment were collected and are shown in Fig. 2a. Gauss peak fitting of the E 1 2g mode had determined that this vibration mode maintains its location at ≈234 cm −1 throughout the experiment, which suggests the 2H-MoTe 2 did not undergo a thickness change after 60 min of 100 °C heat treatment in air. However, the E 1 2g peak exhibits a continuous intensity decline from its max value at room temperature, and the rate of this decline is strongly affected by the presence of the BN cap layer. The Raman analysis did not show the presence of MoO 3 or MoO 2 vibration modes for both uncapped and a-BN capped 2H-MoTe 2 flakes, suggesting that no individual phases of molybdenum oxide crystal structure was formed. However, XPS analysis discussed below shows that heating in air had resulted in surface oxidation. Normalizing the E 1 2g peaks to the Raman peak of silicon at 520 cm −1 allows for comparison between uncapped and BN capped flakes. Figure 2b clearly shows that the uncapped MoTe 2 exhibits a much greater rate of decay of the E 1 2g Raman mode. Such is likely brought by damping the inter-layer oscillation modes with increased lattice imperfections and internal atomic stresses from Te vacancies and oxygen substitutions. In contrast, normalized Raman spectra of the BN-capped MoTe 2 sample maintained its E 1 2g peak intensity after 60 min of heating, see Supplementary data Figure S1. Applying a linear fit to the data, the slopes of the fitting lines suggest that the decay rate for the E 1 2g mode in 2H-MoTe 2 material with 100 °C exposure time is an order of magnitude higher than that for the 2H-MoTe 2 /a-BN, (8.3 × 10 −3 min −1 versus 8.1 × 10 −4 min −1 , respectively). The results clearly demonstrated the benefit of the BN capping layer for the preservation of 2H-MoTe 2 structure at elevated temperatures in air.
In order to investigate in more detail, the chemical state changes of MoTe 2 after exposure to elevated temperatures in air, XPS analysis was conducted on exfoliated flakes with and without BN layers after heating at 100 °C for 60 min. The results are presented in Fig. 3, and XPS spectra of MoTe 2 flakes before heating can be found in Supplemental Figure S2. The uncoated MoTe 2 flake shows three distinct peaks for the Mo 3d orbital, which can be seen in the top spectrum of Fig. 3a. The recorded spectra deconvolutions with 3d doublets reveal a clear presence of two chemical states with Mo 3d 3/2 peak locations at 234.6 eV and 231.5 eV. The higher binding energy doublet belongs to Mo 6+ in MoO 3 bonding, and the lower one to Mo 4+ in MoTe 2 38 . This suggests partial oxidation of the MoTe 2 surface. In a sharp contrast, similar analysis of the MoTe 2 /BN sample after 60 min exposure to 100 °C in air shows only one doublet with Mo 3d 3/2 peak location at 231.4 eV, indicating the absence of molybdenum oxidation when the BN capping layer was applied. Similar observation of the BN-capping layer preventing oxidation can be made from corresponding XPS spectra of Te 3d, shown in

Electrical Measurements of Uncapped and BN-capped devices.
To determine the effect of an a-BN capping layer on the retaining charge transport characteristics and MoTe 2 based FET device performance upon exposure to elevated temperature in air back-gated FET devices were fabricated from few-layer thick MoTe 2 flakes on SiO 2 Figure 2. (a) Raman spectra of the E 1 2g peak intensity variation for few monolayers thick 2H-MoTe 2 flakes exposed to 100 °C in air for the shown time intervals. (b) Comparison of E 1 2g peak intensities changes as a function of 100 °C in air holding time for uncapped MoTe 2 and BN-capped MoTe 2 flakes; to facilitate the comparisons and avoid errors due to absolute peak intensity changes over time, the peak intensity was normalized to the Si substrate peak (520.5 cm −1 ). The shown linear fitting was used to calculate E 1 2g peak intensity decline due to oxidation processes for 2H-MoTe 2 and 2H-MoTe 2 /a-BN.
ScIEntIFIc REPORTS | (2018) 8:8668 | DOI:10.1038/s41598-018-26751-4 using electron beam lithography. This amorphous BN layer was previously determined to be pin-hole free dielectric material with breakdown characteristic and dielectric constant exceeding CVD grown h-BN films, approaching that of mechanically exfoliated h-BN 30,31 . Such deposition method is ideal for a cost-effective application of dielectric BN layer for 2D heterostructures as compared with other 2D BN growth methods 26 . In the recent studies of PLD grown amorphous BN it was found that the dielectric properties were close to that of h-BN and be as smooth as underlying substrates, where graphene, MoS 2 , WS 2 and a variety of metal and ceramic substrates were tested with an equal result in a-BN smoothness and pin-hole free morphology over several cm 2 areas 30,31,39 . While there were no systematic studies to determine a-BN inertness, the earlier explorations with a-BN/graphene heterostructures had shown a significant increase of mobility in graphene monolayers when it was sandwiched with a-BN, relating this to a-BN inertness, similar to monolayer thin h-BN 32 . Stabilization of charge transport characteristics of epitaxially grown graphene in air by the application of a capping a-BN layer was also reported 40,41 . We then reasonably expected a similar effect with MoTe 2 devices, and such is explored for the first time in this study at elevated temperatures in air. Back-gated FET device schematic made from MoTe 2 and including a BN capping layer is depicted in Fig. 4a and an example of such fabricated device is shown in an optical image of Fig. 4b.
Electrical performance of back-gated FET devices annealed from room temperature to 300 °C is shown in Fig. 4. MoTe 2 /BN and MoTe 2 as-exfoliated devices underwent the side by side elevated temperature in air exposure and were held for 1 min at the target temperature in atmospheric conditions before cooling down and measuring their performance. Figure 4c presents a semi-log plot of drain current (I dc ) versus gate voltage (V g ) for a MoTe 2 device as the annealing temperature was systematically increased from room temperature to 300 °C at 50 °C increments. The device initially exhibits n-type behavior as indicated by the positive slope, with the maximum current decreasing slightly as the temperature is increased to 100 °C. At 150 °C (pink curve) the polarity switches to p-type behavior as indicated by the negative slope. At 200 °C, the device operates as a p-type but also exhibits semi-metallic behavior as indicated by the linear slope. For few-layer materials, this temperature is sufficient to sublime Te directly into the vapor state, leaving behind Mo atoms which would contribute to metallic properties.
A semi-log I dc -V g plot for a MoTe 2 /BN device is presented in Fig. 4d. The drain current curve remains very stable up to 100 °C. At 150 °C there is a sharp decline in current, however the device remains n-type, as indicated by the drain current curve slope for positive gate voltages. However, after heat treatment at 200 °C, the drain current switches to negative gate voltages and decreases for positive gate voltages, indicating a switch to p-type semiconductor operation. Once temperatures reached 250 °C we noticed the channel lost semiconductor properties and was unresponsive to gate voltage variation, which suggests degradation of MoTe 2 . At 300 °C, the Ti/Au contacts themselves failed and further temperature increase was not performed. Scanning electron microscopy studies of the contacts had shown dewetting of these thin metal films (Supplementary data Figure S3), resulting in a failure of conductive pathways. Such dewetting is expected for thin electrode materials due to the low melting point of gold, and experiments at higher temperatures may require other electrode materials. Field effect mobility was calculated from the I dc -V g curves and is shown at various annealing temperatures in Table 1. Before performing any temperature exposure, the measurements of as fabricated devices (23 °C in Table 1) clearly indicate that the capping MoTe 2 with BN has a significant effect on the mobility improvement as the measured values for MoTe 2 /BN devices were five times greater than that of uncapped MoTe 2 . The absolute values of the room temperature mobility measurements in the literature vary between 0.03 and 20 cm 2 V − 1 s −1, 11,12,14,42,43 . While our results are within this reported range, our study demonstrates the effect of a-BN capping layer on the device stability during heating as discussed below.
The field-effect mobility of BN capped MoTe 2 (red) is maintained around 0.8-1 cm 2 (V s) −1 until 150 °C where it experiences a sharp drop to 0.2 cm 2 (V s) −1 . Furthermore, after 200 °C the mobility values are shown to switch polarity from n-to p-type. The un-coated MoTe 2 in contrast, has a much lower initial mobility and undergoes its polarity switch already at 150 °C. Finally, at 200 °C the mobility absolute value increases slightly. The switch from initial n-type behavior to p-type can be linked to the oxygen incorporation in MoTe 2 layers found from XPS studies (Fig. 3). Previous reports of DFT simulations of oxygen interactions with MoTe 2 and formation of Mo-O and Te-O bonds can lead to the formation of deep-level trap states located 0.4 eV below the conduction band minimum 44 . Such states may act as acceptor sites, effectively resulting in a switch to the p-type behavior as the number of oxygen substitutes for tellurium in MoTe 2 lattice is increased at higher temperatures in air. This is further supported by the delayed polarity switching of our experiments with BN covered device which protects MoTe 2 from oxygen diffusion and subsequent substitution in Te sites.
To study the effects of the BN layer for improved endurance of MoTe 2 device characteristics in oxidizing environments, the FET devices were held at a constant temperature of 100 °C for incrementally increased times from 5 min to up to one hour. I dc -V g curves displaying the change in electrical properties of MoTe 2 and MoTe 2 /BN over increased time of exposure to 100 °C in air are displayed as semi-log plots in Fig. 5a,b, respectively. The uncapped   Fig. 4. Mobility units are in cm 2 (V s) −1 .
MoTe 2 devices (Fig. 5a) exhibit a significant decrease in maximum drain current after the first 5 minutes, suggesting that the surface oxidation occurs almost immediately. Initially the device is active in the positive gate voltage region, indicating n-type behavior. However, as the heating time increases, the maximum drain currents decrease and transitions to the negative gate voltage region, indicating p-type behavior. Along this n-to p-type transition (about 10-15 min) the devices exhibit ambipolar behavior. For MoTe 2 /BN devices, (Fig. 5b), the maximum drain current also declines with greater heating times but more steadily and remains non-zero even after 60 minutes exposure to 100 °C in air. In addition, the MoTe 2 /BN devices do not undergo a polarity switch to p-type and never lose the n-type behavior (Fig. 5b). In Fig. 5c, the field effect mobility determined after heating to 100 °C in air is plotted to investigate MoTe 2 FET device endurance and the protective effect of the a-BN capping layer. The initial (before heating to 100 °C in air) values for electron-based mobility were used to normalize the data plot in order to demonstrate the rate of change of the device performance over time. Both MoTe 2 and MoTe 2 /BN devices undergo a rapid mobility decline relative to their initial mobility after first 10 minutes before leveling off at 20 min to 60 min, suggesting that surface changes to MoTe 2 occur very quickly. The uncapped MoTe 2 devices show a larger decline in field effect mobility compared to MoTe 2 /BN devices. In addition, the MoTe 2 /BN devices retain their n-type performance over the entire test period without failing, confirming that the a-BN capping is a good choice for improving long-term device performance in addition to the improved overall mobility values discussed earlier. The on/off ratio change with exposure to 100 °C in air is presented in Fig. 5d. The MoTe 2 /BN device maintains a significantly higher ratio of around 10 2 throughout the entire 60 min experiment time, while the on/off ratio for uncapped MoTe 2 device rapidly decreases.
In conclusion, durability of future 2D TMD devices will rely on solutions to prevent material degradation during their fabrication and operation. Moderate heating of uncapped exfoliated 2H-MoTe 2 in air shows a considerable oxidation of few-layer thick MoTe 2 surface resulting in deterioration of the FET device performance. The oxidation can be substantially mitigated by capping the 2H-MoTe 2 with a 10 nm thick a-BN layer, applied by pulsed laser deposition. For uncapped 2H-MoTe 2 , the temperature induced oxidation leads to MoTe 2 structural defects with oxygen incorporation and appearance of Mo-O and Te-O bonding in XPS spectra. However, the absence of MoO 3 structure in the Raman data signals that partial oxidation occurred in the form of oxygen defects in Te vacancy sites or physio-absorption to Te surface sites. This is accompanied by the reduction in the field effect mobility and an eventual switch from an n-to p-type semiconductor behavior at around 150 °C. Devices consisting of 2H-MoTe 2 with an a-BN passivation layer were shown to have improved field effect mobility characteristics and significantly suppress material degradation via oxidation from heating in air. For fabricated MoTe 2 /BN FET devices, the oxidation induced polarity switch was delayed to a temperature of 200 °C and as such, retained their n-type mobility and stabilized on/off ratio in extended exposure to 100 °C in air. Hence, amorphous BN was found to be an effective approach for preventing 2H-MoTe 2 oxidation and improving the 2D FET device endurance and overall performance. Considering a large area scalability, room temperature growth and versatility of the PLD produced a-BN passivation layer, the explored 2H-MoTe 2 /a-BN structures can be beneficial for electronic and optoelectronic device applications.

Methods
Preparation of 2H-MoTe 2 . 2H-MoTe 2 single-crystals were produced by the chlorine-assisted chemical vapor transport (CVT) method. In brief, a vacuum-sealed quartz ampoule with polycrystalline MoTe 2 powder and a small amount of TeCl 4 (4 mg/cm 3 ) was placed in a furnace containing a temperature gradient so that the MoTe 2 charge was kept at 800 °C, and the temperature at the opposite end of the ampoule was about 710 °C. The ampoule was slowly cooled after 7 days of growth. The 2 H phase of the obtained MoTe 2 flakes was confirmed by powder X-ray diffraction and transmission electron microscopy studies.
MoTe 2 Device Fabrication. 2H-MoTe 2 flakes were then later mechanically exfoliated onto heavily p-doped silicon wafers with a 270 nm silicon oxide surface layer using conventional adhesive tape methods, the details of which have been reported previously 45 . Candidate 2D flakes were identified using optical microscopy since the SiO 2 /Si substrate provides a clear optical contrast for single, double, and few-layer flakes compared to bulk 10 . Flake thickness was estimated using Raman spectroscopy and further confirmed using atomic force microscopy (AFM). Back-gated FETs with channel widths of 2 µm were fabricated by evaporating Ti/Au (20 nm/20 nm) contacts onto the MoTe 2 flakes defined by electron beam lithography (EBL) and lift-off. a-BN Deposition. Some of the devices were covered, post device fabrication, by an ultrathin amorphous BN layer. This a-BN capping layer of 10 nm thickness, was deposited by pulsed laser deposition using a BN target (99.999%), ablated with a focused 248 nm KrF laser in nitrogen background gas under processing parameters optimized for a fully dense, pin-hole free, and stoichiometric BN film growth as reported previously 30,39 . The a-BN capping layer growth covered the entire device and was deposited at room temperature. We did not notice an effect to the crystallinity of the MoTe 2 or deposited Ti/Au electrodes.
Electrical Measurements of Heated Devices. The electrical measurements were performed using an Agilent B2902A system equipped with a probe station. The 10 nm a-BN layer was completely pierced by the contact probes to make good metal contact with source and drains electrodes of the devices.
Heat treatment experiments were conducted using a hotplate in laboratory air of about 50% relative humidity. Two sets of heating experiments were performed: i) temperature dependent measurements where samples were heated from room temperature (≈23 °C) to 300 °C with a 50 °C increment and held for 1 min before cooling, and ii) time dependent measurements where samples were held at 100 °C with 5-min time increments up to 60 min before cooling down. Back-gated FET devices with both open and a-BN-capped MoTe 2 surfaces were heated simultaneously to ensure experimental consistency. Before taking I-V characteristics, the devices were allowed to cool to room temperature to prevent thermal effects on the measured results and allow direct comparisons of device performances after heat treatment.