Long-term aging of CVD grown 2D-MoS2 nanosheets in ambient environment

A chemically vapor deposited MoS2 nanosheets (NSs) is aged in the laboratory at ambient and at 40% average humidity for ~36 months. Nanorods of few microns in length and few nanometers in diameter are found to grow from the MoS2 seeds. They have been growing as a result of the chemical reaction between the MoS2 NSs and ambient O2 and moisture, they exhibit an amorphous phase structure in the stoichiometric form of MoO3. Density functional theory simulations further reveal the role of H2O and O2 in the transformation of the MoS2 NSs. The adsorption energy of O2 molecules on the MoS2 sites is Ead = −1.09 eV as compared to lowest absolute Ead = −0.10 eV of H2O indicating the favorable adsorption of O2 and subsequent Mo oxidation. This study provides valuable insight into the aging phenomenon of MoS2 exposed to O2 and moisture which might limit their application.


INTRODUCTION
Molybdenum disulfide (MoS 2 ) has emerged in the last decade as one of the most promising two-dimensional (2D) transition metal dichalcogenide (TMD) materials for its excellent mechanical and unique optical properties. This has spurred the interest around MoS 2 making it highly desirable for optoelectronic applications [1][2][3][4] . Several attempts are currently in progress to build devices based on MoS 2 such as sensors, photodetectors, flash memories, batteries, field emitters, rectenna for energy harvesting, etc. [5][6][7][8][9][10] . Additionally, the low frictional coefficient of MoS 2 triggered its use as a lubricant in aerospace-related applications 11 . Nonetheless, prior to its implementation in realistic applications, it is necessary to evaluate its mechanical, chemical, and structural stability over a long period of time, mimicking the actual use of devices in ambient environment. In a study conducted by Budania et al. 12 , the long-term (~7-8 months) stability of mechanically exfoliated MoS 2 flakes was inspected in air and vacuum. Small patches were observed on the flakes left in ambient air for about 55 days, whereas no change was identified for samples stored in vacuum. White patches were ascribed to oxidized MoS 2 sites with higher Mo concentration. In a separate study by Gao et al. 13 , the stability of MoS 2 and WS 2 exposed to ambient air (6-12 months) was investigated, suggesting the presence of a gradual oxidation of these compounds along with an extensive cracking and changes in their morphology. These transformations were attributed to the presence of oxygen and moisture in the environment leading to the oxidation of the flakes. Other authors ascribed the degradation of chemical vapor deposition (CVD) grown monolayer MoS 2 flakes to the presence of S vacancies 14 . These authors proposed the control of S defect density by tuning the CVD growth condition to ensure a long-term stability of the material.
It was also proposed that the aging effect can be avoided by capping the material with a protecting polymer layer. Besides, Yao et al. 15 have reported that the preheating of a monolayer MoS 2 sample can accelerate its degradation process. In addition, the surface oxidation of MoS 2 was reported to evolve during the sample exposure to moisture 16 . In this case, the oxidation process led to partial etching of the MoS 2 surface, which constitutes a very important finding since MoS 2 is considered as a potential water splitting catalyst. Thus, its instability in water could compromise its implementation in water splitting devices. Indeed, as reported earlier, the degradation of the film/flake quality can eventually affect the performance of their electrical, optical or tribological properties 13,17,18 . Nonetheless, it is worth noting that the degradable nature of MoS 2 (or in general the TMD materials) could be held as a potential candidate for bioabsorbable electronics 19 .
Although reported investigations on the degradation of MoS 2 revealed very interesting findings, challenging the initial assumption of its stability in ambient environment, their extent has few limitations in different aspects. First, all reported modifications induced through aging on MoS 2 were studied for a relatively shorter amount of time. Even though changes have been identified, affecting the applicability of MoS 2 , the aged material could be beneficial for other applications such as in bioabsorbable electronics 19 or could be a potential candidate for charge storage 20 , gas sensing 21 or electromagnetic wave absorption 22 . Second, most of the available studies have mainly assessed the surface dependent structural modifications of MoS 2 layers. That is, the morphological state of surfaces showed the formation of cracks and oxidation-related patches determined by scanning probe microscopy methods. In addition, Raman spectroscopy and imaging, as well as XPS investigations were used to comprehend chemical changes associated with those observed in morphology. Nonetheless, comprehensive structural studies investigating the microscopic origins of aging mechanisms through depth profiles of MoS 2 structures remain lacking. In the light of these limitations, our work presents a longer-term aging (~36 months) study carried out on vertically oriented MoS 2 material grown by CVD. Lately, the vertically aligned MoS 2 NSs are gaining great interest, their sharp edges offer more exposed sites for increasing physical-chemical interactions, making them suitable for various applications such as optoelectronics 23 , gas sensing 24 , and field emission devices 9 . Moreover, the vertically oriented MoS 2 sheets are found to exhibit similar optoelectronic response as monolayer MoS 2 4 . Interconnected MoS 2 nanosheets (NSs) were obtained by sulfurization process 4,25 and then exposed to the laboratory environment for a period of~3 years. Substantial modifications in the structural nature of the sample were observed, namely the "natural" growth of MoO x nanorods formed over time. The microscopic mechanism behind such a remarkable aging effect has been investigated by cross-sectional transmission electron microscopy (TEM). Raman and XPS characterizations were also performed to reveal the chemical changes associated with this aging mechanism. We used density functional theory (DFT) calculations to investigate the origins of the aging process, which has been successfully correlated to experimental results. Based on these understandings, we have finally undergone an artificially induced aging of the MoS 2 NSs using electron beam irradiations to enhance chemical oxidation reactions. Interestingly, we succeeded in reproducing an aging effect in nanorods demonstrating the structural and crystalline changes. This work offers valuable means for the control of very long MoS 2 stability in ambient environment, which is intended to help improving functional devices' nanoengineering strategies.

Structural characterization
Pristine MoS 2 NSs are shown in the SEM image in Fig. 1a. The pristine sample consisted of continuous and densely packed vertically oriented MoS 2 NSs, with a typical thickness of 5-20 nm, forming an interconnected network with a high degree of crystallinity, as demonstrated in our previous TEM study 9 . Exposed edges of the vertically grown NSs presented a tip-like morphology exhibiting high performing field emitting properties 9 .
Micrographs depicted in Fig. 1b, c showcase the top and side views, respectively, of the same sample after being exposed to ambient environment conditions for a period of~36 months. Imaging results clearly show overall morphology changes induced on the sample, where polygonal-shaped nanorods nucleate and grow from the MoS 2 NSs surface.
Nanorods are few tens of nm in diameter and few hundreds of nm to few microns in length. They are observed to nucleate and grow from the MoS 2 surface in a random orientation. Figure 2a shows the starting growth seeds of few typical nanorods, indicated by arrows pointing out their nucleation sites.
A cross-sectional view of a nanorod prepared using FIB milling process is shown in Fig. 2b. The measured height of the vertically oriented rod is 900 nm from the surface. The image provides a closer view on the intact nature of the nanorods with the MoS 2 surface.
To examine the nanorods composition, an ultra-thin crosssectional sample was prepared and further investigated in a TEM microscope. The side view of a nanorod lying on the MoS 2 surface is exhibited in Fig. 3a. Four regions can be clearly distinguished in the TEM image including the SiO 2 substrate (region 1), Mo thin film (region 2) with a sulfurized portion (region 3), and the nanorod portion (region 4). A magnified image collected from the nanorod region 4 is shown in Fig. 3b. The corresponding FFT image is presented in the inset. The HRTEM and FFT images reveal a mostly amorphous form of the nanorods with negligible appearance of randomly distributed nanocrystals. The degree of nano crystallinity increases with the electron beam fluence, which is further discussed later. Additionally, energy dispersive spectroscopy (EDS) was performed on the four regions and results are shown in Fig. 3c. The EDS analysis suggests that molybdenum   Moreover, Raman spectroscopy was conducted on the pristine and aged samples. Strong Raman peaks E 1 2g and A 1g , corresponding to common MoS 2 vibration modes, were observed for both samples (Fig. 4).
However, for the aged sample, additional small peaks at various locations are also recorded. Based on the typical vibration modes of MoS 2 , the obtained extra vibrational peaks are ascribed to MoO 3 structure (Table 1).
Interestingly, it is also observed that the difference in Raman shifts between the main MoS 2 peaks, i.e., Δω (E 1 2g -A 1g ), for the pristine sample is 25.2 cm −1 , whereas for the aged sample it turns to Δω = 26.8 cm −1 . This infers that the thickness of MoS 2 NSs has increased, which is coherent with the vertical shape growth. A throughout theoretical and experimental investigation on the Raman shifts and MoS 2 orientation relationship can be found here 26 .
In addition, a blue shift is also observed for the aged specimen, indicating a strain relaxation occurring in the sample due to the long-term aging. This indicates that during the aging, MoS 2 NSs have released some of their internal stresses that could be at the origin of the nanorods formation after long-term aging. The changes in Raman shifts induced by thermal annealing have been systematically investigated in more details on the MoS 2 monolayer 27 , these authors linked the changes in Raman shifts to the increasing defects. Similar observations have been also reported for GaN 28 and SiGe 29 strained samples. Using Tsang's model 30 , we have utilized the Raman shifts values to determine the induced strains, respectively in porous GaN and in aged SiGe. The calculated strains by this method were supported by other investigations such as X-ray diffraction and nano-indentation. Additionally, the intensity ratio of E 1 2g /A 1g is found to increase with aging, indicating the higher vibrations in the out-of-plane direction, which is coherent with the nanorods formation from MoS 2 NSs.
Finally, XPS measurements were carried out to analyze induced changes in the sample's surface chemistry following the long-term exposure to air and humidity. Figure 5 shows the XPS results obtained for the pristine and aged samples.
A small peak at 235.7 eV, corresponding to Mo 3d , suggests the presence of MoO 3 traces in the pristine sample. Its intensity is significantly increased in the aged sample, which correlates with the observed MoO 3 Raman peaks. S 2p spectra for both, pristine and aged samples, are the common characteristic peaks at~162 and~163 eV. Nevertheless, an additional peak appears at 168.7 eV for the aged sample, which corresponds to the S-O group. O 1s spectra in both cases show the characteristic peak at~532 eV with an additional peak at~530 eV corresponding to MoO 3 . The intensity of this peak is highly increased for the aged sample. XPS results clearly point toward a pronounced role of oxygen reaction with Mo and S in the aged sample. As it can be seen, the intensity of both 235.6 and 530.1 eV peaks associated with MoO 3 is higher for the aged MoS 2 sample, suggesting the oxidation of MoS 2 during the long-term aging. Similarly, it is also recorded in the aged sample that the intensities ratio of MoO 3 and S-O groups is high, which indicates a favored oxidation of MoS 2 NSs during the aging process.

Nanorods growth mechanism
It has been reported that MoS 2 flakes could undergo an oxidation leading to the formation of white speckles when exposed to ambient air containing oxygen and moisture 12 . It is also assumed that the oxidation process of the MoS 2 sites can take place at structural defects. Indeed, grain boundaries and point defects usually exhibit higher amount of dangling bonds promoting the oxidation and degradation of MoS 2 samples 13,19 .
In this study, the vertically grown MoS 2 NSs have a large number of exposed edges. Those edges are likely to retain a high degree of dangling bonds highly susceptible for oxidation. In general, the oxidation process results in the formation of SO 2 and MoO 3 , as shown by the equation below: In fact, SO 2 is a volatile compound, which can easily evaporate and disappear from oxidation sites. Nonetheless, XPS results showed a trace of few remaining S-O bonds on the aged sample. On the other hand, MoO 3 compound is a nonvolatile product, which remains on nucleation sites as a solid material, leading to the growth of nanorods. Their growth up to few microns in length during the 36 months aging period is an indication of the continuous quality and consistency of the oxidation process. Based on experimental evidence, we analyze the microscopic origin of the nanorods formation as a major consequence of the

DFT calculations
To further examine the stability and chemical surface reactivity of MoS 2 monolayers with O 2 and H 2 O under ambient conditions, we have performed thermodynamic and DFT calculations on oxidized MoS 2 in dry and humid air using Eq. (1) and the following equation 31 : The geometrical configuration of adsorbents on top of the MoS 2 monolayer plays an important role in favoring Mo oxidation processes. To evaluate this effect, we have positioned O 2 and H 2 O as adsorbents at four different potential initial sites of the MoS 2 supercell, as shown in Fig. 6. Adsorbent was considered as follow: on top of the Mo at site A, on top of the Mo-S bond at site B, on top of S at site C, and at the center of the hexagon at site D. For each configuration we let the system relax until reaching the minimum energy.
After stabilization of absorbents (O 2 and H 2 O gas molecules) at the considered positions with minimum energy requirements, the obtained configurations are further used to conduct a theoretical study of their adsorption behaviors on pristine MoS 2 monolayer via first-principles calculations 32 .
Adsorbed O 2 , and H 2 O molecules are constructed by positioning a specific gas molecule vertically and horizontally (parallel) to the surface of the compound, as detailed in Tables 2 and 3, respectively. The distance between the MoS 2 surface and first atoms of the gas molecules was set to 2.5 Å before optimization.
Using first-principles calculations, we examine the reaction of MoS 2 surface with the considered gas molecules, where adsorption energies E ad are calculated using the following relation 33,34 : where E MoS2þmolecule is the total energy of MoS 2 with the adsorbed gas molecule, E MoS2 , and E molecule are the total energies of pristine MoS 2 and isolated gas molecule, respectively. A positive value of the adsorption energy indicates that the adsorption is endothermic, meaning that the reaction is energetically unfavorable. In contrast, a negative value of the adsorption energy corresponds to an exothermic reaction, which is energetically favorable.  Table 4). After adsorption on the MoS 2 surface, molecules show some deformations at the vicinity of Mo atoms leading to the formation of MoO 3 at almost all possible positions. Hence, our DFT   34 , supports our analysis and interpretation of the microscopic aging mechanism. That is, the aging-dependent oxidation is a continuous process at intact Mo nanoscopic sites resulting from the initial growth mechanism of MoS 2 NSs. The originality in this work consisted in observing a random growth of merely MoO 3 nanorods, which has been made possible by the large extension of the aging period beyond the 18 months cap, as previously reported in literature. The present results suggest the possibility of even further structural and morphological changes to be induced on MoS 2 samples for even longer aging periods, which would result from prolonged oxidation processes. Furthermore, we have conducted accelerated aging experiments inside electron microscope (see Supplementary Note). The experiments consist of time-lapse assessment of the MoS 2 film/ MoO 3 nanorods degradation under electron beam irradiation. These experiments provide an important outlook of a potential temperature induced aging/degradation of MoS 2 -MoO 3 when used in harsh environments.
It is worth noticing that the MoS 2 -MoO 3 composite materials has various potential applications in charge storage 20 , gas sensing 21 , and electromagnetic wave absorption 22 . Given the promising nature of this composite and these associated technological applications, its stability under various environments is of significant impact. In this regard, our accelerated aging experiments provide further valuable insights into the unstable nature of such composites when used under harsh environments.

DISCUSSION
Alternatively, it is necessary to look for solutions aiming to mitigate the MoS 2 degradation. In this work, O 2 and H 2 O are identified to have a deleterious effect on MoS 2 long-term stability. An obvious approach to slow down the aging process, consists to handle the MoS 2 samples in inert atmosphere at reduced relative humidity, such as in glove box during the manipulation and in a desiccator for storage. The samples/devices could also be stored in vacuum to avoid the aging process. In certain applications, where there will be an inherent exposure of the MoS 2 flakes/edges to the environment, an encapsulating layer such as, polymer layer with 10-20 nm could be deposited to avoid such degradation as reported elsewhere 13 . Although, a conformal deposition throughout the sample could be a challenge due to the presence of  to induce chemical changes in the material. However, reactions with O 2 were found more favorable. Additionally, we demonstrated that nanorods could undergo a transformation from amorphous to crystalline form under intense electron beam irradiation and heating, which indicated the possibility of accelerated aging. Our work offers an original understanding of the extended aging mechanism of MoS 2 nanostructures exposed to ambient and humid conditions. The findings presented here provide fundamental assets for nano-engineering strategies of MoS 2 -based devices in view of their stability over time and operating life cycle.

Sample growth
MoS 2 was grown on SiO 2 /Si substrate using a double sulfurization process. Initially a~50 nm Mo thin film was deposited on the substrate using a sputter system. Sulfurization is performed at ambient pressure in a CVD system. Details of the sample growth are discussed in our earlier article 9 . The sample was then exposed to laboratory ambient conditions, i.e., room temperature at an average relative humidity of 40%, for a period of 36 months.

Microstructural characterization
To investigate the morphology of the MoS 2 aged samples, a dual beam system consisting of a scanning electron microscope (SEM) and a focused ion beam (FIB) Helios 650 Thermo Fisher Scientific™ was used. A typical voltage of 5 kV and beam current of 100 pA was used for the SEM imaging. The Elstar electron column was operated in ultra-high-resolution with magnetic immersion and field free mode. The tool was also used to prepare cross-sectional samples and thin lamellas for TEM measurements. Lamellas were prepared using standard FIB lift-out technique 35 . Initially, a protective layer of Pt material was made on top of the MoS 2 flakes using successively electron beam and ion beam deposition. The dimension of the deposited layer was 20 μm × 2 μm × 2 μm in size. Subsequently, regular cross section and cleaning cross section methods were implemented and a rectangular slice of 20 μm × 1 μm × 6 μm was prepared. Next, the prepared thin slice was fetched (lift out) using a micromanipulator and placed on a Cu TEM grid for further thinning. Ion beam having energy of 30 keV and current of 0.43-0.79 nA was used for rough thinning and beam energy of 5 keV and beam current of 41pA was implemented for final thinning. At the end of the thinning process, the thickness of TEM lamella was below 80 nm.
Analytical characterization TEM investigations and EDS analyses of the samples were conducted using an image corrected TEM system Titan G2 Thermo Fisher Scientific™. The operating voltage was set to 300 kV throughout the investigation. Raman spectroscopy Hor-iba™ was used to characterize the chemical bonds and vibrational states using a green laser excitation (532 nm). X-ray photoelectron spectroscopy (XPS) study was carried out using a Thermo Fisher Scientific™ K-alpha spectrometer and a PHI VersaProbe III scanning XPS microprobe. The spectroscopes possess a monochromatic and microfocused Al K-Alpha X-ray source (1486.6 eV). The X-ray beam power was kept at~50 W. During the experiment, E-neutralizer (1 V), and I-neutralizer (0.11 kV Ar + ion) were implemented. The obtained peaks were corrected using the C 1 s as reference at 284.6 eV. The peaks were further analyzed using MultiPak and OriginPro programs.

Density functional theory simulations
The computational calculations are based on the DFT using generalized gradient approximation with the Perdew-Burke-Ernzerhof exchange-correlation functional 36 , implemented in the open-source software Quantum ESPRESSO 37,38 . A 2 × 2 supercell of monolayer MoS 2 was used as the computational model to understand the nature of the interaction of O 2 and H 2 O molecules with the surface of the two-dimensional monolayer. To address the interaction between two adjacent MoS 2 monolayers, a vacuum space of 10 Å along the c direction was implemented in calculations. All structures were fully relaxed using Broyden-Fletcher-Goldfarb-Shanno method until the maximum Hellmann-Feynman force acting on each atom was smaller than 10 −3 Ryd/Bohr for all structures. The cutoff wave function and cutoff charge densities were 60 and 600 Ryd, respectively. These cut-off energies were determined after a convergence analysis. A (2 × 2 × 1) k-point was used in the structure optimization and total energy calculations.

DATA AVAILABILITY
Data would be made available upon request to the corresponding author.