Selective oxidation of metallic contacts for localized chemical vapor deposition growth of 2D-transition metal dichalcogenides

Chemical vapor deposition (CVD) is the most common fabrication method for transition metal dichalcogenides (TMDs) where direct chemical vapor phase reaction between an oxide transition metal and chalcogen powder results in formation of high-quality crystals of TMDs. However, in this method the nucleation is often random with incomplete nucleation and non-uniform thickness. In this work we studied the formation of a localized transition metal oxide which resulted in controllable growth of mono- to few-layer MoS2 around the formed oxide region. Bulk molybdenum patterns were irradiated with a 532 nm continuous wave laser creating a localized hot-spot which, under ambient conditions, resulted in the formation of molybdenum oxide. The characteristics of the subsequent MoS2 growth depended on the type and thickness of the MoOx which was determined by the power and duration of laser exposure. The resulting MoOx and MoS2 growth around the localized oxide regions were investigated by Raman and photoluminescence spectroscopy. Our studies have shown that exposing bulk molybdenum patterns to 10 mW of laser power for about 2s results in the minimal formation of MoO2 which coincides with high quality mono- to few-layer MoS2 growth.


Introduction
Transition metal dichalcogenides (TMDs), are a class of layered materials with properties that range from metallic to semiconducting depending on their composition, structure, and dimensionality [1,2].Among these are semiconducting TMDs such as MoS 2 , MoSe 2 , WS 2 , and WSe 2 that demonstrate a progressive shift from indirect-to direct-gap semiconductor as they approach their monolayer limit [1][2][3][4].This layer dependence results in exceptional electronic and optical properties including high carrier mobility and photoluminescence, which offers the possibility of many electronic and optoelectronic applications [3][4][5][6][7].
One of the major challenges in this field is synthesizing and processing of 2D-TMDs into devices reproducibly and with scalability to facilitate their application in industry.Predominantly, 2D-TMD based devices are fabricated via a standard procedure which starts by synthesizing 2D material via mechanical exfoliation or chemical vapor deposition (CVD) followed by isolation of the material, and device fabrication using e-beam lithography and metal deposition.The traditional method for isolation of monolayer TMDs is mechanical exfoliation which produces high-quality crystalline materials.However, given the small size and low yield of material produced, this method is not suitable for large scale production [8].To overcome this issue of scale, chemical vapor deposition (CVD) has become one of the most common fabrication methods of mono-to few-layer TMDs.In this method direct chemical vapor phase reaction between a transition metal oxide precursor and chalcogen powder, can result in the formation of high-quality crystals of TMDs in large scale (tens to hundreds of square micrometers), though the nucleation is random with often incomplete coverage and nonuniform thickness [9][10][11][12][13].To achieve large area coverage a CVD reaction of chalcogen precursor with a uniformly deposited thin film of transition metal has been shown to result in the growth of a thin film TMD over an entire substrate [14][15][16][17][18].However, the quality of the resulting material depends on the quality of the predeposited metal, typically leading to growth of TMD films that inherits the disordered nature of the predeposited layer of metal.
In earlier work, we demonstrated a versatile, simple, and scalable method for creating as-grown 2D-TMD based devices in which material is found to grow around lithographically defined patterns of bulk transition metals [19][20][21].In this technique, the controllable oxidation of the bulk transition metal serves as the oxide precursor as well as nucleation site where the TMD material forms and then migrates outward along the substrate, resulting in highly crystalline films with domain sizes of the order of tens of micrometers.By tuning the growth parameters, a thin oxide layer may be formed on the surface of the initial transition metal, leaving the underlying bulk unoxidized.As the oxide layer is consumed during the growth stage, the resulting thin film TMD material producing may be naturally contacted to the remaining bulk metal.In a previous study, MoS 2 based metal-semiconductor-metal photodetectors displayed above average responsivities up to 15 A W −1 and record response times as low as 2 μs indicating good quality electrical contact between the TMD and remaining metal [20].Additional studies of this technique have confirmed that the controlled oxidation of the bulk transition metal pattern results in a metal-oxide layer that serves as the sole precursor for the TMD growth [22], eliminating the need for a separate powder precursor.Specifically, Raman analysis indicated that the formation of a thin layer of MoO 2 leads to an optimal growth of 2D MoS 2 .However, since the entire metal pattern is oxidized, the TMD material typically grows completely around all the metal pattern, complicating subsequent device fabrication.To provide additional control over the location and quality of the TMD growth, this work investigates the use of localized heating through laser irradiation to produce suitable MoO 2 precursor at specific locations on the bulk transition metal [23][24][25][26].As shown in figure 1 selectively producing the oxide precursor using a focused laser can result in localized growth of TMD material.Furthermore, combining such local precursor formation with previous demonstrations of in situ laser sulfurization or selenization of metal oxide films [27,28] may lead to an overall process that directly produces high-quality, localized, TMD based devices within the thermal budget of typical CMOS processes.

Results and discussion
Laser modification of materials is an extensive and ongoing area of research.For transition metal oxides, such as MoO x , results have been reported on the formation of, or transformation between, different phases of molybdenum oxide using various pulsed [24,25,29] and continuous wave (CW) [30,31] laser systems.One potential advantage of using laser irradiation is the wide range of parameters available through the proliferation of modern laser systems.Here, we choose to investigate the formation of different phases of molybdenum oxide by laser direct writing using a standard 532 nm Diode-Pumped Solid State under ambient conditions.This has the advantage of being one of the most common, affordable, and easy to use laser systems available.This also has the added advantage of potentially allowing for in situ monitoring of the oxide formation through concurrent Raman spectroscopy [32].Using this laser, we demonstrate the controlled formation of metal oxide phase as well as the quantity of oxide over the metallic pattern, which can be used to modify and control the quality, scale, and location of subsequent thin-film TMD growth.
The formation of molybdenum oxides on DC sputtered patterns was studied as a function of laser power and exposure time.As described in the supplementary material, and summarized in figures 2 and 3, an initial series of measurements were performed over a range of powers and times to determine a subset of parameters to study in detail.For all measurements, an 80x objective lens (NA = 0.9) was used to focus the laser on the sample and the laser power was varied from 5 to 20 mW.For a diffraction limited spot size with a Gaussian profile this translates to a power density of 1.2 to 4.9 MW cm −2 while exposure time was varied from 2 s to 10 min resulting in laser fluences of 2.4 MJ/cm 2 to 2.9 GJ/cm 2 .The trend observed is that at low power and exposure time the dominant oxide tends to be MoO 2 and as the power and exposure time are increased the resulting material transitions through phases of various sub-oxides, with the higher powers and exposure times resulting in the formation of MoO 3 .For example, as shown in figure 3(c) and in the supplemental material, a laser power of 20 mW resulted in the formation of predominantly Mo 4 O 11 to MoO 3 with little variation as a function of exposure time, whereas the 10 mW series displayed the widest variety of oxide type as a function of exposure time.Raman spectra collected from the 10 mW series are shown in figure 2. For an exposure time of 2 s, peaks associated with MoO 2 are observed at 364, 498, 571, 746 cm −1 (figure 2(a)), while after 1 min of exposure, Mo 4 O 11 Raman peaks at 426, 448, 791, 836, 904 cm −1 were observed [33].
The samples were then subject to a CVD growth process to characterize the quality, quantity, and localization of MoS 2 formation around the oxidized regions.The schematics of the oxidation process via laser annealing followed by the sulfurization process is shown in figure 1, and explained in detail in supplementary material.Following our previous study, the oxygen content inside the growth environment was purposefully minimized by flushing the growth environment with UHP Argon gas at high rate (0.5 LPM) for 1 h and 30 min to avoid any further oxidation of the metallic pattern during the growth.The samples were brought straight to the growth temperature of 760 °C and then vaporized sulfur was introduced for 5 min, which is the proper temperature and time for formation and deposition of monolayer MoS 2 on the substrate.The sample was allowed to naturally cool down and optical images of the regions exposed to laser irradiation were taken and are shown in the rightmost panels of figures 2 and 3. We observe mono-to few-layer growth around the regions where the formation of MoO 2 was detected, while the growth gets bulkier and less extended around the other phases of metal oxide.Raman spectra were taken from the region indicated by the circle on the resulting MoS 2 growth and are shown alongside their respective pre-growth Raman spectra of the corresponding oxide (figures 2 and 3).The separation between the A 1g and E 2g 1 peaks, a standard measure of the number of layers of MoS 2 , indicates the thinnest mono-to-few layer MoS 2 growth coincides with the formation of MoO 2 .Longer laser exposures resulted in bulk MoS 2 growth, indicated by the increased separation between the A 1g and E 2g 1 peaks.The shift in the two Raman modes of MoS 2 corresponds to the variation in the number of layers in the growth result.It has been shown that the A 1g mode blue-shifts while the E 2g 1 mode red-shifts as the number of layers increase in the MoS 2 film, which leads to the variation between the peak's separation.As the laser power and/or exposure time are reduced the Raman signatures of oxide formation become undetectable in our system.However, we do continue to see growth localized to the region exposed to the laser.This indicates that a thin but sufficient layer of oxide is being formed even down to 5 mW for 2 s.
To explore the potential for larger area or patterned growth, a sample was placed on a (x-y-z) motorized stage which was used to raster a square region under the focused laser spot with the raster rate of about 105 μm/s.Here, the delivered power to the sample was fixed at 40 mW (laser fluence of 93.2 kJ/cm 2 ).This would be analogous to exposure of 10 mW power for 3.8 ´10 -2 s at which we detected the clear appearance of MoO 2 Raman bands shown in figure 4(c).After performing the same growth process as described above, we observed localized growth of mono-to few-layer MoS 2 adjacent to the laser processed region (figure 4(a)).The formation of localized crystalline MoS 2 is larger in size when compared to the growth results where a single spot has been oxidized (figure 4    density, perhaps arising from the fact that the laser oxide formation was performed in ambient conditions [34,35].

Conclusion
We have presented an approach for selective oxidation of the bulk transition metal through CW laser processing which may be used to produce deterministically placed mono-to few-layer, high quality TMD films.Local heating through laser irradiation results in the formation of transition metal oxide films which then serve as the precursor and nucleation site for subsequent TMD growth.The Raman analysis revealed that subjecting bulk Mo patterns to a laser power range of 5-10 mW for a brief duration of about 2s under an 80x objective lens, which translates to a power density of 1.2-2.4MW cm −2 and a total energy of 2.4-4.8MJ/cm 2 , yielded minimal MoO 2 formation which then leads to the subsequent growth of high-quality mono-to few-layer MoS 2 with robust photoluminescence response.Furthermore, CW laser can be used as a localized and affordable heating source enabling selective and localized growth of MoS 2 .

Figure 1 .
Figure 1.Schematic of the experimental procedure is shown in this figure.(a) Schematic of Laser irradiation annealing performed on different regions of lithographically defined Mo wires using CW green laser is shown.The laser power was varied between 10 mW to 20 mW for different exposure times between 2s to 10 min.Optical images of three different regions after laser irradiation and oxide formation is shown on the right-hand side.(b) Schematic of the growth procedure.The sample then was placed within a furnace where the oxygen content was minimized by flushing the tube for one hour and a half to avoid metal oxidation at high temperature.The growth was happened at 750 °C after introducing sulfur vapor to the sample.On the right-hand side optical images of three different localized MoS 2 growth is shown.The scale bar in each image is showing 10 mm.

Figure 2 .
Figure 2. Raman characterization and optical images of controlled oxide created on Mo patterns using laser irradiation along with the corresponding growth of MoS 2 .Oxide formation was studied as a function of laser power and exposure time.Exposing the Mo wires to 10 mW of laser power for (a) 2s shows Raman peaks of MoO 2 with the corresponding growth of monolayer MoS 2 , for (b) 10s shows Raman peaks of MoO 2 with the corresponding growth of few layer MoS2, for (c) 30s shows Raman peaks of MoO 2 with the corresponding bulk MoS 2 growth, for (d) 1 min exposure time we can observe Mo 4 O 11 Raman peaks along with MoO 2 peaks and the corresponding bulk growth, for (e) 5 min and (f) 10 min exposure time MoO x Raman peaks dominate the spectrum and the growth results corresponding to these regions are bulk MoS 2 .The oxide Raman spectra were all taken with P 0 = 2 mW and integration time of 15 s.The Raman spectra of MoS 2 were all taken with P 0 = 0.25 mW and integration time of 1s.The scale bar in each image is showing 10 mm.
(b)).The optical image in figure 5(a) as well as the Raman map and selective Raman spectra in figures 5(c) & (f) respectively, show a mono-to-few layer growth of MoS 2 extending approximately 50 mm outward and around the bulk Mo pattern where MoO 2 was formed.Photoluminescence characterization of this region shows the expected A and B peaks of MoS 2 from mono-to few-layer regions (figures 5(b) and (e) with a relatively high B/A ratio as can be seen in figures 5(d) and (e), which could indicate a relatively high defect

Figure 3 .
Figure 3. Raman characterization of controlled oxide formation on Mo patterns were studied as a function of laser power along with their corresponding growth of MoS 2 .The optical images of oxidized region and the growth results are shown on the righthand side of each panel.For 2s exposure time of the Mo wires to laser power of (a) 10 mW shows formation of MoO 2 , and the corresponding large area growth of monolayer MoS 2 , of (b) 15 mW shows Raman peaks of MoO 2 and the corresponding few layers as well as bulk growth of MoS 2 , and of (c) 20 mW shows Mo 4 O 11 Raman peaks and the corresponding bulk growth of MoS 2 .The oxide Raman spectra were taken with P 0 = 2mW and integration time of 15s.The MoS 2 Raman spectra were all taken with P 0 = 0.25 mW and integration time of 1s.The scale bar in each image is showing 10 mm.

Figure 4 .
Figure 4. Comparison of the MoS 2 growth size taken from (a) a region where a large square was irradiated with green laser, and (b) a region where a single spot was irradiated with green laser.(c) Raman spectrum indicating the formation of MoO 2 over the irradiated square in panel (a) before growth.

Figure 5 .
Figure 5. (a) Optical image of a localize mono-to few layer MoS 2 growth. (b) Normalized PL intensity map of the region shown in panel (a) indicating significant PL response.(c) The Raman map of the same region as a function of A 2g and E 2g 1 separation is shown for a better understanding of the quality of the growth.(d) The B peak over A peak intensity ratio map was extracted from the PL intensity map which shows a high B/A intensity ratio indicating high defect density in this region.(e) Representative PL Spectra from the regions indicated by star (monolayer MoS 2 ), triangle (bilayer MoS 2 ), and circle (Bulk MoS 2 ). (f) Representative Raman Spectra from the regions indicated by star (monolayer MoS 2 ), triangle (bilayer MoS 2 ), and circle (Bulk MoS 2 ).