Anomalous photoluminescence thermal quenching of sandwiched single layer MoS_2

: We report an unusual thermal quenching of the micro-photoluminescence (µ-PL) intensity for a sandwiched single-layer (SL) MoS 2 . For this study, MoS 2 layers were chemical vapor deposited on molecular beam epitaxial grown In 0.15 Al 0.85 N lattice matched templates. Later, to accomplish air-stable sandwiched SL-MoS 2 , a thin In 0.15 Al 0.85 N cap layer was deposited on the MoS 2 /In 0.15 Al 0.85 N heterostructure. We confirm that the sandwiched MoS 2 is a single layer from optical and structural analyses using µ-Raman spectroscopy and scanning transmission electron microscopy, respectively. By using high-resolution X-ray photoelectron spectroscopy, no structural phase transition of MoS 2 is noticed. The recombination processes of bound and free excitons were analyzed by the power-dependent µ-PL studies at 77 K and room temperature (RT). The temperature-dependent micro photoluminescence (TDPL) measurements were carried out in the temperature range of 77 – 400 K. As temperature increases, a significant red-shift is observed for the free-exciton PL peak, revealing the delocalization of carriers. Further, we observe unconventional negative thermal quenching


Introduction
Defects in semiconductors are either fatal or vital on the transport and optical emission properties [1,2].However, molybdenum disulfide (MoS 2 ) in a single layer form of the group VI transition metal dichalcogenides (TMDs) has recently emerged as a novel atomic layered material with promising electronic and optoelectronic properties due to its direct bandgap and its prominent transport properties [3][4][5][6].The role of defects on various properties of MoS 2 single layers has remained elusive, and a common consensus has not arrived yet.It has been widely reported that the unintentionally formed lattice point defects such as sulfur vacancies and foreign impurities occupied on these vacancy sites can have a remarkable effect on the structural, electrical transport and optical properties of a single layer MoS 2 [7][8][9][10][11].These defects can be formed in the 2D crystal lattice by several ways.For instance, thermal annealing induces intermediate defect states while the thermodynamically non-equilibrium growth processes result in unintentional defects [10,12].Due to the Li intercalation and the exposure of high energetic N-plasma radicals to the MoS 2 layer, a structural phase change from the thermodynamically most favorable form of trigonal prismatic 2H-MoS 2 to a metastable octahedral 1T-MoS 2 takes place which results in an enormous quenching of photoluminescence intensity [7,13,14].In contrast, the intrinsic and extrinsic point defects significantly enhance the PL intensity of 2D single layer MoS 2 owing to the stronger interaction of excitons with the localized states [10,11].Besides the defect-induced enhancement of the PL intensity, the direct optical transitions associated with the excitons and bound excitons can contribute to an anomalous negative thermal quenching (NTQ) of the PL intensity in semiconductors.NTQ has been reported for various inorganic compound semiconducting materials such as ZnS, GaAs, Cu doped ZnO nanorods, ZnO-Al 2 O 3 core-shell structures [15][16][17][18].Recently abnormal thermal quenching of PL has been observed for Te doped GaInP alloys somewhat in the high-temperature range 175 -300 K, which was attributed to the reduced energy separation between direct and indirect valleys due to the bandgap narrowing effect [19].Till date, there is no preliminary experimental report on NTQ for the case of sandwiched SL-MoS 2 or any other TMDs.
In x Al 1-x N alloys of group III nitrides with a relatively low In composition of 15-20% are promising for the growth of strain-free MoS 2 layers because of their lattice matching unit cells [20].Thereby, it is important to point out that the In composition (x) can be chosen in such a way that the in-plane lattice mismatch between In x Al 1-x N and MoS 2 is vanishingly low (<0.5%).Thus the lattice matched InAlN substrates can be used for MoS 2 growth.To accomplish such lattice matched In x Al 1-x N thin films several kinetically controlled growth parameters were utilized having a prior idea of the growth conditions [21,22].
In this study, we perform temperature-dependent µ-PL measurements on a single layer MoS 2 sandwiched between lattice-matched In 0.15 Al 0.85 N epitaxial films to achieve air-stable SL-MoS 2 .In 0.15 Al 0.85 N cap layers were grown by plasma assisted molecular beam epitaxy (PAMBE) on chemical vapor deposited MoS 2 /InAlN/Si heterostructure at relatively lower nitrogen flow rate and plasma power than out previous reports to preserve the structural and optical properties of sandwiched SL-MoS 2 .It has been reported that the defects like Nitrogen di-molecule (N 2 ) can occupy the lattice positions of MoS 2 even in the ambient conditions [11].In order to reduce such incorporation of defects we used InAlN cap layer for this study.Micro-photoluminescence (µ-PL) and µ-Raman and scanning transmission electron microscopies were employed to confirm the sustainability of SL-MoS 2 after the growth of In 0.15 Al 0.85 N as a cap layer.Further, the power-dependent and the temperature-dependent µ-PL studies were performed to understand the recombination processes and the role of defects on the emission properties of the sandwiched SL-MoS 2 .

Experimental
The large area MoS 2 layers were prepared on lattice matched In 015 Al 0.85 N/Si(111) substrate using a chemical vapor deposition (CVD) system.The molybdenum trioxide (0.3 g) powder was placed in a sapphire boat located in the center of the furnace.The substrates for growing MoS 2 were kept at the downstream side, next to MO 3 powder quartz boat.The sulfur powder was placed at the upstream side of the furnace and maintained at the temperature of 155 °C.The furnace environment was kept at a pressure of 50 Torr with a carrier gas of argon (70 sccm).The furnace was heated up to 755 °C as a process temperature and kept for 15 minutes for MoS 2 growth [23].The growth experiments of In 0.15 Al 0.85 N on MoS 2 /Si and In 0.15 Al 0.85 N on MoS 2 / In 0.15 Al 0.85 N/Si were carried out by Veeco GEN 930 PAMBE system at a substrate temperature of 500 °C.The ion and cryo pumps were aided in attaining a base pressure of 3 × 10 −11 Torr and oxygen partial pressure was <10 −11 Torr, as obtained by a residual gas analyzer (RGA).The Si(111) substrates were chemically cleaned using HF:H 2 O (1:3) to remove the surface oxide and rinsed with acetone, isopropyl alcohol, and deionized water before performing the MoS 2 growth.The MoS 2 /Si and MoS2/InAlN/Si substrates were thermally outgassed in the preparation chamber at 300 °C for 30 min and in the growth chamber at 400 °C for 30 min.For In 0.15 Al 0.85 N growth, the nitrogen plasma source was operated with a flow rate of 0.5 sccm and a radio frequency (RF) power of 200 W.The bandgap of In 0.15 Al 0.85 N of is 4.76 eV and the other details of In 0.15 Al 0.85 N are reported elsewhere [22].High-angle annular dark field scanning transmission electron microscopy (HAADF-STEM) was utilized by operating a probe-corrected FEI Titan at an acceleration voltage of 300 kV.A crosssectional TEM specimen of sample B was prepared by using FEI's Helios Dual Beam focused ion beam (FIB)/SEM equipped with an Omniprobe.Electron energy loss spectroscopy (EELS) acquisition was performed by using Gatan's GIF Quantum of Model 966.Using Horiba Aramis room temperature (RT) µ-Raman and µ-PL measurements were performed on the sandwiched SL-MoS 2 with excitation line of 473 nm.Different objective lenses such as Olympus 100 × /NA 0.9 and Leitz Wetzlar 50 × /NA 0.4 were used to focus the laser on the samples.The high-resolution x-ray photoelectron spectroscopy (HRXPS) measurements were carried out using a Kratos Axis Ultra DLD spectrometer equipped with a monochromatic Al K α X-ray source of 1486.6 eV.A clean copper (Cu) foil was electrically connected to the sample surface so as to compensate the photoemission induced positive charge shifts.Further, binding energies were referenced to the adventitious carbon signal.

Results and discussion
The high-angle annular dark field -scanning transmission electron microscopy (HAADF-STEM) studies were performed on the sandwiched MoS 2 sample (In 0.15 Al 0.85 N /MoS 2 /In 0.15 Al 0.85 N) to confirm the existence and thickness of the MoS 2 layer.Figure 1(a) shows the cross sectional HAADF-STEM image recorded across the stack of In 0.15 Al 0.85 N /MoS 2 /In 0.15 Al 0.85 N layers which was prepared by FIB.This reveals that the thickness of MoS 2 well layer is ≈0.85 nm while thickness values of the top and bottom barrier layers are of ≈10 and ≈40 nm, respectively.The scanning transmission electron microscopy -electron energy loss spectroscopy (STEM-EELS) imaging data sets were acquired to generate the elemental maps of Mo, S, Al and N elements by employing Mo-M (227 eV), S-L (2472 eV), Al-K (1560 eV), and N-K (401 eV), EELS edges, respectively.The elements Mo, S, Al and N across the junction are represented by blue, green, red, and black colored pixels, respectively.Figure 1(c) shows the elemental mapping of MoS 2 well that sandwiched between In 0.15 Al 0.85 N barrier layers.This mapping obviously demonstrates the existence of both Mo and S atoms in SL-MoS 2 region including nitrogen atoms (black pixels) which can be in the form of point defects.This confirms that the nitrogen atoms are diffused into the sandwiched MoS 2 during the growth of the over-grown In 0.15 Al 0.85 N epitaxial layer by keeping its structural and optical properties unaffected.Figure 1(d) shows the µ-Raman spectrum acquired on the sandwiched SL-MoS 2 having the two characteristic peaks observed at 385.5 ± 0.5 and 405.0 ± 0.5 cm −1 corresponding to 1   2 g E and A 1g phonon modes.These modes indicate the in-plane vibration of Mo and S atoms ( 12 g E ) and out-of-plane vibration of S atoms ( 1g A ) in MoS 2 .The separation between 1   2 g E and A 1g phonon modes is observed to be 19.5 ± 0.5 cm −1 , which reveals that the sandwiched MoS 2 layer deposited by CVD is a single-layer [24,25].Fig. 1(e) shows µ-Raman mapping of A 1g phonon mode in the range of 400 -410 cm −1 collected from the sample surface area of 15 × 15 µm 2 .This confirms that the sandwiched SL-MoS 2 is not damaged during the growth of the cap layer and exists throughout the sample.Thus, HAADF-STEM and µ-Raman measurements confirm that the MoS 2 is a single layer with unaltered structural and optical properties.High-resolution XPS measurements were extensively employed to evaluate the existence of several phases of 2D material from the variation in the electronic structure of a constituent atom [7,14].Since the escape depth of photo-emitted electrons in HRXPS is significantly low, over grown In 0.15 Al 0.85 N layer of the sample has to be thin enough so that the electrons removed from MoS 2 layers can be easily probed [26].Figure 2  It should be noted that even with low power of plasma, a significant amount of defects, such as sulfur vacancies can be generated in MoS 2 [27].These defects may not be revealed by HRXPS analysis.Figure 2(b) is a deconvoluted S 2p core level with S-Mo chemical state at 163.00 ± 0.10 eV.The Mo 3d 3/2 and S 2p 1/2 core-levels of Mo and S have similar deconvolutions at higher binding energy value differed by ≈3.10 and ≈1.2 eV, from Mo 3d 5/2 , S 2p 3/2 states, respectively.Thereby, deconvolutions of Mo 3d and S 2p core levels reveal that after the growth of In 0.15 Al 0.85 N cap layer the trigonal prismatic-2H phase of MoS 2 is unaltered and phase transition to octahedral-1T is not found.The observation of none other than the 2H phase of MoS 2 is due to the low nitrogen flow rate and power of plasma source than previous reports under oxygen free MBE growth conditions [8].Thus, the deconvolution of Mo 3d and S 2p core-levels of the sample infers the absence of the octahedral phase in MoS 2 .Thereby we confirm that there is no change in the structural properties of sandwiched SL-MoS 2 .To investigate the optical quality of the sandwiched MoS 2 sample, the temperaturedependent µ-PL (TDPL) measurements were carried out.Figure 3(a) shows the normalized TDPL spectra of the sandwiched MoS 2 where the individual emission spectrum exhibits two distinguishable emissions based on the temperature.We notice a broad peak at a lower energy of ≈1.76 eV appearing up to the temperature of 120 K, whereas the higher energy free exciton emission peak (X 0 ) observed for the entire temperature range (77 K -400 K).The lower energy peak results from the radiative recombination of bound excitons (X B ) which are the deep level defect states that bound to the neutral excitons.These bound excitons can arise from the presence of impurities such as either sulfur vacancies (V s ) in the region of single layer MoS 2 [11].The large width of this bound exciton peak is a consequence of the spreading in the energy of defect states that are available for excitons [28].The TDPL emission peak positions of free excitons exhibited remarkable red shift while increasing the temperature.These are fitted with the Varshni relation  Moreover, to understand the recombination processes involved in µ-PL spectra of the sandwiched SL-MoS 2 , the dependency of the excitation power on the free and bound excitons was studied at 77 K and RT.The µ-PL peaks at ≈1.91 and ≈2.05 eV in Fig. 4(a) are the exciton resonances corresponding to the transitions from the bottom of the conduction band to the spin-split valence sub-bands that are resulting from the broken inversion symmetry [7,30].As discussed, the PL spectrum exhibits an additional broad peak of ≈1.76 eV at 77 K.The PL spectra in Fig. 4(b) doesn't exhibit this broad peak at RT while it has free exciton peaks for all the values of excitation power.The insets of Figs.4(a) and 4(b) show the excitation power (P) dependence on the µ-PL intensity (I).This is commonly studied using a power law I ~Pα , where the exponent (α) depicts the type of carrier recombination.The linear dependence indicates the free exciton recombination with the exponent value close to unity (α≈1), whereas nonlinear dependence with exponent less than unity (α<1) indicates free to bound exciton recombination or donor acceptor pair transitions [31].The integrated intensity of free excitons at 77 K and RT varies linearly (α≈1 and 0.95) with laser excitation power as shown in the insets of Figs.4(a) and 4(b).On the other hand, the intensity of bound excitons shows a sub-linear (α≈0.5)behavior with the excitation power, as shown in the inset of Fig. 4(a), and tends to saturate at higher excitation power values resulting from the filled density of states associated with the defects [11,31].Here, the integrated intensity of µ-PL peaks was obtained by deconvolution process.Consequently, the µ-PL spectrum of the sandwiched SL-MoS 2 at 77K is mainly contributed by the defect peak at low excitation power, whereas the free exciton peak is predominant at high excitation power as described in Fig. 4(a).From Figs. 4(a) and 4(b), the observed bound exciton defect peak at low temperature is dominated at low excitation power density over the free exciton peak.Figure 5(a) shows the variation of µ-PL integrated intensity with the temperature for sandwiched SL-MoS 2 , where we observe negative thermal quenching (NTQ) of µ-PL.The NTQ of PL for sandwiched SL-MoS 2 can be understood in terms of the hopping relaxation mechanism of the carriers.In particular, at very low temperatures, the PL emission arises primarily from the recombination of the free excitons and the excitons trapped in localized states.Here, the carriers can either directly recombine or perform phonon-assisted hopping transitions between neighboring localized states before they recombine and thus results in two peaks at low temperature.When the temperature increased the thermally activated carriers mobile and consequently hopping takes place between localized defect states thereby ultimately reach the delocalized states and causing single peak with enhanced intensity.The reduction of µ-PL intensity after RT is due to the non-radiative recombination or saturation of the carrier hopping.Since no S-shape behavior of the TDPL peak energy observed.Presumably, a significant amount of the carriers are delocalized even at low temperatures or the S-shape could occur in the temperature range below 77 K.It has been reported that the PL emission from MoS 2 can be enhanced due to the presence of defects [9,10].Thus, the anomalous thermal quenching of µ-PL observed in this study can be attributed either to the extrinsic point defects such as oxygen impurities during the CVD growth of MoS 2 and physisorbed nitrogen occupied on sulfur vacancies (N S ) during the growth of cap layer.The existence of nitrogen and oxygen related defects in the sandwiched layer is corroborated by STEM-EELS elemental mapping and HRXPS as described in Figs.1(c) and 2(a), respectively.Such defects are expected for the case of thermodynamically non-equilibrium growth processes such as MBE and CVD which create shallow states having low activation energies close to the band edges [12,32].In contrast, V s and N 2 defects can establish deep level density of states spreading in energy up to 150-300 meV causing bound exciton peak as reported by Tongay et al. [11] Thus, SL-MoS 2 can have both deep and shallow level defects [33].Consequently, the enhancement of the µ-PL intensity of the sandwiched SL-MoS 2 with increasing temperatures is due to the raising radiative recombination rate of the excess delocalized carriers created from the carrier hopping from the shallow defect states to the band edges.
Further, the variation of µ-PL integrated intensity I T for free excitons with temperature is explained by a model developed by Shibata et al. [34], which can be expressed as where k B is the Boltzmann's constant, T is the temperature and I 0 is the integrated intensity at 0 K. E n and E d depict the activation energies for the processes that increase the µ-PL intensity with increasing temperature and for the nonradiative channels, respectively, with corresponding weight factors N and D. The denominator and numerator of Eq. ( 1) represent the normal temperature quenching and the negative thermal quenching (NTQ) processes, respectively.The experimental µ-PL intensity data of free excitons were fitted by a model described by Eq. ( 1).We obtained activation energies E d and E n of ≈350 meV and ≈36 meV, respectively, for normal and negative thermal quenching processes that involved in TDPL transitions.The high energy 350 meV corresponds to the activation energy for resonant free excitons associated with the conventional thermal quenching, whereas the low energy 36 meV is the thermal activation energy for excitons that bound to the defects to be activated which results in negative thermal quenching of µ-PL intensity.Therefore, E d represents the thermal activation energy for the dissolution of free excitons.It has been observed that when the temperature increases above the RT, the radiative quantum efficiency of the excitons reduces which causes the conventional PL thermal quenching [18].

Conclusions
In summary, we prepared MoS 2 sandwiched structure by growing an In 0.15 Al 0.85 N cap layer on a SL-MoS 2 / In 0.15 Al 0.85 N lattice matched heterostructure template.The STEM, HRXPS and µ-Raman spectroscopy studies confirmed that the structural and optical properties of the sandwiched SL-MoS 2 remain unaltered during the preparation process.This is ascribed to the low operation conditions of nitrogen plasma source during the MBE growth of In 0.15 Al 0.85 N cap layer.Further, we analyzed the optical emission properties of the air-stable sandwiched SL-MoS 2 using the temperature and power-dependent µ-PL measurements.While increasing temperature, a significant red-shift is observed for the free-exciton PL peak exhibiting the delocalization of carriers.The excitation power-dependent µ-PL acquired at 77 K and RT reveal that the PL of sandwiched SL-MoS 2 consists of the transitions associated with the free excitons and bound excitons.The bound excitons are associated with the deep level defects.Furthermore, we observed an anomalous negative thermal quenching behavior in a certain temperature range which is a consequence of the carrier hopping between shallow defect states to the band edges.Thus, this study demonstrates the role of defect states in understanding the fundamental emission properties of the sandwiched MoS 2 single layer.

Fig. 1 .
Fig. 1.(a) cross-sectional image of high-angle annular dark field -scanning transmission electron microscopy (HAADF-STEM) and 1(b) schematic model for the sandwiched SL-MoS 2 sample by taking account of the van der Waals epitaxy.1(c) STEM-EELS elemental color maps (Mo-blue, S-green, Al-red, and N-black).1(d) µRaman spectrum collected using 473nm excitation line.1(e) µRaman mapping of A 1g phonon mode acquired on the sample.
presents the Mo 3d and S 2p core levels with an inset of a typical STEM cross section image of the In 0.15 Al 0.85 N/SL-MoS 2 heterostructure sample.Figure 2(a) depicts the peak deconvolution process of Mo 3d core level where the Voigt (mixed Lorentzian-Gaussian) line shapes employed to fit the Mo-S, S-Mo (trigonal prismatic-2H) and M-O chemical states, respectively, centered at 230.00 ± 0.10, 227.50 ± 0.10 and 233.00 ± 0.10 eV.The Mo-O impurity state with extremely low-intensity results from MoO 3 precursor at the interface of MoS 2 /Si during the high-temperature CVD growth process.

Fig. 2 .
Fig. 2. The deconvolution of HRXPS (a) Mo 3d and (b) S 2p core-levels acquired on In 0.15 Al 0.85 N/SL-MoS 2 hetero-junction samples.The peak positions of core-levels are shown in parentheses.The inset shows the typical STEM cross-section image of the sample.
the reduction of the bandgap in semiconductors with increasing temperature[29].Figure 3(b) shows the Varshni fit to the data points with the fitting parameters 0 1.92 g E =eV (band gap at 0 K), α = 0.001 eV/K and β = 1090 K.It is worth noticing that no S-shape behavior of the temperature-dependent µ-PL peak energy is observed.Presumably, a significant amount of the carriers are delocalized even at low temperatures or the S-shape could occur in the temperature range below 77 K.

Fig. 5 .
Fig. 5. (a) Temperature-dependent integrated µ-PL spectra for sandwiched SL-MoS 2 and (b) Shibata model fit to the integrated µ-PL intensity data points.Here, PL integrated intensity is obtained by peak deconvolution process.