Defect Passivation of 2D Semiconductors by Fixating Chemisorbed Oxygen Molecules via h‐BN Encapsulations

Abstract Hexagonal boron nitride (h‐BN) is a key ingredient for various 2D van der Waals heterostructure devices, but the exact role of h‐BN encapsulation in relation to the internal defects of 2D semiconductors remains unclear. Here, it is reported that h‐BN encapsulation greatly removes the defect‐related gap states by stabilizing the chemisorbed oxygen molecules onto the defects of monolayer WS2 crystals. Electron energy loss spectroscopy (EELS) combined with theoretical analysis clearly confirms that the oxygen molecules are chemisorbed onto the defects of WS2 crystals and are fixated by h‐BN encapsulation, with excluding a possibility of oxygen molecules trapped in bubbles or wrinkles formed at the interface between WS2 and h‐BN. Optical spectroscopic studies show that h‐BN encapsulation prevents the desorption of oxygen molecules over various excitation and ambient conditions, resulting in a greatly lowered and stabilized free electron density in monolayer WS2 crystals. This suppresses the exciton annihilation processes by two orders of magnitude compared to that of bare WS2. Furthermore, the valley polarization becomes robust against the various excitation and ambient conditions in the h‐BN encapsulated WS2 crystals.


Introduction
−4] Furthermore, the valley-dependent optical selection rules given by the broken inversion symmetry enable the selective generation of excitons in the particular valley (+K or −K) using circularly polarized light, [5] providing the opportunity for applications toward valleytronic devices.However, the external disorders in close proximity of monolayer TMDs such as substrate-induced surface roughness and absorbates can significantly alter the excitonic properties of two-dimensional (2D) TMD materials, hindering the observation of the unique properties in monolayer TMDs. [6,7] an attempt to reduce the disorder from substrates, it has been proposed to encapsulate TMD materials using hexagonal boron nitride (h-BN) layers. [8,9]Recently, it has been shown that h-BN encapsulation enables to observe the intrinsic optical properties of monolayer TMDs, including the excitonic linewidth with homogeneous broadening limit [9] and the suppression of exciton annihilation processes. [10]As the origin, the reduced substrate disorders have often been suggested in the previous works, [9,10] but the excitonic properties of TMDs on h-BN substrates show large discrepancies from those of TMDs encapsulated by h-BN. [11,12]−18] The adsorption of the oxygen molecules occurs on the defects with relatively lower kinetic barrier rather than the perfect sites of TMD materials, [18] and the oxygen molecules unlike other molecules can only be chemisorbed at chalcogen vacancies due to isovalent valence electrons (two unpaired electrons) with the chalcogen atom. [14,15]This chemisorption of oxygen molecules at the defect sites removes the defect-related gap states without significantly altering the electronic band structures of TMD materials. [13,14]Thus, the oxygen molecules, that are supplied during the exposure of TMDs into the atmosphere, can significantly change the properties of defect-related states through the chemical adsorption process.In this regard, h-BN encapsulation can play a crucial role in the defect states of TMDs, in which the h-BN layers fixate the adsorbed oxygen molecules on the TMD defects and facilitate the interaction between the oxygen molecules and the defect states.However, the role of h-BN encapsulation in relation to the defects of TMDs remains unexplored.
In this work, we found that h-BN encapsulation stabilizes the chemisorbed oxygen molecules on the defect sites of monolayer WS2 crystals, which greatly passivates the defectrelated gap states along with the decrease in the free electron density.Electron energy loss spectroscopy (EELS) combined with theoretical analysis clearly reveals that the oxygen molecules are chemisorbed onto the defects of WS2 crystals and are fixated by h-BN encapsulation, that excludes a possibility of oxygen molecules trapped in bubbles or wrinkles formed at the interface between WS2 and h-BN.Optical spectroscopic studies show that h-BN encapsulation prevents the desorption of oxygen molecules over various excitation and ambient conditions, resulting in a greatly lowered and stabilized free electron density in monolayer WS2 crystals.This suppresses the exciton annihilation processes by two orders of magnitude compared to that of bare WS2.Furthermore, due to the stabilized free electron density in the h-BN encapsulated WS2 crystals, the valley polarization becomes robust against the elevated excitation condition.

Results and Discussion
Figure 1a shows a schematic illustration showing that the fixated oxygen molecules by the h-BN layers effectively passivate defects of the WS2.−15] However, in the case of the WS2 used in our study, the kinetic barrier for the O2-chemisorption (0.56 eV) is lower than that for the O2-dissociative chemisorption process (0.76 eV).It is estimated that the probability of the O2chemisorption is 1000 times higher than that of the dissociative chemisorpiton (see Figure S1 in Supporting Information).Thus, the oxygen chemisorption on the monolayer WS2 crystals would have the final configuration of the O2-chemisorption rather than the O2-dissociative chemisorption. [14]The major molecules such as N2, O2, and H2O in air can be weakly physisorbed at both the pristine surface and defect sites of WS2.However, this physisorption has virtually no influence on the electronic and optical properties of the WS2 monolayer due to easy desorption of physisorbed molecules. [14]In addition, our first-principle calculations demonstrate that the oxygen molecules can only be chemisorbed onto the defects (sulfur vacancy) and attain a fully stable chemisorption state, indicating that the oxygen molecules can be majorly adsorbed onto the WS2 in the air.The detailed theoretical calculation results on molecular interactions with the sulfur vacancy and pristine surface of WS2 are provided in Supporting Information S2 and S3.To investigate the role of oxygen fixation in the excitonic properties of monolayer WS2 with excluding the effects of disorders induced by the substrates, we studied h-BN encapsulated WS2 crystals suspended on line trenches with a linewidth of 1.8 μm in comparison with bare WS2, as shown in Figures 1b,d.Scanning electron microscope images confirm the suspended structures for both the bare (Figure 1b) and h-BN encapsulated (Figure 1d) WS2 crystals on the line trenches (see Figure S4 in Supporting Information).The monolayered WS2 crystals used in this study were grown on sapphire substrates using a chemical vapor deposition (CVD) method. [19]Mechanically exfoliated h-BN flakes with a thickness of ~40 nm were used as the encapsulating layers in the h-BN/WS2/h-BN structures.
The detailed sample preparation processes are described in the methods.Figures 1c,e display the spatial photoluminescence profiles of the bare (Figure 1c) and h-BN encapsulated (Figure 1e) WS2 suspended on the line trenches, respectively.The steady-state photoluminescence measurements were carried out at a low level of excitation (~0.065 kW/cm 2 ) to rule out the heating effect.It is worth noting that the photoluminescence intensity becomes stronger in the suspended regions than in the supported regions for both the bare and h-BN encapsulated WS2 crystals due to the enhanced local field effect by optical interference in the trench region. [7]To confirm the exciton species, the photoluminescence spectra were measured at a cryogenic temperature of 77 K under a vacuum level of ~1 × 10 −5 Torr, as shown in Figure 1f.For the h-BN encapsulated WS2, the neutral exciton (X 0 ) and the trion (X − ) are identified at energies of 2.042 and 2.001 eV, respectively, while the bare WS2 shows three species of the neutral exciton (X 0 ), trion (X − ), and defect-related trapped exciton (L) at 2.087, 2.042, and 2.026 eV, respectively.The energy of neutral exciton was assigned by measuring the differential reflectance spectra (see Figure S5 in Supporting Information), and those of trion and defectrelated trapped exciton were identified by the energy differences from the neutral exciton. [20,21]mpared to the bare WS2 showing the emission prevailed by the trion and the defect-related trapped exciton, the h-BN encapsulated WS2 exhibits the predominant emission from the neutral exciton with homogeneous linewidth of the 6 meV, [9] indicating that the defect-induced free electrons and inhomogeneous broadening are substantially reduced by the encapsulating WS2 with h-BN layers.These results indicate that the oxygen fixation by h-BN encapsulation can play a crucial role in the defect removal with reducing the free electron density.To directly confirm the fixation effect of the adsorbed oxygen molecules on the defects, the change in excitonic spectra was monitored under the different ambient conditions of air and vacuum, as shown in Figure 1g (see Figure S6 in Supporting Information).Striking features are observed for the bare WS2 crystals, showing that the spectral weight of neutral excitons is predominant over that of trions under ambient air condition, whereas that of trions becomes larger than that of neutral excitons under vacuum.These results indicate that the oxygen adsorbates on the defect sites are released by changing the ambient condition from air to vacuum, raising the density of free electrons in the bare WS2 crystals under vacuum. [16,17]As shown in Figures S7, the h-BN encapsulated WS2 samples fabricated under an inert (N2) environment exhibit much stronger trion intensity (higher free electron concentration) compared to that of the h-BN encapsulated WS2 fabricated in the air.The spectral feature is very similar to that of the bare WS2 measured in the vacuum environment (see top panels in Figures 1f and 1g).Furthermore, the exfoliated monolayer WS2 and WSe2 with a lower density of chalcogen vacancies give rise to a less change in the free electron density against the variation of the ambient conditions, implying that the oxygen molecules are mostly adsorbed on the chalcogen vacancies (see Figure S8 in Supporting Information).In contrast, the h-BN encapsulated WS2 crystals exhibit almost the same spectra prevailed by the neutral excitons regardless of the ambient conditions, highlighting that the h-BN encapsulation effectively removes the internal defects by stabilizing the oxygen molecules adsorbed onto WS2 crystals.The black, olive, and blue dashed lines represent the neutral exciton (X 0 ), trion (X − ), and defectrelated trapped exciton (L) states, respectively.
The EELS analysis for oxygen K-edge confirms that the h-BN encapsulation anchors the chemisorbed oxygen molecules onto the defects of monolayer WS2 crystals.Figure 2a displays the EELS spectra of the oxygen K-edge for the h-BN encapsulated WS2, bare WS2, and h-BN flake crystals.The h-BN encapsulated WS2 crystals show the oxygen K-edge peaks centered at 538 and 556 eV, whereas any oxygen-related features are not observed for the bare WS2 and the h-BN crystals.This suggests that the adsorbed oxygen molecules on WS2 are fixated by the h-BN encapsulation (see Figure S9 in Supporting Information).Figures 2b-e show the EELS maps for the boron (b), nitrogen (c), oxygen (d) K-edge, and the sulfur (e) L-edge meausred from the h-BN encapsulated WS2.Note that the elemental signal displayed for a pixel in the EELS map is the sum of the elemental signals detected through 2D scanning using an electron beam with a spatial resolution of 1 Å over an area of 20 nm × 20 nm.The spatial EELS maps confirm that the oxygen K-edge signal is markedly weak compared to those of other elements, and is randomly distributed in the 2D plane of h-BN encapsulated WS2 samples.This strongly indicates that the oxygen molecules are chemisorbed on the randomly distributed local defects in WS2 crystals (see Figure S10 in Supporting Information).
To discover the adsorption types of the adsorbed oxygen molecules by h-BN encapsulation, we demonstrate the theoretical EELS spectra of oxygen K-edge for the physisorbed oxygen molecule on the pristine WS2 (Figure 2f) and chemisorbed oxygen molecule at the sulfur vacancy site of the WS2 (Figure 2g).To understand the underlying physics for the EELS results, we performed theoretical EELS calculations based on the firstprinciples calculations implemented in the full-potential linearized plane wave (FLAPW) + local orbitals with the ELK code.Here, we used the pseudo core-hole method, where the selfconsistent calculation is made in terms of one of the oxygen nuclei constrained to be positively charged (+1e) and an additional electron (−1e) is simultaneously constrained to occupy the conduction orbitals.After the self-consistent calculation, the Kohn-Sham orbitals become well defined, and then we compute the photon-absorption matrix elements between the core orbital (s-orbital) of the oxygen nuclei and the unoccupied conduction bands, which corresponds to the dielectric function for EELS spectra (see Supporting Information S11 for more detail).
The calculated EELS spectrum for the physisorbed oxygen molecule on the pristine WS 2 indicates the two main peaks at 530 eV and 539 eV, which originate from the transition of the core electrons to two kinds of trivial hybridized states in the oxygen molecule, featured as antibonding orbital  * (top) and  * (bottom) distributions, respectively (see the inset of Figure 2f), whereas the calculated oxygen K-edge peaks for the chemisorbed oxygen molecule appear at the energy loss positions of 538 and 555 eV (labelled as A and B on the spectrum), respectively.
Since the core-hole excitation makes the oxygen molecule to have an asymmetric potential, the  * orbital distribution in the inset reflects asymmetric densities.As shown in Figure 2g, the calculated EELS spectrum (red dashed) for a chemisorbed oxygen molecule is in good agreement with that of the h-BN encapsulated WS2 (blue line), indicating that the fixated oxygen molecules by the h-BN encapsulation are chemisorbed at the defect sites of the WS2.
The anti-bonding orbital  * peak (530 eV) is absent for the chemisorbed oxygen molecule in the calculated EELS spectrum.The disappearance of -bonding is commonly interpreted as a transformation of bonding sequences. [22]In our study, it is found that the hybridizations between the oxygen molecule and the surrounding tungsten atoms directly suppress the  * peak.As shown in the top inset of Figure 2g, the real-space orbital distribution for the peak A resembles an orbital shape of the σ* peak for the physisorbed oxygen molecule, indicating that the hybridizations between the oxygen molecule and the tungsten atoms induce σ* antibonding energy state similar to that of the physisorbed oxygen molecule.In contrast, for the peak B, the real-space orbitals show highly delocalized distribution for both the oxygen molecule and the WS2 crystal (the bottom inset of Figure 2g), implying that the peak B is due to the transition of the core electrons to continuum bands contributed from both the oxygen molecules and the WS2 crystals.The investigation of the exciton recombination processes against the excitation power for the bare and h-BN encapsulated WS2 crystals reveals that the oxygen fixation suppresses the nonradiative decay for the neutral excitons by stabilizing the free electron density of the WS2.
Figures 3a,b show the double-logarithmic plots of the neutral exciton (X 0 ) emission intensity of the bare and h-BN encapsulated WS2 crystals as a function of the excitation power density under ambient air and vacuum conditions at room temperature, respectively.The bare WS2 crystal in air exhibits a sublinear increase with an exponent of 0.71 in a power-law (I ∝ P α , P: excitation power), while that in vacuum shows a smaller exponent of 0.32.The power dependence can be understood by introducing a simple rate equation model for the steady-state photoluminescence.The rate equation for exciton generation and recombination can be described by [23,24] where   is the neutral exciton density,   is the exciton lifetime,   is the electron density,  is the trion formation coefficient,  is the exciton-electron Auger coefficient, and  is the exciton-exciton annihilation coefficient.At our excitation power range, the exciton density is estimated to be from the low 10 10 cm −2 to the low 10 11 cm −2 .In this range of exciton density, it has been known that the exciton-exciton annihilation process becomes negligible. [25]rthermore, the recent work has shown that the photogenerated excitons are mostly converted to the trions in this range, indicating that the Auger process can also be neglected. [24]Thus, the predominant nonradiative decay channel of the neutral excitons is the exciton-to-trion conversion process (see Figure S13 in Supporting Information).When the free electron density increases with the power dependence of P µ , the neutral excitons are converted to the trions in proportional to P µ , and the neutral exciton emission intensity follows the power-law of P 1−µ (see Supporting Information S12).The sublinear increase in the exciton emission in bare WS2 indicates that the free electron density increases with increasing excitation power density by releasing oxygen molecules chemisorbed on defect sites, promoting the trion conversion process.The smaller exponent further confirms the easier desorption under ambient vacuum condition.Interestingly, the h-BN encapsulated WS2 shows a linear increase with an exponent of 0.99 for both the air and vacuum ambient conditions, as shown in Figure 3b.This indicates that h-BN encapsulation results in predominant neutral exciton emission over the decay channels without generating free electrons by desorption. [16,24]Figure 3c shows the photoluminescence intensity ratio (X − /X 0 ) of the trion to the neutral exciton as a function of the excitation power density.The X − /X 0 of bare WS2 measured under ambient vacuum condition steeply increases from 0.38 to 4.18, while the X − /X 0 under the air environment gradually increases from 0.25 to 0.50 over the excitation power density range from 0.03 to 0.66 kW/cm 2 .
This further evidences that the desorption process, which induces free electrons and exciton-totrion conversion, is much facilitated in vacuum than in air.In striking contrast, the X − /X 0 of h-BN encapsulated WS2 is kept almost constant at approximately 0.02 over the excitation power densities under both the ambient air and vacuum conditions, which is greatly reduced by more than one order of magnitude compared to that of the bare WS2.This clearly indicates that h-BN encapsulation prevents the desorption of oxygen molecules by the laser illumination, resulting in predominant recombination by neutral excitons.By considering the intensity weight (  − /  ) of the trion in the photoluminescence spectra, we estimated the free electron density as a function of the excitation power density by using the mass action law (see Supporting Information S14). [26,27]As shown in Figure 3d, the free electron density exhibits a very similar trend as X − /X 0 .For bare WS2, the free electron density drastically increases and reaches 2.66 × 10 13 cm −2 at an excitation power density of 0.66 kW/cm 2 in vacuum, which is close to the Mott density (~1 × 10 14 cm −2 ), [28] while gradually increasing to 7.38 × 10 12 cm −2 at an excitation power density of 4.99 kW/cm 2 in air.However, regardless of the ambient conditions, the free electron density of h-BN encapsulated WS2 is maintained at ~9 × 10 10 cm −2 with increasing excitation power, resulting from the fixation of chemisorbed oxygen molecules by h-BN layers.
Additonally, using the bare and h-BN encapsulated WS2 capacitor devices, we estimated the free electron densities in the bare and h-BN encapsulated WS2 at the vacuum and air environments.For the bare WS2, the electron densities were estimated to be 1.15 × 10 13 cm −2 (vacuum) and 3.69 × 10 12 cm −2 (air) at   ≅ 0  (gate voltage), respectively, while those of h-BN encapsulated WS2 were estimated to be 2.05 × 10 11 cm −2 at both the vaccum and air environments.These electron densities were in good agreement with those estimated using the mass-action law.From the difference in free electron densities in-between the vacuum and air environments, we estimated the number of desorbed/adsorbed oxygen molecules on the sulfur vacanies to be 7.17 × 10 12 cm −2 (see Supporting Information S15).The nonradiative decay channel of neutral excitons can be attributed to exciton-to-trion conversion at our excitation power range.From the rate equation (1), the total recombination rate (   ) can be described by   =  0 +     , where  0 is the density-independent recombination rate and   is the exciton annihilation rate constant due to the exciton-to-trion conversion process.The role of the oxygen fixation can be quantitatively understood by estimating the exciton annihilation rate constant (  ).To measure the exciton lifetime with increasing excitation power, we carried out time-resolved photoluminescence spectroscopy.For Information.The increase in exciton lifetime in the h-BN encapsulated WS2 strongly suggests that the nonradiative decay by the trion formation, which occurs on a very fast time scale of a few ps, is significantly inhibited due to the passivation of defects by the oxygen fixation. [23,29,30] quantitatively evaluate the exciton annihilation rate constant (RA), the exciton densityinduced recombination rate (τ −1 ) was plotted as a function of the exciton density using the measured exciton lifetimes (see Supporting Information S17). [30]Note that the exciton density was estimated by calculating the net absorption in the monolayer WS2 for the pump fluence. [31,32]As presented in Figures 4c,d, the h-BN encapsulated WS2 crystals exhibit an exciton annihilation rate constant of 8.3 × 10 −3 cm 2 s −1 (7.8 × 10 −3 cm 2 s −1 ), while the bare WS2 crystals show a value of 8.0 × 10 −2 cm 2 s −1 (3.2 × 10 −1 cm 2 s −1 ) under air (vacuum) ambient conditions.As expected, the exciton annihilation rate constant of the h-BN encapsulated WS2 is remarkably reduced by approximately two orders of magnitude compared to that of the bare WS2.These results are attributed to the suppression of exciton-to-trion conversion process in the h-BN encapsulated WS2 due to the greatly lowered and stabilized free electron density.This fact can be additionally verified by investigating the decay dynamics of neutral excitons with an electrostatic doping in the h-BN encapsulated WS2 capacitor devices at a fixted exciation power.Figure 4e shows the gate-voltage-dependent photoluminescence spectral map in the h-BN encapsulated WS2 capacitor devices, showing that the charge neutral point is determined at   ≅ 0 .The increase in the gate voltage (  > 0 ) gives rise to the gradual decrease in the emission intensity for the neutral exciton and, simultaneously, the increase in the emission intensity for the trion.This indicates that the increase in free electron density facilitates the exciton-to-trion conversion process, leading to the nonradiative decay of neutral excitons.The gate-voltage-dependent photoluminescence decay curves of neutral excitons also clearly show that the increase in the free electron density promotes the exciton annihiliation.As shown in Figure 4f, the decay time of neutral excitons in the h-BN encapsulated WS2 capacitor devices becomes steeply shorten from 139 ps to 36 ps for increasing the gate voltage from 0.5 V to 0.9 V. The exciton annihilation rate constant with the electrostatic doping is estimated to be 1.3 × 10 −1 cm 2 s −1 (see Supporting Information S18), which is similar to that of bare WS2 under the vacuum ambient condition.The h-BN encapsulation gives rise to an almost constant level of the free electron density in WS2 crystals for elevated excitation powers, which can provide a stable and robust valley polarization against various excitation conditions.In contrast, for bare WS2, the drastic increase in free electron density with increasing excitation would cause a large variation in the valley polarization due to the change in decay dynamics of the neutral excitons caused by the excitonto-trion conversion. [33,34]To investigate the effect of h-BN encapsulation on the valley polarization, we carried out circular polarization-resolved photoluminescence measurements as a function of the excitation power density at 77 K. Figure 5a show the circularly polarized photoluminescence spectra measured from the bare (top panel) and h-BN encapsulated (bottom panel) WS2, respectively.As a result, the degree of valley polarization with increasing excitation power shows a very different trend for the bare and h-BN encapsulated WS2 crystals, as shown in Figure 5b.An important distinction is that the h-BN encapsulated WS2 exhibits a stable valley polarization ratio at a constant level, whereas that of the bare WS2 shows a large variation by changing the excitation power.The valley polarization can be described by the , [35] where  0 is the initial valley polarization,   is the valley exciton lifetime, and   is the valley relaxation time.The initial valley polarization ( 0 ) given by the optical selection rules can be assumed to be unity, [5] meaning that the valley polarization (  ) is then mainly governed by the competition between the exciton lifetime and the valley relaxation time. [35]As shown in Figure 5b, the degree of valley polarization of bare WS2 is 1.5 times higher than that of h-BN encapsulated WS2 at the lowest excitation power density of 0.17 kW/cm 2 .This can be attributed to the fact that in bare WS2, the neutral excitons decay rapidly within the valley, rather than an intervalley scattering, due to a shortened exciton lifetime caused by a higher exciton-to-trion conversion rate.In addition, the large variation in the valley polarization of bare WS2 can be understood as a result of the change in the exciton lifetime by the accelerated trion conversion process with increasing excitation power, as shown in Figures 4a and c. [33,34] Accordingly, the valley polarization can be increased in the bare WS2 for elevated excitation powers.For the h-BN encapsulated WS2, however, the exciton-to-trion conversion and the exciton lifetime are maintained at almost constant levels, resulting in a stable valley polarization ratio for elevated excitation powers.The large variation in the valley polarization is also observed in the h-BN encapsulated WS2 capacitor devices with the electrostatic doping (Figure 5c), showing the drastic decrease in the exciton lifetime as the gate voltage increases (see Supporting Information S19).This also confirms that the change in the free electron density leads to the large variation in the valley polarization.

Conclusion
In conclusion, we have demonstrated that h-BN encapsulation greatly removes the defectrelated gap states by stabilizing the chemisorbed oxygen molecules onto the defects of monolayer WS2 crystals, that are provided during the interactions between WS2 and atmosphere.
It is clearly shown that the oxygen molecules are chemisorbed onto the defects of WS2 crystals and are fixated by h-BN encapsulation with excluding a possibility of oxygen molecules trapped in bubbles or wrinkles formed at the interface between WS2 and h-BN, as confirmed by the EELS study.Optical spectroscopic studies show that h-BN encapsulation prevents the desorption of oxygen molecules over various excitation and ambient conditions, resulting in a greatly lowered and stabilized free electron density in monolayer WS2 crystals.This suppresses the exciton annihilation processes by two orders of magnitude compared to that of bare WS2.
Furthermore, due to the stabilized free electron density in the h-BN encapsulated WS2 crystals, the valley polarization becomes robust against the various excitation and ambient conditions.
Our findings provide insight into the role of h-BN encapsulation and open up the possibility to control the defect states in 2D semiconductors through adsorbate-engineered 2D heterostructures.

Experimental Section
Sample preparation: The monolayer WS2 crystals were grown on a sapphire substrate by chemical vapor deposition methods, [19] and the h-BN flakes with a thickness of ~40 nm were prepared by mechanical exfoliations from bulk h-BN single crystals.The h-BN encapsulated WS2 structure was fabricated by sequential pick-up processes using the dry van der Waals stacking method.Elvacite resin or polycarbonate (PC) was utilized as a polymer stamp for the pick-up of layered materials.The assembled h-BN/WS2/h-BN structures and picked-up WS2 crystals were released onto the line trenches by melting the polymer stamp, where the line trenches were fabricated through conventional photolithography and reactive ion etching processes using 400-nm-thick SiO2-coated Si substrates.The polymer stamps were removed by immersing the fabricated sample in chloroform.Finally, both the h-BN encapsulated and the bare WS2 onto line trenches were annealed at 350 °C to improve the coupling between the stacked layers and remove transfer residues.
Optical measurements: The steady-state and time-resolved photoluminescence measurements were performed using a home-built confocal microphotoluminescence system.Using a 40× (0.6 NA) objective (Nikon), the excitation beam was focused, and the signal from the samples was collected through an optical fiber on the focal image plane.For the steady-state photoluminescence measurements, an argon-ion laser with a wavelength of 457.9 nm (continuous wave) was used as an excitation source.The photoluminescence spectra were resolved by a spectrometer (Acton SpectraPro 500i with 0.

Scanning transmission electron microscopy (STEM) and EELS analysis: STEM was used by
Monochromated ARM-200F (NEO-ARM) in Korea Basic Science Institute (KBSI) operated at 200 kV.Gatan imaging filter (GIF) Continuum HR-1066 spectrometer was used to collect electron energy loss spectra.

S1. Kinetic barriers for O 2 -chemisorption and O 2 -dissociative chemisorption processes
We performed density functional theory calculations implemented in the quantum espresso code, employing Optimized Norm-Conserving Vanderbilt (ONCV) pseudopotentials and the nudged elastic band (NEB) methods for 3 by 3 superstructures of monolayer WS2.These calculations, based on the grimme-D2 van der Waals corrections, reveal the relative energy evolution for each reaction path from the initial configurations to the final configurations.
Figures S1a and S1b show the kinetic barriers for the O2-chemisorption and O2-dissociative chemisorption processes in the monolayer WS2 (a) and WTe2 (b) with the sulfur vacancy (SV).
In the case of the WS2, the kinetic barrier for the O2-chemisorption (0.56 eV) is lower than that for the O2-dissociative chemisorption process (0.76 eV), as shown in Figure S1a.From the transition state theory, the reaction rate is given by  ≅ exp(−  /  ) , where  is the reaction rate,  is the attempt frequency,   is the barrier,   is the Bolzmann constant, and  is the temperature.The attempt frequency can be approximated by the value of 10 12 s −1 . [1]The reaction rate (T = 300 K) for the O2-chemisorption and dissociative chemisorption processes is estimated to be approximately 180 s −1 (Eb = 0.56 eV) and 0.17 s −1 (Eb = 0.76 eV), respectively.
This result indicates that the probability of the O2-chemisorption is 1000 times higher than that of the O2-dissociative chemisorption.Thus, the oxygen chemisorption on the monolayer WS2 crystals would have the final configuration of the O2-chemisorption rather than the O2dissociative chemisorption. [2] the other hand, in the case of the WTe2 (Figure S1b), there are no kinetic barriers (0.00 eV) for both the O2-chemisorption and O2-dissociative chemisorption processes, indicating that the oxygen chemisorption spontaneously occurs toward the O2-dissociative chemisorption process in the case of WTe2.These results are in good agreement with previous theoretical studies, showing that the type of the oxygen chemisorption depends on TMD materials. [2]

S2 and S3. Molecular interactions with the pristine surface and sulfur vacancy of WS 2
Figures S2 and S3 reveal the relative energy evolutions as a function of the reaction path for N2, O2, and H2O with both the pristine WS2 and SV of WS2.In this reaction path, the decrease in relative energy indicates that the molecular adsorption proceeds toward a physisorption, corresponding to an exothermic process. [3]Meanwhile, an increase in relative energy exhibits that the molecular adsorption proceeds toward a chemisorption, which leads to an activation barrier in the chemical bonding sequence.
As shown in Figure S2, the N2 molecule does not achieve a structurally favorable state when chemisorbed onto the SV.However, the N2 exhibits a preference for physisorption (−0.08 eV) just prior to forming a chemical bond with surrounding tungsten atoms.Conversely, the O2 starting from a weakly physisorbed state displays a minor repulsive barrier of ~0.56 eV just before chemically bonding.Subsequently, it attains a fully stable chemisorption state with the SV site, resulting a favorable energy of −3.28 eV.On the other hand, the H2O prefers a physisorption-like configuration with the SV of WS2, which does not show a chemisorption configuration.
Figure S3 shows the first-principles NEB calculation results for the molecular interaction with pristine WS2.In case of the physisorption, all three molecules can be weakly physisorbed with the pristine WS2.However, the weak adsorption energies in the final states lead to unstable physisorption configurations, resulting in the easy desorption of physisorbed molecules on pristine WS2 surface.Thus the physisorption has virtually no influence on the electronic and optical properties of the WS2 monolayer.reflectance measurements, the exciton energy for the bare and h-BN encapsulated WS2 was assigned to be 2.087 and 2.042 eV, respectively.By considering the peak separation between the exciton species based on the literatures, [4,5] the photoluminescence spectrum for the bare WS2 is deconvoluted to three excitonic species corresponding to the neutral exciton (X 0 ), trion (X − ), and defect-related trapped exciton (L) states at 2.087, 2.042, and 2.026 eV, respectively.
For the h-BN encapsulated WS2, the two exciton species are assigned to be the neutral exciton (X 0 ), trion (X − ) at 2.042 and 2.001 eV, respectively.Note that the redshift of the exciton and trion energy in the h-BN encapsulated WS2 compared with the bare WS2 is due to the increase in the dielectric constant of the environment by the h-BN encapsulation. [6]6 as widely reported in the previous works. [7,8] The peak A (black) at 538 eV can be decomposed into the two independent peaks at 537.7 eV (blue) and 538.7 eV (yellow).The physical origin of these decomposed peaks is found to be similar as depicted in the (b) and (c).However, the peak B at 555 eV is mostly contributed from that of the  2 (;  2 ).

S12. Power-law for the neutral exciton emission intensity at room temperature
The exciton generation and recombination in the steady-state photoluminescence can be described by the following rate equation: [12] G =     +     +     +   2 (1)   where  is the exciton generation rate,   is the neutral exciton density,   is the exciton lifetime,   is the electron density,  is the trion formation coefficient,  is the exciton-electron Auger coefficient, and  is the exciton-exciton annihilation coefficient.Note that the contribution of localized excitons and dark excitons can be neglected by the thermal activation effect at room temperature. [13,14]The exciton-electron Auger (    ) and exciton-exciton annihilation (  2 ) processes can be neglected at a low level of excitation, and thus the excitonto-trion conversion process becomes a dominant nonradiative decay at our experimental conditions.When the neutral exciton and free electron density increase with increasing the excitation power density (for the bare WS2), the equation ( 1) can be simplified as the following equation: When the free electron density increases with the power dependence of   (  ~   ), the neutral exciton density can be written by where the exciton generation rate is proportional to the excitation power ( ∝ ).The neutral exciton emission intensity for the excitation power can be expressed as the following relation (4), showing that the neutral exciton emission intensity follows the power-law of  1− .

S14. Estimation of the free electron density using mass action law
To determine the free electron density in the bare and h-BN encapsulated WS2, we used the mass action law describing the formation of the trion ( − ) from the neutral exciton ( 0 ) and the free electrons ( ).The formation of trion can be expressed by the following chemical equation: [15,16]  0 +  ⇔  − From the chemical equation, the density of each carrier can be related by where   − ,   0 , and   correspond to the density of trions, neutral excitons, and electrons.
() is the temperature-dependent equilibrium constant defined by the following equation: where   − ,   0 , and   are the effective mass of the trion, neutral exciton, and electron.The effective mass of the trion and neutral exciton was calculated to be 1.33 0 (  − = 2  +  ℎ ) and 0.89 0 (  0 =   +  ℎ ) , where  0 is the electron mass,   (0.44 0 ) is the effective mass of the electron,  ℎ (0.45 0 ) is the effective mass of the hole,   is the Boltzmann constant,  is the temperature, and    − is the trion binding energy (~34 meV). [17,18] using the equations ( 2) and (3), we can obtain the following parameter: The intensity weight (  − /  ) of trions in the photoluminescence spectra measured from the bare and h-BN encapsulated WS2 crystals can be written by the following equation: where   0 and   − are the decay rate of neutral excitons and trions, respectively.  −   0 ⁄ was estimated by measuring the lifetimes of neutral excitons and trions.
By comparing the parameter (4) and the equation ( 5), we can obtain the following equation: From the photoluminescence spectra measured as a function of the excitation power density, we estimated the free electron density in the bare and h-BN encapsulated WS2 crystals.

S15. Quantitative estimate for the number of desorbed/adsorbed oxygen molecules on sulfur vacancies in-between the vacuum and air envrionments
To investigate a quantitative estimate of the adsorption/desorption of oxygen molecules on WS2, we fabricated the bare WS2 capacitor device (Figure S15a), enabling the electrostatic control of electron cencentration as a function of gate voltage (  ).The gate-voltage-dependent PL measurements were performed at an excitation power density of 0.196 kW/cm 2 .As shown in Figures S15c and S15d, the charge neutral poins of the bare WS2 device were determined at   ≅ −14  and   ≅ −9  under the vacuum (Figure S15c) and air (Figure S15d) ambient conditions, respectively.We calculated the electron densities in the bare WS2 device at the vacuum and air environments using the equation ∆  2 =  × (  −   )  ⁄ , where ∆  2 indicates the electron density injected by the gate voltage,  is the capacitance of the used h-BN layer (6.55 × 10 −4 F•cm −2 ),   is the gate voltage,   is the onset voltage determined by the neutral point, and  is the electronic charge. [19]The electron densities were estimated to be 1.15 × 10 13 cm −2 (vacuum) and 3.69 × 10 12 cm −2 (air) at   = 0 , respectively, while those estimated using mass-action law were 1.11 × 10 13 cm −2 and 2.29 × 10 12 cm −2 at the vacuum and air environments, respectively, showing a good agreement.In-between the vacuum and the air environments, the free electron density induced by the desorption of oxygen molecules is estimated to be 7.77 × 10 12 cm −2 by considering the difference in the free electron densities at the vacuum and air environments.Since an oxygen molecule gains 1.083 electrons from WS2, [2] the number of desorbed/adsorbed oxygen molecules on the sulfur vacanies were estimated to be 7.17 × 10 12 cm −2 .
On the other hand, we also fabricated the h-BN encapsulated WS2 capacitor devices (Figure S15b).As shown in Figures S15e and S15f, the charge neutral points of the h-BN encapsulated WS2 device were determined at   ≅ 0.5  for both the vacuum and air ambient conditions, resulting in the electron densities of 2.05 × 10 11 cm −2 .where A1 and τ1 are the amplitude and the lifetime for the single exciton component.
Table S1.The lifetime of the neutral excitons measured as a function of the energy fluence for the bare WS2 under the vacuum and air ambient conditions.

Figure 1 .
Figure 1.a) Schematic illustration showing the chemisorbed oxygen molecules anchored by the h-BN encapsulation.Right inset image represents the detailed atomic configuration of the chemisorbed oxygen molecule at the sulfur vacancy.b) Scanning electron microscope image of the bare WS2 crystals on the line trenches.c) Spatial photoluminescence profile measured from the bare WS2 on the line trench.d) Scanning electron microscope image of the h-BN encapsulated WS2 crystals on the line trench.e) Spatial photoluminescence profile measured from the h-BN encapsulated WS2 on the line trenches.The yellow and black dashed lines marked in (b,d) and (c,e) indicate the boundary of the suspended and supported regions.The scale bars of (b,d) are 1 μm.f) Photoluminescence spectra for the bare (top panel) and h-BN encapsulated (bottom panel) WS2 measured at a cryogenic temperature of 77 K under a vacuum level of ~1 × 10 −5 Torr.g) Photoluminescence spectra for the bare (top panel) and h-BN encapsulated (bottom panel) WS2 measured under different ambient conditions of air and vacuum (T = 300 K).Each photoluminescence spectrum was fitted by Lorentzian functions.

Figure 2 .
Figure 2. a) Experimental oxygen K-edge EELS spectra for the h-BN encapsulated WS2, bare WS2, and h-BN flake crystals.b-e) The EELS maps for the boron (b), nitrogen (c), oxygen (d) K-edge, and the sulfur (e) L-edge measured from h-BN encapsulated WS2.The scale bars are 100 nm.f) The calculated oxygen K-edge EELS spectrum  2 () for a physisorbed oxygen molecule on the pristine WS2.g) The calculated oxygen K-edge EELS spectrum  2 () (red dashed) for a chemisorbed oxygen molecule on the sulfur vacancy (O2-SV) of the WS2.The experimental oxygen K-edge spectrum (blue line) for the h-BN encapsulated WS2 is in good agreement with the calculation result (red dashed).The insets of (f,g) represent the real-space orbital distributions ()  of the  * ( ≈ 530 eV) and  * ( ≈ 539 eV) for (f) and the A ( ≈ 538 eV) and B ( ≈ 555 eV) for (g), respectively.

Figure 3 .
Figure 3. a,b) Neutral exciton (X 0 ) emission intensity of the bare (a) and h-BN encapsulated (b) WS2 crystals as a function of the excitation power density at room temperature.The blue diamond and pink circle symbols indicate the power-dependent exciton emission intensity measured under ambient air and vacuum conditions.c) Photoluminescence intensity ratio (X − /X 0 ) of the trion to the neutral exciton as a function of the excitation power density.d) Free electron density as a function of the excitation power density by using the mass action law.The pink square (blue triangle) and olive diamond (purple circle) symbols represent the X − /X 0 and free electron density of the bare and h-BN encapsulated WS2 under vacuum (air) ambient conditions, respectively.The error bars in (a−d) indicate the standard deviation of the measured data.
bare WS2, the photoluminescence decay curves of neutral excitons under air ambient condition show a steep shortening of the decay time from 74 ps to 43 ps with increasing pump fluence from 27 to 959 nJ/cm 2 , as shown in Figure4a.However, the exciton decay curves for h-BN encapsulated WS2 show a slight decrease in the lifetime above the pump fluence of 421 nJ/cm 2 , as shown in Figure4b.It is noteworthy that the exciton lifetime in h-BN encapsulated WS2 becomes longer than that of bare WS2, showing exciton lifetime of 136 ps in air at a pump fluence of 27 nJ/cm 2 .The decay spectra at vacuum are shown in FigureS16in Supporting

Figure 4 .
Figure 4. a,b) Photoluminescence decay curves of the neutral excitons measured as a function of the energy fluence for bare (a) and h-BN encapsulated (b) WS2 under ambient air condition.c,d) Exciton density-induced recombination rate (τ −1 ) for the bare (c) and h-BN encapsulated (d) WS2 under ambient air and vacuum conditions, resulting in the exciton annihilation rate constant (RA) due to exciton-to-trion conversion process by a linear fit.e) Gate-voltagedependent photoluminescence spectral map of the h-BN encapsulated WS2 capacitor devices.The black arrow indicates the charge neutral point of the h-BN encapsulated WS2 capacitor device.f) Gate-voltage-dependent photoluminescence decay curves of the neutral excitons in the h-BN encapsulated WS2 capacitor device.

Figure 5 .
Figure 5. a) Circularly polarized photoluminescence spectra for the bare (top panel) and h-BN encapsulated (bottom panel) WS2 measured at 77 K. b) Degree of valley polarization for the exciton emission in the bare and h-BN encapsulated WS2 taken as a function of the excitation power density.c) Degree of valley polarization for the neutral excitons in the h-BN encapsulated WS2 capacitor devices taken as a function of the gate voltage.The error bars marked in (b,c) exhibit the standard deviation of the measured valley polarization values.
5 m focal length and 1200 grooves/mm grating) equipped with a charge-coupled device (CCD) detector (Princeton Instruments, 512 × 2048 pixels).For the circular polarization measurements, a combination of linear polarizer and quarter waveplate was used to generate circularly polarized excitation light, while another pair of linear polarizer and quarter waveplate was set for a polarization analyzer before collecting the signal through the slit of the spectrometer.The degree of valley polarization is defined as  = [(( + ) − ( − )] [( + ) + ( − )] ⁄ , where ( ± ) are the photoluminescence intensity for  + and  − polarized light components under excitation with a  + or  − polarized laser beam.The time-resolved photoluminescence measurements were carried out using a picosecond pulsed diode laser (PicoQuant, LDH-P-FA-355) with a wavelength of 355 nm (FWHM = 56 ps) and repetition rate of 40 MHz.The exciton lifetimes were measured using a hybrid photomultiplier detector (PicoQuant, PMA hybrid series) and a time-correlated single photon counting system (PicoQuant).

Figure S1 .
Figure S1.(a,b) Calculated reaction path and kinetic barriers for the O2-chemisorption and O2dissociative chemisorption processes in the monolayer WS2 (a) and WTe2 (b) with the sulfur vacancy, respectively.

Figure S2 .
Figure S2.First-principles NEB calculations for molecular interactions with SV of WS2.(a−c) The relative energy evolution of the N2 (a), O2 (b), and H2O (c) for each reaction path from the initial configurations to the final configurations.The relative energy is defined as the total energy difference compared to that of the initial.Adsorption of a molecule at given surface has two different forms, physisorption and chemisorption.The physisorption reflects an exothermic process, resulting in a decrease in relative energy.The chemisorption, on the other hand, exhibits an activation barrier in the chemical bonding sequence, leading to an increase in relative energy.(a) The N2 molecule does not achieve a structurally favorable chemisorption state when adsorbed onto the SV.(b) The O2 molecule can be only chemisorbed onto the SV and attain a fully stable chemisorption state (−3.28 eV).(c) The H2O molecule prefers a physisorption-like configuration with the SV of WS2, which does not show a chemisorption.

Figure S3 . 2 Figure
Figure S3.First-principles NEB calculations for molecular interactions with pristine WS2.(a−c) The relative energy evolution of the N2 (a), O2 (b), and H2O (c) for each reaction path from the initial configurations to the final configurations.

.
Figure S6.a,b) Photoluminescence spectra for the bare (a) and h-BN encapsulated (b) WS2 measured under the ambient conditions of air, vacuum, and then air again.The photoluminescence spectral features of bare WS2 are significantly altered according to the change in ambient condition, while those of h-BN encapsulated WS2 are kept almost constant regardless of the change in ambient condition.

S10. Electron energy loss spectroscopy maps for h-BN encapsulated WS 2 Figure
Figure S10.a) ADF-STEM image for the h-BN encapsulated WS2 crystals.b-e) The EELS maps for the boron (b), nitrogen (c), oxygen (d) K-edge, and the sulfur (e) L-edge measured at the region marked by the yellow square box in (a).f) EELS spectrum measured at the region indicated by the green square box in (e).

Figure S11 .
Figure S11.Core-hole dependent EELS spectra and orbital distributions.As long as chemisorbed, the two oxygens are not equivalent to each other and oxygen-dependent core-hole states become distinguishable.Here we assume that the experimental EELS is equally induced from the two kinds of oxygen core-holes.a) core-hole decomposed EELS,  2 () =  2 (;  1 ) +  2 (;  2 ), and the  2 () is displayed in the Figure 2(c) of the main text.b,c) The (,   ) for each peak, A and B, are displayed.White arrows indicate the oxygen atom with core-hole, i.e.,  =1 (b) and  =2 (c).The peak A (black) at 538 eV can be decomposed into the

Figure S13 .
Figure S13.Photoluminescence intensity ratio (X − /X 0 ) of the trion to the neutral exciton with increasing the excitation power density.In a power-law ( − / 0 ∝   ), for the bare WS2, the neutral excitons are converted to the trions in proportional to P µ with the exponent  of 0.71 and 0.29 in the ambient vacuum and air conditions with increasing the excitation power, while showing the exponent of 0.06 and 0.04 for the h-BN encapsulated WS2 in the vacuum and air ambient conditions, respectively.As shown in Figures3(a,b), the exciton emission intensity in the bare and h-BN encapsulated WS2 increases with the exponent  ≅ 1 −  under the vacuum and air conditions, indicating that the neutral exciton-to-trion conversion is the predominant process in the nonradiative decay of the excitons.Thus, Auger recombination process can be neglected under our excitation conditions.

Figure S15 .
Figure S15.a,b) Schematic illustration for the bare (a) and h-BN encapsulated (b) WS2 capacitor devices.c,d) Gate-voltage-dependent photoluminescence spectral maps for the bare WS2 device under the vacuum (c) and air (d) ambient conditions.e,f) Gate-voltage-dependent photoluminescence spectral maps for the h-BN encapsulated WS2 device under the vacuum (e) and air (f) ambient conditions.

Figure S18 .
Figure S18.Recombination rate (τ −1 ) as a function of free electron density for the h-BN encapsulated WS2 capacitor devices.A linear fit for the τ −1 results in the exciton annihilation rate constant (RA) due to the exciton-to-trion conversion process.

Table S2 .
The lifetime of the neutral excitons measured as a function of the energy fluence for the h-BN encapsulated WS2 under the vacuum and air ambient conditions.

Exciton lifetime as a function of free electron density and annihiliation rate constant in the h-BN encapsulated WS 2 capacitor devicesTable S3 .
The lifetime of the neutral excitons in the h-BN encapsulated WS2 capacitor device measured as a function of the gate voltage at a fixed pump fluence of 27 nJ/cm 2 .