W ithin the expanding family of two-dimensional (2D) materials, layered transition metal dichalcogenides (TMDs) have emerged as leading candidates for electronics and optoelectronics applications, and provide a versatile platform for realizing novel correlated quantum states. Following the paradigm of foundational materials such as silicon and GaAs, minimizing defect density is essential for advancing both applications and fundamental research. Moreover, the electronic and optical properties are particularly sensitive to atomic defects due to their extreme 2D nature and reduced dielectric screening. Recent advances in self-flux synthesis have dramatically reduced defect densities in bulk TMD crystals by orders of magnitude, 1 enabling enhanced device functionality and access to new quantum phenomena. [2][3][4][5][6][7] These ultrapure materials provide an essential benchmark close to intrinsic optical and electronic properties, facilitating direct correlations between defect density, macroscopic device performance, and quantitative theoretical modeling. 4,6,[8][9][10][11] However, further advancement hinges on identifying the residual defects in the monolayers derived from these crystals, enabling refined synthesis, optimized performance, and deeper insights into defect-driven phenomena.

In previous studies of defects in self-flux TMDs we have used STM and conductive atomic force microscopy (cAFM) to quantify the defects at the surface of freshly cleaved bulk crystals. 1,12 Two classes of defects are found: sparse (10 9 -10 10 cm -2 ) charged defects that induce band bending and appear as prominent features extending over nanometers, and more numerous (>10 11 cm -2 ) isovalent defects that are subtle and more localized. Identifying the nature of these prevalent isovalent defects is essential for developing defect mitigation strategies, and will facilitate theoretical and experimental analysis of how each type affects bulk properties such as carrier mobility and excited-state lifetime.

The current prevailing consensus in the literature regards chalcogen vacancies as the dominant intrinsic defect in TMDs, [13][14][15][16][17] owing to their low formation energy [18][19][20] and frequent observation in electron microscopy studies. [21][22][23] However, experimentally observed defect densities are orders of magnitude higher than thermodynamically predicted values. 24,25 Scanning transmission electron microscopy (STEM) defect studies can be misleading, as chalcogen vacancy defects are easily formed by radiation damage during measurements, 26,27 and substitutional impurities of light elements can be difficult to distinguish from vacancies. 28 Moreover, different synthesis methodsincluding film growth by chemical vapor deposition (CVD) and molecular beam epitaxy (MBE), and crystal growth by chemical vapor transport (CVT) and self-flux synthesisinvolve distinct physical processes that can potentially give rise to different defect types. A recent study concluded that the dominant defects in CVD-grown WS 2 and MBE-grown MoSe 2 monolayers are substitutional oxygen. 29 The presence of O in CVD-grown TMDs is not surprising as WO 3 and MO 3 precursors are commonly used, [30][31][32] and O passivation of chalcogen vacancies in TMDs is theoretically predicted to be energetically favorable. [33][34][35] Nevertheless, this finding motivates careful investigation of defects in self-flux crystals.

In this work we use STM/S and complementary theoretical calculations to characterize the isovalent defects in the model system of pristine self-flux WSe 2 . We identify the defects in stacked monolayer WSe 2 , determining they are Se-site substitutional O impurities (O Se ). These O Se defects are present at the same density in bulk WSe 2 , indicating that oxygen is introduced during synthesis and that mechanical exfoliation and stacking do not change the nature or density of the defects.

We first examine defects in monolayer WSe 2 by creating a sample consisting of an exfoliated monolayer of WSe 2 on a graphite flake, as shown in Figure 1a (details of sample fabrication in Methods). Extending the previous studies of defects in flux-grown WSe 2 to monolayers brings two advantages: first, it permits direct comparison to previous studies of CVD-grown monolayers; 29 second, it provides understanding as to whether mechanical exfoliation and stacking of vdW heterostructures modifies or introduces defects.

Figure 1b presents an STM image obtained at the edge of WSe 2 , the inset revealing the atomic lattice without observable contamination. The bright contrast at the edge corresponds to cleaved WSe 2 bonds. Figure 1c shows a larger (50 × 50 nm 2 ) STM image, which also reveals a pristine surface without polymer residue. The surface remains remarkably free of adsorbates even after exposure to atmosphere during exfoliation and dry stacking, in contrast to prior STM studies of MBE-grown TMDs exposed to atmosphere. 36 This uniformly clean character is observed for even larger areas, where we also observe the previously reported charged defects (see Supporting Information). STM images of the monolayer also reveal subtle localized isovalent defects as previously observed in bulk self-flux crystals, 1 which are seen as dark points in Figure 1c. The low density (∼ 5 × 10 11 cm -2 ) of these defects confirms that even after exfoliation and dry stacking, self-flux synthesis yields higher-purity monolayer TMDs than direct synthesis methods. 37,38 Figure 2 compares the appearance and density of the defects (circled to guide the eye) in the monolayer vs the parent bulk crystal from which the monolayer was derived. Whereas the defects appear dark in Figure 1c, the sample bias in these images is tuned deeper into the band where the defects appear as bright features. We note that multiple STM studies of CVD grown WSe 2 have reported defects with similar bias-dependent behavior, hinting that the defects are identical 35,39,40 (see Supporting Information for discussion of bias dependence).

In both the monolayer (Figure 2a) and the bulk (Figure 2b), the defects appear as bright rings with dark centers, consistent with prior observations of isovalent defects in flux-grown TMDs. 1,12 Furthermore, the defect density is identical within experimental uncertainty. We measured a density of 5.24 ± 0.55 × 10 11 cm -2 (101 defects over 1.925 × 10 -10 cm 2 ) in the monolayer and 5.30 ± 0.49 × 10 11 cm -2 in the bulk (136 defects over 2.562 × 10 -10 cm 2 ), confirming that the exfoliation and stacking process does not introduce additional defects.

To determine whether the bulk defects correspond to vacancies that could be passivated by oxygen, we exposed freshly cleaved WSe 2 to pure oxygen for 1 week. As shown in Figure 2c, oxygen exposure had no observable effect on the appearance or density of the isovalent defects, indicating their chemical stability. Since vacancies are highly reactive and should be modified by oxygen exposure, 33,34 this suggests that the defects are substitutional impurities. Prior STM and theoretical investigations of TMDs defects determined that chalcogen vacancies induce in-gap states near the conduction

Letter band edge, 33,34,41 while O Se impurities slightly perturb the valence band edge. 29,[33][34][35] Scanning tunneling spectroscopy (STS) reveals the electronic structure of the defects 2d, showing that they do not introduce in-gap states in the monolayer (top panel) or bulk (bottom panel). Taken together, the chemical stability, electronic characteristics, and consistent appearance of the monolayer and bulk defects strongly supports their identification as substitutional impurities rather than Se vacancies.

Having quantitatively shown that exfoliation, stacking, and oxygen exposure do not introduce or alter isovalent defects in exfoliated TMD monolayers, we now examine the defect structure in detail. Figure 3a presents a high-resolution STM image of monolayer WSe 2 that shows two types of defect. These defects are strikingly similar to the prevalent defects reported by numerous STM studies of CVD grown monolayer TMDs. 29,35,37,39,40 Both types are centered between the bright points in the STM image, which have been commonly attributed to surface chalcogen atoms. [42][43][44] Thus, some studies have identified the defects as W vacancies 39,40 or interstitials, 35 whereas a study combining STM with noncontact AFM identified them as substitutional oxygen at the chalcogen site. Our recent study showed that the charge density of the valence band edge of WSe 2 is actually localized at the honeycomb center, as depicted by the simulated STM image in Figure 3b, 45 clarifying the difficulty in defect identification. With this knowledge, closer examination of the two defect types with an overlay of the atomic lattice (Figure 3c,d) reveals that they are in fact located at the chalcogen (Se) sites, as indicated by the rotational alignment of the neighboring bright honeycomb centers. The density of these two defects is nearly identical (2.70 ± 0.43 and 2.54 ± 0.35 × 10 11 cm -2 ), indicating that they are the same defect species located at the top and bottom Se sites.

Our interpretation of the STM images is consistent with the assignment of the defects as O Se at the top (Figure 3c) and bottom (Figure 3d) Se sites, as shown schematically in Figure 3e. Moreover, we previously showed that the isovalent defect density in bulk crystals closely matches the oxygen impurity concentration as determined by elemental analysis. 1 Having established above that the bulk and monolayer have identical defects (Figure 2), this provides direct evidence that the isovalent defects are O Se introduced during synthesis, likely by the W source powder (see Supporting Information for details). We performed similar studies of bulk self-flux grown MoSe 2 and observed identical defects, indicating O Se are common in other self-flux TMDs (see Supporting Information). To support our identification, we perform first-principles STM simulations of top and bottom O Se (Figure 3f,g, respectively). We find the simulation of the top O Se defect agrees well with the STM images (Figure 3c,f). The simulated bottom O Se defect shows less agreement, likely due to the difference in dielectric environment between the freestanding WSe 2 used in the calculation and the experimental condition of WSe 2 supported by a graphite substrate (Figure 3d,g). Taken together, our results show that O Se are the most common defect in self-flux TMDs, contradicting the widely accepted theory that chalcogen vacancies are the most prevalent defects.

We next investigate the O Se defects in bulk self-flux WSe 2 observed in Figures 2b,c. The STM image in Figure 4a shows that upon closer examination, we observe three distinct defect types: two with a similar bright appearance which we denote types 1 and 2 (blue and green arrows) and a third with a more subtle dark contrast denoted type 3 (red arrow). Although the bright defects might be mistaken for a single type, closer inspection (Figure 4b,c) reveals that types 1 and 2 have a subtle difference in the dark center, which can be distinguished by the line profiles Figure 4a (inset). We find that the defect density of all three types is equal within experimental uncertainty (2.61 ± 0.27, 2.69 ± 0.47, and 2.68 ± 0.48 × 10 11 cm -2 ). Since only two defect types are present in the monolayers, the most straightforward interpretation of these observations is that the three defect types are O Se in the uppermost three Se planes, shown schematically in Figure 4d. Further, we note that these values agree well with the measured densities of the bright defects from 2a,b,c, as they correspond to the defect density of only the top layer of WSe 2 .

Examining defect types 2 and 3 more closely, we observe that they are both positioned between the bright features (Figure 4c,e), similar to the case of monolayer WSe 2 (Figure 3). However, the bright features near the center of the two defects are rotated by 180°with respect to each other, indicating that defect type 3 is centered at the W site of the top layer (see Supporting Information). In 2H-stacked WSe 2 the Se site of the subsurface layer is centered at the W site of the top layer (see Figure 4d schematic). Thus, we attribute defect type 3 to a buried O Se in the layer below. Figure 4f shows a simulated STM image of the buried O Se type 3 defect that shows a similar appearance, supporting our interpretation.

High-purity TMD crystals grown by the self-flux method have been widely used in optical and transport studies of both monolayer and heterostructure devices. 2,[4][5][6]10,[46][47][48] However, previous STM studies have examined only bulk crystals, and the identity of the point defects has remained unclear. This study confirms that self-flux TMD monolayers have the same defect profile as bulk counterparts, and identifies oxygen substitutional impurities (O Se ) as the dominant defect species. O Se defects are chemically stable and likely introduced during synthesis; their negative formation energy suggests that their density is governed by oxygen availability rather than thermodynamic constraints. [33][34][35] Therefore, eliminating possible sources of oxygen contamination during synthesis will be crucial for achieving further improvements in the purity of TMD crystals.

This work corroborates the previous identification of substitutional oxygen as the primary point defect in TMD monolayers synthesized by chemical vapor deposition (CVD) and molecular beam epitaxy (MBE). 29 The consistent incorporation of oxygen across these distinct growth and preparation methods, including vapor phase, epitaxial, and bulk self-flux synthesis, strongly suggests that substitutional oxygen is a pervasive point defect in TMDs. Theoretical studies support this picture, showing that oxygen substitution is exothermic and that chalcogen vacancies are readily passivated upon exposure to air. Nevertheless, the prevailing view that chalcogen vacancies are the primary point defects in TMDs continues to guide much of the literature, shaping interpretations of many optical and electronic phenomena. [13][14][15][16][17] Our results highlight the need to reassess these assumptions and consider oxygen impurities and other possible defect species in future analyses.

Recent studies have linked high densities of oxygen substitutionals to a variety of macroscopic effects, including changes in optical spectra, doping behavior, and charge transport, suggesting that O Se may play a role in device performance. [49][50][51] At the low densities we observe in self-flux WSe 2 , O Se defects are likely relatively benign for most applications as they do not introduce midgap states. However, their presence might strongly impact phenomena observed at low temperatures, where transport is dominated by short-range scattering of isovalent defects, 4 and weak local potential perturbations could suppress correlated phases. 11,52 Therefore, minimizing substitutional oxygen could be crucial for unlocking low-energy states and enabling the exploration of emergent quantum phenomena in TMDs.

The presence of multiple coexisting defect species within a material under study can complicate efforts to understand the effects of any given defect type. The availability of ultrapure bulk crystals with a well-identified and relatively benign dominant defect offers an ideal platform for isolating and directly probing the influence of individual defect species. Introducing known defects into these crystals (e.g., by substitutional doping) will enable targeted investigations of their effects on optical and electronic properties toward future defect engineering applications.

■ METHODS TMD Synthesis. WSe 2 crystals were synthesized using a two-step self-flux method with a metal to chalcogen molar ratio of 1:100 following the method of ref. 1 Tungsten powder of 99.999% purity and selenium shot of 99.999+% purity were put in a quartz ampule that was evacuated and sealed at 10 -5 Torr. The ampules were heated to 1000 °C for 24 h, held for 2 weeks, then cooled at a rate of 1 °C/h to 500 °C, before increasing the cooling rate to 5 °C/h until reaching room temperature. The ampule contents were transferred to a new quartz ampule with a quartz wool filter, evacuated, sealed, and heated to 285 °C to melt and separate the excess selenium by centrifuge. The isolated crystals were removed and annealed across a temperature gradient for 24 h with the hot end at 285 °C, and the cold end at room temperature.

Fabrication of Monolayer TMD Samples. Bulk WSe 2 and highly oriented pyrolytic graphite (HOPG) crystals were mechanically exfoliated using the conventional tape technique onto 285 nm SiO 2 /Si substrates. The monolayer WSe 2 / graphite heterostructures were assembled by the layer-by-layer dry stacking method with poly(propylene carbonate) (PPC). 53 The stacks were transferred to a SiO 2 wafer by melting the polymer at 100 °C with the vdW stack on top. Indium contacts were then made to the graphite before the wafer was mounted on the STM sample holder using conductive silver paste (EPO-TEK H20E).

STM Characterization. STM measurements were obtained using a commercial Omicron ultrahigh vacuum STM system. A chemically etched W STM tip was cleaned and calibrated using an Au(111) single crystal prior to the measurements. Measurements were obtained at room temperature (300 K).

Letter Density-Functional Theory (DFT) Calculations. DFT calculations were performed using the Quantum ESPRESSO software package, 54 employing optimized norm-conserving Vanderbilt (ONCV) pseudopotentials 55 and the Perdew-Burke-Ernzerhof (PBE) 56 exchange-correlation functional. A kinetic energy cutoff of 85 Ry was applied to the plane-wave basis set to optimize the geometric structure and obtain the electronic structure. For defect-state STM simulations, a 5 × 5 × 1 supercell of monolayer WSe 2 was used, with a 3 × 3 × 1 kpoint mesh sampling the first Brillouin zone. STM data analysis was carried out using critic2. 57,58