Covalent nitrogen doping in molecular beam epitaxy-grown and bulk WSe₂

Transition metal dichalcogenides (TMDs), a family of two-dimensional (2D) materials with layered structure, have gained much attention recently due to their unique material properties like atomic-scale thickness, sizeable bandgap, and potential applications in various electronic and optoelectronic devices such as tunneling field-effect transistors (FETs), p-n diodes, and photovoltaics.^1–4 Tungsten diselenide (WSe₂) is among the most promising candidates for implementation of a TMD-based channel in next generation FETs due to its indirect-to-direct bandgap transition from bulk (∼1.2 eV) to monolayer (∼1.8 eV) and high p-type mobility up to 500 cm²/V-s at room temperature.^5–7 Controllable and reliable doping of TMDs is essential for realizing commercially viable TMD-based devices. Various techniques have been employed in efforts to dope WSe₂ such as a “mild” He or H₂ plasma treatment to create anion Se vacancies,⁸ Octadecyltrichlorosilane (OTS) and hexane solution treatment,⁹ SiN_x deposition on WSe₂,¹⁰ potassium doping of WSe₂ using a K dispenser,¹¹ and Au decoration of WSe₂ using thermal evaporation,¹² none of which involve substitutional doping. Other methods such as adsorption of NO₂ molecules on WSe₂ lead to substitutional doping;¹³ however, the formation of various WSe_2−x−yO_xN_y species with different oxidation states was detected. A more manufacturable approach, using a remote N₂ plasma, has also been employed previously to introduce nitrogen into the graphene lattice (among other materials).^14,15 Recently, this remote N₂ plasma method was reported as a successful method for controllable doping of exfoliated MoS₂.¹⁶ In this work, we extend and compare the effect of similar N₂ plasma treatment on bulk WSe₂ and also on a few monolayers of WSe₂ grown by molecular beam epitaxy (MBE).

WSe₂ samples (thickness ≤ 7 layers) were grown via MBE on highly ordered pyrolytic graphite (HOPG) substrates provided by Momentive Performance Materials. The HOPG substrates were freshly exfoliated prior to being loaded into vacuum, followed by annealing at the 550 °C growth temperature for 1 h. Once the HOPG surface preparation was finished, the 2 h MBE growth of WSe₂ commenced with a W flux pressure of 7 × 10⁻⁹ mbar and a Se flux pressure of 1 × 10⁻⁶ mbar. Samples were slowly cooled down under a Se environment, and RHEED was performed in situ resulting in streaky patterns indicative of high-quality, smooth films.¹⁷ The MBE samples were then removed from the growth chamber for the ex situ nitrogen doping experiments. Bulk WSe₂ crystals, grown by chemical vapor transport (CVT) and purchased from HQ Graphene, were exfoliated in air using Scotch® Magic™ tape. Immediately following bulk WSe₂ exfoliation, both non-exfoliated MBE and exfoliated bulk WSe₂ samples were mounted on the same Mo-plate and loaded into an ultra-high vacuum (UHV) system described elsewhere¹⁸ in less than 10 min. Samples were heated to 300 °C under UHV and held at this temperature for 2 h (base pressure ∼ 10⁻⁹ mbar) to remove physisorbed species induced by air exposure following exfoliation. Subsequently, the samples were exposed to a N₂ plasma, remotely generated via 13.56 MHz Radio Frequency (RF) at a power of 100 W under 45 SCCM nitrogen gas flow, for different sets of exposure time with an operation pressure of ∼7 × 10⁻³ mbar. To study the surface chemistry after each treatment, the samples were transferred through a UHV transfer tube to an analysis chamber with a base pressure of ∼10⁻¹⁰ mbar equipped with X-ray photoelectron spectroscopy (XPS) described elsewhere.¹⁸ The software AAnalyzer was used for XPS peak analysis. Active Shirley background and Voigt line shape was employed for peak fitting.¹⁹ Raman and photoluminescence (PL) spectroscopy were employed to investigate the characteristic vibrational modes and optical bandgap of monolayer and bilayer WSe₂ flakes and films before and after 10 min N₂ plasma treatment. For Raman and PL measurements, few layers WSe₂ samples were prepared by mechanically exfoliating the bulk WSe₂ crystal and transferring them to thermally grown 270 nm SiO₂/Si substrates. Raman and PL spectra were acquired via 532 nm wavelength laser with a spot size of ∼500 nm and 0.22 mW power under ambient conditions. An atomic force microscope (AFM), Veeco, Model 3100 Dimension V, operating in the tapping mode under ambient conditions was also used to characterize the surface morphology ex situ before and after the N₂ plasma treatment. In a separate UHV system equipped with an Omicron variable temperature scanning tunneling microscope (VT-STM),²⁰ room temperature STM and scanning tunneling spectroscopy (STS) were employed to compare the surface morphology and electronic structure of the MBE WSe₂ films before and after the N₂ plasma treatment. AFM and STM images were processed with the WSxM software.²¹ 

Figures 1(a)–1(c) show the W 4f, Se 3d, W 5p_3/2, and N 1s core level spectra obtained from as-grown MBE WSe₂ and as-exfoliated WSe₂. Only spectral features associated with WSe₂ in the W 4f and Se 3d core level spectra are detected, which indicates initial WSe₂ surface oxidation is below detectable limits despite 10 min air exposure prior to loading into the UHV cluster tool. A small amount of adventitious carbon and oxygen are detected on the surface due to the short air exposure prior to loading into UHV as evidenced by the low-intensity chemical states detected in C 1s and O 1s in Fig. S1 of the supplementary material. In order to remove these undesired species and also prevent the formation of CN_x during N₂ plasma treatment, samples were annealed at 300 °C for 2 h under UHV conditions.¹⁶ It is important to note that prior to the N₂ plasma treatment, the N 1s core level region is convoluted with a broad Se feature at 400.0 eV, as shown in Fig. 1(c). Nitrogen remains below the XPS detection limit following N₂ plasma treatment for less than 10 min. For exposure times ≥10 min, the binding energy of the chemical state detected in the N 1s core level (397.9 eV) is consistent with previous reports of a transition metal-nitrogen bond.^22,23 An additional chemical state is detected in the W 4f core level obtained from both MBE and bulk WSe₂ at ∼0.3 eV higher binding energy to that of W–Se after 10 min N₂ plasma exposure. The binding energy of the chemical state detected in the N 1s core level, in conjunction with the additional chemical state detected in the W 4f core level after 10 min N₂ plasma exposure, suggests the formation of covalent W–N bonds during the N₂ plasma treatment. Previous reports on W–N chemical states have shown similar binding energies to those reported here.^22–24 The detection of chemical states consistent with covalent W–N bonding in the W 4f and N 1s core level spectra obtained from both MBE and bulk WSe₂ samples suggests nitrogen incorporation into WSe₂ through Se substitution.

No additional chemical state was detected in the Se 3d core level aside from that of WSe₂ through all N₂ plasma exposure times investigated in this work, which indicates Se–N bond formation, if any, is below the limit of detection by XPS. After 20, 30, and 60 min exposure time, the intensity of the chemical state detected in the N 1s core level obtained from both MBE and bulk WSe₂ gradually increases (Fig. S2 of the supplementary material). Increasing the N₂ plasma exposure time results in an increased nitrogen to tungsten ratio and corresponding decrease in the selenium to tungsten ratio in WN_xSe_2−x. After 60 min N₂ plasma exposure, x = 0.17 for MBE WSe₂ and x = 0.16 for bulk WSe₂ as calculated from the XPS data. It should be noted that the W–N bond represents the only nitrogen species detected in either sample after N₂ plasma exposure of any duration. According to angle resolved XPS (ARXPS) performed at a bulk sensitive angle (75°) versus a surface sensitive angle (45°), the remote N₂ plasma exposure introduces nitrogen to the surface of WSe₂ samples (Fig. S3 of the supplementary material). Taken together, the ARXPS and detected surface chemical states are consistent with Se replacement by N from the remote plasma process, analogous to that reported for MoS₂.

The XPS spectra also shows a clear shift to lower binding energy by the same amount for all core levels, indicative of a Fermi level shift. The valence band maximum (VBM) was extrapolated from linear fits of the valence band spectra obtained throughout the experiment [Fig. 1(d)]. A valence band edge shift to lower binding energy (toward the Fermi level) is an indicator of p-type doping induced by N₂ plasma treatment.⁹ P-type doping behavior upon Se substitution by nitrogen is expected considering the electronegativity of nitrogen comparing to selenium.²⁵ After 10 min exposure to N₂ plasma, the valence band edge shifts detected from MBE WSe₂ (0.2 eV) and bulk WSe₂ (0.5 eV) were consistent with the shifts in the W 4f and Se 3d core levels. Fermi level shifts toward the valence band edge (p-type) were larger in bulk WSe₂ compared to MBE WSe₂ in all sets of samples investigated in this work. The difference in the Fermi level shift detected between MBE and bulk WSe₂ samples could be explained by the initial VBM position of the respective samples. The as-grown MBE film is initially p-type as indicated by the VBM located 0.7 eV below the Fermi level while the VBM initially detected from bulk WSe₂ is 1.2 eV below E_F. These differences could be due to extrinsic impurities in the bulk crystals.⁶ 

In comparison with the study on MoS₂,¹⁶ nitrogen incorporation into WSe₂ occurs at a significantly slower rate as nitrogen is detectable after 10 min N₂ plasma exposure compared to just 2 min in the case of the same treatment applied to MoS₂. In addition, after 60 min N₂ plasma exposure, the nitrogen to tungsten ratio in WN_xSe_2−x is much lower than the nitrogen to molybdenum ratio in MoN_xS_2−x obtained after the same duration of plasma exposure.

Similar to the covalent doping of MoS₂ by N₂ plasma,¹⁶ the incorporation of nitrogen into WSe₂ samples and formation of the W–N bond could result in compressive strain due to a smaller atomic radius of nitrogen compared to that of selenium. Raman spectroscopy is known to be a powerful tool to reveal the effect of doping on the sample structure. Raman spectroscopy was employed to qualitatively investigate the effects of N₂ plasma on lattice strain in exfoliated monolayer (ML) and bilayer (BL) WSe₂ flakes. Figures 2(a) and 2(d) show optical images of ML and BL WSe₂ flakes, respectively. The separation between the peak convoluted with degenerate E¹_2g/A_1g vibrational modes and the higher wavenumber peak corresponding with the second order Raman mode due to LA phonons at the M point in the Brillouin zone [2LA(M)]²⁶ is a fingerprint of number of layers of WSe₂.^26,27 The separation between the degenerate E¹_2g/A_1g peak and 2LA(M) peak detected from the flakes shown in Figs. 2(a) and 2(d) is ∼11 cm⁻¹ and ∼6 cm⁻¹, consistent with ML and BL WSe₂, respectively [Figs. 2(b) and 2(e)]. After 10 min N₂ plasma exposure, the E¹_2g/A_1g peak blue shifts 2.3 cm⁻¹ and 1.4 cm⁻¹ for the monolayer and bilayer samples, respectively. The blue shift in the E¹_2g/A_1g peak indicates compressive strain from a reduction in the lattice constant²⁸ and also p-type doping. P-type doping has also been correlated with an intensity increase of the A_1g phonon.⁹ 

To examine the optical bandgap, PL spectroscopy was employed on the same exfoliated WSe₂ flakes as were probed with Raman spectroscopy. The peak position at 1.66 eV detected by PL from the ML WSe₂ flake [Fig. 2(c)] corresponds with the direct bandgap of monolayer WSe₂.²⁸ Monolayer WSe₂, of course, exhibits the highest PL intensity among ML, BL, Tri-layer, and quad-layer WSe₂ flakes,²⁶ with a maximum intensity over the limit of the detector employed in this work. Figure 2(f) shows the PL spectrum obtained from bilayer WSe₂ with peak intensity at 1.52 eV, consistent with a previous report.²⁸ A red shift in PL peak is detected after 10 min N₂ plasma exposure in the case of monolayer WSe₂ [Fig. 2(c)], consistent with a change in band structure due to induced strain caused by the covalently incorporated nitrogen.²⁸ Desai et al.²⁸ report a crossover from indirect to direct bandgap in bilayer WSe₂ is practically possible by applying strain due to the small difference between the direct and indirect bandgaps of bilayer WSe₂. As strain increases, a second peak, corresponding with the direct bandgap, appears at higher energy in the PL spectrum and increases in intensity until the indirect bandgap peak is no longer resolvable. Further increase in strain results in a red shift of the direct bandgap peak of bilayer WSe₂. In this work, the PL spectra detected from bilayer WSe₂ after 10 min N₂ plasma exposure exhibits characteristics reflecting a lesser magnitude of strain than that corresponding with a PL spectrum dominated by the direct bandgap peak as observed by Desai et al. Both Raman and PL provide evidence of successful nitrogen incorporation at the surface of mono and bilayer WSe₂ flakes resulting in p-type doping and inducing strain.

This work demonstrates covalent p-type doping of both MBE and bulk WSe₂ via a remote N₂ plasma treatment. However, surface roughening and/or etching could be an undesirable side effect of the N₂ plasma exposure to WSe₂ in a device fabrication procedure.²³ The WSe₂ surface topography after N₂ plasma exposure was evaluated by AFM and STM. Figure 3(a) shows the flat surface of bulk WSe₂ before N₂ plasma treatment. After N₂ plasma exposure for 15 min, there are no obvious topographical changes [Fig. 3(b)] in the surface of bulk WSe₂ consistent with an analogous N₂ plasma treatment on bulk MoS₂ for short exposure times (15 min).¹⁶ After 60 min of the N₂ plasma treatment, significant changes in the surface structure appear in the form of cracks with depth of ∼0.7 nm, the height of one layer of WSe₂ [Fig. 3(c)]. Well-defined, mono and few-layer triangular WSe₂ domains with sharp (oxidation-free) edges are observed on the as-grown MBE WSe₂ surface [Fig. 3(d)]. After 15 min N₂ plasma exposure, the AFM image in Fig. 3(g) indicates a strong edge decoration on all WSe₂ domains suggesting that nitrogen preferentially reacts with the edges.

The contrast at the as-grown WSe₂ domain edges is comparable with that of the terraces suggesting the surface potential is similar across the probed region. However, N₂ plasma exposure for 30 min results in a sharp contrast between the terraces (dark) and domain edges (bright, difference in the electron density, and/or the height), which corresponds to significant change in surface potential in the vicinity of the domain edges [Figs. 3(e) and 3(h)]. Edge roughening observed in AFM images concomitant with the edge localized contrast variation detected with STM suggests that nitrogen preferentially nucleates at WSe₂ edges, which are known to be more reactive than the WSe₂ basal plane.²⁹ The contrast variation observed in STM images could relate to the metallic character of the W–N bond, and the tunneling current is higher in these areas compared with the adjacent semiconducting WSe₂. Figures 3(f) and 3(i) present STM images obtained from the MBE WSe₂ film, showing a high density of WSe₂ domains at HOPG step edges with small domain size as previously reported.¹⁷ Following 30 min remote N₂ plasma treatment, there is a noticeable degradation of the surface as evidenced by a large increase in surface roughness [Fig. S4(b) of the supplementary material]. Damage is observed across the entire surface of smaller MBE WSe₂ domains (Fig. S4 of the supplementary material). In contrast, large MBE WSe₂ domains appear to only be affected on the edges [Fig. 3(h)]. Both AFM and STM results suggest that nitridation occurs preferentially at the domain edges prior to affecting the WSe₂ terraces. Figure 4 displays the STS spectra obtained from an MBE WSe₂ monolayer prior to and following N₂ plasma treatment. A 1.83 eV bandgap is measured from as-grown monolayer WSe₂ with no significant changes induced by the N₂ plasma treatment (1.81 eV after 30 min exposure). The Fermi level moves toward the valence band maximum (VBM) by 0.3 eV in agreement with the Fermi level shift observed by XPS. This is further evidence that N₂ plasma exposure induces p-type doping in MBE WSe₂.

In summary, the effects of N₂ plasma exposure on the electronic and structural properties of MBE thin films and bulk WSe₂ are evaluated using XPS, Raman, PL, AFM, and STM. The appearance of an additional chemical state in the W 4f core level in conjunction with increased nitrogen concentration above the limit of XPS detection following N₂ plasma exposure indicates that nitrogen forms a covalent bond with tungsten. Shifts detected in characteristic WSe₂ vibrational modes and the PL peak after N₂ plasma exposure suggest that incorporating nitrogen into the WSe₂ lattice in the form of covalent W–N bonds induces strain in both MBE and bulk WSe₂. In addition, AFM and STM images indicate that N₂ plasma exposure damages the WSe₂ surface after longer exposure times eventually resulting in the formation of cracks in the case of bulk WSe₂, while the remote N₂ plasma process employed here is more destructive to few-layer MBE WSe₂ compared to bulk WSe₂. Tuning the duration of N₂ plasma exposure offers control over the amount of strain induced in the WSe₂ structure, both bulk and MBE, and can pave the ground for engineering the bandgap and electronic properties of WSe₂.³⁰ 

See supplementary material for the O 1s and C 1s core levels and Valence Band Maximum of MBE and exfoliated WSe₂ prior to load them into UHV system and after annealing, the sequential N₂ plasma exposure and gradual intensity increase of W–N peak is demonstrated, the angle resolved XPS in bulk sensitive angle from both samples is provided, XPS spectra from HOPG as a substrate after N₂ plasma exposure, and STM image from MBE WSe₂ after 30 min N₂ exposure is included.

This work is supported in part by the Center for Low Energy Systems Technology (LEAST), one of six centers supported by the STARnet phase of the Focus Center Research Program (FCRP), a Semiconductor Research Corporation program sponsored by MARCO and DARPA. Also, by the Southwest Academy on Nanoelectronics (SWAN) sponsored by the Nano-electronic Research Initiative and NIST and in part by NSF Award No. 1407765 under the US/Ireland UNITE collaboration. The authors acknowledge Dr. Lanxia Cheng for useful discussions on the Raman and PL study.