www.acsnano.org
Interplay Between Cr Dopants and Vacancy Clustering in the Structural and Optical Properties of WSe2 Ching-Hwa Ho,*,†,‡ Wei-Hao Chen,†,§ Kwong K. Tiong,§ Kuei-Yi Lee,‡ Alexandre Gloter,∥ Alberto Zobelli,∥ Odile Stephan,∥ and Luiz Henrique Galvaõ Tizei*,∥ Downloaded via UNIV OF WINNIPEG on June 27, 2018 at 21:03:43 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.
†
Graduate Institute of Applied Science and Technology, National Taiwan University of Science and Technology, Taipei 106, Taiwan Graduate Institute of Electro-Optical Engineering and Department of Electronic and Computer Engineering, National Taiwan University of Science and Technology, Taipei 106, Taiwan § Department of Electrical Engineering, National Taiwan Ocean University, Keelung 202, Taiwan ∥ Laboratoire de Physique des Solides, University of Paris-Sud, CNRS UMR 8502, F-91405 Orsay, France ‡
S Supporting Information *
ABSTRACT: Here, we analyze the effect of Cr doping on WSe2 crystals. The topology and the chemistry of the doped samples have been investigated by atom-resolved scanning transmission electron microscopy combined with electron energy loss spectroscopy. Cr (measured to have formal valence 3+) occupies W sites (formal valence 4+), indicating a possible hole doping. However, single or double Se vacancies cluster near Cr atoms, leading to an effective electron doping. These defects organization can be explained by the strong binding energy of the CrW−Vse complex obtained by density functional theory calculations. In highly Cr-doped samples, a local phase transition from the 2H to the to 1T phase is observed, which has been previously reported for other electron-doped transition-metal dichalcogenides. Cr-doped crystals suffer a compressive strain, resulting in an isotropic lattice contraction and an anisotropic optical bandgap energy shift (25 meV in-plane and 80 meV out-of-plane). KEYWORDS: transition-metal-dichalcogenide, defects, doping, EELS, STEM, optical properties ransition-metal dichalcogenides (TMDs) 1−3 with composition MX2 (M = Mo, W, Re, etc. and X = S, Se) have attracted substantial interest because of their mechanical, chemical, electrical, and optical properties. Specifically, they have been used as lubricants,4,5 as electrodes for photoelectrochemical cells,6,7 catalysts for sulfur-tolerant hydrogenation and hydrodesulfurization,8,9 field-effect transistors,10,11 flexible electronics,12 polarization-sensitive photodetectors,13,14 light emitters,15 and phototransistors.16 Singlelayer WSe217 has been shown to contain single photon sources. TMDs with the 2H-phase may be exfoliated into monolayers, which do not have inversion symmetry, enabling carriers valley control in the reciprocal space using light.18,19 The top of their valence band is constructed from metal d-orbitals, giving rise to strong spin−orbit coupling and spin-split valence bands.19
T
© 2017 American Chemical Society
Valley and spin control opens the possibility to valleytronic and spintronics in the same material.20,21 Production of monolayers is possible because of their layered structure loosely bound by van der Waals forces. This is evident from the easy cleavage perpendicularly to the c-direction, along which three atomic layers (X-M-X, where M is a transition metal and X is a chalcogenide) are stacked to form the crystal. Doping 2D TMD semiconductors leads to changes in structural, physical, and chemical properties. The presence of Nb-doped WSe2 at interfaces with WxNb1−xSe2 significantly reduces the potential barrier height and improves the Received: July 31, 2017 Accepted: October 31, 2017 Published: October 31, 2017 11162
DOI: 10.1021/acsnano.7b05426 ACS Nano 2017, 11, 11162−11168
Article
Cite This: ACS Nano 2017, 11, 11162-11168
Article
ACS Nano performance of the corresponding WSe2-based field-effecttransistor devices.22 Mn doping of MoS2 monolayers on sapphire leads to a redshift of ∼90 meV of the A-exciton emission.23 Re doping of WS224 and MoS225−27 results in a transition from a two-layer hexagonal (2H) phase to a threelayer rhombohedral (3R) structure and in a redshift of the A and B excitons. Density functional theory (DFT) calculations indicate that magnetism may be induced on MoS2 by doping with different transition metals.28 An experimental realization of magnetism in TMDs through doping would create a different class of diluted magnetic semiconductors, with applications in spintronic devices.29 In this work, WSe2 single crystals incorporated with various Cr amounts were grown by chemical vapor transport (CVT)30,31 using I2 as transport agent (see Methods section). Millimeter-wide monocrystals (optical images in Figure SI1 in the Supporting Information, SI) have been characterized using an ensemble of macroscopic and microscopic techniques (Methods) to verify their composition, crystal structure, Cr doping sites, and optical band gap. We have observed that substitutional Cr atoms occur at the metal W site, which are systematically accompanied by a single or a double Se vacancy. Even though Cr atoms are in the formal 3+ state, the combination with Se vacancies leads to effective electron doping. Local structural changes between the 2H and possibly the 1T phases have been observed in Cr-rich regions, an effect attributed to electron doping.
Figure 1. (a) STEM HAADF image of a monolayer region in pure 2H-WSe2. The circles point to a single (orange) and a double (purple) Se vacancy, which are shown in detail in (b) and (c). In (b) two single Se vacancies are seen. (d) Comparison between experimental profiles of a Se-vacancy (orange), a double Se Vacancy (purple), and of the same defects in simulated images (dashed lines). Simulated images are shown in Figure SI4 of the SI.
RESULTS As-produced WSe2 crystals show flat surfaces with step edges, as evidenced by scanning electron microscopy (SEM) images (Figure SI2 in the SI). Five samples with nominal Cr content x = 0.00, 0.01, 0.05, 0.10, and 0.20 (x the atomic ratio in the initial mix for crystal production, W1−xCrxSe2) were produced. Qualitative energy dispersive X-ray spectroscopy (EDS) in a SEM indicates that the actual Cr concentration in WSe2 crystals is much lower than the nominal content. For the x = 0.20 sample, a concentration of about 2% was determined, with local fluctuations at scales of tens of micrometers. The presence of Cr in the target material (semiconductor 2H-WSe2) was confirmed by the observation of individual dopant atoms using electron energy loss spectroscopy (EELS) in a scanning transmission electron microscope (STEM) operated at 60 kV. WSe2 bulk crystals are stable in the standard hexagonal 2H phase, as confirmed by X-ray diffraction (XRD) experiments (Figure SI3 in the SI). Moreover, the same structure is observed in atomically resolved high-angle annular dark-field (HAADF) STEM images of exfoliated monolayers (Figure 1a). In a first approximation, the HAADF image intensity is proportional to Z1.6 (see for example ref 32). Therefore, in a monolayer (Figure 1a), W atoms (Z = 74) appear with higher intensity than double Se columns (Z = 34). In pure crystals the most common defects observed are single and double Se vacancies, which are shown in Figure 1b,c. Simulated and experimental profiles are compared in Figure 1d, and simulated images are shown in Figure SI4 of the SI. These defects may be native or created by electron beam irradiation. Single W vacancies were not observed, while more complicated defects, such as voids, can be created by electron irradiation. This behavior is compatible with previously calculated sputtering thresholds for the acceleration voltage (60 kV) at which experiments were performed.33
WSe2 containing Cr also crystallizes with the 2H-WSe2 phase, as confirmed by XRD (SI) and HAADF images (Figure 2a). In these images, darker regions (roughly the size of a single atomic site) are observed in the metal sublattice. As a first guess they could be associated with W vacancies or more complex defects. However, closer inspection at higher magnification shows that these features correspond to W sites occupied by lighter atoms. Spatially resolved EELS permits their identification as individual Cr atoms. The spectrum of a single substitutional Cr atom at a W site (the CrW defect) is shown in Figure SI5 in the SI, along with the accompanying spectrum image. The data are shown for a bilayer of WSe2 due to its longer stability under electron beam illumination. EELS data have been analyzed using the HyperSpy Python library.34 In addition to single CrW defects, we have observed more complex defective structures. In Figure 2a a distinct cross-like intensity distribution (marked by red squares) can be seen. Closer inspection indicates that these features are substitutional-vacancy complexes with a Cr substitutional atom at the metal site and a single or double Se vacancy at a neighboring chalcogen site (Figure 2b). In fact, the proximity between CrW and a VSe defects is quite recurrent. In Figure 2c a different configuration is shown, where the Se vacancy is located at the third neighbor site in respect to the Cr atom (trans position, atomic structure shown in Figure 3). This attribution is confirmed by comparing profiles without defects (blue), containing a single Se vacancy (green), containing a single substitutional Cr atom (red), and a simulated one containing these two defects (dash line). The Cr (Z = 24) atom can be distinguished from double-Se (2.Z = 68), single-Se (Z = 34), and single-W (Z = 74) columns. The positions of the Cr atom and the single Se vacancy are pointed by dashed red and green 11163
DOI: 10.1021/acsnano.7b05426 ACS Nano 2017, 11, 11162−11168
Article
ACS Nano
the Se chemical potential has been derived from the Se α phase, and chemical potentials for Cr and W have been derived using the following expressions: μW = E(WSe2)−2·μSe and μCr = [E(Cr2Se3)−3·μSe]/2. In Table 1 we report formation energies Table 1. DFT-Derived Formation Energies for a Cr Substitutional Atom, a Selenium Vacancy, and Various Substitutional-Vacancy Complexes in a WSe2 Monolayera EF (eV) CrW VSe CrW−VSe1st CrW − VSe3rd‑cis CrWVSe3rd‑trans CrWVSe5th‑1 CrWVSe5th‑2 a
Figure 2. (a) Low-magnification HAADF image of the x = 0.10 sample. Darker regions are associated with Cr dopants. Red squares mark positions where a Cr dopant is next to a single or double Sevacancy. There are three possible orientations for this defect, of which two are observed in (a). (b) High-magnification HAADF image of a Cr/Se vacancy complex. (c) HAADF image of another region of the sample, showing a single Cr atom with a Se-vacancy as a third nearest neighbor (trans site). The positions of the Cr atom and vacancy in (c) can be confirmed in the profiles shown in (d) taken along the colored dashed arrows in (c). The positions of the Cr atom and the Se vacancy are pointed by dashed red and green circles, respectively, in (c) and (d). The intensities for CrW and VSe are compared to a simulated profile (dashed black line in (d)). The simulated image for this profile is shown in Figure SI4 of the SI.
0.55 3.49 3.51 3.48 3.46 3.49 3.50
The atomic configurations used are shown in Figure 3.
for a chromium-substitutional defect, an individual selenium vacancy, and various substitutional-vacancy complexes for which the position of the Se vacancy with respect to the Cr atom is indicated in Figure 3. We have observed a strong binding energy for all considered substitutional-vacancy complexes, on the order of ∼0.5 eV. This tendency corresponds to a lowering of formation energies for VSe close to Cr atoms, explaining the increased abundance of native vacancies for the most Cr doped samples. In view of these observations, the effective behavior of CrW atoms in the WSe2 lattice will depend on their valence, which can be formally 2+, 3+ and 6+. XPS measurements (Figure SI3 in the SI) indicate the presence of Cr3+and Cr6+. In spatially resolved EELS the energy position of the L2,3 edge allows the determination of the CrW valence state.36 For these experiments, we have used Cr2O3 (Cr3+) and K2CrO4 (Cr6+) as references. In all of the measurements from exfoliated crystals the Cr L3 peak (575.5 eV, Figure 4) is consistently observed at lower energy than those of Cr2O3 (577.0 eV) and K2CrO4 (580.0 eV), but higher than those expected for CrSe (574 eV for Cr2+).36 Therefore, the Cr atoms in W sites have a 3+ formal valence. The L3/L2 intensity ratio is also a possible indication of the valence state, although more sensitive to data noise.36 The XPS result corroborates our valence state
Figure 3. Atomic structure configuration for the Cr w −V Se complexes. Experimentally, the complexes with VSe as a first neighbor and a third neighbor on the trans configuration were observed.
circles in Figure 2c,d. We have observed no experimental evidence pointing to the existence of CrSe defects in our materials. A spatial correlation between substitutional atoms and chalcogen vacancies has been recently observed in Mo-doped WS2,35 with Mo having the same valence as W atoms (W4+). This clustering of S-vacancies around dopants leads to a possible electron doping (as the MoW-VS complex adds electrons to the conduction band). To obtain further insights on defect clustering, we have computed the formation energy of various substitutionalvacancy complexes in the framework of the DFT. Formation energies have been calculated using chemical potentials corresponding to selenium-rich conditions, a choice justified by the Se excess introduced at the synthesis. According to it,
Figure 4. EEL spectra of the x = 0.10 and 0.20 samples along with those of Cr2O3 and K2CrO4 measured using the same experimental conditions. The L3 peaks for Cr in WSe2 samples match the energy position for Cr in the 3+ formal valence state. Spectra have been measured for a large number of Cr atoms by acquiring data, while the electron beam was scanning a ∼100 nm2-wide area. 11164
DOI: 10.1021/acsnano.7b05426 ACS Nano 2017, 11, 11162−11168
Article
ACS Nano assignment. The lower L2,3 energy compared to Cr3+ in Cr2O3 occurs due to the lower electronegativity of Se (ligand) compared to O. In CrCl3, Cr is also 3+, even though the L3 peak appears at 575 eV.36 Our Cr2O3 and K2CrO4 L3 peak energy measurements are about 1 eV higher than those previously reported in the literature.36 The 3+ formal valence observed would indicate that the Cr substitution at W site leaves extra holes in WSe2, as the W atoms have a 4+ formal valence. However, the systematical observation of single or double Se-vacancies next to these atoms implies an overall electron surplus, a behavior similar to that observed in Mo-doped WS2.35 In fact, n-doping may be induced by native chalcogen vacancies in pure TMD materials.37 Cr6+ atoms, only observed in XPS measurements, are probably located in a secondary phase observed in XRD measurements but not in electron microscopy experiments. As the Cr concentration increases, the crystals are progressively more resistant to mechanical exfoliation. In fact, production of wide monolayers of the x = 0.2 crystals was not achieved. Images from Cr-rich regions are consistent with the 1T-WSe2 phase (Figure 5a,b), which can be clearly distinguished from
quantified directly from HAADF images of thick specimens, one can infer that the concentration is below the nominal 20% for the x = 0.20 sample, as indicated by EDS. The 1T phase is a minor fraction of the whole volume even in the x = 0.20 sample, which explains why it was not observed in the XRD data. The 2H-1T phase transition can be understood in view of our structural and electronic results for the Cr dopants: The Cr3+ coupled with Se vacancies indicate an electron doping, which is known to promote this phase transition in TMDs.39−41 Electron doping through Li intercalation has also been shown to induce a 2H to 1T transition in MoS2.42 We should note that the stable phase of CrSe2 is 1T, and therefore apart from purely electronic considerations, a steric effect could contribute in the stabilization of Cr doped 1T-WSe2 domains. The role of multiple and concomitant effects in the 2H to 1T transition has been extensively discussed in ref 43. Along with this microscopic structural change, Cr incorporation influences the global crystal structure. XRD (Figure SI3 in the SI) shows a lattice contraction as a function of Cr nominal content. The decrease is approximately linear with Cr content (Figure 6a). The lattice parameters decrease from a =
Figure 6. (a) Lattice parameters as a function of Cr content, showing an anisotropic contraction. (b) Energy of transition features observed in reflectance spectra. The out-of-plane transition, C, suffers a much larger shift, indicating a larger optical gap increase.
Figure 5. (a) STEM HAADF image of a x = 0.20 crystal showing a higher concentration of Cr sites and a different crystal structure, which is shown in detail in (b). (c) HAADF image simulations showing that the contrast observed is consistent with the 1T phase of WSe2. However, a similar intensity distribution is a match to the 3R phase.
3.297 Å, c = 13.024 Å for x = 0.00 to a = 3.295 Å, c = 13.020 Å for x = 0.20. This is consistent with the presence of 1T phase domains which have a lattice parameter compression of about 0.6% with respect to the 2H phase as demonstrated by DFT calculations. DFT calculations reported in the literature44,45 also indicate that for small compressive strain (
