Rapid and Low-Temperature Molecular Precursor Approach toward Ternary Layered Metal Chalcogenides and Oxides: Mo1–xWxS2 and Mo1–xWxO3 Alloys (0 ≤ x ≤ 1)

Metal sulfide and metal oxide alloys of the form Mo1–xWxS2 and Mo1–xWxO3 (0 ≤ x ≤ 1) are synthesized with varying nominal stoichiometries (x = 0, 0.25, 0.50, 0.75, and 1.0) by thermolysis of the molecular precursors MoL4 and WS(S2)L2 (where L = S2CNEt2) in tandem and in various ratios. Either transition-metal dichalcogenides or transition-metal oxides can be produced from the same pair of precursors by the choice of reaction conditions; metal sulfide alloys of the form Mo1–xWxS2 are produced in an argon atmosphere, while the corresponding metal oxide alloys Mo1–xWxO3 are produced in air, both under atmospheric pressure at 450 °C and for only 1 h. Changes in Raman spectra and in powder X-ray diffraction patterns are observed across the series of alloys, which confirm that alloying is successful in the bulk materials. For the oxide materials, we show that the relatively complicated diffraction patterns are a result of differences in the tilt angle of MO6 octahedra within three closely related unit cell types. Alloying of Mo and W in the products is characterized at the microscale and nanoscale by scanning electron microscopy–energy-dispersive X-ray spectroscopy (EDX) and scanning transmission electron microscopy–EDX spectroscopy, respectively.


■ INTRODUCTION
Inorganic layered materials have attracted the attention of researchers in the last decade. Transition-metal dichalcogenides (TMDs) are of particular interest and importance because of their intrinsic 2D layered structures that are formed by bonding two chalcogen planes with a transition-metal plane. The resulting materials can be semiconducting or metallic depending on the coordination geometry of the metal centers by the chalcogen atoms in the crystal structure. For example, 2H-MoS 2 (trigonal prismatic metal coordination) is a semiconductor, while 1T-MoS 2 (octahedral metal coordination) is metallic. Similar to graphite, the weak van der Waals interaction between planes allows for their exfoliation as atomic layers and introduces thickness-dependent emergent properties in the 2D limit. The group VI metal chalcogenides of the type MX 2 (MoS 2 , WS 2 , MoSe 2 , and WSe 2 ) in particular can be tuned from indirect band gap (Γ to Q point) to direct band gap at the K point of the Brillouin zone with increases in band gap energy due to perpendicular confinement. For example, when MoS 2 is thinned down to a monolayer from the bulk, the excitonic band gap energy increases from approximately 0.85 to 1.8 eV and the quantum yield for the conduction band photoluminescence emission is dramatically increased. 1−5 Hence, MoS 2 and WS 2 have been extensively studied, 6−17 with applications including novel lubricants, 18 45,46 Novel properties can be imparted to transition-metal dichalcogenides via adding new atoms or introducing different stacking modes into their host crystal structures. For example, main group (B, P, and Cl)-doped MoS 2 is a suitable candidate for gas detection and semiconductors, 47,48 whereas transitionmetal-doped MoS 2 (Fe, Co, Ni, or Cu dopants) increases the rate of the hydrogen evolution reaction. 49,50 Introduction of paramagnetic ions (Mn, Fe, Co) can potentially bestow dilute magnetic semiconducting properties amenable to spintronic applications. 51−53 Heterostructures of MoS 2 and TiO 2 exhibit enhanced photocatalytic capability, 54 while MoS 2 and WS 2 heterostructures may possess ultrafast charge-transfer characteristics. 6 MoS 2 , hexagonal boron nitride, and graphene heterostructures can also be used as high-performance fieldeffect transistors. 55 Our group has recently demonstrated successful doping of W, 56 Cr, 57 and Re 58 into MoS 2 via aerosol-assisted chemical vapor deposition (AACVD), which can be used to produce thin films of these interesting materials.
Transition-metal oxides, such as MoO 3 and WO 3 , are wide band gap semiconductors (3.0−3.2 eV), which have also been widely researched because of their electronic, 59−62 gas sensing, 63−65 photocatalytic, 66−70 electrochromic, 71−74 and photochromic properties. 74−77 Various alloys have been studied in order to produce gas sensors: MoO 3 and WO 3 alloy thin-films produced by sol−gel methods can be used for O 2 gas sensing; 78 1 wt % MoO 3 was doped into the WO 3 structure for NH 3 and NO gas detection; 79 the mixture of MoO 3 (5 wt %), Au, and WO 3 (94.2 wt %) was reported to achieve a selective detection of NH 3 , while the sensitivity toward NO is eliminated. 80 WO 3 /MoO 3 films (7 at. % Mo) produced by atmospheric chemical vapor deposition (CVD) have been used as electrochromic devices. 81 However, inhomogeneous films were reported as a challenge encountered during production. 82 Traditionally, bulk transition-metal chalcogenides and oxides have been accessed either by physical vapor deposition or from solid-state reactions, respectively. Although these processes are effective at producing binary metal chalcogenides, they require high temperatures and often have extended reaction times because of the need for diffusion of atoms through solid-state crystalline host matrices. The production of ternary materials from these routes can therefore also be problematic because of unequal solid-state diffusion rates for different atoms, leading to the formation of subphases of binary materials in the final products. For oxides, this has been addressed by the introduction of sol−gel routes based on the hydrolysis of metal alkoxide precursors, while for metal chalcogenides, a range of molecular precursors have been developed, e.g., metal dithiocarbamate and metal xanthate complexes, that decompose under thermal stress to the corresponding metal sulfides. Because the mixing of precursors occurs at molecular scales and metal sulfur bonds are preformed within the complexes, the production of ternary and even quaternary materials can be achieved with excellent control of dopant homogeneity and often crystalline phase purity, and the approach can be extended to a wide range of metal chalcogenide materials. 83 We have recently been interested in exploring the use of single-source precursors for the production of group VI B TMDCs. We previously reported a solventless thermolysis approach for the production of MoS 2 from the decomposition of MoL 4 in an inert atmosphere. Interestingly, in the same experiments, we also observed that MoL 4 was decomposed under identical conditions, but in air, the corresponding metal oxide, that is, α-MoO 3 , was produced, leading us to the conclusion that alkyl dithiocarbamato molybdenum(IV) complexes could act as "masked" precursors toward oxides as well as chalcogenide materials with low processing temperatures (<500°C) and comparatively short reaction times (∼1 h). 84 We have also shown, subsequently, that MoL 4 can be chemically decomposed to produce nanocrystalline MoS 2 at room temperature at the liquid−liquid interface. 85 We therefore hypothesized that through a dual precursor reaction of MoL 4 and a structurally similar tungsten precursor, WS(S 2 )L 2 , we could also potentially access a range of layered ternary metal chalcogenide alloys of the form Mo 1−x W x S 2 as well as produce the family of Mo 1−x W x O 3 (0 ≤ x ≤ 1) alloys at relatively low temperatures (Scheme 1). In this paper, we show that this is indeed possible and characterize the alloyed materials using diffraction data combined with electron microscopy, Raman spectroscopy, and X-ray photoelectron spectroscopy (XPS). This approach represents a step change in producing ternary transition-metal dichalcogenide alloys at a fraction of the time compared to traditional solid-state reactions and with control of alloying achieved at the nanoscale.
■ EXPERIMENTAL SECTION General Considerations. All materials were purchased from Sigma, Merck, or Fisher and used without purification.
Synthesis of Tetrakis(diethyldithiocarbamato)molybdenum(IV) (MoL 4 ). The title complex was synthesized following the method reported previously by Decoster et al. 88 Molybdenum hexacarbonyl (1.0 g, 3.8 mmol) and tetraethylthiuram disulfide (2.2 g, 7.6 mmol) were mixed in degassed acetone (30 mL), and the mixture was heated at 80°C under reflux for 2 h. The reaction was cooled to room temperature and a black microcrystalline precipitate formed, which was collected by vacuum filtration. The solid collected was washed three times with pentane (3 × 30 mL).  . The title complex was synthesized using the method reported by Lewis et al. 56 Ammonium tetrathiomolybdate (2.9 g, 8.4 mmol) and sodium diethyldithiocarbamate (7.6 g, 34 mmol) were mixed in deionized water (300 mL) and diluted HCl (2 M) was added dropwise until the pH of the solution reached ca. ∼2. The dark green precipitate which formed was collected by vacuum filtration and was washed three times with deionized water (3 × 100 mL). The crude product was dried in a vacuum oven for 1 h and dissolved in 250 mL of acetone, followed by vacuum filtration to remove solid impurities. The dark green solution was collected, and the solvent was removed by a rotary evaporator before being dried overnight in a vacuum oven. Anal in a Carbolite Gero tube furnace under an argon flow (flow rate 300 sccm) for 1 h and left to cool down to room temperature before the powder products were collected. For oxides, the mixture was heated to 500°C in air for 1 h, followed by cooling to room temperature, and the solid products were then collected.
Synthesis of Nanoscale Mo 1−x W x S 2 and Mo 1−x W x O 3 . Mo 1−x W x S 2 (x = 0, 0.25, 0.5, 0.75, and 1) powders were dissolved in ethanol/water (50/50) and sonicated for 20 min using an ultrasonic bath (30 W/37 Hz), followed by centrifugation at 1500 rpm for 45 min to remove bulk materials. The supernatants were resuspended in deionized water and were further centrifuged at 12,000 rpm for 45 min. The precipitates after 12,000 rpm were collected and diluted in deionized water before being drop-cast onto SiO 2 /Si substrates or transmission electron microscopy (TEM) grids for further characterization.
Material Characterization. Raman spectroscopy was performed in HORIBA LabRAM Evolution HR, with a 488 nm excitation wavelength laser (25% ND filter, acquisition time 120 s, accumulation 2). The silicon peak was calibrated to 520 cm −1 as a reference. Powder X-ray diffraction (PXRD) measurements were conducted on a PANalytical X'Pert Pro, which has a θ/θ geometry with a copper line focus X-ray tube (Kα = 1.5406 Å). Zero background sample holders were used for all the samples to reduce noise. XPS was performed using a Thermo Scientific Kratos Axis Ultra Hybrid system equipped with a monochromatic Al Kα anode (150 W) equipped with a charge neutralizer. CasaXPS software was used for data analysis and curve fitting, with all peaks charge-corrected via fitting the C 1s peak to a standard (284.8 eV), whose binding energy relates to the surface adventitious carbon. To analyze the oxidation state and surface chemical composition of the compounds, a library (Kratos) was employed for relative sensitivity factors and the peaks were fitted with 70% Gaussian and 30% Lorentzian. Scanning electron microscopy (SEM) and energy-dispersive X-ray spectroscopy (EDX) were run on a Zeiss Sigma VP FEG-SEM (SE) with accelerating voltages of 1−1.5 and 30 kV, respectively. High-resolution TEM (HRTEM) and scanning TEM (STEM) with EDX were performed using either a Talos F200X (200 kV, FEG) microscope equipped with a Super-X EDS detector and fitted with a fast-acquisition 4k CMOS camera or an FEI Tecnai F30 (300 kV, FEG) fitted with an Oxford Instruments X-Max 80 silicon drift detector (SDD) EDX detector and T20 (200 kV, LaB 6) equipped with an Oxford Instruments X-Max N 80TLE SDD or a probe side aberration-corrected FEI Titan G2 ChemiSTEM (200 kV, FEG) with a beam current of ∼90 pA. EDX spectra and spectrum images were processed using OI Aztec or Bruker ESPRIT software. Atomic force microscopy (AFM) was performed using a Bruker MultiMode 8 microscope in ScanAsyst mode for automatic image optimization. SiO 2 /Si substrates were used (lateral size 8 × 8 mm). Elemental analysis was conducted by the microanalytical laboratory at the University of Manchester using a Thermo Scientific Flash 2000 organic elemental analyzer.

■ RESULTS AND DISCUSSION
We reasoned that by using mixtures of group VI B metal dithiocarbamate complexes in tandem, we could produce either metal sulfide or metal oxide materials of the form Mo 1−x W x S 2 or Mo 1−x W x O 3 where ratios of the precursors were chosen so that nominally x = 0, 0.25, 0.50, 0.75, or 1.0. The molecular precursors MoL 4 and WS(S 2 )L 2 (Scheme 1) have been previously used to produce Mo 1−x W x S 2 by AACVD, and for this study, these precursors were produced via the methods reported by Lewis et al. 56 These precursors have similar thermal profiles and codecompose in the solid state to give alloyed materials, which was previously shown by Lewis et al. to produce the target metal chalcogenide alloys in AACVD experiments to produce thin films. 56 Here, a mixture of both precursors in the desired mole fraction of W (i.e., x) was heated at a relatively low temperature (450 or 500°C) in a tube furnace for 1 h under either an argon flow atmosphere (for sulfides) or in air (for oxides). 84 The metal chalcogenide and oxide alloys produced as crystalline powders from these short (1 h) thermolysis reactions were studied using a range of characterization techniques.
Structural Analysis of Mo 1−x W x S 2 and Mo 1−x W x O 3 Alloys. PXRD was performed to structurally characterize the materials. The PXRD patterns of pristine MoS 2 , WS 2 , and their sulfide and oxide alloys are shown in Figure 1. We first consider the materials produced in an argon atmosphere. The almost identical lattice constants of MoS 2 and WS 2 result in an overlapping of diffraction peaks. The diffraction peak at 2θ ∼ 33.5°can be ascribed to the (100) and (101) planes and 2θ ∼ 59°corresponds to the (110) plane of 2H-MoS 2 (JCPDF no. 37-1492) and 2H-WS 2 (JCPDS card no. . For the alloys, Mo 1−x W x S 2 , no significant change is observed in these patterns upon alloying except that a minor peak becomes noticeable at about 70°as the level of W increases to 75%, which corresponds to the (200) plane. For pure WS 2 , a small peak appears at 2θ ∼ 23.5°, which might be caused by oxidation. Almost no diffraction from the (002) planes (2θ ∼ 15°) is observed, which is an indication of the ultrathin nature of the alloyed materials formed. Indeed, the pattern we observe with these systematic absences of the usually intense (00l) reflections is remarkably similar to that of single-layer MoS 2 . 89 In addition, the peaks are relatively broad, which suggests that the materials are also nanoscale in their ab plane, and this is a feature that is observed for every Mo−W sulfide alloy investigated.
For materials derived from the decomposition of WS(S 2 )L 2 in air at 500°C, only the diffraction pattern for monoclinic WO 3 is observed (JCPDF card no. 43-1035). The crystal structure of tungsten trioxide is temperature-dependent, adopting a tetragonal structure at temperatures above 740°C , orthorhombic from 330 to 740°C, monoclinic from 17 to 330°C, triclinic from −50 to 17°C, and monoclinic again at temperatures below −50°C. Therefore, in this case, we conclude that the most thermodynamically stable structure at room temperature is adopted in the product of decomposition, 90,91 although we do note that triclinic WO 3 is the most thermodynamically stable polymorph overall. 92 Similarly, for materials derived from the decomposition of MoS(S 2 )L 2 in air at 500°C, only the diffraction pattern for orthorhombic MoO 3 is observed (JCPDF card no. 05-0508).
In contrast to the WO 3 and MoO 3 end compounds, which are single phases, the Mo 1−x W x O 3 alloys produced adopt a range of interrelated crystalline structures. With 25 and 50% Mo doping in WO 3 , we observed the emergence of new peaks for the (120) and (112) Bragg planes, with peaks shifted from 24.1 to 25.2°(marked "*" in Figure 1b). This is caused by another phase of WO 3 and it is known that mixtures of phases can potentially be formed during alloying of this system, which we explore later. 92−97 A summary of the shifts in the 2θ peak position of individual Bragg planes as a function of the Mo mole fraction in Mo 1−x W x O 3 alloys using the Miller indices previously designated by Stevenson and co-workers 98 is presented in Figure S6 (Supporting Information). The vanishing (020) reflection among the cluster of three peaks around 23°is a result of Mo doping into WO 3 ; 99 this is due to preferred orientation in the (00l) planes with respect to the (0k0) planes. 100,101 Hence, for this quite complicated metal oxide system, we employed Rietveld refinement of the diffraction data to identify and quantify the constituent crystalline phases. Refinement of the diffraction data for Mo 1−x W x O 3 (0.25 ≤ x ≤ 0.75, Supporting Information) reveals that in all cases the alloys comprise three highly related WO 3 structures: orthorhombic, monoclinic, and triclinic WO 3 , which in all cases is accompanied by a varying amount of α-MoO 3 . The variations in the crystal structure away from the parent monoclinic WO 3 structure are strain-induced; as the amount of Mo is reduced, the amount of monoclinic structure is increased. Hence, from this analysis, we may conclude that the system is therefore best described as a paracrystal with varying short to medium local order in the WO 3 phase, leading to highly related crystal systems that can be envisaged as being structurally related by varying the tilt in MO 6 octahedra (M = Mo or W), where the local variation in the M−O tilt angle and thus crystal structure adopted is governed by the amount of Mo dopant (Figure 2). 93,102,103 A similar phenomenon has been noted by Morandi et al. for Mo 1−x W x O 3 produced from a sol−gel type pathway with diffraction patterns of the solid state materials analyzed by Rietveld refinement. 104 The transformation between phases requires relatively low energy because of their structural similarity; for example, conversion of monoclinic WO 3 to the triclinic phase has previously been reported by grinding at room temperature. 92,97 Other phase changes that are elicited upon doping in crystalline WO 3 have been reported by a number of groups. 93,98,105,106 It is also worthwhile noting that the phase of the Mo 1−x W x O 3 produced is often dependent on the synthetic pathway employed. Rao   Chemistry of Materials pubs.acs.org/cm Article produced. 107 Ghiotti and co-workers produced five stoichiometry-dependent phases from a sol−gel processing route to Mo 1−x W x O 3 where the phases observed seem to be dependent on the stoichiometry of the material produced, with one stoichiometry (x = 0.2) showing predominantly mixed phase products. 104 Hibble and Dickens reported the synthesis of Mo 1−x W x O 3 with various stoichiometries from three various routes (sol−gel, solid-state reaction, and co-crystallization); yet, only single phase products were produced, which adopted the monoclinic and the ζ structures, 108 the latter described by Salje et al. 93 Similarly, Figlarz et al. produced Mo 1−x W x O 3 alloys with the hexagonal (x < 0.2), orthorhombic (0.2 < x < 0.6), and cubic ReO 3 (0.6 < x < 1.0) structures. 109 Stevenson and co-workers have noted that the initial stoichiometry of the material may have a major influence over which phases are formed as final products after sintering is performed. 98 In the context of the solid-state reactions described in this paper, it seems that because we are rapidly producing the alloys that we trap out most of the possible phases as metastable products as in all cases for the Mo 1−x W x O 3 alloys we see a range of interrelated phases produced. Raman analysis provides complementary structural information to diffraction methods. Raman spectra of MoS 2 and WS 2 (i.e., x = 0, x = 1) have two signature optical phonon modes: the in-plane optical mode (E 2g 1 ) and the out-of-plane optical mode (A 1g ) (Figure 3a). Accordingly, intense peaks are observed for the MoS 2 (x = 0) sample at 381 and at 403 cm −1 , which we attribute to the E 2g 1 in-plane and the A 1g out-ofplane vibrational mode, respectively. The difference between both peaks has previously been reported to be indicative of layer number (Δν = 18 cm −1 for monolayer, Δν = 21 cm −1 for bilayer molybdenum disulfide). 110−112 For the MoS 2 sample, the E 2g 1 peak appears at 382 cm −1 and the A 1g mode is centered at 404 cm −1 (Δν = 22 cm −1 ), with the peak corresponding to the silicon substrate located at 520 cm −1 . 113 For Mo 0.75 W 0.25 S 2 , both the E 2g 1 and A 1g phonons appear at positions similar to MoS 2 (381 and 404 cm −1 ). However, when the amount of tungsten in the sample is increased (mol % Mo is reduced) in Mo 0.5 W 0.5 S 2 , the E 2g 1 shifts from ∼381 to 378 cm −1 and the A 1g remains at 403 cm −1 (Figure 3c). By further increasing the tungsten content in the Mo 0.25 W 0.75 S 2 , the E 2g 1 phonon softens and is further shifted to 375 cm −1 , while the A 1g phonon is blue-shifted (403−408 cm −1 ) and a new peak appears at 350 cm −1 , which is characteristic of WS 2.112 . For the WS 2 sample (i.e., x = 1.0), the Raman spectrum includes the longitudinal acoustic mode [LA(M)] located at 171 cm −1 , with the E 2g 1 and A 1g phonon modes observed at 353 and 414 cm −1 , respectively ( Figure 2a). The difference in the peak maxima wavenumbers (Δν = 61 cm −1 ) indicates that WS 2 could potentially be on average somewhere between a bilayer and a few layers in thickness (Δν = 60 cm −1 for monolayer and 65 cm −1 for bulk). 114−117 This is also evidenced by the appearance of the 2LA(M) mode, which is reported to monotonically increase in intensity as a function of decreasing layer number. 114 The Raman spectra of Mo 1−x W x O 3 samples also give insight into their structure and bonding. All spectra recorded have two clear peaks that correspond to the vibrational modes of WO 3 with some noticeable minor peaks below 100 cm −1 , which also appear in pristine WO 3 (Figure 3b). For undoped WO 3 , two sharp peaks are located at 354 and 416 cm −1 , indicating that the WO 3 is highly crystalline. 118,119 Additionally, two broad peaks located at 704 and 804 cm −1 can both be assigned to O− W−O stretching vibration modes, 120,121 with the latter arising from doubly bound corner shared octahedra in the structure. 98 These modes are the strongest signals among all the phonon modes observed and are upshifted (713 → 718 and 810 → 819 cm −1 ) upon increasing at. % of Mo in the alloys. The increase of peak intensities of the O−W−O stretching mode relative to the other peaks is concomitant with the increase of Mo dopant atoms in the WO 3 structure. This has previously been ascribed   (Figure 3d) may be caused by the phase transition of WO 3 . A higher wavenumber of this Raman stretching mode could indicate a more distorted structure; similar shifts associated with phase transitions from orthorhombic (Pnma) to monoclinic (P21/n) structures in WO 3 have been reported. 105 Hence, the various phases that were determined from refinement of X-ray diffraction (XRD) data (vide supra) in this system could be the main reason for these observed peaks' shifts. 93,105,123 This is also supported by the new peak that appears at 319 cm −1 in the Mo 0.5 W 0.5 O 3 alloy, which indicates that the lattice incorporation has affected this molecular vibration mode. For MoO 3 , the vibrational modes for Mo chains are located at 112, 125, and 154 cm −1 , and the oxygen wagging modes are observed as peaks at 281 and 288 cm −1 (B 2g , B 3g ). 124 The symmetric and asymmetric stretching vibrational phonon modes located at 664 cm −1 (B 2g , B 3g ), 816 cm −1 (A g , B 1g ), and 993 cm −1 (A g , B 1g ) are characteristic peaks of α-MoO 3 . 124−126 Overall, the Raman results indicate that alloying has been achieved with a predominant crystalline structure of MoS 2 in the sulfide alloys, while the WO 3 structure is primarily adopted in the oxide alloys within the reported concentration range. Elemental Analysis of Alloying in Mo 1−x W x S 2 and Mo 1−x W x O 3 in the Bulk and at the Surface. XPS was performed to measure elements at the surface upon doping. High-resolution measurements of Mo 3d and W 4d peaks were run in one scan in order to minimize charging effects. Atomic concentrations derived from XPS were plotted as a function of the nominal concentrations for Mo and W individually ( Figure  4e). The ratios of Mo 4+ or W 4+ to S 2s are very close to 1:2 with the full width at half-maximum (fwhm) range ∼1.3−1.5 (±0.2) eV. The high-resolution scan for sulfides in Figure 4a shows that the fitted Mo 6+ (surface oxidation) binding energies appear at 233.  Chemistry of Materials pubs.acs.org/cm Article content increases to 75%, a peak for the Mo 6+ binding energy disappears along with a noticeable decrease of peak intensity for Mo 4+ . The binding energy for Mo 4+ 3d 5/2 also shifts from 230.7 to 230.4 eV. This effect is potentially caused by the reduced electron attraction provided by S and increased electron attraction by W due to the increase in electronegativity and orbital radii with the increase of W dopant. 127−129 Both the Mo 4+ and S 2s peaks present a noticeable intensity difference arising from the W substitution, as the fwhm for Mo 4+ changes from 1.1 to 0.88 to 1.3 eV and fwhm for S 2s changes from 3.11 to 2.57 and to 2.78 eV. The sharpening of both peaks for Mo 0.5 W 0.5 S 2 indicates a higher binding energy for this composition. Figure 4b presents the binding energies for the tungsten doublet peak W 4d 5/2 changing from 247.5 to 246.7 and to 245.7 eV, which could be assigned to W 4+ ions in WS 2 . 49 The binding energies for S 2p are located at about 163.5 eV for all the sulfide alloys. The majority of oxide alloys show only one clear Mo oxidation state; the observed peaks for Mo 3d 5/2 located at 233.0−233.2 eV can be attributed to Mo 6+ in MoO 3 ( Figure  4c). Only minor peaks for Mo 5+ are shown in the Mo 0.5 W 0.5 O 3 alloy; the insignificant intensities indicate a relatively pure oxide state for samples in all the ratios. The binding energies are in good agreement with that in the literature. 130,131 While the W 4d 5/2 and W 4d 3/2 peaks are at around 248 and 260 eV, respectively (Figure 4d), a peak difference of ∼12 eV was observed which we attribute to the spin−orbit splitting, suggesting the existence of W 6+ (WO 3 ). 132 The nominal at. % of the Mo atom is compared with elemental analysis by XPS and SEM−EDX (Figure 4e   Chemistry of Materials pubs.acs.org/cm Article SEM−EDX spectroscopy is due to the depths profiled; for XPS, this is around 6 nm and hence these quantifications reflect the chemical composition found at the surface of the material, while SEM−EDX spectra arise from emission of photons from within the bulk of the sample and the sampling depth is dependent on beam voltage. Analysis of Alloying in Mo 1−x W x S 2 and Mo 1−x W x O 3 at the Microscale. Inspection of surface morphology was conducted by SEM in the secondary electron imaging mode. We have previously reported MoS 2 as nanostructured powders. 84 After doping, a similar fine scale surface morphology was observed for all the Mo 1−x W x S 2 alloys ( Figure  5a), which is consistent with few-layer MoS 2 reported previously. 133 (Figure 5b).
Analysis of Alloying in Mo 1−x W x S 2 and Mo 1−x W x O 3 at the Nanoscale. We were interested in studying the elemental distributions at the nanometer scale in order to fully demonstrate alloying over a range of length scales. We used ultrasonication in water/ethanol and centrifugation to produce nanoscale materials. TEM was used to study the nanoscale particle morphology (Figure 6a−f) and crystallinity of the samples ( Figure S2). Interestingly the sulfide alloys all contained both small particles and nanosheets with size distributions shown in Figure 6g,h. Tilt experiments confirmed that these were indeed separate populations. The nanoparticles have diameters peaked around 4−5 nm and size distributions in the range of 3−9 nm, while nanosheets are typically larger laterally, in the range 5−17 nm, but with typical thicknesses of less than three atomic layers. The presence of both nanosheets and nanoparticles is consistent with the X-ray data collected and the van der Waals layered structures of the group VI B sulfides, with both types of morphology having been previously reported for MoS 2 . 84 TEM investigation of the alloyed oxides showed them to consist of roughly spherical nanoparticles with mean sizes of between 3 and 5 nm. These simple 0D structures are consistent with the WO 3 crystal structure, which is less anisotropic than the sulfides. A few larger nanoparticles >10 nm were also observed in the oxide sample and these showed some evidence of surface faceting, as is expected for surface energy minimization in larger crystals (see Figure S3). AFM was also applied to investigate the morphology of the alloyed sulfide and oxide nanomaterials, giving results that are broadly Chemistry of Materials pubs.acs.org/cm Article consistent with TEM size analysis. However, the ultrasmall size of the nanoparticles and their tendency to form aggregates make quantitative interpretation of such data difficult ( Figure  S1a,b). High-angle annular dark-field (HAADF) STEM and STEM−EDX was carried out in order to use nanoscale elemental mapping to confirm the composition and elemental homogeneity of the alloyed nanomaterials (Figure 7a−f). HAADF-STEM images showed aggregated clusters of Mo 1−x W x S 2 nanosheets and Mo 1−x W x O 3 nanoparticles. STEM−EDX spectrum imaging revealed a fairly even distribution of the constituent elements Mo, W, S, and O at the nanometer scale (Figure 7a−f), supporting the successful alloying of these nanoscale materials. The sensitivity of these samples to damage by the high-energy STEM electron probe prevents their compositional analysis at the atomic scale. Nonetheless, we can exploit the atomic number sensitivity of the HAADF-STEM imaging mode to probe local compositional difference, and no large differences were observed. In nanoparticles and thicker nanosheets, this is to be expected as atomic scale compositional differences are averaged along the electron beam direction. However, in the few-layer sulfides viewed along [001], there are only a few atoms superimposed and the large variability in HAADF intensity we observe for neighboring atomic columns suggest that Mo and W are alloyed at the atomic scale ( Figure 7g). As limited information on alloying can be extracted from diffraction data on the sulfide materials, line profiles were extracted through STEM−EDS elemental maps (Figure 8). These demonstrate uniform alloying in the samples at the nanoscale, with quantification at the expected level based on the Cliff Lorimer absorption corrected Mo/W composition ratio.

■ CONCLUSIONS
We have shown that both Mo 1−x W x S 2 and Mo 1−x W x O 3 alloys can be produced via a rapid, low-temperature molecular precursor method. Control of alloying is readily achieved with resulting variations in morphological and crystallographic changes as demonstrated by Raman spectroscopy, XRD, SEM−EDX, and TEM. The sulfide Mo 1−x W x S 2 alloys generally all adopt the MoS 2 structure with 2H-hexagonal layered polytype. In the Mo 1−x W x O 3 oxides, the WO 3 and MoO 3 structures are dominant; however, for alloys midway between the end compounds (i.e., 0.25 ≤ x ≤ 0.75), the picture is complicated and Rietveld refinement reveals that a series of structurally related unit cells (orthorhombic, triclinic, and monoclinic WO 3 and orthorhombic MoO 3 ) exist for these materials that are related by tilt in MO 6 octahedra induced by strain introduced by Mo doping and hence we conclude that these solids possess a high degree of local disorder and may be better considered as paracrystalline materials. For all alloys, we reveal that the doping of materials is coherent across a range of length scales covering 6 orders of magnitude. The new synthetic pathway that we present here allows simple access to these interesting alloy materials at relatively low temperatures and with excellent control of elemental composition and access to potentially interesting metastable structures. The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.chemmater.0c02685.
Rietveld refinement data and fittings for metal oxides; details of AFM experiments, images and associated histograms for nanomaterials; additional TEM and HRTEM images of nanomaterials; XPS spectra; comparison of products from different reaction times by thermogravimetric analysis, differential scanning calorimetry and Raman spectroscopy; and overview of diffraction peak shifts in metal oxide alloys as a function of metal mole fraction (PDF)