Model oxide supported MoS₂ HDS catalysts: structure and surface properties

Abstract

Supported hydrodesulfidation (HDS) MoS₂/SiO₂, MoS₂/γ-Al₂O₃ and MoS₂/MgO catalysts having a model character have been synthesized by using CS₂ as the sulfiding agent and deeply investigated by means of several techniques. XRPD, HRTEM, Raman and UV-Vis methods have been applied to obtain information on the morphology and the structure of the catalysts as well as on the vibrational and spectroscopic properties. It is shown that, when compared with HRTEM results, XRPD, Raman and UV-Vis data give realistic information on the stacking degree, on the particle size distribution and on the heterogeneity of supported MoS₂ particles on the various supports. (S K-, Mo L₃- and K- edges) EXAFS and XANES spectroscopies have been also used to set up the best sulfidation procedure. UV-vis analysis under controlled atmosphere has been performed to understand the presence of sulfur vacancies and the valence state of Mo ions associated with them. To explore the structure of coordinatively unsaturated Mo sites after reducing or sulfiding treatments (with CS₂ or, occasionally, with H₂S), in situFTIR of adsorbed CO has been performed. It is demonstrated that CO is a sensitive probe for coordinatively unsaturated sites and that the formation of sulfur vacancies on the MoS₂ surface upon reduction in pure H₂ at 673 K is accompanied by an increase of the coordinative unsaturation and a decrease of the valence state of a fraction of surface Mo cations, mainly located on corner and edge sites. Furthermore, it is demonstrated that this process can be reversed upon interaction with the sulfiding agent and that this reversible behavior is really mimicking some of the elementary acts occurring in the HDS process. The complexity of the IR results suggests that the adopted reduction procedure in pure H₂ at 673 K induces the formation of several types of sulfur vacancies, presumably located in different crystallographic positions. It is also concluded that the sulfiding steps are strongly involving the surface of the support and that reductive treatments at high T in H₂ are causing sulfur depletion not only from supported MoS₂ particles, but also from the supporting phase. The involvement of the support is particularly relevant for Al₂O₃ and MgO.

---

1. Introduction

Molybdenum sulfide compounds find many applications in several fields, such as chemical sensors,¹ solar cells,² low-friction surface applications³ and in catalysis.^4–7 In particular, MoS₂ is a widely used catalyst in hydrotreatment processes, like hydrodesulfidation (HDS) and CO hydrogenation, for the production of cleaner fuels in the oil refining industry.

Hydrotreatment catalysts are probably the best described among heterogeneous catalysts, because there is a large amount of experimental and theoretical research devoted to these strategic materials.^8–10

It is usually accepted that the active phase of HDS catalysts is constituted by small nanoparticles (average diameter 5 nm) of lamellar MoS₂, in the form of sandwiched S–Mo–S layers (slabs), stacked on top of each other and separated by van der Waal's gaps. It has been shown that in many alumina-supported systems, the MoS₂ phase is highly dispersed and single slab structures might dominate.^11,12

Unsupported MoS₂ nanoparticles constituted by a single layer were studied by atomically resolved scanning tunneling microscopy (STM), in order to achieve atomic-scale insight into the mechanism of interaction with hydrogen and sulfur-containing molecules (such as thiophene).^13,14 It was shown that, depending on the synthesis and sulfidation conditions, MoS₂ nanoparticles adopt triangular or hexagonally truncated morphologies. The MoS₂ triangular slabs are terminated by dimer-saturated Mo edges, while the hexagonal MoS₂ structures exhibit both Mo-edge and S-edge terminations. Furthermore, STM measurements showed that thiophene molecules adsorb and react on the fully sulfided edges of triangular single-layer MoS₂ nanoplatelets, where special brim sites with a metallic character exist. As a consequence, the activity in HDS reactions has been attributed to sites located at particle edges and has been thought to be originated from sulfur vacancies or so-called coordinatively unsaturated sites (cus sites),^9,15–17 which are created by stripping off one or more sulfur atoms from the MoS₂ nanoparticles edges, during the treatment in hydrogen. The behavior of MoS₂ platelets towards hydrogen has been studied by means of the TPR technique^16,18 and several sulfur species have been distinguished. The most weakly bonded species were assigned to the so called “extra sulfur” in MoS₂₊x samples, i.e. samples sulfided in the absence of hydrogen. The depletion of these species originates a TPR peak at about 520 K.¹⁶ The “stoichiometric” sulfur reacts with H₂ forming H₂S and sulfur vacancies in the 473–873 K range only^16,18 and it is ascertained that small particles are much more reducible than the larger ones. No definite conclusion has been advanced on the relative propensity of sulfur anions located on defects, corners, edges and on extended faces to undergo depletion upon hydrogen treatment, although there is a complete agreement that sulfur ions located on extended faces will be the last to be depleted. In this regard, spectroscopic data have indicated that S–H species are present at the MoS₂ edges,^17,19 and it was proposed that S–H groups may play a key role in supplying the hydrogen during HDS reaction. The matter of sulfur coverage level on edge sites has been discussed by several other authors^8,10,20–22 and the overall conclusion that can be drawn is that the sulfur coverage on the edges decreases with increasing reduction conditions (high hydrogen pressure and temperature).

Notwithstanding the large amount of literature on the topic, up to now the exact nature of the active sites for HDS on MoS₂ platelets dispersed on metal oxide supports and characterized by a low degree of structural definition is still not fully characterized. It is a matter of fact that the overall trends in catalytic activities proposed so far, also for promoted catalysts, come from model structures which are very far from real catalysts, where the MoS₂ particles are located on the nanostructured metal oxide support (often not inert towards sulfiding and reduction treatments) and are characterized by a multiplicity of structural and valence states. In particular: (i) the morphology of the MoS₂ particles is far from the triangular or hexagonal model shape considered in some deeply studied model systems;¹³ (ii) the particles can be either in the form of single slabs or variably stacked; (iii) defect sites (edges, corners, sulfur vacancies, stacking faults and others) can be abundantly present and (iv) defects induced by the interaction with the support can also contribute to the overall defectivity. In relation to the problem of the importance of the interaction with the support, it is worth to recall that the role of the support as “chemical ligand” of the active MoS₂ dispersed phase has been specifically discussed by Raybaud et al.²³ On this basis, a further detailed investigation of the catalysts morphology, of the heterogeneity of active sites structure, and interaction with the support is desirable.

Herein, the preparation and characterization of the MoS₂ HDS catalyst supported on three different metal oxides (SiO₂, γ-Al₂O₃ and MgO) are described. Following our procedure, usually involving CS₂ as the sulfiding agent, the supported MoS₂ phases were obtained in situ in the experimental cells, where the spectroscopic characterization measurements were performed. In order to have catalysts characterized by similar surface concentration of the active phase, MoS₂ loadings roughly proportional to the surface area of the supports were chosen. The structure of the supported MoS₂ phase was characterized by means of several complementary techniques: X-ray Powder Diffraction (XRPD), X-ray absorption (XAS), Raman and UV-Vis-NIR spectroscopies, which gave information on the structural, vibrational and electronic properties of MoS₂, whereas Transmission Electron Microscopy (TEM) was adopted to determine the morphology of the supported particles. Furthermore, a detailed FTIR analysis of CO adsorbed at low temperature on the in situ prepared samples is reported. To the best of our knowledge, few authors have reported studies of hydrotreating catalysts by means of FTIR of adsorbed CO.^24–27 The most recent contribution on the sulfided molybdenum/alumina catalyst comes from the work of Travert et al.,²⁸ in which an IR absorption band in the 2100–2110 cm⁻¹ range on samples treated in a H₂S/H₂ atmosphere was attributed to CO adsorbed on the catalytically active sites located at the edges of MoS₂ slabs. The preparation of the catalyst was similar to that adopted in this study, the main difference being represented by the sulfiding step (H₂S was used instead of CS₂) and by the reducing conditions (H₂/H₂S was used instead of pure H₂). For this reason in some experiments, we used H₂S as the sulfiding agent in order to compare our results with those obtained by other researchers and to ascertain the possible role of carbonaceous residues deriving from the sulfidation with CS₂ in determining the surface structures of MoS₂.

Most of the experiments have been performed on catalysts obtained following a sequence of processes mimicking the activation, deactivation and rejuvenation steps of an HDS catalyst. In order to close the gap between the model and the real catalysts, the reactivity of supported MoS₂ particles towards hydrocarbons (which are main components of the industrial reaction mixture) has been also investigated. By comparing the IR spectra of adsorbed CO at the various treatment stages, we were able to follow clearly the changes in the coordination and valence states of Mo sites induced by the different treatments. The presence of several types of reduced Mo^x+ (x < 4) species upon prolonged hydrogen treatment at 673 K has been substantiated by parallel UV-Vis experiments. The results obtained for MoS₂ supported on the three different metal oxide supports (SiO₂, γ-Al₂O₃ and MgO) are compared in detail to get information on the role of the support in determining the surface properties of the active phase. To this end, the careful comparison of the IR spectra of adsorbed CO on the three systems and the associated morphological details obtained by TEM represent a relevant part of the discussion.

2. Material and methods

2.1 Samples preparation

We have prepared MoS₂ catalysts supported on SiO₂, γ-Al₂O₃ and MgO, which are three high surface area metal oxides having well known properties. The samples have been synthesized by wet impregnation of SiO₂ (Aerosil: 200 m² g⁻¹), γ-Al₂O₃ (Degussa-Alon C: 100 m² g⁻¹) and Mg(OH)₂ (home made: 80 m² g⁻¹) with an aqueous solution of (NH₄)₆Mo₇O₂₄·4H₂O (Merck, Art. 1182). All the impregnated samples have been subjected to the following preparation steps.

Step 1: drying at 373 K overnight followed by calcination at 723 K for 12 hours and outgassing at 673 K for 1 hour under controlled atmosphere. Then, in order to compensate for possible oxygen loss during thermal activation, 40 mbar of oxygen were dosed at 673 K. Finally, oxygen in excess was outgassed at the same temperature. Under these conditions we have been able to transform the catalyst precursors into the MoO₃ supported phase. For the MgO support, during the calcination at 723 K, Mg(OH)₂ is fully transformed into MgO.

Step 2: sulfiding procedure, where the catalyst was subjected to a treatment in CS₂ atmosphere at 673 K overnight, followed by outgassing at the same temperature for 2 hours under dynamic vacuum, to remove the remaining CS₂. We wish to underline that, although H₂S and thiophene are the most used sulfur containing molecules used in this step,^15,29CS₂ can be used as the sulfiding agent as well. The simplicity of the molecular structure, its weak interaction with the surface of the support and the negligible acidic character are in favor of the choice made in this investigation. To be sure that the adoption of CS₂ as the sulfiding agent was not substantially altering the final results, in few cases H₂S was used as the sulfiding agent and the results compared with those obtained following the standard procedure.

Step 3: reduction in H₂ atmosphere (two doses of H₂: P_H2 = 100 mbar at 673 K; total reduction time: 2 h). After each reduction step the gas phase is removed at 673 K under dynamic vacuum.

Step 4: re-sulfidation in CS₂ atmosphere for 2 h (P_CS2 = 20 mbar at 673 K). Before probing the surface with CO, the gas phase is removed under dynamic vacuum at 673 K for 20 min.

It must be underlined that, after the above mentioned steps the systems are outgassed under high vacuum at 673 K. This can have some consequences on the sulfided systems (step 3). In fact, as mentioned in ref. 10, MoS₂ tends to lose sulfur under high vacuum conditions with the consequent alteration on the surface stoichiometry.

Three samples have been analyzed, differing in the Mo loading and in the metal-oxidic support:

- MoS₂/SiO₂ (7% wt MoS₂)

- MoS₂/γ-Al₂O₃ (4.5% wt MoS₂)

- MoS₂/MgO (3% wt MoS₂)

To obtain a comparable surface concentration of the active phase, the adopted MoS₂ loadings were roughly proportional to the surface area of the supports.

2.2 Samples characterization

The morphology of the sulfided samples has been investigated by means of a JEOL 3010-UHR HRTEM microscope operating at 300 kV, equipped with a 2 k × 2 k pixels Gatan US1000 CCD camera. The samples were deposited on a copper grid covered with a lacey carbon film. The stacking number and the sizes of basal planes of the MoS₂ slabs were evaluated by analysing about 100 particles in all systems (MoS₂/SiO₂, MoS₂/γ-Al₂O₃, MoS₂/MgO).

X-Ray Powder Diffraction patterns have been collected with a PW3050/60 X'Pert PRO MPD diffractometer from PANalytical working in Bragg–Brentano geometry, using as source the high power ceramic tube PW3373/10 LFF with a Cu anode (λ = 0.541 Å) and equipped with an Ni filter to attenuate K_β. Scattered photons have been collected by a RTMS (Real Time Multiple Strip) X'celerator detector: data were collected in the 5 ≤ 2θ ≤ 70° angular range, with 0.02° 2θ steps.

X-Ray absorption (XAS) experiments at the S K-edge (2472 eV) and Mo L₃-edge (2520 eV) were performed simultaneously at the Swiss Synchrotron Radiation Facility (SLS, Zurich, CH) on LUCIA beamline.³⁰ A Si(111) monochromator was adopted and harmonic rejection was made by using Ni coated mirrors. Due to the particular configuration of the beamline, it was possible to collect in parallel both fluorescence yield (FY) and total electron yield (TEY), allowing us to choose a posteriori the spectrum with best signal to noise ratio between the two yields. I₀ was measured via TEY on a 300 nm Ni film sputtered on mylar, I_1FY was measured using a Si drift diode detector cooled at −20 °C by Peltier effect, and I_1TEY was measured with a pico-amperometer. The whole experiment was performed in high vacuum to prevent sample contamination: a special vacuum pipe and a manipulator were used to transfer the samples (activated ex situ) from the glove-box to the measurement chamber. Mo K-edge (20 keV) XAS spectra have been measured at the European Synchrotron Radiation Facility (ESRF, Grenoble, F) on BM26A beamline.³¹ All the spectra were collected in the transmission mode and the set-up, from source to detector, was as follows: I₀ measured on a first ionization chamber (1 bar 40% Ar, 60% He), sample, I₁ measured on a second chamber (2 bar 100% Ar), reference, I₂ measured on a third chamber (0.5 bar 100% Ar). This set-up allows a direct energy/angle calibration for each spectrum avoiding any problem related to little energy shifts due to small thermal instability of the monochromator crystals.³² The samples have been measured inside a home-made quartz cell, equipped with kapton windows, that allows thermal treatments to be performed in controlled atmosphere.

Raman spectra have been recorded at room temperature (RT) both in air and under controlled atmosphere (using a home made quartz cell) equipped with windows in optical quartz, by using a Renishaw inVia Raman Microscope spectrometer and an Ar⁺ laser emitting at 514 nm. A 20× “long working distance” magnification objective has been adopted.

Diffuse reflectance (DR) UV-Vis-NIR spectra of the samples after the sulfidation procedures have been performed on a Cary 5000 Varian spectrophotometer equipped with a diffuse reflectance sphere. To obtain reasonable Kubelka–Munk values the samples were diluted in BaSO₄. Before and after reduction in hydrogen (step 4) spectra were recorded on the powdered sample under vacuum in a home made quartz cell equipped with windows in optical quartz.

For FTIR experiments, thin self-supported pellets were used. The spectra were collected on a Bruker IFS 28 Fourier transform spectrophotometer, equipped with a cryogenic MCT detector with 2 cm⁻¹ resolution. To investigate the state of the surface, the CO probe was dosed on the samples by means of a gas manifold permanently connected to the home made IR cell allowing to perform thermal treatment under vacuum and gas dosage. The spectra have been collected on samples in contact with CO pressures in the 0–40 mbar range. The temperature of the pellet could be varied gradually from 77 to 300 K.

3. Results and discussion

3.1 MoS₂/SiO₂ (7% wt MoS₂)

3.1.1 XRPD, Raman and HRTEM results. XRPD measurements have been performed to demonstrate the effective synthesis of the molybdenum sulfide phase on silica, upon sulfidation of the supported Mo oxide for prolonged time. Upon calcination at 723 K (step 1) the supported heptamolybdate phase is entirely transformed into MoO₃ (pattern a in Fig. 1).

Fig. 1 XRPD patterns of the heptamolybdate/SiO₂ after calcinations at 723 K (pattern a) and after the consequent treatment in CS₂ (∼20 mbar) at 673 K for 1 h (b), 6 h (c) and 12 h (d), as compared to the XRD line positions of standard MoO₃ (PDF code number: 005-0508), MoO₂ (PDF code number: 032-0671) and MoS₂ (PDF code number: 37-1492), respectively. Phases are marked by asterisks (red: SiO₂, gray: MoO₃, magenta: MoO₂ and blue: MoS₂).

Further treatments in CS₂ atmosphere (P = 30 mBar) at 673 K for increasing time (1, 6 and 12 h, respectively) transform the supported MoO₃ into MoO₂ (patterns b and c in Fig. 1) and finally into MoS₂ (patterns c and d in Fig. 1).

More in detail, the peaks at 2θ ≈ 12.8°, 23.4°, 25.7° and 27.4° in the pattern a in Fig. 1 (calcined heptamolybdate phase) are associated to (020), (110), (040) and (021) reflections of the MoO₃ phase. The pattern b in Fig. 1 (sulfidation for 1 h) shows, in addition to MoO₃ features, also minor reflections at 2θ ≈ 26°, 37° and 53°, associated to plane reflections of the MoO₂ phase. These features become the dominating XRD signals of pattern c in Fig. 1 (sulfidation for 6 h). In this pattern, broad XRD signals at 2θ ≈ 14° and 33° due to the hexagonal MoS₂ appearing. Finally, in pattern 4 shown in Fig. 1, only the peaks of the MoS₂ phase are observed. The presence of these peaks (characteristic of MoS₂) and the concomitant disappearance of the MoO₂ peaks indicate that the sulfidation process is completed after heating the sample for 12 h in CS₂ at 673 K. Similar results have been obtained for the formation of MoS₂ on the other supports (vide infra).

The XRPD technique used for the control of the sulfidation process was also adopted to get information on the structure of supported MoS₂. In fact, as the XRD peak broadening depends on the coherent scattering domains, it is possible to determine the crystallite dimensions of the MoS₂ slabs in the c-axis direction by applying the Sherrer's equation (D₀₀₂ = 0.76λ/β₀₀₂cos θ, where λ is the wavelength of the X-rays, β is the angular full-width-at-half-maximum FWHM and θ is the diffraction angle),³³ to the (002) XRD diffraction line (see the inset of Fig. 2a). A mean value of about 4 nm was obtained by considering an interlayer distance of 6.17 Å, which indicates a stacking number of ∼6 layers.

Fig. 2 XRPD patterns (a) and Raman spectra (b) of the MoS₂/SiO₂ sample (red line) after sulfidation at 673 K compared to the reference MoS₂ (black line). HRTEM images (c) and (d) of MoS₂/SiO₂ sample. In the inset of (a) the (002) XRD diffraction peak of MoS₂/SiO₂ in the 2θ ≈ 10°–18° range. The arrows in (c) and (d) indicate single and double layers and defective stacking of MoS₂ slabs. In the insets of (d) distribution of the layer numbers and basal lengths of MoS₂ are reported.

As the presence of single MoS₂ layers and the defectivity (such as imperfect stacking and bending of layers) affects the scattering profile and the (002) XRD diffraction peak as well,^5,18,33–37 the evaluation of the average number of slabs following this method is quite approximate and overestimated, as already observed for well dispersed MoS₂ catalysts.⁵ The presence of heterogeneous MoS₂ particles characterized by an average stacking value of about 6 layers and the high heterogeneity emerging from the analysis of the 2θ ≈ 30°–50° range will be substantially confirmed by the Raman and TEM analyses.

The vibrational properties of the supported MoS₂ phase have been investigated by Raman spectroscopy. The spectrum of the MoS₂ reference sample is reported in Fig. 2b (black curve). According to the group theory, four first-order Raman transitions are predicted: E²_2g (at 32 cm⁻¹), E_1g (at 287 cm⁻¹), E¹_2g (at 383 cm⁻¹), and A_1g (at 409 cm⁻¹).³⁸ E²_2g is too low in energy to be detected by our instrument, E_1g is low in intensity (due to the extinction of this mode when c-axis is parallel to the scattering path), while E¹_2g (vibrations in the basal plane) and A_1g (vibrations along the c-axis, i.e. the stacking direction) have stronger intensities. It has been recently shown³⁹ that the E¹_2g and A_1g vibrational bands can be used as very informative indicators of the particle stacking. In fact, the A_1g mode shifts upward from 403 to 408 cm⁻¹ on passing from n = 1 layer to n → ∞ (bulk) (Δν = +5 cm⁻¹), while the E¹_2g undergoes a smaller downward shift from 384 to 382 cm⁻¹ (Δν = −2 cm⁻¹). On this basis the frequency difference between these two characteristic Raman bands is considered as an indicator of the thickness of the particles. In Fig. 2b the Raman spectrum of the MoS₂/SiO₂ sample in the 450–350 cm⁻¹ range (red curve) is compared to that of MoS₂ from Fluka (black curve) considered as a prototype of a perfect bulk counterpart. The frequencies of the characteristic MoS₂ bands in the spectrum of the reference sample closely correspond to those (narrower) reported in the literature for single crystals and consequently their use as reference is justified. The Raman spectrum of the MoS₂/SiO₂ sample shows clearly two vibrational bands, which correspond to the characteristic E_2g and A_1g modes of bulk MoS₂. Following ref. 39, a frequency difference between the two bands of 24 ± 1 cm⁻¹ is indicative of particles constituted in average by 6 ± 2 layers. This average stacking degree of the MoS₂ particles is in pretty good agreement with the XRD results discussed above and will be confirmed later by transmission electron microscopy.

From Fig. 2b it is also evident that the E_2g and A_1g absorption bands in the spectrum of supported particles are broader than the corresponding bands in the spectrum of bulk MoS₂. This might be due to several factors: (i) a broad distribution of the stacking degree (thickness); (ii) variable size and shape of the platelets (either single or clustered) and (iii) presence of vacancies and other defects influencing the size and planarity of the ordered domains inside the single platelets. A clear cut between these factors cannot be performed.

In conclusion, the Raman spectrum is a useful indicator of the stacking and of structural disorder of supported MoS₂ particles. This simple spectroscopic method looks very general and hence will be used also to investigate the dispersion of MoS₂ on Al₂O₃ and MgO.

HRTEM images of the MoS₂/SiO₂ sample (Fig. 2c and d) show that the majority of the SiO₂ surface is covered by structures with high contrast (deep grey zones), which indicates the presence of MoS₂ particles with elongated shape predominantly exposing the basal plane parallel to the electron beam (either single or stacked). As a large number of particles expose the lateral face, the stacking analysis is possible. The MoS₂ particles with stacking degree ranging from about 8 slabs to single slab are observed. The distribution displayed in the inset shows that most of the particles are characterized by stacking values comprised in the 4–6 range. This result is in good agreement with those obtained from XRD and Raman.

Many particles, irrespective of the stacking degree, show clear bending (arrows in Fig. 2c and d), a fact which might be ascribed mainly to the interaction of MoS₂ with the support, although vacancies and dislocations could be invoked as well. Interaction with the support and presence of vacancies are not mutually exclusive. The stacking of slab particles is far from parallel perfection: this also confirms that many types of defects are also present on extended faces, which interfere with the perfect alignment of the slabs. Finally, as for the size of the lamellae is concerned, the inset of Fig. 2d shows that it is mainly comprised in the 5–10 nm range, the lamellae with maximum size (about 10 nm) being observed in the particles with highest stacking degree.

3.1.2 EXAFS and XANES results. The supported sulfide phase was analyzed also by means of XAS spectroscopy at S K-edge and Mo L₃ and K-edges. While a wide number of work deals with determination of the MoS₂ local structure by EXAFS analysis, the electronic configuration and both, metal and ligand atom valence as obtained by means of XANES analysis on both Mo and S edges, are less discussed. Fig. 3a and c shows the XANES spectra of bulk MoS₂ (black curve) and of MoS₂/SiO₂ sample (red curve) at the three different edges, whereas Fig. 3d displays the modulus of the k³-weighted Fourier transform of the corresponding Mo K-edge EXAFS spectra.

Fig. 3 XAS spectra of the MoS₂/SiO₂ sample (red curves) and bulk MoS₂ reference (black curves): (a) S K-edge XANES spectra; (b) Mo L₃ edge XANES spectra; (c) Mo-K-edge XANES spectra; (d) k³-weighted, phase-uncorrected, |FT| of EXAFS spectra collected at the Mo K-edge.

It is important to evidence that S K- and Mo L₃-edges spectra were collected in fluorescence yield (FY) and thus are more sensitive to surface or near-surface structures and electronic configurations, whereas Mo K-edge spectra were measured in the transmission mode that is mainly informative on bulk properties.

Starting from the soft X-ray spectra, the S K-edge spectra (Fig. 3a) reflect the dipole allowed transition 1s → 3p of a core electron of S atoms. The XANES spectrum of MoS₂ bulk (black curve, Fig. 3a) shows a well defined component at 2470 eV, with an evident shoulder at 2473 eV and first oscillation at 2481 eV. The position of the 1s → 3p peak is characteristic of a 2-formal valence state. The spectrum of the supported MoS₂/SiO₂ (red curve, Fig. 3a) is very similar to that of bulk MoS₂, but it is characterized by a decrease of the white-line intensity, whereas the shoulder remains unchanged and the first oscillation slightly changes. This behavior can be explained in terms of a preferential orientation of the MoS₂ platelets, enhanced by the polarized nature of X-ray synchrotron radiation, as already discussed in the work of Guay et al.⁴⁰ In fact, anisotropy in bonding direction for S atoms produces differences in orbital contribution to absorption when slabs are disposed with the stacking axis parallel or perpendicular to the X-ray beam. Moreover, empty p_z and p_ysulfur orbitals are slightly lower in energy than p_x one, explaining why the white line absorption shows a shoulder at higher energy.⁴¹ Therefore, a change in the orientation of the slabs involves a variation in the white line intensity, which decreases when beam hit slabs perpendicular to their stacking.⁴⁰ Since the sample holder was maintained fixed during the whole experiment, differences in the white line structure might be the consequence of a different preferential orientation for pure and supported MoS₂. In particular, it is known that MoS₂ slabs assume preferential orientation when pressed, with stacking axis normal to the holder. This explains why the XANES of bulk MoS₂ resembles those reported in the literature and collected with the X-ray beam parallel to the c-axis direction,⁴⁰ whereas the XANES spectrum of the supported catalyst presents contribution coming from differently oriented platelets. This phenomenon is easily explained by considering that platelets in supported MoS₂ follow the morphology of the support in which they are dispersed. This finding fits with TEM results where both dark contrast zone, from flat slabs, and lamellae, from perpendicularly disposed slabs, are present. Unfortunately, this polarization effect does not allow to distinguish the changes induced in the XANES spectrum by the nanometric scale domains.

A few differences between bulk MoS₂ and the silica supported catalyst are also present in the Mo L₃-edge spectra, which reflect mainly the 2p → 4d transition of Mo core electrons (Fig. 3b). In this case, due to the symmetry of the Mo site, no polarization effect occurs. Bulk MoS₂ (black curve, Fig. 3b) exhibits an intense and sharp white line without crystal field induced splitting, typical of a Mo species in a trigonal prismatic environment,^42,43 and a well defined feature centered at 2530 eV. It is worth noticing that the edge is 2.0 eV downward shifted with respect to that of MoO₃ (not reported),⁴⁴ as expected by going from a Mo(VI) to a Mo(IV) absorbing species. The spectrum of the MoS₂/SiO₂ sample (red curve, Fig. 3b) is very similar to that of bulk MoS₂, but broader and less intense. These differences are ascribed to the nanosized nature of the supported particles.

Coming to hard X-rays, Mo K-edge XANES spectra are shown in Fig. 3c. The spectrum of the bulk MoS₂ sample (black curve, Fig. 3c) has an edge located at 20 001 eV, in agreement with an average 4+ oxidation state of Mo cations. At high energy, the spectrum is dominated by a sharp feature at 20 015 eV and by a single broad band in the white line region. The spectrum of the supported MoS₂ sample (red curve, Fig. 3c) differs from that of the bulk for a partial erosion of the feature at 20 015 eV and, at the same time, for an increase of the intensity in the tail of the edge just before the white line (20 025 eV). Since the spectra of molybdenum oxides (MoO₃ and MoO₂)⁴⁵ reference samples are characterized by a similar behavior, one could interpret the spectrum of MoS₂/SiO₂ as due to the presence of a small residual amount of Mo oxide.⁴⁶ However the XRPD and Raman results discussed above and the similitude of the XANES spectrum of our supported MoS₂ XANES sample with that shown by Leliveld et al.⁴⁶ for the cobalt–molybdenum oxide/γ-Al₂O₃ system let us interpret this result in terms of a weak interaction between MoS₂ and the support.

For Mo K-edge data, the EXAFS part of the spectrum was also collected and the |FT| of the k³-weighted χ(k) function of bulk MoS₂ and of silica supported catalyst are reported in Fig. 3d. The |FT| of the EXAFS signal for bulk MoS₂ (black curve, Fig. 3d) presents an intense first shell signal, due to six degenerate Mo–S single scattering (SS) paths at 2 Å and a more intense second shell peak around 3 Å due to six Mo–Mo SS contributions. A less intense and more distant single path contribution and a strong signal at 6.3 Å, due to collinear multiple-scattering Mo–Mo paths enhanced by the focusing effect, are also present, reflecting a high crystalline order (a small difference in the relative Mo–Mo distances is enough to kill these contributions).

The |FT| of the EXAFS function for the MoS₂/SiO₂ sample (red curve, Fig. 3d) is less intense than that of bulk MoS₂ and shows an inversion in the relative intensities of the first and second shell signals. Moreover, the first shell peak has a lower intensity than in the bulk case, meaning that, in average, the Mo absorbing atoms are surrounded by less S neighbours. This behavior is well documented in the case of nanosized and disordered MoS₂ particles.^47–49 As a matter of fact, the debate on the ability of EXAFS to give unambiguous information on the dimensions of MoS₂ slabs is still open. Shido et al.⁴⁹ demonstrated by simulating the experimental data with various clusters that a structural disorder can induce the same weakening of shell contributions. Other works explain the low value of local order (around 1–3 nm) as due more likely to the extension of ordered sub-domains inside a single slab, instead of the whole extension of the particle.^18,47–49 Following this interpretation an order extension of 2–3 nm in our case is inferred. A direct comparison with TEM images, which show lateral distribution reaching 10 nm, strongly suggests the hypothesis that MoS₂ platelets in the catalyst are far from being regular and that defects are abundantly present.

3.1.3 UV-Vis results. The DRUV-Vis-NIR spectrum of the MoS₂/SiO₂ system after complete sulfidation is reported in Fig. 4a (black curve). In the same figure the spectra obtained on the other systems (MoS₂/Al₂O₃ and MoS₂/MgO) are reported for comparison. In Fig. 4b the spectrum of a thin layer of MoS₂ deposited on a quartz lamina,⁵⁰ collected in the transmission mode, is also reported. The latter is characterized by two bands at around 16 160 and 14 730 cm⁻¹, which are the two so-called excitonic features associated to the direct transition between d states separated by crystal field effect and split by spin orbit interaction.^51–54 These absorption bands are known as the A₁ and B₁ bands, respectively.⁵⁵ The frequency of these bands is influenced by several factors, the major one being represented by the size of the particles, as demonstrated in ref. 56. In particular the A₁ and B₁ transitions shift to higher frequency upon decreasing the size of the ordered domains. Also the intensity changes with decreasing size of defect-free domains and is expected to decline or vanish in favor of discrete molecular-like transitions when the MoS₂ particle size becomes very small. On the basis of the literature results⁵⁶ and from the observed upward shift on passing from the film to the MoS₂/SiO₂ system we can conclude that the size of the MoS₂ particles on the supported system is definitely lower than that of the reference material. However it is difficult to obtain from the UV-Vis spectra a quantitative information concerning the particle size and the stacking value.

Fig. 4 (a) UV-Vis-NIR spectra of MoS₂/SiO₂, MoS₂/Al₂O₃ and MoS₂/MgO samples (black, green and blue curves, respectively) collected in the diffuse reflectance mode in air (a.u. = absorbance units); (b) UV-Vis-NIR spectrum of a MoS₂ thin film deposited on an optical quartz window collected in the transmission mode in air; (c) UV-Vis spectra of MoS₂/SiO₂ outgassed (curve 1) and H₂ reduced (curve 2) at 673 K, collected in controlled atmosphere in the diffuse reflectance mode.

The half width of the A₁–B₁ doublet in the spectrum of supported MoS₂ is much higher than that found in the spectrum of the reference film obtained from the Fluka sample (which is assumed to be characterized by n = ∞ stacking). This is due to the heterogeneity of the particles in the supported MoS₂ and to the contribution of surface states. In fact both phenomena are increasing the width of the excitonic transitions. Another point which merits a specific mention is the significant alteration of the intensity ratio of the A₁–B₁ doublet with respect to the reference film. We think that also this effect is associated with the small particle size of supported MoS₂.

The UV-Vis-NIR spectrum of the MoS₂/SiO₂ sample outgassed at 673 K and collected in controlled atmosphere (vacuum) is illustrated in Fig. 4c. This spectrum is similar to that obtained in air (Fig. 4a), the main relevant difference being represented by an alteration of the intensity ratio of the A₁–B₁ doublet. This effect is plausibly associated with the modification of the interaction with the support due to the different atmosphere surrounding the particles.

The effect of reduction in hydrogen at 673 K on the UV-Vis-NIR spectrum of MoS₂/SiO₂ (Fig. 4c) is important and can be summarized as follows: (a) the absorbance of the sample uniformly increases in the whole visible range; (b) the A₁–B₁ doublet can be still observed on the top of the above mentioned continuous absorption, although with reduced intensity. Both effects can be explained by assuming the formation of a MoS_2−x phase as a consequence of the reductive removal of stoichiometric sulfur and formation of sulfur vacancies on the corner and edges of the particles (although other defects on basal planes could contribute). In fact sulfur vacancies are necessarily associated to a multiplicity of reduced states of Mo and hence to the appearance of new d-states in the band gap of MoS₂⁵⁶ (similar to those of isoelectronic Cr²⁺ and Cr³⁺ ions).⁵⁷ The formation of a continuum of states is well understood. These states, considered as surface defects and shallow traps,^58,59 are typical of nanosized semiconductors and explain the observed optical features.⁶⁰ Notice that the presence of sulfur vacancies is also reducing the extension of the ordered domains necessary for the formation of the excitonic features. This justifies the observed intensity decrease of the A₁–B₁ doublet. In conclusion the UV-Vis-NIR spectra collected under controlled atmosphere give reasonable proof of the reductive elimination of sulfur and formation of surface vacancies. These sulfur vacancies, being associated with coordinatively unsaturated Mo^x+ (x < 4) sites, will be probed by CO molecule (vide infra).

3.1.4 FTIR spectroscopy of adsorbed CO. As it is widely accepted, information about the properties of coordinatively unsaturated (cus) surface sites can be obtained from the IR spectra of adsorbed probe molecules (such as H₂, CO and N₂).^57,61–63 For the scope of this investigation, CO molecule was chosen because of its ability to probe the coordinatively unsaturated (cus) M^x+ species (M = transition metal). In fact, the CO interaction with coordinatively unsaturated metal sites having incomplete d-shell leads to formation of M(CO)_n (n = 1–3) complexes, characterized by strong IR bands in the carbonyl stretching region (1700–2200 cm⁻¹). The frequency of these bands depends upon the balance between σ donation and π back donation between the metal center and the CO ligand. In very general terms, there is broad consensus that the ν(CO) of carbonyls characterized by prevailing σ donation is higher than that of carbonyls where π back donation is relevant. Another point to be stressed is that the intensity of the ν(CO) bands goes in the opposite direction.^64,65 In conclusion, the interaction of CO with positively charged centers leads to carbonyls characterized by reduced π back donation (weak bands), while the opposite occurs for CO in interaction with electron rich centers (strong bands). This is the reason why CO is an excellent probe of the coordination and valence states of surface metal centers. A relevant contribution concerning the interaction of CO with Mo^x+ centers is that of Guglielminotti and Giamello,⁶⁶ where the ν(CO) frequency of carbonyls formed on Mo(V), Mo(IV), Mo(III), Mo(II) and Mo(0) species is considered and determined experimentally. The data reported in ref. 66 are in agreement with the literature on homogeneous molybdenum complexes (see for instance ref. 67–71).

Fig. 5a reports the evolution of the FTIR spectra (in the CO stretching region) of CO adsorbed at 77 K on the sulfided MoS₂/SiO₂ system (i.e. after the sulfidation step), upon decreasing coverage (θ).

Fig. 5 FTIR spectra of CO adsorbed at 77 K on the MoS₂/SiO₂ sample outgassed at 673 K (a), H₂ reduced at 673 K (b) and treated in CS₂ atmosphere at 673 K (c). The evolution of the spectra upon decreasing coverages (θ) is reported (θ_max = 20 mbar, bold grey curve).

At high coverage values (bold gray curve) the spectrum is dominated by an intense and out of scale band at 2157 cm⁻¹, which has been already attributed to CO adsorbed on silanols present on the surface of the SiO₂ support, and by a strong band at 2138 cm⁻¹, due to physically adsorbed (liquid-like) CO. These IR absorption bands are easily reversible upon outgassing at 77 K and both can be ascribed to CO adsorbed on the support. We cannot exclude a contribution due to CO adsorbed on the planar faces of MoS₂. In fact, due to the absence of coordinatively unsaturated centers, these faces can only interact with COvia physical (dispersive) forces. In the 2130–2000 cm⁻¹ range, other significant (although weak) IR absorption bands are present. In particular, at θ(CO)_max, a band at 2068 cm⁻¹ (with a shoulder at 2073 cm⁻¹) is well evident, accompanied by two minor features at 2115 and 2028 cm⁻¹. Although the species responsible for all these absorptions are more resistant to outgassing than CO species interacting with the support and with the planar faces as well, the corresponding IR bands decrease in intensity upon decreasing CO pressure. These bands must be attributed to a small fraction of carbonylic species formed on coordinatively unsaturated Mo centers.

Fig. 5b and c report the evolution of the FTIR spectra of CO adsorbed at 77 K on the MoS₂/SiO₂ sample after H₂reduction and after re-sulfidation at 673 K, respectively. Also in this case, the spectra are dominated by the characteristic bands due to the adsorption of CO on the SiO₂ support (2157–2138 cm⁻¹). The constant intensity of the 2157 cm⁻¹ band upon different activation and sulfidation conditions indicates that no appreciable transformation of silanols into –SH groups and that no appreciable sulfidation of the silica surface are occurring. However, it is evident that the bands in the 2130–2000 cm⁻¹ range are significantly affected by the activation treatment. In particular, the IR absorption band at 2115 cm⁻¹ becomes the most prominent one upon reduction (Fig. 5b), whereas it almost disappears upon re-sulfidation (Fig. 5c). All these bands, which are red-shifted with respect to the free CO molecule (ν_CO = 2143 cm⁻¹), can be attributed to the interaction of CO with coordinatively unsaturated Mo^x+ cations. Following ref. 66–68,70,71 the involved valence states should be lower than IV. The shift with respect to the 2143 cm⁻¹ value of the ν(CO) gas can be ascribed to the effect of π back-donation of the d electrons of Mo towards the CO anti-bonding orbitals, as observed in classical linear metal carbonyls. These coordinatively unsaturated molybdenum ions are associated with sulfur vacancies preferentially formed on samples reduced in hydrogen. It is however difficult to give a more detailed assignment of all the single bands, since, beside the σ–π balance in linear species, other additional factors can affect the IR frequency of adsorbed CO oscillators. In particular, we have to consider that CO adsorbed at a sulfur vacancy position can interact in a bridged form with more than one Mo ion, a fact that can have profound effect on the stretching frequency (as well known when the frequencies of linear and bridged carbonyls are compared). On this basis the only safe conclusion that can be derived from the analyses of the IR spectrum is that the presence of more than one absorption band reflects a complex surface situation, most probably associated with different oxidation states and different coordinative unsaturation of Mo centers located at the sulfur vacancies positions on the edges and corners of the disordered particles (possibly not excluding also some defect present on the basal faces).

The hypothesis that coordinatively unsaturated Mo^x+ (x < 4) centers could be responsible for the CO absorption bands in the 2115–2000 cm⁻¹ range has been already discussed in the literature. In particular, it has been reported that CO adsorption on MoS₂ catalysts (treated in H₂/H₂S atmosphere) supported on Al₂O₃ gives rise to a main absorption band at 2110 cm⁻¹ and a broad band at 2070 cm⁻¹, with a tail extending to 2000 cm⁻¹. The main band has been ascribed to CO adsorbed on the catalytically active sites located at the edges of MoS₂ slabs.^25,28,72,73 The activity of these sites is thought to be associated with sulfur ion vacancies, i.e. coordinative unsaturation. For example, Elst et al. attributed two IR absorption bands at 2106 and 2065 cm⁻¹ to active edge and corner sites of MoS₂ slabs, respectively.⁹ Our data are in good agreement with this assignment. It is worth noticing that the observed frequencies can be justified only if CO interacts with Mo^x+ species in reduced states (x < 4). The hypothesis that reduced Mo ions are involved in the formation of carbonyls seems in conflict with the presence of low intensity of the IR bands in the 2068–2028 cm⁻¹ range on the sulfided sample not treated in hydrogen at 673 K (Fig. 5a). About this point let us however recall that the sulfided samples were always outgassed at 673 K under high vacuum before CO dosage. As it is well known that under high vacuum MoS₂ loses sulfur (a fact which have prevented so far the study of MoS₂ single crystals under ultra high vacuum conditions), the obtained results are not surprising. This hypothesis is confirmed by the results of experiments (not reported for brevity), where CO is dosed on samples outgassed at lower temperature (no carbonylic bands are present). It is quite conceivable that under high vacuum conditions at 673 K sulfur vacancies can be formed on the most exposed sites, where sulfur anions show the lowest coordination. We do not exclude that the weak band at 2028 cm⁻¹ could be due to a small amount of Mo⁰(CO)_n species formed by reduction of Mo ions in very exposed corner positions. Due to the well known increment of the extinction coefficient of IR absorption bands due to carbonyls characterized by increasing back donation, the species absorbing in the 2068–2028 cm⁻¹ range could be associated with very low concentrated species.

In conclusion, on the basis of the experimental results and of the literature data concerning the IR spectroscopy of molybdenum carbonyls, it can be inferred that: (i) the IR absorption bands observed in the spectra shown in Fig. 5 are associated with coordinatively unsaturated Mo species in a valence state lower than 4+ and that these sites become exposed to the interaction of with CO, because of the formation of sulfur vacancies; (ii) the IR absorption band at 2115 cm⁻¹, which grows mostly after reduction in H₂ at 673 K, is due to carbonyls formed on Mo^x+ species (x < 4) located on edge sites (because, considering the inertness of the basal planes, they are the most abundant reducible structures), being the formation of these reduced species in pure H₂ at 673 K in agreement with TPR results (data not shown); (iii) the 2068 cm⁻¹ absorption band, which is formed by simple activationin vacuo at 673 K, is assigned to reduced Mo^x+ (x < 4) species associated with surface sulfur vacancies located on very exposed sites (for instance corner sites); (iv) we do not exclude that the weak band at 2028 cm⁻¹ could be due to a small amount of Mo⁰(CO)_n species formed by reduction of Mo ions in very exposed corner positions; (v) whatever is the assignment of each IR absorption band, the complexity of the IR spectrum demonstrates that on highly dispersed MoS₂ particles characterized by a different size, irregular shape, variable stacking degree and presence of a variety of defects, several types of sulfur vacancies (and hence of coordinatively unsaturated, Mo^x+ (x < 4)) centers can exist, which once probed with CO give a variety of IR signals. We also note that, under the adopted treatment conditions at 673 K (prior to CO contact) no clear IR absorption band at about 2600 cm⁻¹, indicative of the formation of SH species on the MoS₂ phase, was observed (spectra not shown).

To verify whether the sulfidation with CS₂ is equivalent to the most commonly performed process that makes use of H₂S, we have studied the adsorption of CO on samples sulfided with H₂S under equivalent conditions. As the obtained spectra (reported in the ESI†) are nearly equivalent to those illustrated in Fig. 5a–c, we can conclude that the spectra obtained on samples sulfided using CS₂ are fully representative of the catalyst system. Another consequence of this result is that carbonaceous impurities possibly deriving from the use of CS₂ instead of H₂S do not sensibly influence the surface chemistry of MoS₂. The formation of a small amount of carbon was detected by Raman (data not shown for sake of brevity).

As final consideration let us underline that the picture emerging from the IR spectroscopy results is fully confirming the results obtained by UV-Vis spectroscopy.

3.2 MoS₂/γ-Al₂O₃ (4.5% wt MoS₂)

It is known that by changing the oxide support of the industrial HDS catalysts a significant change in their activity is induced. Many scientific works have highlighted these experimental results and key questions about the influence of the support on industrial HDS catalysis have been formulated. Furthermore, it is known that a control of acidic–basic character of the supports can promote a high and stable dispersion of MoS₂ and should also inhibit coke formation. In particular, γ-Al₂O₃ has received great attention as the catalytic support because of its outstanding structural and morphological properties and its relatively low cost. As for MoS₂/SiO₂, the XRD, HRTEM and Raman results will give information on the dispersion of the supported MoS₂ phase. FTIR of adsorbed CO will be discussed in detail because it has been demonstrated to be the most sensitive technique for the detection of surface coordinatively unsaturated Mo species.

3.2.1 XRD, HRTEM, Raman and UV-Vis results. In Fig. 6a, the X-ray diffraction patterns of γ-Al₂O₃ (gray line) and of MoS₂/γ-Al₂O₃ (red line) samples are reported, together with that of the standard MoS₂ (Fluka) (black line).

Fig. 6 XRPD patterns (a) and Raman spectra (b) of MoS₂/γ-Al₂O₃ (×2, red line), γ-Al₂O₃ (gray line, only in the XRD part), and reference MoS₂ (black line); HRTEM images (c) and (d) of the MoS₂/γ-Al₂O₃ sample after sulfidation at 673 K. In the inset of (a), the (002) XRD diffraction peak of the MoS₂/γ-Al₂O₃ (×4) in the 2θ ≈ 10°–18° range. The arrows in (c) and (d) indicate single and double slabs, defective stacking and bended layers. In the insets of (d) distributions of the layer numbers and basal lengths of MoS₂ layers are reported.

The XRPD pattern of MoS₂/γ-Al₂O₃ reveals the presence of three broad peaks at 37.6°, 45.7° and 66.7°, which can be assigned to the γ-Al₂O₃ support, while the five peaks labeled with the asterisks are ascribed to the MoS₂ phase (PDF no. 037–1492). As discussed for MoS₂ /SiO₂, the application of the Sherrer's equation to the (002) XRD peak allows to determine the average crystallite dimensions of the MoS₂ slabs in the c-axis direction and hence the stacking degree. A mean value of about 5 layers is obtained, thus indicating that the dispersion of MoS₂ on Al₂O₃ is better than on SiO₂, considering also the fact that few nanometres single stacked particles escape from XRD detection.

Likewise, Raman spectrum of MoS₂/γ-Al₂O₃ is reported in Fig. 6b (red curve) and shows only the two vibrational E_2g and A_1g modes of bulk MoS₂ associated with the sulfide phase. The frequency difference between the two peaks is 24 ± 1 cm⁻¹, which is indicative of the fact that, in average, the particles are constituted by 6 ± 2 layers, as for MoS₂/SiO₂. Moreover, the E_2g and A_1g bands in the spectrum of supported particles are broader than the corresponding bands of bulk MoS₂, and even broader than those observed on the MoS₂/SiO₂ system. This is likely due to a larger distribution of the stacking degree (including a major fraction of single platelets) and is indicative of a larger heterogeneity.

Representative TEM images of the MoS₂/γ-Al₂O₃ sample are reported in Fig. 6c and d. Multi-layered MoS₂ platelets, with a distribution of the stacking degree up to 5–7 layers, and numerous single dark lines, corresponding to single-layered MoS₂ platelets, are observed. The monolayer MoS₂ platelets are curved and follow the curvature and shape of the γ-Al₂O₃ support (arrows in Fig. 6c and d). This behavior is more pronounced with respect to that observed for MoS₂/SiO₂, so suggesting that in MoS₂/γ-Al₂O₃ a larger interaction of the MoS₂ platelets with the substrate is present, a fact which favors dispersion. In the inset, the stacking distribution as derived from HRTEM shows that most of the particles are characterized by stacking comprised in the 1–4 layer range, a datum which is in agreement with XRD and Raman results. In the same figure an inset concerning the size distribution is also presented which shows that the size of most platelets is 5 nm.

Compared to the MoS₂/SiO₂ the MoS₂ platelets on γ-Al₂O₃ look definitely smaller, hence justifying the broader character of the Raman bands and the UV-Vis spectrum of the MoS₂/γ-Al₂O₃ (Fig. 4a) showing a broadening and a shift to higher frequency of the excitonic peaks.

3.2.2 The FTIR of adsorbed CO. Fig. 7a reports the evolution of the FTIR spectra of CO adsorbed at 77 K on the MoS₂/γ-Al₂O₃ sample, upon decreasing coverage (θ) in the 2300–1950 cm⁻¹ range.

Fig. 7 FTIR spectra of CO adsorbed at 77 K on the MoS₂/γ-Al₂O₃ sample outgassed at 673 K (a), H₂ reduced at 673 K (b) and treated in CS₂ atmosphere at 673 K (c). The evolution of the spectra upon decreasing coverages (θ) is reported (θ_max = 20 mbar, bold grey curve).

The MoS₂/γ-Al₂O₃ catalyst was subjected to the same in situ pre-treatment procedures carried out for MoS₂/SiO₂ systems (see Section 3.1.2). FTIR spectra of CO adsorbed on MoS₂/γ-Al₂O₃ outgassed at 673 K are dominated by three main IR absorption bands centered at 2184, 2157 and 2140 cm⁻¹ at θ = θ_max, which decreases in intensity and shift in frequency upon decreasing the CO coverage. These IR absorption bands are entirely due to CO in interaction with the γ-Al₂O₃ support, although CO physically adsorbed on the basal planes of the MoS₂ platelets can contribute as well. It must be underlined that the spectrum in the 2184–2140 cm⁻¹ range is not identical to that observed when CO is dosed on pure γ-Al₂O₃. On the reason of this difference we shall return in the following. As already observed for the MoS₂/SiO₂ system, additional significant IR absorption bands are present in the 2130–2000 cm⁻¹ range, which are attributed to the interaction of CO with Mo^x+ (x < 4) cations on the surface of MoS₂ crystallites. The frequency of each band is practically the same as observed in the spectrum of MoS₂/SiO₂, indicating the presence of the same species. The major difference is represented by the bands intensity, which is definitely higher on MoS₂/γ-Al₂O₃. This confirms the increased dispersion of MoS₂ on the γ-Al₂O₃ support, already shown by TEM and other techniques. Notice that the IR absorption bands associated with CO on Mo^x+ (x < 4) centers undergo an evolution upon H₂reduction (Fig. 7b) and re-sulfidation in CS₂ (Fig. 7c) at 673 K in the same way as observed for MoS₂/SiO₂. Also in this case the formation and removal of sulfur vacancies upon sulfiding and reducing conditions is then demonstrated.

As a final comment, unlike MoS₂/SiO₂, the IR absorption bands attributed to CO adsorbed on the γ-Al₂O₃ support are altered by the treatment conditions. In particular, only the spectrum recorded after treatment in H₂ at 673 K is identical to that of CO adsorbed on pure γ-Al₂O₃. Although it is outside the scope of this work, we mention that preliminary FTIR experiments performed on pure γ-Al₂O₃ treated with CS₂ show that a partial sulfidation of the γ-Al₂O₃ surface occurs and that the effect on the IR spectrum in the 2184–2140 cm⁻¹ range is similar to that observed in Fig. 7. This process is mainly due to O²⁻/S²⁻ substitution on a fraction of the most reactive sites. The treatment of MoS₂/SiO₂ in H₂ at 673 K removes the sulfur species from the Al₂O₃ surface and consequently the spectrum of adsorbed CO becomes identical to that obtained by dosing CO on pure γ-Al₂O₃. Notice also that under the adopted treatment conditions at 673 K no IR absorption bands clearly due to SH groups were observed, on both the support and on MoS₂ phase.

3.3 MoS₂/MgO (3% wt MoS₂)

MgO, having basic character, differs from other common metal-oxidic supports, which are usually acidic or neutral. Its basicity might be favourable for the MoS₂ dispersion, because of increased interaction between acidic MoO₃ anhydrides and MgO surfaces. A reduced tendency to form coke is also expected for basic support materials. Hydrodesulfurization activity of MgO-supported MoS₂ catalysts has been found to be higher than that of γ-Al₂O₃-supported MoS₂ systems. The higher activities were attributed to the increased dispersion of MoS₂ on the MgO surface.

3.3.1 XRD, HRTEM, Raman and UV-vis results. Fig. 8a shows the X-ray diffraction patterns of MgO pretreated with CS₂ (blank experiment, S-doped MgO: gray line) and of MoS₂/MgO (red line) samples, as well as the pattern of the standard MoS₂ (black line).

Fig. 8 XRPD patterns (a) and Raman spectra (b) of S-doped MgO (gray line, only in the XRD part), MoS₂/MgO (red line) and reference bulk MoS₂ (black line); HRTEM images of the MoS₂/MgO sample, as outgassed at 673 K (c) and (d). In the inset of (a), the (002) XRD diffraction peak of the MoS₂/MgO (×4) in the 2θ ≈ 10°–18° range. The arrows in (c) and (d) indicate single slabs and defective stackings of the MoS₂ layer. In the insets of (d) distributions of the layer numbers and basal lengths of MoS₂ layers are reported.

The XRPD pattern of the MoS₂/MgO system reveals, together with the characteristic peaks of MgO, also four reflections (labeled with the asterisks) due to MoS₂ (PDF no. 037–1492).

The pattern of S-doped MgO shows two broad peaks (2θ ≅ 42.6° and 61.8°), which are ascribed to the (200) and (220) crystalline planes of the MgO phase, and three narrow peaks (2θ ≅ 29.7°, 34.4° and 49.4°), labeled with the circles, that can be assigned to (111), (200) and (220) reflection planes of the MgS phase. This result clearly demonstrates that the pure MgO support can react with sulfur containing molecules not only altering the surface O²⁻/S²⁻ ratio, but also forming crystalline MgS. However, the latter phase is not present on the MoS₂/MgO system. The reason for the inhibitory effect of MoS₂ on the formation of bulk MgS is not known. As made for MoS₂/SiO₂ and MoS₂/γ-Al₂O₃ systems, the Sherrer's equation has been applied to the (002) XRD peak to determine the average crystallite dimensions of the MoS₂ slabs in the c-axis direction and, hence, the stacking degree. A mean value of about 2 layers is obtained, thus indicating that on MgO the MoS₂ dispersion is higher than that observed on both γ-Al₂O₃ and SiO₂.

The Raman spectrum of MoS₂/MgO fully confirms this result (Fig. 8b). In fact, the frequency difference between the E_2g and A_1g vibrational modes of bulk MoS₂⁸ is now 22 ± 1 cm⁻¹, which indicates that the contributing particles are constituted by 3 ± 1 layers in average. This mean stacking degree of the MoS₂ particles is in pretty good agreement with that obtained from XRD results. The broad character of the bands indicates that the system is highly heterogeneous.

The morphological properties of the MoS₂/MgO system are shown in Fig. 8c and d. HRTEM images reveal that many single slabs are present, together with platelets having a distribution of the stacking degree between 3 and 8 layers. The lateral dimension of the platelets is in the 5–10 nm range. Single and double layered MoS₂ platelets are often curved, because they conform to the curvature of the support particles (arrows in Fig. 8c and d). This fact suggests that MoS₂-support interaction is enhanced even with respect to MoS₂/γ-Al₂O₃. In most cases the curvature of the platelets is so evident that the presence of defects induced by interaction with the MgO surface must be invoked. These defects interrupt the planarity and regularity of the platelets and hence increase the system heterogeneity. We speculate that the enhanced interaction of MoS₂ platelets with the MgO surface and the tendency of the support to undergo O²⁻/S²⁻ exchange are intimately connected and that the role of the support to act as “chemical ligand” of the MoS₂ platelets is influenced by the presence of S²⁻ ions on the matrix.⁶⁴ Following ref. 74 stronger edge wetting effects should be considered as well.

As evidenced in the stacking distribution derived from HRTEM (inset of Fig. 8d), most of the particles are characterized by stacking comprised in the 1–2 layer range, in agreement with XRD results. The basal size distribution (right inset in Fig. 8d) is centered around 5 nm, although the distribution does not consider the effect of the curvature. The small size of MoS₂ platelets can justify both, the broad character of the Raman bands and the UV-vis spectrum of the MoS₂/MgO system illustrated in Fig. 4a. In fact, the broadening of the excitonic peaks and the shift to higher frequency can be fully explained on the basis of a decrement of the platelets size, of the presence of curvature and of a lowered stacking degree.

3.3.2 FTIR spectra of adsorbed CO. The evolution of the FTIR spectra of CO adsorbed at 77 K on the MoS₂/MgO sample, upon decreasing coverage (θ) is reported in Fig. 9.

Fig. 9 FTIR spectra of CO adsorbed at 77 K on the MoS₂/MgO sample outgassed at 673 K (a), H₂ reduced at 673 K (b) and treated in CS₂ atmosphere at 673 K (c). The evolution of the spectra upon decreasing coverages (θ) is reported (θ_max = 20 mbar, bold grey curve).

The strong IR absorption bands observed in the 2180–2150 cm⁻¹ range are associated with the stretching modes of CO species physically adsorbed on the supporting matrix.⁶⁴ Similarly to what observed for the MoS₂/γ-Al₂O₃ system, this part of the spectra changes as a function of treatment conditions. A detailed discussion about this point is again outside the scope of the paper and we only mention that this effect is due to a change in the surface structure and composition of MgO upon sulfidation. Only the spectrum of CO adsorbed on the sample reduced in H₂ at 673 K is identical to that of CO adsorbed on the pure MgO matrix. These results demonstrate that sulfidation and reduction in H₂ have no effect not only on MoS₂ but also on the MgO support. As already discussed for MoS₂/SiO₂ and MoS₂/γ-Al₂O₃ systems, the IR absorption bands in the 2170–2030 cm⁻¹ range are attributed to CO interacting with Mo^x+ (x < 4) species of MoS₂ platelets. These bands show an evolution induced by the treatments substantially similar to that observed for MoS₂/SiO₂ and MoS₂/γ-Al₂O₃ samples.

From the spectral evolution upon H₂reduction (Fig. 9b), it can be highlighted that: (i) the intensity of the band at 2118 cm⁻¹ increases; (ii) a new broad band grows at 2085 cm⁻¹, (iii) the band at 2068 cm⁻¹ is substantially unaffected and (iv) the IR bands in the 2120–2000 cm⁻¹ range are broader than those observed on the other systems. As for the sample re-sulfided in CS₂ atmosphere at 673 K (Fig. 9c), we highlight that the IR absorption band at 2118 cm⁻¹ is preferentially affected and that this result is again similar to those obtained for MoS₂/SiO₂ and MoS₂/γ-Al₂O₃ systems. All these facts confirm that molybdenum sites can be reduced, with reversible production and saturation of sulfur vacancies.

4. Conclusions

Supported MoS₂ model systems were obtained from the supported oxide precursor phase on three different metal oxides (SiO₂, γ-Al₂O₃ and MgO); structural, optical and vibrational techniques confirmed the validity of our synthesis to obtain well dispersed MoS₂ nanoparticles. The adopted procedure, which is based on the use of CS₂ instead of the classical H₂S/H₂ mixture in autoclave, is very simple and leads to complete sulfidation of the supported MoO₃ phase, suggesting that CS₂ has the dual function of sulfiding and reducing agent. XRD measurements performed at increasing sulfidation stages have been used to set up the best procedure for the in situ synthesis of MoS₂/silica. This procedure was successively adopted also to obtain the MoS₂/γ-Al₂O₃ and MoS₂/MgO systems. XANES analysis confirms the expected S and Mo valences for the MoS₂ structure, whereas S K-edge measurements revealed that MoS₂ platelets adopt a preferential orientation to follow the support structure. From both, EXAFS signal and TEM images we verified that the supported MoS₂ platelets are highly defective.

By examining the XRD, Raman and UV-Vis the stacking and the size of the supported MoS₂ platelets have been determined. The application of the Sherrer's equation to the (002) XRD peak of MoS₂ allowed to determine the average crystallite dimensions of the MoS₂ slabs in the c-axis direction, hence the stacking degree on all samples is decreasing from ∼6 and 4 (MoS₂/SiO₂ and MoS₂/γ-Al₂O₃) to ∼2 (MoS₂/MgO). Also the Raman spectra of supported MoS₂ gave valuable information about the structure of supported MoS₂. In fact, the frequency difference between the two E_2g and A_1g vibrational modes of supported MoS₂ resulted to be a sensitive probe of the stacking degree, while the broad character of the corresponding bands was a qualitative indicator of particle heterogeneity. Although the general trend is the same, stacking figures obtained from Raman data are slightly higher than those obtained by the other methods, because the Raman signal from particles characterized by highest stacking is more intense. The application of UV-vis spectroscopy has demonstrated a great utility in the (qualitative) determination of two parameters: particle size and presence of sulfur vacancies induced by hydrogen reduction. In particular, it has been ascertained that reduction in hydrogen at 673 K causes the formation of a variety of coordinatively unsaturated Mo^x+ sites characterized by x < 4 and that the presence of these reduced states profoundly alters the optical properties of the systems. It is worth underlining that the aforementioned physical methods are of simple application and therefore can be utilized to monitor the dispersion of the active phase in industrial catalysts in a fast semi-quantitative way.

HRTEM pictures confirmed the coexistence and the dispersion degree of single and stacked MoS₂ slabs having basal dimension never exceeding 10 nm. However, great differences among the three supports can be highlighted: in fact, a decrease of the average stacking degree from 4–6 in MoS₂/SiO₂ to 2–4 in MoS₂/γ-Al₂O₃ and to 1–2 in MoS₂/MgO is inferred. The effect on the basal size of platelets is less important as it decreases from about 10 nm (MoS₂/SiO₂) to 5 nm (MoS₂/γ-Al₂O₃, MoS₂/MgO). From these data and from the observation that slabs are intimate curved following the profile of the support particles, it is inferred that the interaction between the MoS₂ and the support increases by moving from SiO₂ to γ-Al₂O₃ to MgO and reflects the increasing dispersion of catalyst particles. In this regard, we mention the fact that the sulfidation process has its influence also on the supporting matrix. In fact, MgO incorporates sulfur in the structure, γ-Al₂O₃ shows surface reactivity, while SiO₂ does not show reactivity at all.

Considering that both the HDS activity and the tendency of the three investigated supports to undergo O²⁻/S²⁻ exchange follow the order MoS₂/MgO > MoS₂/γ-Al₂O₃ > MoS₂/SiO₂ and that the same order is verified for the particle dispersion, it is hypothesized that all these factors are strongly linked and that the incorporation of S²⁻ ions in the surface of the support is influencing the MoS₂/support interaction.

The results obtained from the FTIR spectroscopy of CO adsorbed at 77 K on MoS₂/oxide samples allowed to obtain direct information on the nature of the cus sites involved in the HDS reaction, showing that: (i) outgassing the completely sulfided particles at 673 K under high vacuum is enough to reduce a small number of Mo sites probably located on corners and other exposed positions. Evidence of the formation of a low amount of Mo⁰ species is also found. On the contrary, samples outgassed only at 373 K do not show any evidence of reduced Mo sites; (ii) treatment in hydrogen at 673 K causes the formation of new and abundant low valence Mo^x+ (x < 4) species (the corresponding carbonyls giving an IR absorption band at 2115 cm⁻¹) most likely located on edges. The reduction of Mo ions located on flat surfaces in correspondence of defects is however not excluded. CO is only physically adsorbed on defect free portions of the flat MoS₂ surfaces which are not exposing coordinatively unsaturated sites and are very resistant to hydrogen reduction; (iii) thermal treatment in CS₂ at sufficiently high temperature (673 K) is able to restore almost completely the starting sulfide phase, with the consequent disappearance of the Mo sites associated with sulfur vacancies on edge sites. The complexity of the IR spectra of adsorbed CO demonstrates that on a highly dispersed catalyst, MoS₂ particles characterized by large size dispersion, irregular shape, variable stacking degree, presence of a variety of defects, and several types of sulfur vacancies (and hence of coordinatively unsaturated Mo^x+ (x < 4)) centres coexist, which originate a variety of surface carbonyls and IR absorption bands once probed with CO. Finally, let us underline that the similarity of the results of IR of adsorbed CO on MoS₂ supported on the SiO₂, γ-Al₂O₃ and MgO supports excludes that the carbonyl bands can be due to Mo^x+ species directly anchored to the supports.

The whole set of data reported herein directly face the problem of the nature of the active sites in MoS₂-based HDS catalysts and bring evidence that the surface concentration of sulfur vacancies associated with low valence Mo^x+ (x < 4) species can be reversibly formed by changing the activation conditions (reduction or sulfidation). These vacancies are mostly located on corner and edge sites. However, considering the highly disordered nature of the samples and the remarkable curvature of the MoS₂ platelets particularly evident on γ-Al₂O₃ and MgO, the presence of sulfur vacancies also on flat surfaces is not excluded. Finally, the participation of the γ-Al₂O₃ and MgO supports in the sulfidation/reduction steps is also demonstrated.

Acknowledgements

We are indebted with the whole staff of the beamlines LUCIA, at the SLS, and BM26 DOUBBLE, at the ESRF, where XAFS experiments have been performed; in particular with M. Janousch (LUCIA) and with S. Nikitenko (BM26) for their competent and friendly support.

This work was supported by MIUR (Ministero dell'Istruzione, dell'Università e della Ricerca), INSTM Consorzio and Eni S.p.A.

References

1. H. L. Zhang, K. P. Loh, C. H. Sow, H. R. Gu, X. D. Su, C. Huang and Z. K. Chen, Langmuir, 2004, 20, 6914–6920.
2. N. Barreau and J. C. Bernede, J. Phys. D: Appl. Phys., 2002, 35, 1197–1204.
3. L. Rapoport, Y. Bilik, Y. Feldman, M. Homyonfer, S. R. Cohen and R. Tenne, Nature, 1997, 387, 791–793.
4. J. P. Wilcoxon, J. Phys. Chem. B, 2000, 104, 7334–7343.
5. M. P. De la Rosa, S. Texier, G. Berhault, A. Camacho, M. J. Yacaman, A. Mehta, S. Fuentes, J. A. Montoya, F. Murrieta and R. R. Chianelli, J. Catal., 2004, 225, 288–299.
6. M. Kouzu, K. Uchida, Y. Kuriki and F. Ikazaki, Appl. Catal., A, 2004, 276, 241–249.
7. R. R. Chianelli, M. H. Siadati, M. P. De la Rosa, G. Berhault, J. P. Wilcoxon, R. Bearden and B. L. Abrams, Catal. Rev. Sci. Eng., 2006, 48, 1–41.
8. P. Raybaud, Appl. Catal., A, 2007, 322, 76–91.
9. L. E. Elst, S. Eijsbouts, A. D. van Langeveld and J. A. Moulijn, J. Catal., 2000, 196, 95–103.
10. J. F. Paul, S. Cristol and E. Payen, Catal. Today, 2008, 130, 139–148.
11. N. Topsoe, J. Catal., 1980, 64, 235–237.
12. J. Grimblot, P. Dufresne, L. Gengembre and J.-P. Bonelle, Bull. Soc. Chim. Belg., 1981, 90, 1311.
13. J. V. Lauritsen, M. V. Bollinger, E. Lagsgaard, K. W. Jacobsen, J. K. Norskov, B. S. Clausen, H. Topsoe and F. Besenbacher, J. Catal., 2004, 221, 510–522.
14. J. V. Lauritsen, M. Nyberg, J. K. Norskov, B. S. Clausen, H. Topsoe, E. Lagsgaard and F. Besenbacher, J. Catal., 2004, 224, 94–106.
15. H. Topsoe, B. S. Clausen and F. E. Massoth, Science and Technology in: Hydrotreating Catalysis, Springer, Berlin, 1996.
16. N. Dinter, M. Rusanen, P. Raybaud, S. Kasztelan, P. da Silva and H. Toulhoat, J. Catal., 2009, 267, 67–77.
17. H. Topsoe, B. S. Clausen, N. Y. Topsoe and E. Pedersen, Ind. Eng. Chem. Fundam., 1986, 25, 25–36.
18. C. Calais, N. Matsubayashi, C. Geantet, Y. Yoshimura, H. Shimada, A. Nishijima, M. Lacroix and M. Breysse, J. Catal., 1998, 174, 130–141.
19. N. Y. Topsoe and H. Topsoe, J. Catal., 1993, 139, 631–640.
20. P. Raybaud, J. Hafner, G. Kresse, S. Kasztelan and H. Toulhoat, J. Catal., 2000, 189, 129–146.
21. S. Cristol, J. F. Paul, E. Payen, D. Bougeard, S. Clemendot and F. Hutschka, J. Phys. Chem. B, 2000, 104, 11220–11229.
22. S. Cristol, J. F. Paul, E. Payen, D. Bougeard, S. Clemendot and F. Hutschka, J. Phys. Chem. B, 2002, 106, 5659–5667.
23. C. Arrouvel, M. Breysse, H. Toulhoat and P. Raybaud, J. Catal., 2005, 232, 161–178.
24. A. A. Tsyganenko, P. Y. Storozhev and C. O. Arean, Kinet. Catal., 2004, 45, 530–540.
25. B. Mueller, A. D. van Langeveld, J. A. Moulijn and H. Knoezinger, J. Phys. Chem., 1993, 97, 9028–9033.
26. J. B. Peri, J. Phys. Chem., 1982, 86, 1615–1622.
27. M. I. Zaki, B. Vielhaber and H. Knoezinger, J. Phys. Chem., 1986, 90, 3176–3183.
28. A. Travert, C. Dujardin, F. Mauge, S. Cristol, J. F. Paul, E. Payen and D. Bougeard, Catal. Today, 2001, 70, 255–269.
29. M. L. Vrinat, Appl. Catal., 1983, 6, 137–158.
30. A. M. Flank, G. Cauchon, P. Lagarde, S. Bac, M. Janousch, R. Wetter, J. M. Dubuisson, M. Idir, F. Langlois, T. Moreno and D. Vantelon, Nucl. Instrum. Methods Phys. Res., Sect. B, 2006, 246, 269–274.
31. (a) S. Nikitenko, A. M. Beale, A. M. J. van der Eerden, S. D. M. Jacques, O. Leynaud, M. G. O'Brien, D. Detollenaere, R. Kaptein, B. M. Weckhuysen and W. Bras, J. Synchrotron Radiat., 2008, 15, 632–640; (b) G. Silversmit, B. Vekemans, S. Nikitenko, W. Bras, V. Czhech, G. Zaray, I. Szaloki and L. J. Vincze, J. Synchrotron Radiat., 2009, 16, 237–246.
32. C. Lamberti, S. Bordiga, F. Bonino, C. Prestipino, G. Berlier, L. Capello, F. D'Acapito, F. X. Xamena and A. Zecchina, Phys. Chem. Chem. Phys., 2003, 5, 4502–4509.
33. G. Berhault, M. P. De la Rosa, A. Mehta, M. J. Yacaman and R. R. Chianelli, Appl. Catal., A, 2008, 345, 80–88.
34. P. Joensen, E. D. Crozier, N. Alberding and R. F. Frindt, J. Phys. C: Solid State Phys., 1987, 20, 4043–4053.
35. P. Joensen, R. F. Frindt and S. R. Morrison, Mater. Res. Bull., 1986, 21, 457–461.
36. Y. Y. Peng, Z. Y. Meng, C. Zhong, J. Lu, W. C. Yu, Z. P. Yang and Y. T. Qian, J. Solid State Chem., 2001, 159, 170–173.
37. K. S. Liang, R. R. Chianelli, F. Z. Chien and S. C. Moss, J. Non-Cryst. Solids, 1986, 79, 251–273.
38. T. J. Wieting and J. L. Verble, Phys. Rev. B: Solid State, 1971, 3, 4286.
39. C. Lee, H. Yan, L. E. Brus, T. F. Heinz, J. Hone and S. Ryu, ACS Nano, 2010, 4, 2695–2700.
40. D. Guay, W. M. R. Divigalpitiya, D. Belanger and X. H. Feng, Chem. Mater., 1994, 6, 614–619.
41. L. F. Mattheiss, Phys. Rev. B: Solid State, 1973, 8, 3719–3740.
42. R. Coehoorn, C. Haas and R. A. de Groot, Phys. Rev. B: Condens. Matter, 1987, 35, 6203.
43. G. N. George, W. E. Cleland, J. H. Enemark, B. E. Smith, C. A. Kipke, S. A. Roberts and S. P. Cramer, J. Am. Chem. Soc., 1990, 112, 2541–2548.
44. T. Ressler, O. Timpe, T. Neisius, J. Find, G. Mestl, M. Dieterle and R. Schlogl, J. Catal., 2000, 191, 75–85.
45. T. Ressler, J. Wienold, R. E. Jentoft and T. Neisius, J. Catal., 2002, 210, 67–83.
46. R. G. Leliveld, A. J. van Dillen, J. W. Geus and D. C. Koningsberger, J. Catal., 1997, 171, 115–129.
47. S. P. Bouwens, R. Prins, V. H. J. Debeer and D. C. Koningsberger, J. Phys. Chem., 1990, 94, 3711–3718.
48. E. J. M. Hensen, P. J. Kooyman, Y. van der Meer, A. M. van der Kraan, V. H. J. de Beer, J. A. R. van Veen and R. A. van Santen, J. Catal., 2001, 199, 224–235.
49. T. Shido and R. Prins, J. Phys. Chem. B, 1998, 102, 8426–8435.
50. K. M. Garadkar, A. A. Patil, P. P. Hankare, P. A. Chate, D. J. Sathe and S. D. Delekar, J. Alloys Compd., 2009, 487, 786–789.
51. J. P. Wilcoxon, P. P. Newcomer and G. A. Samara, J. Appl. Phys., 1997, 81, 7934–7944.
52. K. K. Kam and B. A. Parkinson, J. Phys. Chem., 1982, 86, 463–467.
53. A. A. Akl, H. Kamal and K. Abdel-Hady, Appl. Surf. Sci., 2006, 252, 8651–8656.
54. G. L. Frey, S. Elani, M. Homyonfer, Y. Feldman and R. Tenne, Phys. Rev. B: Condens. Matter Mater. Phys., 1998, 57, 6666–6671.
55. A. Splendiani, L. Sun, Y. B. Zhang, T. S. Li, J. Kim, C. Y. Chim, G. Galli and F. Wang, Nano Lett., 2010, 10, 1271–1275.
56. B. L. Abrams and J. P. Wilcoxon, Crit. Rev. Solid State Mater. Sci., 2005, 30, 153–182.
57. E. Groppo, C. Lamberti, S. Bordiga, G. Spoto and A. Zecchina, Chem. Rev., 2005, 105, 115–183.
58. L. Brus, J. Phys. Chem., 1986, 90, 2555–2560.
59. L. Brus, J. Phys. Chem. Solids, 1998, 59, 459–465.
60. J. Ramirez, L. Cedeno and G. Busca, J. Catal., 1999, 184, 59–67.
61. A. Zecchina, S. Coluccia and C. Morterra, Appl. Spectrosc. Rev., 1985, 21, 259–310.
62. A. Zecchina and C. O. Arean, Chem. Soc. Rev., 1996, 25, 187.
63. A. Zecchina, D. Scarano, S. Bordiga, G. Spoto and C. Lamberti, Advances in Catalysis, Academic Press Inc, San Diego, 2001, vol. 46, pp. 265–397.
64. A. Zecchina, D. Scarano, S. Bordiga, G. Ricchiardi, G. Spoto and F. Geobaldo, Catal. Today, 1996, 27, 403–435.
65. C. Lamberti, E. Groppo, A. Zecchina and S. Bordiga, Chem. Soc. Rev., 2010, 39, 4951–5001.
66. E. Guglielminotti and E. Giamello, J. Chem. Soc., Faraday Trans. 1, 1985, 81, 2307–2322.
67. W. S. Tsang, D. W. Meek and A. Wojcicki, Inorg. Chem. (Washington, DC, U. S.), 1968, 7, 1263–1268.
68. M. H. Chisholm, J. A. Connor, J. C. Huffman, E. M. Kober and C. Overton, Inorg. Chem. (Washington, DC, U. S.), 1984, 23, 2298–2303.
69. C. C. Williams and J. G. Ekerdt, J. Phys. Chem., 1993, 97, 6843–6852.
70. T. Szymanska-Buzar, T. Glowiak and I. Czelusniak, Inorg. Chem. Commun., 2001, 4, 183–186.
71. X. Z. Sun, S. M. Nikiforov, A. Dedieu and M. W. George, Organometallics, 2001, 20, 1515–1520.
72. B. M. Vogelaar, N. Kagami, T. F. van der Zijden, A. D. van Langeveld, S. Eijsbouts and J. A. Moulijn, J. Mol. Catal. A: Chem., 2009, 309, 79–88.
73. Z. L. Wu, F. X. Sun, W. C. Wu, Z. C. Feng, C. H. Liang, Z. B. Wei and C. Li, J. Catal., 2004, 222, 41–52.
74. D. Costa, C. Arrouvel, M. Breysse, H. Toulhoat and P. Raybaud, J. Catal., 2007, 246, 325–343.

---

Footnote

† Electronic supplementary information (ESI) available: FTIR spectra of CO adsorbed at 77 K on the MoS₂/SiO₂ sample outgassed at 673 K. See DOI: 10.1039/c0cy00050g

---