MoS₂/TiO₂ heterostructures as nonmetal plasmonic photocatalysts for highly efficient hydrogen evolution

Abstract

In this study, we report nonmetal plasmonic MoS₂@TiO₂ heterostructures for highly efficient photocatalytic H₂ generation. Large area laminated MoS₂ in conjunction with TiO₂ nanocavity arrays is achieved via carefully controlled anodization, physical vapor deposition, and chemical vapor deposition processes. The broad spectral response ranging from ultraviolet-visible (UV-vis) to near-infrared (NIR) wavelengths and finite element frequency-domain simulations suggest that this MoS₂@TiO₂ heterostructure enhances photocatalytic activity for H⁺ reduction. A high H₂ yield rate of 181 μmol h⁻¹ cm⁻² (equal to 580 mmol h⁻¹ g⁻¹ based on the loading mass of MoS₂) is achieved using a low catalyst loading mass. The spatially uniform heterostructure, correlated with plasmon-resonance through the conformal MoS₂ coating that effectively regulates charge transfer pathways, is proven to be vitally important for the unique solar energy harvesting and photocatalytic H₂ production. As an innovative exploration, our study demonstrates that the photocatalytic activities of nonmetal, earth-abundant materials can be enhanced with plasmonic effects, which may serve as an excellent catalytic agent for solar energy conversion to chemical fuels.

---

Broader context

A novel MoS₂@TiO₂ heterostructured film with periodically patterned morphology, exhibiting enhanced light absorption within UV-vis-NIR wavelengths, was developed. Both experimental studies and theoretical modeling demonstrate that S-vacancies in the multilayered MoS₂ nanoflakes are responsible for the localized surface plasmon resonance (LSPR, visible light absorption) and tunable band gaps (near infrared absorption). An outstanding H₂ yield rate of 181 μmol h⁻¹ cm⁻² (equal to 580 mmol h⁻¹ g⁻¹ based on the loading mass of MoS₂) was then achieved. The present study also shows that the developed catalyst can even split sea water under solar light irradiation, which will open a new paradigm for direct solar energy harvesting.

1. Introduction

Recently, intensive research has focused on solar energy conversion to provide clean chemical fuels (H₂, CH₄, etc.) as a result of an impending global energy crisis as well as hazardous environmental pollution conditions largely induced by fossil fuel consumption.^1–3 Solar-to-fuel conversion significantly depends on the semiconductor materials that can harvest photon energy across the wide solar spectrum (from the ultraviolet (UV) to near-infrared (NIR) region) and simultaneously generate charge-carriers at suitable energy levels for H⁺ reduction.^4,5 Among various materials and technologies, low-cost, nontoxic, and chemically stable TiO₂ has long been studied as a promising photocatalyst for solar-driven water splitting.^6,7 However, the main disadvantages (e.g., wide band gap and sluggish charge transfer kinetics) of TiO₂ have limited its feasibility within the visible light region, which accounts for approximately 43% of the solar spectrum.⁶ It is possible to overcome these issues by a variety of techniques, including heavy doping (nitrogen), integrating narrow band gap semiconductors, and decorating with noble metals and co-catalysts (CdS, CdSe, Pt, Au, Ag, etc.), to broaden the passive oxide semiconductor absorption range to include visible or NIR, creating a more efficient photocatalyst.^8–10 However, the widespread use of scarce noble metals is not ideal due to their high cost and potential to be environmentally toxic. Therefore, the quest for materials that are solar light sensitive and earth-abundant is of utmost importance.

Molybdenum disulfide (MoS₂) belongs to the two-dimensional (2D) layered transition metal dichalcogenide (TMD) family that has a sandwich-like structure of Mo atoms between two layers of hexagonally packed sulfur atoms. The weak van der Waals bonding between these 2D layers often gives rise to single- or few-layer nanosheet architectures.^6,11,12 Recent work shows that MoS₂ can be a promising electrocatalyst for the hydrogen evolution reaction (HER), owing to the nanosized MoS₂ edge defects that are preferential for hydrogen adsorption. Therefore, few-layered nanoscale MoS₂ flakes serve as a valuable strategy for improving the efficiency of hydrogen evolution.^13,14 Chemical exfoliation and solvothermal methods are usually employed for the fabrication of nanostructured MoS₂. However, these methods usually promote irregular particle formation or undesirable stacking of multilayer MoS₂ deposit products.^15,16 Multilayer stacking exposes more catalyst basal planes than edges, rendering much of the material catalytically inert. Additionally, without proper dispersion and immobilization, powdered nanomaterials can suffer from particle aggregation, leading to performance degradation.¹⁷ Until now, an efficient scheme or design principle for the integration of ordered, nanostructured MoS₂ with wide band gap semiconducting behavior had eluded researchers.

Compared with metallic 1T-MoS₂, 2H-MoS₂ exhibits high stability and semiconducting properties at room temperature, allowing it to be used as a co-catalyst coupled with other wide band gap semiconductors for H₂ evolution in a photoelectron reactive medium.^18,19 However, there is a lack of discussion in the present literature that highlights MoS₂ as the main photocatalyst, especially those that are attributed to NIR intrinsic absorbance. Theoretical calculations demonstrate that the band gap of MoS₂ can be modulated through control of particle size. In this case, the band gap broadens from 1.2 eV to 1.9 eV when the MoS₂ architecture changes from bulk to monolayer. This characteristic can be ascribed to the quantum confinement of nanomaterials. The energy band structure can also be affected by the metal-chalcogenide stoichiometric ratio.^16,20 A recent theoretical study showed that nonstoichiometric metal-chalcogenides (e.g., WO_3−x, Cu_2−xS, and MoO_3−x) displayed evidence of an indirect plasmonic absorption, which is distinctly different from previously reported band gap transitions.^21–23 Therefore, local surface plasmonic resonance (LSPR) can be used to describe the light harvesting phenomena, which is mainly deduced from charge collective oscillation on the metal-chalcogenide surface propagated by numerous anion (O or S) vacancies within the crystal lattice. It is conceivable that nonmetal MoS₂ with plasmonic absorption may be a solution to fill the visible-light harvesting gap vacated by wide band gap semiconductors, effectively improving the solar-to-energy efficiency of HER photocatalysts.^22,24 Lastly, MoS₂@TiO₂ hybrid catalysts have been recently reported that aim to enhance photocatalytic efficiency by modulating the TiO₂ energy level with inter-band coupling.⁶

Herein, we demonstrate a combined physical vapor deposition (PVD) and chemical vapor deposition (CVD) strategy to coat few-layered MoS₂ nanoflakes conformally on the inner surface of anodized TiO₂ nanocavity arrays (referred to as MoS₂@TiO₂) with a highly-ordered 3D hierarchical configuration. The stoichiometric ratio of Mo and S atoms within the MoS₂ lattice and vertically contacting facets of MoS₂ and TiO₂ can be controlled by tuning the S source in the Na₂S_x solution and altering the CVD reaction rate (please refer to experimental details). This highly localized growth enables conformal and uniform MoS₂ nanoflakes to be formed on the surface of TiO₂ nanocavities. Such heterostructures have shown powerful photon harvesting abilities in the UV-vis-NIR range by tethering the TiO₂ substrate to a plasmonic/intrinsic MoS₂ coating. The facilitated electron transfer pathway and appropriately tuned energetic position of the conduction band results in facile charge-carrier separation and dramatically enhanced H₂ evolution efficiency. UV-vis spectroscopy analysis, finite element method simulation (FEM), and the monochromatic light irradiated H₂ generation rate conclude that LSPR, mainly excited in the wavelength from 400–600 nm, substantially contributes to photocatalytic activity. Minimal red light (600–700 nm) induced H₂ production is observed from the lower photoelectron energetic level of the MoS₂ inter-band excitation and excessive photoelectron–hole recombination. This nonmetal plasmonic heterostructure is also expected to be applicable to other 2D material systems, which can serve as a new design protocol for highly efficient photocatalysts.

2. Experimental section

2.1 Material preparation

Anodic growth of TiO₂ nanocavities. Titanium foils (25.4 × 25.4 × 0.05 mm thick, 99.7% purity, MTI Corporation) were ultrasonically cleaned in acetone, ethanol, and deionized (DI) water for 30 min and dried in air. Sample size is rationally chosen but not limited to a 1 inch square by this fabrication method. Anodization was carried out in 3 M HF/H₃PO₄ (98%, Alfa Aesar, USA). A constant voltage of 10 V was applied to a two-electrode setup for 4 h with Pt foil as a counter electrode. After anodization, the TiO₂ films were rinsed with ethanol and dried in air.

Mo deposition and sulfurization. An e-Beam evaporator (Thermo Scientific) was used to deposit Mo layers into anodized TiO₂ films with a thickness of 10, 20, and 30 nm, respectively. The Mo layer thickness was controlled by an automated quartz crystal film-thickness monitor. The as-prepared Mo@TiO₂ hybrid was then put in the center of a quartz-tube furnace (MTI Corporation) together with a 0.5 M Na₂S solution containing 1 M sulfur (placed in a crucible at the upstream side, at a temperature of 100 °C, with a ramping speed of 2 °C min⁻¹). The MoS₂@TiO₂ heterostructure was fabricated at 400 °C for 10 min with a heating rate of 5 °C min⁻¹ under vacuum. The obtained MoS₂ mass loading was estimated by depositing Mo on compact Ti foils with the same e-beam rate and time mentioned above, and the thickness of the obtained Mo was confirmed by SEM and TEM. Loading of Mo was estimated by multiplying the volume and density of Mo. The deposited Mo is supposed to be changed to MoS₂ (10.2, 20.4 and 30.6 μg cm⁻² for 10, 20, and 30 nm Mo deposits, respectively).

2.2 Material characterization

The morphologies of the MoS₂@TiO₂ films were observed with a field-emission scanning electron microscope (FE-SEM, ZEISS ultra 55) and a high-resolution transmission electron microscope with EDS mapping (Cs corrected JEM ARM200F STEM). The cross-section sample for TEM was cut off by a Tescan LYRA-3 Model GMH focused ion beam microscope and pasted onto a Cu ring holder. X-ray diffraction (XRD) was obtained using an X’pert Powder (PANalytical, equipped with a Panalytical X’celerator detector using Cu Kα radiation, λ = 1.54056 Å). The chemical composition was characterized by X-ray photoelectron spectroscopy (XPS, Physical Electronics 5400 ESCA). Raman spectra were measured using a Renishaw InVia Microscope Raman (532 nm laser excitation). Absorption spectra were obtained on a Cary Win UV-visible spectrometer from 300 to 700 nm. Incident angle (θ, the angle of excitation light to normal through the center of TiO₂ nanocavity) and excitation polarization angle (ϕ, excitation polarization around the normal through the center of TiO₂ nanocavity) dependent absorption spectra were tested by rotating the sample from 0–180° with a fixed incident angle or rolling over the sample from a vertical direction to a parallel one. Fourier transform infrared spectroscopy (spectrum 100 FT-IR spectrometer, PerkinElmer) was employed for testing absorbance in the infrared region. The photoluminescence spectra were collected by a NanoLog Spec fluorescence spectrometer excited by a helium–cadmium lamp at 400 nm.

XPS fitting. All spectra were analyzed with the CasaXPS software (version 2.3.15). The samples were conductive so the binding energies were not charge referenced. Shirley background subtraction was used for all spectra. Peak models for each photoelectron line were generated using nonlinear least-squares Gaussian/Lorentzian curve-fitting. The S 2p line was fitted with multiple sets of doublets and constrained to have the same FWHM within each pair as well as a 2 : 1 intensity ratio and 1.16 eV separation between the 2p_3/2 and 2p_1/2 peaks. Similarly, the Mo 3d line was fitted with multiple sets of doublets as well as a 3 : 2 intensity ratio and 3.3 eV separation between the 3d_5/2 and 3d_3/2 peaks.

2.3 Photocatalysis characterization

Photocatalytic H₂ evolution. The photocatalytic H₂ production was performed in a sealed 20 ml quartz reactor. MoS₂@TiO₂ films with a size 7 × 7 mm were submerged in 15 ml of mixed solution made of DI water (Sea water) and methanol (8 : 2 by volume). Subsequently, the reactor was illuminated by a solar light simulator (AM 1.5, 300 W Xe, 100 mW cm⁻²) or monochromatic light (Zahner CIMPS-QEIPCE system with a monochromator and light source from 350 to 800 nm). The gas produced from the upper space above the solution in the quartz reactor was periodically analyzed.

2.4 Computational method

Finite element method simulation (FEM). Comsol Multiphysics (version 5.2) was used for FEM Solution. The 3D simulation model was designed as a simplified heterostructure, with three MoS₂ nanorods (diameter: 10 nm; length: 20 nm, parallel to y-axis) standing vertically on TiO₂ (diameter: 50 nm; length: 70 nm, parallel to z-axis). The incident laser wavelength was set from 300–700, with the polarization direction along the x-axis. The electric field distributions of the hybrid nanocavity arrays were monitored across the middle of the TiO₂ nanotube in the x–z and y–z planes.

Band structure calculation of MoS₂. First-principles calculation based on density functional theory (DFT) was used to estimate the electronic property of the MoS₂ layer upon S-vacancy deviation (Vienna ab initio simulation program, VASP). The ion electron interactions were depicted by the projector-augmented wave method. The generalized gradient approximation (GGA) was adopted with the Perdew–Burke–Ernzerhof (PBE) exchange–correlation function. 4 × 4 unit cells were employed as a calculation model with S-vacancies of 0%, 8%, and 16% respectively. The band structure was relaxed by a 6 × 6 × 1 special k-point mesh grid with an 880 eV plane-wave cut off energy.

3. Results and discussion

3.1 Morphology, microstructure and component analysis

A typical fabrication route of a MoS₂@TiO₂ plasmonic heterostructure is schematically illustrated in Fig. 1a. Highly-ordered honeycomb-shaped TiO₂ nanocavities are obtained through Ti anodization (Fig. S1a, ESI† with an average pore size of 50 nm and a wall thickness of 10 nm). E-beam evaporation was performed to deposit 10, 20, and 30 nm of Mo onto the anodized TiO₂ nanocavity arrays. Obvious cavity wall thickening (increased from 10 nm to 20 nm) is found after Mo coating (Fig. S1b, ESI†).

Fig. 1 (a) Schematic illustration of the fabrication process of the MoS₂@TiO₂ heterostructure. (b–d) Top-view SEM images of MoS₂₍₁₀₎@TiO₂, MoS₂₍₂₀₎@TiO₂ and MoS₂₍₃₀₎@TiO₂, respectively (scale bar: 100 nm).

CVD sulfurization was carried out at 400 °C for 10 min on TiO₂ coated with different Mo thicknesses (abbreviated as MoS₂₍₁₀₎@TiO₂, MoS₂₍₂₀₎@TiO₂, MoS₂₍₃₀₎@TiO₂, respectively). It was observed that the pore size significantly shrunk after CVD treatment when the deposited Mo thickness increased from 10 to 20 and 30 nm (Fig. 1b–d and Fig. S1c, ESI†). Transmission electron microscopy (TEM) of the MoS₂@TiO₂ nanocavity arrays shows that MoS₂ nanoflakes are grown inside TiO₂ nanocavities with a highly-ordered 3D laminate structure (Fig. 2a–c). These MoS₂ nanoflakes typically consist of less than 10 layers, which was verified by Raman spectra as shown in Fig. S4b (ESI†). Some of the MoS₂ nanoflakes stand vertically on the TiO₂ surface or connect with the TiO₂ nanocavity walls with a big intersection angle (Fig. 2d, e, Fig. S2d and e, ESI†).

Fig. 2 TEM images of (a and d) MoS₂₍₁₀₎@TiO₂, (b and e) MoS₂₍₂₀₎@TiO₂, and (c) MoS₂₍₃₀₎@TiO₂. (f) HRTEM image of MoS₂₍₁₀₎@TiO₂ showing few-layer MoS₂ nanoflakes erect on TiO₂ cavity walls. Inset (e) shows the Fast Fourier transform (FFT) pattern taken from the red square in (e), corresponding to the (002) facets of 2H-MoS₂. (g) Filtered and colored atomic-resolution BF-STEM image of the MoS₂ nanoflakes. The dark and light dots correspond to Mo and S atoms, respectively, where the S atom depletion area is observed with weaker contrast (indicated by blue arrows and dash circles). An atomic model with S-vacancies along the b-axis of 2H-MoS₂ is inserted beside (g). (a–d: scale bar is 20 nm; e and f: scale bar is 5 nm; g: scale bar is 1 nm).

High resolution TEM (HR-TEM) identifies lattice fringes of 0.62 nm and 0.35 nm (Fig. 2f), corresponding to the (002) hexagonal facets of MoS₂ and the (101) facets of anatase TiO₂, respectively. It is also revealed that the MoS₂ nanoflakes are bound seamlessly to the TiO₂ nanocavity wall surfaces, indicating a perpendicular growth of MoS₂ nanoflakes on the TiO₂ surface (refer to Fig. S2d, e and S3, ESI,† for more TEM images showing the perpendicular growth of MoS₂). Spherical aberration corrected high angle annular-dark-field (HAADF) scanning transmission electron microscopy (STEM) and bright-field (BF) STEM were performed to clarify the atomic structure of MoS₂. As shown in Fig. 2g, the image contrast of BF-STEM exhibits a relationship with respect to the atomic number (Z), therefore the sandwich structure of MoS₂ is clearly observed. In addition, the S-vacancies can be identified, as marked by blue dashed circles and blue arrows.²⁵ Interestingly, it seems that the S-vacancies exhibit local ‘staging’ structures, as seen for most S atoms throughout the structural layers. Where the blue arrows are absent, only a small number of S-vacancies are observed, indicated by the red arrows marked in Fig. 2g. The HAADF-STEM image and corresponding filtered image are shown in Fig. S2a and b (ESI†). Conversely, S-vacancies are displayed as darker spots at the edge of the S–Mo–S layer (red square). Fig. 3a and b show the cross-sectional STEM images of MoS₂@TiO₂ heterojunctions, demonstrating that a fully filled MoS₂ laminated network inside the TiO₂ nanocavities was successfully fabricated. Energy dispersive X-ray spectrometry (EDS) mapping analysis shows the distribution of each element at the interface of the junction (Fig. 3c–f). The separation assignment of Ti to Mo and S further proves the isolated vertical connection of MoS₂ and TiO₂ at the heterojunction.

Fig. 3 (a and b) Cross-sectional HAADF-STEM images of the MoS₂@TiO₂ heterostructure. (c–f) EDS elemental mapping taken from (b). The distribution of Ti is separated from Mo and S (a: scale bar 200 nm, b: scale bar 50 nm).

In a normal S steam based CVD process, Mo conversion takes place almost spontaneously upon S arriving at the metal surface, making it difficult to control the MoS₂ morphology.¹⁹ One method to allow more morphological control is to slow S evaporation using a liquid precursor containing S_x²⁻ ions (see more details in the Experimental section) and combined with a low temperature ramping rate (2 °C min⁻¹). The S source derived from the Na₂S_x solution first arrives at the hollow TiO₂ nanocavities and then will transmit gradually along the cavity walls.⁷ As it is understood, the growth direction of MoS₂ nanoflakes is synchronous with S diffusion, from the middle of the TiO₂ cavity to the edges.²³ Eventually, a vertically laminated MoS₂@TiO₂ heterostructure was formed inside the TiO₂ cavities. A photograph of the MoS₂₍₁₀₎@TiO₂ heterostructure provided in Fig. S1g (ESI†) shows uniform morphologies at different sample locations, indicating highly uniform MoS₂@TiO₂ throughout the entire sample. Control experiments were carried out by using a normal solid S source, flat TiO₂ film or Mo foil in duplicate reactions. Larger MoS₂ flakes with disordered architecture were obtained for all control experiments (Fig. S1d–f, ESI†), further verifying the significant impact made by the addition of TiO₂ nanocavity arrays and using liquid S source toward the inhibited MoS₂ overgrowth that promotes the formation of a MoS₂@TiO₂ heterostructure configuration.

XRD patterns of as-prepared MoS₂@TiO₂ films are shown in Fig. S4a (ESI†). The main diffraction peaks are indexed to the Ti substrate (PDF file No. 44-1294) because of the thin thickness of the TiO₂ film and low MoS₂ content. Noticeably, a weak peak at 14.5°, corresponding to the c-plane (002) of 2H-MoS₂ (S–Mo–S graphene-like layer, PDF file No. 86-2308), can be observed and intensifies with Mo thickening. Diffraction patterns of Ti substrates are a little different by the orientation diversity of commercial Ti foil, depending on the manufacturing process. Raman spectroscopy is more distinct for demonstrating the MoS₂ layered structure (Fig. S4b, ESI†). Peaks around 378 and 402 cm⁻¹, corresponding to in-plane E_2g and out-of-plane A_1g vibration modes of 2H-MoS₂, are dominant in the spectra. Red-shift of E_2g and blue-shift of A_1g occur with increasing Mo thickness from 10 to 30 nm as well as prolonging sulfurization time from 10 to 50 min for the MoS₂₍₃₀₎@TiO₂ film. The frequency difference monotonically increased from 21.25 cm⁻¹ to 24.24 cm⁻¹, which is in excellent agreement with the literature report indicating a thickness of less than 10 stacking layers of MoS₂ per flake.¹⁵ The greater intensity ratio of A_1g to E_2g is due to the large amount of surface edge exposure and strong interlayer restoring force interactions acting on misalignments or defects in the MoS₂ layers.²⁶ It is very clear that no Mo oxide can be found from both XRD and Raman, indicating a complete conversion from Mo to MoS₂ after CVD treatment. Experimental tests also indicate that the effect of Mo oxide on the optical and photocatalytic properties can be ignored, which will be further discussed in the following sections.

Surface defects, chemical states and the stoichiometric ratio of Mo and S in the MoS₂@TiO₂ heterostructure were investigated by X-ray photoelectron spectroscopy (XPS) analysis. 2H-MoS₂ and TiO₂ were confirmed by XPS results (Fig. S5, ESI†). The high-resolution S 2p spectrum demonstrates that there are three overlapping chemical states of S (Fig. S5b and c, ESI†). The doublet centered at 161.7 and 163.2 eV are assigned to S 2p in a Mo–S configuration, and the upshifted peaks at 163.4, 164.9, 166.6, 168.1 and 169.9 eV are due to the existence of S, Mo–O–S bands or oxidized S on the surface of MoS₂.²⁷ More oxidized S was found with an increase in the amount of MoS₂. Binding energies for Mo 3d_3/2 and Mo 3d_5/2 are fitted to a pair of doublets at 227.1/230.4 eV and 228.9/232.2 eV, respectively, which confirms the Mo³⁺ and Mo⁴⁺ accordingly (Fig. S5d and e, ESI†).²⁸ The existence of Mo³⁺ ions implies the formation of S defects in MoS₂ and accounts for about 20–30% of the total Mo element semi-quantitatively estimated from the XPS spectral area in the MoS₂₍₁₀₎@TiO₂ sample (S-vacancies are in the range of 5–8% of S atom deficiency). Mo⁶⁺ ions can be slightly detected because of surface oxidation during in-air storage. Combining the aforementioned investigations, it can be concluded that the S-vacancy and nonstoichiometric features of MoS₂ nanoflakes are where free electrons and plasmonic resonance processes originate. The XPS spectral states of Ti and O are in agreement with crystalline TiO₂ (Fig. S5f and g, ESI†).²⁹

3.2 Spectral response and photocatalytic performance

The solar light harvesting of MoS₂@TiO₂ heterostructures was investigated through UV-vis absorption spectra, as shown in Fig. 4a (incident angle: 0°). The intrinsic absorption edge located at about 360 nm belongs to the TiO₂ substrate, suggesting a band gap of 3.2 eV. Furthermore, very broad peaks ranging from 400–600 nm can be detected on the MoS₂@TiO₂ heterostructures. This visible light response is distinctly far from band gap excitation of both MoS₂ and TiO₂ portions within the heterostructured films. Consequently, a metal-like LSPR is assigned to this absorbance, which arises from collective oscillations of excess charges (electrons) on the edge of MoS₂. Abundant S-vacancies and highly ordered vertically laminated structures may dominate the free charge interactions.²⁷ It has been well recognized that the catalytic activities of both semiconducting 2H-MoS₂ and metallic 1T-MoS₂ generally arise from edge sites and S-vacancies.²⁶ Both edges and S-vacancies are considered responsible for crystal asymmetry of analytes, where the electronic structure is changed slightly. Electrons in a higher Fermi level are easy to transfer to locally collected vacancies on the surface of MoS₂. When the orbital density vibration of these electrons is resonant with incident light, photo-excited local dipoles and charge separation happen, forming a confined local field around the MoS₂ surface. Thereby S-vacancies and surface “hot” electrons are applied to plasmonic antennas, which results in modulation of the Fermi level and drastically changes the initial electronic properties of MoS₂.

Fig. 4 (a) Optical absorption of Mo@TiO₂, MoO₃@TiO₂ and MoS₂@TiO₂ heterostructures. (b) Incident angle-resolved plasmonic absorption of MoS₂₍₃₀₎@TiO₂. (c and d) H₂ evolution rate over pristine TiO₂ and MoS₂@TiO₂ films, and comparison of the H₂ evolution rate to MoS₂ geometric area (photocatalytic activity to seawater is inserted. All testing was carried out under simulated solar light). (e) Modulated H₂ yield rate as a function of the wavelength of monochromatic light irradiation. (f) Recycling photocatalytic H₂ evolution test of MoS₂₍₁₀₎@TiO₂ over 21 h.

Combining the plasmonic effect with intrinsic absorption of MoS₂@TiO₂, full-solar-spectrum harvesting is nearly achieved in this co-catalyst heterostructure. Additionally, the plasmonic resonance is demonstrated to be tunable upon Mo mass loading changes from 10 to 30 nm, with absorption peaks at 420, 480 and 510 nm, accordingly. This process may be similar to the classic size-dependent plasmonic effect.²⁹ The maximal light harvesting cross-section is obtained in the sample with 30 nm of Mo loading. Control experiments carried out on a pure metal based film (30 nm Mo on TiO₂) and an oxidized sample (MoO₃ on TiO₂) give little visible light coverage, but for the UV portion from the TiO₂ substrate (Fig. 4a). Incident angle and excitation polarization angle dependent absorption spectra were also tested on MoS₂₍₃₀₎@TiO₂ (Fig. 4b and Fig. S6a, ESI†). The resonance wavelengths are independent of the polarization angle (Fig. S6a, ESI†). All the absorption peaks of MoS₂₍₃₀₎@TiO₂ locate in the range of 450–500 nm at an incident angle of 15°. The absorbance fluctuation (cross-section of curves) may come from the nonuniformity of MoS₂ flakes inside TiO₂ (as revealed by TEM). Absorption spectra measured at different incident angles are presented in Fig. 4b (from bottom to top: 0, 30, 60, and 90°, respectively). The absorption peak position increases from 490 nm to 600 nm with the increase of θ from 0 to 90°. The absorption peak position can further shift backward when the incident angle decreases to 0°. This dynamic shifting indicates that the resonance has 90° difference in the incident angle dependence, suggesting that the orientations of MoS₂ flakes affect the resonance modes and charge distribution profiles in the hybrid catalyst. The excitation efficiencies of the samples are also dependent on incident angles. A maximum photo-efficiency (absorption cross-section) is achieved at an incident angle of 30°, consistent with the previous findings that the resonance intensity can be optimized at certain angles.³⁰ The relationship of nonstoichiometric features to light harvesting was verified by extending the sulfurization time from 10 min to 50 min on the MoS₂₍₃₀₎@TiO₂ film (Fig. S6b, ESI†). A strong absorbance degeneration can be observed with longer reaction times and complementary to the S source within the heterostructure, implying that S-vacancies are the main plasmonic charge donors.

Surface-enhanced Raman scattering (SERS) was used to further determine the enhanced electromagnetic field in the vicinity of MoS₂@TiO₂. Mercaptobenzoic acid (4-MBA) was used as the probe molecule with laser excitation at 630 nm. The dominant peaks (Fig. S6c, ESI†) at 1099 and 1594 cm⁻¹ are attributed to the vibration mode of the ν8a aromatic ring and the breathing mode of the ν12a ring, respectively. The weak peaks at 1293 and 1188 cm⁻¹ are the stretching mode of ν_(COO–) and deformation mode of C–H, respectively. The stronger SERS signal on the MoS₂@TiO₂ substrate (compared with bare TiO₂ nanocavity arrays using 10⁻⁵ mol L⁻¹ 4-MBA) demonstrates the plasmonic polarization presented on the MoS₂@TiO₂ surface.

The photocatalytic activity of a series of MoS₂@TiO₂ co-catalysts for H₂ evolution was estimated under simulated solar light illumination with methanol as a hole scavenger in DI water (Fig. 4c and d). The photocatalytic reaction kinetics and formula for producing H₂ are shown in Fig. S7 (ESI†). Pure TiO₂, sulfurized TiO₂ and a MoO₃@TiO₂ hybrid are photocatalytically inert toward hydrogen evolution, with a reaction rate of lower than 10 μmol h⁻¹. Noticeably, the photocatalytic activity is sharply enhanced with MoS₂ loading and achieves maximal values on the MoS₂₍₃₀₎@TiO₂, reaching 84 μmol h⁻¹ (≈8 times that of pure TiO₂). The photocatalytic activities of the samples increase non-linearly with increasing MoS₂ loading (from 10 to 30 nm Mo deposition) because of the relatively lower solar energy input via increasing catalyst amount.^31,32 Additionally, a deterioration is found with longer sulfurization time from 10 to 50 min (MoS₂₍₃₀₎@TiO₂), deducing a weakened plasmonic effect from the stoichiometric assignment, according to spectroscopy analysis (Fig. S6b, ESI†). The photocatalytic activities of the samples normalized according to the geometric area are shown in Fig. 4d. An optimal photocatalytic activity of 181 μmol h⁻¹ cm⁻² is observed on the heterostructure coated with 30 nm Mo. Fig. S8a (ESI†) shows the rate of H₂ production normalized by mass loading of MoS₂ (mass-measurement of the MoS₂ catalyst is given in the Experimental section). Samples with 10.2 μg MoS₂ have an activity of 580 mmol h⁻¹ g⁻¹, which is superior to the state-of-the-art photocatalytic systems (Table S1, ESI†). This may be ascribed to the highly-ordered architecture of MoS₂ nanoflakes on TiO₂ nanocavities with a lower mass loading, which reduces the electron–hole recombination. Furthermore, the LSPR induced electron energetic level is more negative in MoS₂₍₁₀₎@TiO₂ because of the blue-shifted light harvesting and more negative lowest unoccupied molecular orbital level that favours effective charge utilization rates in H⁺ reduction.²⁸ Stability and recyclability of the MoS₂₍₁₀₎@TiO₂ co-catalyst were estimated by repeating intermittent H₂ evolution under simulated solar light (Fig. 4f). 90% of incipient production can be kept for the hybrid films even after 21 h. Control experiments were carried out on the samples using MoS₂ deposited on compact TiO₂ and Mo foils (MoS₂₍₃₀₎@TiO_2(compact) and MoS₂@Mo_(compact)) to further confirm the crucial role of TiO₂ nanocavity arrays in the improved photocatalytic stability and recyclability. We found that MoS₂ flakes fell off from these compact substrates during H₂ evolution testing, leading to a considerably deteriorated H₂ evolution. As shown in Fig. S8b and c (ESI†), approximately 30% of H₂ yield (less than 15 μmol h⁻¹) was kept for MoS₂@TiO_2(compact) after 5 cycles of testing. An even lower H₂ yield (about 20%, equal to 10 μmol h⁻¹) was observed on MoS₂@Mo_(compact) after 15 h of testing. These control experiments indicate that TiO₂ nanocavity arrays not only contribute to UV-light absorption but also serve as a host to immobilize the MoS₂ co-catalyst.

The photoluminescence (PL) spectrum is used to detect the charge carrier trapping, migration and transfer mechanism during photocatalytic reactions (Fig. S8d and e, ESI†). The steady-state PL spectra located at 595 nm (Fig. S8d, ESI†) characterize remarkable quenching along with Mo thickening from 10 nm to 30 nm, suggesting either a shorter lifetime or faster trapping of photo-electrons with increasing MoS₂ content. Transient-state PL decay curves of MoS₂@TiO₂ heterostructures are compared in Fig. S8c (ESI†). A bi-exponential function was used here for fitting the PL decay curves mathematically:³³

y = A₁ exp(−x/τ₁) + A₂ exp(−x/τ₂) + y₀, (1)

where A₁, A₂ and y₀ are amplitude coefficient and basal constant. τ₁ and τ₂ are fluorescent lifetimes corresponding to non-radiative recombination and inter-band recombination, respectively. The calculated carrier lifetime is inserted in Fig. S8e (ESI†) accordingly. The average fluorescent lifetime of MoS₂₍₁₀₎@TiO₂ (τ₁: 2.8 ns and τ₂: 14 ns) is longer than the other ones after integration by the equation:³³This indicates that appropriate MoS₂ loading and periodically patterned morphology provide the MoS₂₍₁₀₎@TiO₂ film with less trapping centers and prolonged retention of hot electrons, as well as enhanced H₂ production.

3.3 Computational study

The photo-electron arising mechanism, transfer pathway and consuming profile were computationally studied to assist in explaining the excellent photocatalytic activity of MoS₂@TiO₂ heterostructures (Fig. 5a). Firstly, the UV portion of solar light is mainly absorbed by the TiO₂ substrate, where photo-excited electrons transfer to the MoS₂ basal plane and diffuse to edge active sites. Here, the designed MoS₂@TiO₂ short nanocavity arrays and seamless junctional connection dramatically reduce the charge barriers at active sites. The perpendicular MoS₂–TiO₂ configuration benefits the electron transfer pathways at the basal plane where a lower resistance and suppressed quenching capture are expected. On the other hand, a broad LSPR band (ranging from 400 to 600 nm) makes the MoS₂@TiO₂ heterostructures suitable for visible light-driven H⁺ reduction. This plasmon-enhanced activity is presumably a result of the aforementioned S-vacancies (confirmed by STEM) and provides predominant aspects for H₂ evolution. Finally, photoelectrons excited on the MoS₂ conduction band from NIR-light, may also contribute to H₂ evolution. However, considerable carrier recombination is unneglectable in this narrow band gap semiconductor. MoS₂ edge defects (S-vacancies) play a dominant role in plasmonic resonance and hot electron excitation upon solar light irradiation, but will also capture free carriers and cause unwanted quenching. Moreover, MoS₂ and TiO₂ intrinsic recombination is another route for hot electron consumption in catalytic processes (dashed line between band gap), which may weaken photocatalytic performance.

Fig. 5 (a) Schematic diagram of the energy band structure, plasmonic resonance and electron transfer pathway in the MoS₂@TiO₂ heterojunction. (b) Band structure of monolayer 2H-MoS₂ with 0%, 8% and 16% S-vacancies, with the valence band maximum and conduction band minimum both at the K point. (c) The model used for band gap computation by DFT (4 × 4 unit cell). (d) FEM simulation of the near-field electric field distribution inside MoS₂@TiO₂ heterostructures excited by a 400 nm laser and (e) a 3D-simplified model used for FEM simulation.

To have an insightful understanding of the inter-band excitation of MoS₂ nanoflakes and eliminate the possibility of MoS₂ intrinsic activity in visible-light, the electronic band structure was calculated by density functional theory (DFT, Fig. 5b and Fig. S9a–c, details about the calculation method are given in the ESI†). S-depletion was introduced in the computation models with an atom content of 0%, 8%, and 16%, respectively, mainly at the edge position (Fig. 5c). A fundamental band gap of around 2.07 eV is found here for a perfect MoS₂ crystal, with the highest occupied valence band and the lowest empty conduction band located at 1.63 eV and −0.4 eV, accordingly (Fig. 5b). The band structure is modulated by introduction of a S-vacancy, which induces the narrowing of the band gap to 1.73 (8%) and 1.12 eV (16%). Moreover, defect states appear near to the Fermi level for the model with 16% S-depletion, which improves the metallic feature and charge mobility of S-depleted MoS₂ layers.²⁰ The overall electronic density of states (DOS) for MoS₂ before and after S-depletion are shown in Fig. S9a–c (ESI†). Defect states near to the Fermi level originate mainly from the d orbitals of Mo and p orbital of S atoms (Fig. S9c, ESI†), creating pseudo-ballistic electron transport channels within MoS₂. This result is further supported by other literature and fully compatible with our optical spectra in the IR region (Fig. S9d, ESI†), where the MoS₂ band gap absorption peaks are located around 882 nm (1.40 eV).³⁴ A further conclusion can be deduced from the discussion aforementioned that solar light caused activity of MoS₂@TiO₂ films mainly functionalized by the plasmonic effect surrounding the MoS₂ surface.

To further confirm this assumption, wavelength-dependent photocatalysis and finite element method (FEM) simulations were performed on MoS₂@TiO₂ heterostructures under monochromatic light illumination (350–700 nm, with a 50 nm interval). A spectrum dependent H₂ evolution rate is revealed over the UV-vis-NIR region, with a consistent variation in absorption spectra for the MoS₂@TiO₂ (Fig. 4a and e). The largest H₂ yield rate is about 59, 73, and 86 μmol h⁻¹ mW⁻¹ cm⁻² for Mo loading of 10, 20 and 30 nm, respectively, located at 450, 500, and 550 nm accordingly. The relative consistency of photo-absorption and H₂ production over solar light indicates that the photocatalytic activity enhancement is primarily driven by the plasmonic effect from blue-green wavelengths. Upon excitation by monochromatic light irradiation, the H₂ yield is found to be higher than that with the equivalent solar light (AM 1.5 100 mW cm⁻²) excitation, especially in the 400–600 nm region. This suggests an additional benefit for use of variant colored light for plasmonic material based H₂ generation. A diagrammatic simulation (FEM) of electric field distribution at MoS₂ nanoflake vicinities is shown in Fig. 5d and Fig. S10 (ESI†). A 3D model was designed using a single TiO₂ nanocavity (50 nm in diameter and 70 nm in length) and three vertically loaded MoS₂ nanorods (10 nm in diameter and 20 nm in length) on both sides (Fig. 5e). It was found that light irradiation coupled with LSPR yields strong electric field enhancements at the tip of MoS₂. The field intensity increases with wavelength from 300 to 400 nm and reduces significantly after 500 nm, correlating with the LSPR absorption spectra of the MoS₂₍₁₀₎@TiO₂ film (Fig. 4a). The maximum electric field enhancement is produced at 400 nm illumination (Fig. 5d). Field intensity between each antinode covers a distance longer than 5 nm, achieving zero-gap field distribution between the MoS₂ interspace (<10 nm). Consequently, a 3D electric field distribution is built throughout the networked MoS₂ laminate heterostructure and responds strongly to photocatalytic H₂ production.

Prospectively, H₂ generation from seawater will be highly desirable. However, few studies have focused on practical photocatalysis on natural seawater because of the barrier blocking of dissolved salt for photocatalytic activity and durability of catalysts.³⁵ Here we investigated the effect of salt (mainly NaCl) on the photocatalytic activity of MoS₂@TiO₂ films with natural seawater (pH: 8.4) splitting under simulated solar light. As expected, the activity from seawater decreased markedly compared to that of pure water (Fig. 4c and d), because the isoelectric point of MoS₂ is lower than 7 due to the plasmonic “hot” electrons on the surface. The electrostatic adherence of hydroxyl groups and metal ions and the oxidation of sulfides take place severely in seawater with a weak alkaline environment. Fortunately, 60% of incipient production is kept for samples, where the highly-ordered architecture may be helpful. This control experiment permits our MoS₂@TiO₂ heterostructure as a promising material for H₂ production from seawater.

4. Conclusions

In summary, novel MoS₂@TiO₂ laminate heterostructures were synthesized by a facile process that combines anodization, PVD, and CVD. Such structures displayed intense LSPR throughout regulating the S stoichiometry of the MoS₂ surfaces. The MoS₂@TiO₂ co-catalysts showed nearly full-solar-spectrum absorption and superb photocatalytic activity for H₂ evolution. Samples with a 30 nm Mo coating enabled a H₂ yield rate as high as 181 μmol h⁻¹ cm⁻², based on the synergistic plasmonic effect with homogeneity promoting charge-carrier separation and higher conductivity of “hot” electrons in this highly ordered architecture. FEM simulation and DFT calculations were performed to understand this nonmetal plasmonic system and simultaneously derive light harvesting, charge separation and transfer dynamics. Such materials, structures, and design schemes can have significant impact in using plasmonic materials for cost effective solar energy conversion.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

This work was financially supported by the University of Central Florida through a startup grant (no. 20080741). STEM, EELS, and XPS data analysis were supported by the U.S. Department of Energy (DOE), Office of Science, Office of Basic Energy Sciences, Early Career Research Program under award # 68278. A portion of the work was performed at the W. R. Wiley Environmental Molecular Sciences Laboratory, a DOE User Facility sponsored by the Office of Biological and Environmental Research and located at the Pacific Northwest National Laboratory.

Notes and references

1. J. Zhang, X. Jin, P. I. Morales-Guzman, X. Yu, H. Liu, H. Zhang, L. Razzari and J. P. Claverie, ACS Nano, 2016, 10, 4496.
2. K. Liang, Y. Yan, L. Guo, K. Marcus, Z. Li, L. Zhou, Y. Li, R. Ye, N. Orlovskaya, Y.-H. Sohn and Y. Yang, ACS Energy Lett., 2017, 2, 1315.
3. X. Zhang, Y. L. Chen, R.-S. Liu and D. P. Tsai, Rep. Prog. Phys., 2013, 76, 046401.
4. Y.-C. Pu, G. Wang, K.-D. Chang, Y. Ling, Y.-K. Lin, B. C. Fitzmorris, C.-M. Liu, X. Lu, Y. Tong, J. Z. Zhang, Y.-J. Hsu and Y. Li, Nano Lett., 2013, 13, 3817.
5. L. Guo, Z. Li, K. Marcus, S. Navarro, K. Liang, L. Zhou, P. D. Mani, S. J. Florczyk, K. R. Coffey, N. Orlovskaya, Y.-H. Sohn and Y. Yang, ACS Sensors, 2013, 2, 621.
6. H. He, J. Lin, W. Fu, X. Wang, H. Wang, Q. Zeng, Q. Gu, Y. Li, C. Yan, B. K. Tay, C. Xue, X. Hu, S. T. Pantelides, W. Zhou and Z. Liu, Adv. Energy Mater., 2016, 6, 1600464.
7. W. Zhou, Z. Yin, Y. Du, X. Huang, Z. Zeng, Z. Fan, H. Liu, J. Wang and H. Zhang, Small, 2013, 9, 140.
8. K. Liang, L. Li and Y. Yang, ACS Energy Lett., 2017, 2, 373.
9. S. Gao, G.-D. Li, Y. Liu, H. Chen, L.-L. Feng, Y. Wang, M. Yang, D. Wang, S. Wang and X. Zou, Nanoscale, 2015, 7, 2306.
10. Y. Liu, Y.-X. Yu and W.-D. Zhang, J. Phys. Chem. C, 2013, 117, 12949.
11. H. Fei, Y. Yang, X. Fan, G. Wang, G. Ruan and J. M. Tour, J. Mater. Chem. A, 2015, 3, 5798.
12. Y. Yang, H. Fei, G. Ruan, Y. Li and J. M. Tour, Adv. Funct. Mater., 2015, 25, 6199.
13. X. Liu, Z. Xing, H. Zhang, W. Wang, Y. Zhang, Z. Li, X. Wu, X. Yu and W. Zhou, ChemSusChem, 2016, 9, 1118.
14. M. J. Islam, D. A. Reddy, N. S. Han, J. Choi, J. K. Song and T. K. Kim, Phys. Chem. Chem. Phys., 2016, 18, 24984.
15. L. Liang and V. Meunier, Nanoscale, 2014, 6, 5394.
16. M. Ye, D. Winslow, D. Zhang, R. Pandey and Y. K. Yap, Photonics, 2015, 2, 288.
17. J. Shi, Y. Zhang, Y. Hu, X. Guan, Z. Zhou and L. Guo, J. Alloys Compd., 2016, 688, 368.
18. Q. Li, N. Zhang, Y. Yang, G. Wang and D. H. Ng, Langmuir, 2014, 30, 8965.
19. M. M. Ali and K. N. Y. Sandhya, New J. Chem., 2016, 40, 8123.
20. H. Li, C. Tsai, A. L. Koh, L. Cai, A. W. Contryman, A. H. Fragapane, J. Zhao, H. S. Han, H. C. Manoharan, F. Abild-Pedersen, J. K. Nørskov and X. Zheng, Nat. Mater., 2016, 15, 48.
21. Y. Kang, S. Najmaei, Z. Liu, Y. Bao, Y. Wang, X. Zhu, N. J. Halas, P. Nordlander, P. M. Ajayan, J. Lou and Z. Fang, Adv. Mater., 2014, 26, 6467.
22. U. Ahuja, A. Dashora, H. Tiwari, D. C. Kothari and K. Venugopalan, Comput. Mater. Sci., 2014, 92, 451.
23. I. Minguez-Bacho, M. Courté, H. J. Fan and D. Fichou, Nanotechnology, 2015, 26, 185401.
24. H. Cheng, X. Qian, Y. Kuwahara, K. Mori and H. Yamashita, Adv. Mater., 2015, 27, 4616.
25. L. P. Hansen, Q. M. Ramasse, C. Kisielowski, M. Brorson, E. Johnson, H. Topsøe and S. Helveg, Angew. Chem., Int. Ed., 2011, 50, 10153.
26. Y. Yang, H. Fei, G. Ruan, C. Xiang and J. M. Tour, Adv. Mater., 2014, 26, 8163.
27. Z. Zhang, J. Huang, Y. Fang, M. Zhang, K. Liu and B. Dong, Adv. Mater., 2017, 29, 1606688.
28. L. Lin, N. Miao, Y. Wen, S. Zhang, P. Ghosez, Z. Sun and D. A. Allwood, ACS Nano, 2016, 10, 8929.
29. L. Guo, K. Liang, K. Marcus, Z. Li, L. Zhou, P. D. Mani, H. Chen, C. Shen, Y. Dong, L. Zhai, K. R. Coffey, N. Orlovskaya, Y.-H. Sohn and Y. Yang, ACS Appl. Mater. Interfaces, 2016, 8, 34970.
30. J. Wu, X. Lu, Q. Zhu, J. Zhao, Q. Shen, L. Zhan and W. Ni, Nano-Micro Lett., 2014, 6, 372.
31. K. R. Catchpole and A. Polman, Opt. Express, 2008, 16, 21793.
32. J. H. Heo, D. H. Song, B. R. Patil and S. H. Im, Isr. J. Chem., 2015, 55, 966.
33. H. S. Jung and N.-G. Park, Small, 2015, 11, 10.
34. Y. Han, T. Hu, R. Li, J. Zhou and J. Dong, Phys. Chem. Chem. Phys., 2015, 17, 3813.
35. Y. Li, D. Gao, S. Peng, G. Lu and S. Li, Int. J. Hydrogen energy, 2011, 36, 4291.

---

Footnote

† Electronic supplementary information (ESI) available: Morphological and structure characterization, photo-response and photo-catalysis performance. See DOI: 10.1039/c7ee02464a

---