Operando Raman Spectroscopy of Amorphous Molybdenum Sulfide

Oct 18, 2016 - Hyunho NohChung-Wei KungKen-ichi OtakeAaron W. .... T. Le Mogne , C. Charrin , S. Loridant , C. Geantet , P. Afanasiev , B. Thiebaut...
17 downloads 0 Views 3MB Size
Research Article pubs.acs.org/acscatalysis

Operando Raman Spectroscopy of Amorphous Molybdenum Sulfide (MoSx) during the Electrochemical Hydrogen Evolution Reaction: Identification of Sulfur Atoms as Catalytically Active Sites for H+ Reduction Yilin Deng,† Louisa Rui Lin Ting,†,‡ Perlin Hui Lin Neo,† Yin-Jia Zhang,§ Andrew A Peterson,∥ and Boon Siang Yeo*,†,‡ †

Department of Chemistry, Faculty of Science, National University of Singapore, 3 Science Drive 3, Singapore 117543 Solar Energy Research Institute of Singapore, National University of Singapore, 7 Engineering Drive 1, Singapore 117574 § Department of Chemistry, Brown University, 324 Brook Street, Providence, Rhode Island 02912, United States ∥ School of Engineering, Brown University, 184 Hope Street, Providence, Rhode Island 02912, United States ‡

S Supporting Information *

ABSTRACT: Amorphous molybdenum sulfide (MoSx) is currently being developed as an economically viable and efficient catalyst for the electrochemical hydrogen evolution reaction (HER). An important yet unsolved problem in this ongoing effort is the identification of its catalytically active sites for proton reduction. In this work, cyclic voltammetry (CV), X-ray photoelectron spectroscopy (XPS), and Raman spectroscopy were used to investigate the catalytically active sites and structural evolution of MoSx films during HER in 1 M HClO4 electrolyte. Transformation of anodically deposited MoSx (x ≈ 3) to a structure with MoS2 composition during the cathodic sweep of a CV was demonstrated using XPS and operando Raman spectroscopy. Interestingly, a Raman peak at 2530 cm−1 was recorded at potentials relevant to H2 evolution, which we ascribed to the S−H stretching vibration of MoSx−H moieties. This assignment was corroborated by H/D isotope exchange experiments. Mo−H (or Mo−D) stretching vibrations were not observed, which thus allowed us to rule out Mo centers as catalytic sites for proton reduction to H2. Density functional theory (DFT) calculations were performed on a variety of MoSx structures to capture the heterogeneous nature of amorphous materials and corroborated the assignments of the observed vibrational frequencies. On the basis of these experimental measurements and quantum chemical simulations, we have for the first time directly pinpointed the sulfur atoms in amorphous MoSx to be the catalytically active sites for evolving H2. KEYWORDS: hydrogen evolution reaction, electrocatalyst, amorphous molybdenum sulfide, active site, Raman spectroscopy, X-ray photoelectron spectroscopy, density functional theory

1. INTRODUCTION The electrochemical splitting of water to H2 and O2 gas using renewable solar electricity is one of the sustainable and environmentally friendly ways of producing carbon-neutral transportable fuels.1−3 The most effective cathode material to catalyze the hydrogen evolution (HER, 2H+ + 2e− → H2) is platinum.3 However, the scarcity and high cost of Pt have prevented its large-scale application in industry. Numerous efforts have thus been devoted to the development of cheaper but highly efficient electrocatalysts for the HER. Inspired by the FeMo cofactor in nitrogenase enzymes in which the bridging S ligands of Mo were found active for proton reduction, Mo−Sbased materials have been investigated as possible alternatives.4−8 Among them, amorphous MoSx species have attracted great attention due to their ease of preparation and efficacious activities toward water reduction. At a modest loading of 0.2 mg/cm2 and overpotential (η) of 160 mV, they © XXXX American Chemical Society

can catalyze H2 production with geometric current densities up to 10 mA/cm2.9,10 The structures of amorphous MoSx are generally agreed to be polymeric aggregations of MoIV3 clusters, with bridging S22−, terminal S22−, unsaturated S2−, and apical S2− ligands (Figure 1).11 An important question was unavoidably raised during the development of these catalysts for HER: which sites are active for the reduction of H+ to H2? Mo or S, or both? Hu and coworkers first hypothesized that the coordinatively unsaturated S atoms in MoSx could adsorb hydrogen and promote their recombination to H2.12 Subsequently, in situ X-ray absorption spectroscopy on the MoSx films suggested that their terminal S22− ligands were involved in the HER.13 On the other hand, Received: July 1, 2016 Revised: October 3, 2016

7790

DOI: 10.1021/acscatal.6b01848 ACS Catal. 2016, 6, 7790−7798

Research Article

ACS Catalysis

2.1.1. Anodically Deposited MoSx Films (MoSx-AE). A threeelectrode setup was used. Hg/HgSO4 (saturated K2SO4, CHI 151, CH Instruments) and a platinum (Pt) wire (99.99%, Sigma-Aldrich) were respectively employed as reference and counter electrodes. MoSx-AE films were electrodeposited from a plating solution onto Mo or GC substrates at 0.7 V vs RHE for 1000 s (RHE = reversible hydrogen electrode; all potentials in this work are cited with respect to the RHE). The plating solution contained 2 mM (NH4)2[MoS4] (99.97%, SigmaAldrich) in aqueous 0.1 M KCl (GCE Laboratory Chemicals).18 2.1.2. Cathodically Deposited MoSx Films (MoSx-CE). The parameters were the same as those for preparing MoSx-AE, except at the different deposition potential of −0.4 V. 2.2. Materials Characterization. The surface morphologies and compositions of the MoSx films were examined using a scanning electron microscope (SEM, JEOL JSM-6701F) coupled with an energy dispersive X-ray analyzer (EDX, JEOL JED-2300). The Mo:S ratio presented in this work is an average of 10 independent measurements on the films. Analysis of surface electronic structure by X-ray photoelectron spectroscopy (XPS) was performed in a Kratos AXIS UltraDLD (Kratos Analytical Ltd., chamber pressure 5 × 10−9 Torr) using Al Kα radiation (hν = 1486.71 eV, 5 mA, 15 kV). Calibrations of the binding energies were based on the C1s peak at 285.0 eV. Curve fitting parameters are shown in section S1 of the Supporting Information. 2.3. Electrochemical/Raman Measurements. A round Teflon dish cell was used for electrochemical/Raman measurements (section S2 of the Supporting Information).19,20 Hg/ HgSO4 (saturated K2SO4, CHI 151, CH Instruments) was employed as the reference electrode. Unless specifically stated, a Pt wire rather than a graphite rod (Pine Instrument) was used as the counter electrode. This is because graphite could corrode during oxygen evolution (the counter reaction) to give amorphous carbon, which may diffuse and adsorb onto the working electrode.21,22 Amorphous carbon has a very large Raman scattering cross-section, and its signals would easily overwhelm those from the analyte.23 The working electrodes were sheathed in Teflon. Aqueous 1 M HClO4 (70%, SigmaAldrich) and DClO4 (94.5 atom % D, Cambridge Isotope Laboratories, section S3 of the Supporting Information) were used as electrolytes. Electrochemical data were acquired using a Metrohm Autolab PGSTAT30 potentiostat. Potential error compensation was made using the positive feedback mode (at 90% level for 5 Ω). All currents reported in this work were normalized to the exposed geometric surface area of the working electrode (0.50 cm2). Raman spectroscopy was performed using a confocal Raman microscopy system (Horiba Jobin Yvon).20 It consisted of an epi-illumination microscope (Olympus U-5RE-2) joined to a spectrometer (iHR 320) with a charge coupled device detector (Synapse). An argon laser (Modu-Laser Stellar-Pro) provided the Raman excitation at 514 nm. A water immersion objective (Olympus LUMFL, 60×, numerical aperture 1.10) protected with an optically transparent Teflon film (0.013 mm thin, American Durafilm) was used for operando measurements during electrolysis. An air objective (Olympus MPlan N, 50×, numerical aperture 0.75) was used for measurements in air. 2.4. Calculation of Turnover Frequencies for HER. The procedure for calculating the TOF of H2 evolution has been described in our previous work.15 The number of active sites in the catalysts was estimated by integrating the oxidation peaks in

Figure 1. Amorphous MoSx constructed from Mo3 cluster units. The type of S ligands present are indicated by different colors: green, bridging S22−; yellow, terminal S22−; red, unsaturated S2−; blue, apical S2−.

Tran et al. postulated otherwise. They reported that the terminal S22− ligands of MoSx were first eliminated during reduction at HER-relevant potentials, which then allowed the formation of Mo−H active species.14 Recently, we discovered a linear correlation between the turnover frequency of H2 production and the percentage of S atoms with higher electron binding energies (measured by Xray photoelectron spectroscopy).15 These active S atoms with higher electron binding energies are likely to be bridging S22− atoms, as shown from density functional theory calculations within the same work. We have also demonstrated through in situ Raman spectroscopy of the [Mo3S13]2− model compound that the intensity of its S−S stretching vibration (ν(S−S)) changed during the HER, which thus indicated its involvement in the HER. These results are consistent with studies made using [Mo3S13]2− or [Mo2S12]2− complexes, which invoked S as active sites.16,17 However, while these aforementioned works are highly meritorious, they only provide indirect evidence to demonstrate the involvement of Mo or S in the HER. A direct detection of either Mo−H or S−H moieties has never been made. The objective of this work is to spectroscopically identify the HER-active sites in amorphous MoSx catalysts. To this end, we have designed a combined cyclic voltammetry, X-ray photoelectron, and operando Raman spectroscopy study of these films during the HER. Raman spectroscopy was chosen, as it can give the molecular fingerprints of the species present. H/D isotopic exchange experiments and density functional theory calculations on a large series of Mo−S motifs were further made to corroborate the peak assignments.

2. EXPERIMENTAL SECTION 2.1. Preparation of MoSx Catalysts. For the X-ray photoelectron spectroscopy, cyclic voltammetry (CV), and Raman spectroscopy analyses, the working electrode substrates were Mo disks (diameter 10 mm, 99.9%, Goodfellow), which were polished to a mirrorlike finish using diamond slurries (Struers). Glassy-carbon (GC) disks (diameter 10 mm, Goodfellow) were used as substrates for scanning electron microscopy (SEM), energy dispersive X-ray spectroscopy (EDX), and electrochemical measurements for the turnover frequency (TOF) calculations. The glassy-carbon disks were polished using alumina powder suspension (Struers). Only ultrapure water (18.2 MΩ cm, Barnstead, Thermo-Fisher Scientific) was used for preparing the electrolytes and for washing. 7791

DOI: 10.1021/acscatal.6b01848 ACS Catal. 2016, 6, 7790−7798

Research Article

ACS Catalysis

Figure 2. SEM images of freshly prepared (a(i)) MoSx-AE and (b(i)) MoSx-CE. Cyclic voltammograms (cathodic sweep first) of (a(ii)) MoSx-AE and (b(ii)) MoSx-CE at 10 mV/s in 1 M HClO4. The CV of the bare Mo substrate acquired under the same conditions is included. Insets show enlarged views of the same curves in (a(ii)) and (b(ii)). Cyclic voltammograms (cathodic sweep first) of (a(iii)) MoSx-AE and (b(iii)) MoSx-CE at 10 mV/s in 1 M HClO4 and 1 M DClO4 (containing 94.5 atom % D) electrolytes. Insets show the Tafel plots obtained from CVs collected at 0.5 mV/s (section S6 of the Supporting Information).

MoS2 slabs (Mo edge with 0.5 ML S monomers) after their geometric configurations were optimized. Both the semiconductive 2H phase and the conductive 1T phase slabs were investigated.27−29 As the 1T phase is metastable and tends to convert into a zigzag-distorted 1T′ phase,30−33 we have used the 1T′ phase as the T phase model in the current work. A detailed description of other structures can be found in our previous work.15 Different k-point samplings were used due to different periodic boundary conditions: 1 × 1 × 1 for clusters, 4 × 1 × 1 for polymers, and 4 × 4 × 1 for slabs.

their cyclic voltammograms (section S4 of the Supporting Information). As Mo disks would oxidize during the anodic sweep of the CV, glassy-carbon disks were used instead as substrates for the MoSx films. The TOF of each catalyst was calculated by normalizing its HER current (obtained from its linear sweep voltammogram) with the number of active sites present. The TOF values presented in this work were an average of three independent measurements on each catalyst. 2.5. Electronic Structure Calculations. All density functional theory (DFT) calculations were performed with the GPAW (grid-based projector-augmented wave) calculator,24,25 with the RPBE exchange-correlation functional of Hammer, Hansen, and Nørskov,26 a 0.1 eV Fermi smearing temperature, a 0.18 Å real-space grid spacing, and the eigensolver of RMM-DIIS (residual minimization method− direct inversion in iterative subspace). A quasi-Newton gradient descent method was used to find the optimal geometric configurations until the maximum force on any unconstrained atom was less than 0.05 eV/Å. Vibrational modes of hydrogen adsorbates were investigated in the harmonic approximation by performing normal-mode analyses on the adsorbates, under the assumption that the substrate vibrational modes were independent of the adsorbate modes. The vibrational frequencies of adsorbed H on Mo and S were calculated on MoSx clusters, MoSx 1-D polymers, and

3. RESULTS 3.1. SEM and EDX Characterizations of As-Synthesized MoSx-AE and MoSx-CE. SEM analyses showed that both anodically (MoSx-AE) and cathodically (MoSx-CE) electrodeposited films consisted of 20−30 nm nanoparticles (Figure 2a(i),b(i)). EDX revealed the Mo:S atomic ratios in MoSx-AE and MoSx-CE films to be [1.0(±0.1)]:[3.1(±0.1)] and [1.0(±0.1)]:[1.9(±0.1)], respectively. These observations were consistent with the morphologies and elemental stoichiometries of similarly prepared MoSx catalysts.9,10,15 3.2. Electrochemistry of MoSx-AE and MoSx-CE. Freshly prepared MoSx-AE and MoSx-CE catalysts were subjected to cyclic voltammetry scans in 1 M HClO4 electrolyte (Figure 7792

DOI: 10.1021/acscatal.6b01848 ACS Catal. 2016, 6, 7790−7798

Research Article

ACS Catalysis

Smaller H2 currents of −1.7 and −1.5 mA/cm2 at η = 200 mV were also observed for MoSx-AE and MoSx-CE films, respectively. Furthermore, larger Tafel slope values of 47−49 mV/dec were measured. These observations are in agreement with slower kinetics associated with the H/D isotope effect during the hydrogen evolution reaction.39 3.3. X-ray Photoelectron Spectroscopy of MoSx-AE and MoSx-CE. The XPS spectra of freshly prepared MoSx-AE and MoSx-CE are presented in Figure 3. The Mo 3d peak can

2a(ii),b(ii)). MoSx-AE exhibited a strong reduction peak at −0.07 V during the cathodic scan in the first CV (labeled with R in Figure 2a(ii)).10 This feature has been reported in previous cyclic voltammograms on the same type of films and has been attributed to the irreversible removal of S atoms which changed the chemical composition of MoSx-AE from MoS3 to MoS2.9,10,15 Hu and co-workers have considered this step to be an activation process for the anodically deposited MoSx catalyst to work for the HER.9,10 While the physical and chemical natures of the activation have not been fully elucidated, it is plausible that the reduction/dissolution of HER-inactive surface Mo oxides (formed during the deposition of MoSx) could be involved.10,18 H2 evolution occurred on the MoSx-AE catalyst from potentials cathodic to −0.15 V. At η = 200 mV, a H2 current of −5 mA/cm2 was recorded. During the reverse (anodic) scan, two hitherto-unreported anodic peaks were recorded at −0.07 and ∼+0.2 V (inset of Figure 2a(ii)). Subsequent CV scans further revealed two pairs of broad redox peaks at +0.2 V (labeled with A and A′) and −0.07 V (B and B′). Control experiments using pristine Mo disks demonstrate that these peaks were not due to the underlying Mo substrate. These peaks were also not related to Pt contaminants from the Pt counter electrode, as they were present when a graphite counter electrode was alternatively employed (section S5 of the Supporting Information). Hence, we propose that these redox peaks must originate from electrochemical reactions occurring on the electrodeposited films. Interestingly, all the cyclic voltammograms of MoSx-CE exhibited the aforementioned pairs of broad redox peaks (Figure 2b(ii)). Note that the first scan showed a slightly more negative current than the subsequent scans, especially at −0.07 V, which could be from the reduction of residual Mo oxides.10,18 Since the structure of MoSx-CE is known be stable during HER measurements,9,10,15 we propose that these redox peaks could be assigned to reversible reactions such as H+ discharge and Hads oxidation, rather than irreversible reactions such as the cleavage of the Mo−S polymer chains.14 If this proposition is true, the two pairs of redox peaks measured at different potentials would indicate the existence of reactive sites with different adsorption energies for H binding. In support of this, we highlight that peaks assigned to weak and strong H adsorption on Pt have been previously observed in its voltammogram in H2SO4 electrolyte.34,35 The average TOFs of H2 evolution exhibited by MoSx-AE and MoSx-CE at η = 200 mV were also evaluated and found to be 0.23 and 0.13 s−1, respectively (section S4 of the Supporting Information). These values agreed well with previously reported values of amorphous MoSx at the same overpotential.15,36,37 Tafel slopes of 40 mV/dec (Figure 2a(iii),b(iii)) were measured for both MoSx-AE and MoSx-CE (section S6 of the Supporting Information). These Tafel slope values are consistent with those found previously for amorphous MoSx, and indicate that the HER could have proceeded via the Volmer−Heyrovsky mechanism, with the Heyrovsky step as the rate-determining reaction:15,18,38 H+ + e− + A → AHads

Figure 3. XPS spectra of (a) MoSx-AE and (b) MoSx-CE (i) as prepared and (ii) after the first CV scan at 10 mV/s in 1 M HClO4.

be deconvoluted into two doublets. The first doublet with electron binding energies (BE) at 229.8/233.0 eV was assigned to the MoIV center of MoSx,40−42 while the second at 232.4/ 235.6 eV was attributed to the MoVI center in MoO3.10,40 The oxide was probably formed during the electrodeposition of the MoSx films or during sample transfer for XPS analysis.10,18 The sulfur peak was also resolved into two doublets at 162.5/163.7 and 163.8/165.0 eV. The doublet with the lower electron BE can be ascribed to unsaturated S2− and terminal S22−, while the higher BE doublet can be assigned to apical S2− and bridging

Volmer step

AHads + H+ + e− → H 2 + A

Heyrovsky step

where A denotes the adsorption site. When aqueous 1 M DClO4 (containing 94.5 atom % D) was used as electrolyte, the HER occurred at a more cathodic onset potential of −0.17 V. 7793

DOI: 10.1021/acscatal.6b01848 ACS Catal. 2016, 6, 7790−7798

Research Article

ACS Catalysis

Figure 4. Raman spectra recorded of (a) MoSx-AE and (b) MoSx-CE (i) in air, (ii) during the cathodic half sweep, and (iii) during the anodic half sweep of the cyclic voltammetry at 0.5 mV/s in 1 M HClO4 (shown in section S6 of the Supporting Information). The signals marked with an asterisk originated from a cosmic ray.

S22−.16,40,42−44 The Mo:S stoichiometries of MoSx-AE and MoSx-CE were ascertained to be 1:3.1 and 1:1.8, respectively, which were consistent with the EDX measurements presented in section 3.1. The ratios of low electron BE sulfur atoms to high electron BE sulfur atoms in MoSx-AE and MoSx-CE were respectively 49:51 and 73:27. All of the peak assignments are summarized in section S1 of the Supporting Information. Freshly prepared MoSx samples were subjected to one cycle of CV scan at 10 mV/s in HClO4 (0.40 → −0.22 → 0.40 V) and then analyzed by XPS (Figure 3). Post-CV-scanned MoSxAE exhibited a XPS spectrum similar to that of MoSx-CE, with the percentage of high electron BE sulfur atoms decreasing to 33%, a value close to that of the as-prepared MoSx-CE sample (27%). On the other hand, the ratio of low and high electron BE sulfur atoms of MoSx-CE was maintained after CV. These observations are consistent with the electrochemical data presented in section 3.2 that MoSx-AE had reduced to a MoSx-CE-like structure during CV scans to H2-evolving potentials.9,10,15 3.4. Raman Spectroscopy of MoSx-AE and MoSx-CE during Hydrogen Evolution Reaction. Freshly prepared MoSx samples were first characterized by Raman spectroscopy

(in air, Figure 4a(i),b(i)). MoSx-AE showed four broad peaks at 320, 445, 520, and 550 cm−1, which were assigned respectively to its ν(Mo−S)coupled, ν(Sapical−Mo), ν(S−S)terminal, and ν(S− S)bridging vibrations.11,42,45,46 MoSx-CE exhibited two broad Raman bands at 320 and 415 cm−1, ascribed to its ν(Mo− S)coupled and ν(Mo−S−Mo) vibrations, respectively.11,45,47 Considering that the latter film contained 73% of unsaturated S2− + terminal S22− (revealed by XPS, Figure 3b(i)) but did not show a strong ν(S−S)terminal Raman peak at 520 cm−1, we propose that unsaturated S2− could be a major species in MoSxCE. A broad Raman band at 800−990 cm−1 that can be attributed to MoO3 was also recorded on both MoSx samples.48 It should be further noted that the broad Raman bands of MoSx-AE and MoSx-CE differ from the sharp E12g (380 cm−1) and A1g (404 cm−1) peaks of bulk MoS2 (section S7 of the Supporting Information). This shows that the electrodeposited amorphous MoSx films have chemical structures fundamentally different from that of crystalline MoS2.9,10,37 The Raman spectra of a freshly prepared MoSx-AE film were acquired in real time in 1 M HClO4 electrolyte, while its cyclic voltammogram was being measured (Figure 4a(ii,iii), section S6 of the Supporting Information). We found that the MoO3 7794

DOI: 10.1021/acscatal.6b01848 ACS Catal. 2016, 6, 7790−7798

Research Article

ACS Catalysis peaks at 800−990 cm−1 decreased once the catalyst was in contact with the HClO4 electrolyte. This demonstrated dissolution of the surface MoO3 under strongly acidic conditions.13 The peaks observed at 630 and 934 cm−1 were vibrational bands of the ClO4− anion.49 As the potential was swept cathodically to −0.07 V (corresponding to the presence of the strong cathodic peak, R), the ν(Sapical−Mo) (445 cm−1), ν(S−S)terminal (520 cm−1), and ν(S−S)bridging (550 cm−1) peaks were attenuated. Concurrently, the broad ν(Mo−S)coupled peak at 320 cm−1 became sharper. A peak at 415 cm−1 which can be assigned to the ν(Mo−S−Mo) vibration of MoSx-CE also appeared. The overall changes in Raman spectral pattern, along with our CV and XPS measurements, demonstrate that the structure of MoSx-AE has evolved to one that is similar to that of MoSx-CE. The attenuation of the ν(S−S) bands at 520−550 cm−1 suggested that the S−S bonds of the terminal and bridging S22− were cleaved. This would result in the bridging S22− being transformed to unsaturated S2− species. Previous studies on a similar anodically deposited MoSx catalyst showed that terminal S22− atoms may have been removed during H2 evolution.14 These results are consistent with the XPS data presented in section 3.3, which showed a percentage increase in S atoms with low electron BE (terminal S22− and unsaturated S2−) after the HER. Interestingly, a new peak at 2530 cm−1 appeared at −0.07 V, which was at a remarkably consistent potential with the redox peaks (B and B′) observed in the electrochemical measurements (Figure 4a(ii)). Its intensity increased as the potentials further decreased from −0.07 to −0.16 V (cathodic sweep). We could not extend the Raman measurements to more cathodic potentials, as the evolution of H2 gas bubbles interfered with the collection of the inelastically scattered photons. When the potential sweep was reversed, the intensity of the aforementioned peak decreased and finally disappeared before +0.28 V (Figure 4a(iii)). The observed frequency, 2530 cm−1, matches well with the reported S−H stretching vibration (ν(S−H)) of sulfhydryl adsorbates on Mo(100) surfaces (2500 cm−1), SH− ligands bonded to MoIV molecular complexes (2480−2562 cm −1 ), and S−H on MoS 2 (2500 cm −1 ).50−54 These observations suggest strongly that MoSx−H species were formed during the HER. The high frequency of the 2530 cm−1 band also indicates that the H atom is bonded to a single S atom, instead of a bridging coordination mode. The assignment of the 2530 cm−1 peak to S−H vibration was further corroborated with H/D isotopic substitutions (section S8 of the Supporting Information). In 1 M DClO4 electrolyte, a Raman signal at 1825 cm−1 was observed in lieu of the 2530 cm−1 peak. The decrease in frequency is consistent with an isotopic shift factor of 1.39, predicted using the harmonic oscillator model

reduced to a MoSx-CE-like structure. To address this question, we performed operando Raman spectroscopy on MoSx-CE during a CV sweep (Figure 4b(ii,iii), section S6 of the Supporting Information). MoSx-CE was chosen, as its structure remains intact during the HER.9,10,15 As expected, the Raman spectra of MoSx-CE did not show any obvious changes during CV, which confirmed its robust structure. Similar to the case for MoSx-AE, the 2530 cm−1 Raman peak appeared on the MoSxCE sample as well. We also performed control experiments and verified that this Raman peak was from the MoSx films, not from the underlying Mo substrate (section S9 of the Supporting Information). Therefore, we confidently ascribed the 2530 cm−1 peak to the ν(S−H) vibration and assign S atoms as HER-active sites. The frequencies of the Raman peaks observed in this study accompanied by relevant reference values are summarized in Table 1. Table 1. Assignment of Raman Peaks of MoSx-AE, MoSx-CE and Crystalline MoS2 assignment

Raman shift/cm‑1 (this work)

MoSx-AE

ν(Mo−S)coupled ν(Sapical−Mo) ν(S−S)terminal ν(S−S)bridging ν(S−H(D))

320 445 520 550 2530 (1825)

MoSx-CE

ν(Mo−S)coupled ν(Mo−S−Mo) ν(S−H(D))

320 415 2530 (1825)

crystalline MoS2

E12g

380

317−32511,45 445−46211,42,46 504−52511,42 544−55211,42,46 2480−2562 (1830)50−54 317−32511,45 42547 2480−2562 (1830)50−54 38029

A1g

404

41029

sample

vibrational freq/cm‑1 (ref)

It is notable that the ν(S−H) Raman band appeared on MoSx-CE at +0.18 V during the cathodic CV scan and disappeared correspondingly after +0.18 V during the anodic sweep. This result corroborated the detection of the pair of redox CV peaks at around +0.2 V on MoSx-CE (Figure 2b(ii)). Another pair of redox peaks were reproducibly recorded at −0.07 V. The measurements of the two pairs of reduction/ oxidation peaks and the corresponding intensity increase/ decrease of ν(S−H) Raman band indicate the existence of at least two different active S sites. However, we could not discern any frequency shift of the ν(S−H) Raman peak even at different potential ranges. Thus, H bonded on these different S sites could be sharing similar ν(S−H) frequencies. 3.5. Simulations of Vibrational Frequencies using Density Functional Theory. To shed further light on the assignment of the Raman spectroscopy data, DFT simulations were performed to calculate the vibrational frequencies of ν(S− H), ν(S−D), ν(Mo−H), and ν(Mo−D) on four MoSx clusters, two MoSx 1-D amorphous polymers, and two reference MoS2 slabs. Our structural models provide possible H binding sites including terminal (t), bridging (1b/2b), apical, inter S atoms and Mo atoms. The bridging S atoms are further categorized into 1b, which has only one S atom bonded to the corresponding Mo atom, and 2b, which has another S atom bonded to the same Mo atom. The inter S refers to the S atom connecting two Mo3 units in the amorphous polymers. Both the 1b and inter S atoms are the so-called unsaturated S in this work (see Figure 1).

μS − D / μS − H

where μ is the reduced mass of the S−D or S−H bonds. These observations also agreed well with a high-resolution electron energy loss spectroscopy study of S−H (ν(S−H) 2500 cm−1) and S−D (ν(S−D) 1830 cm−1) adsorbates on Mo(100) surfaces.52 The absence of the Mo−H stretching (ν(Mo−H)) vibration at 1714−1942 cm−1 and Mo−D (ν(Mo−D)) at 1235−1293 cm−1 indicates that Mo atoms are not directly involved in the HER catalysis.55−57 It is important to ascertain if the 2530 cm−1 peak belongs to the ν(S−H) vibration of Hads intermediates on S atoms or if the peak belongs to a side product formed while MoSx-AE is being 7795

DOI: 10.1021/acscatal.6b01848 ACS Catal. 2016, 6, 7790−7798

Research Article

ACS Catalysis The simulated average ν(S−H) frequencies of H attached to the terminal (t), bridging (1b/2b), and inter S atoms lie close to each other at 2458, 2472−2520, and 2533 cm−1, versus an average frequency of 2289 cm−1 for apical S sites (Figure 5). In

different from the amorphous MoSx structures and the edge sites of slabs investigated in this work.

4. DISCUSSION In this study, we have provided spectroscopic evidence in the form of ν(S−H) signals, which allowed us to directly pinpoint S atoms in amorphous MoSx as catalytically active sites for H+ reduction. Our assignment of S atoms as active sites is in agreement with conclusions drawn from previous studies on the same catalysts as well as from similar HER catalysts such as [Mo3S13]2− and [Mo2S12]2− complexes, amorphous CoSx, and the FeMo cofactor in nitrogenase.4,13,15−17,58,60 The overlaps in vibrational frequencies for the ν(S−H) peaks do not allow us to distinguish the type of S atoms involved in the HER catalysis. However, from the Raman spectroscopy data and DFT simulations shown in the earlier sections, we could first eliminate apical S as active sites for HER. Furthermore, on MoSx-AE, the broad +0.2 V proton discharge peak was absent during the first CV scan but appeared in subsequent scans (Figure 2a(ii)). This suggests that the S species responsible for this peak is likely to be absent or present in small quantities in freshly prepared MoSx-AE but was subsequently formed during the latter’s transformation to a MoSx-CE-like structure. Among bridging, terminal, and unsaturated S atoms, unsaturated S2− is the only species that increased in population after CV and hence should contribute to the +0.2 V redox peaks. For the −0.07 V redox peaks, we propose bridging S22− atoms to be responsible. We note that our previous experimental and DFT simulation studies revealed that bridging S atoms have a moderate binding strength to H and are thus likely to be the HER-active sites.15 Many of the terminal S22− atoms are likely to be removed during the first CV scan of the MoSx-AE film and thus should not be heavily involved in the HER.14 The appearance of a ν(S−H) Raman signal as early as +0.18 V (Figure 4b(ii)) shows that proton discharge (Volmer step) had occurred prior to the standard potential for H2 evolution (0 V vs RHE). However, Hads did not desorb as H2 until potentials negative of −0.15 V were reached. This finding indicates that the latter reaction is rate determining, which is also consistent with the Tafel slope analysis shown in section 3.2 (measured Tafel slopes of 40 mV/dec suggest the Heyrovsky step as rate determining). Similar observations have also been found from in situ infrared spectroscopy on Pt electrodes, on which the HER proceeded via a fast Volmer step followed by a ratedetermining H atom combination process (Tafel step).34,61 Adsorbed hydrogen on Pt (Hads), identified as the HER intermediate, was recognized by its characteristic Pt−H stretching band at ∼2090 cm−1 at significant underpotentials of +0.11 V vs RHE. Tran et al. had recently employed Raman spectroscopy to investigate the structural changes of anodically deposited amorphous MoSx during HER in pH 7 phosphate buffer electrolytes.14 The ν(S−S) signal from the terminal S22− atoms of MoSx were found to decrease during the HER, while the ν(MoO) signals increased. It was proposed that the removal of the terminal S22− could have resulted in the formation of unsaturated Mo centers. These Mo centers were then suggested to coexist with MoO species, while also serving as catalytically active sites for HER. The ν(MoO) Raman signal was thus used as a marker for Mo active sites. However, we highlight that these observations do not demonstrate that H+ reduction had occurred on the Mo sites. Direct evidence

Figure 5. H (upper) and D (lower) vibrational frequencies on various S and Mo binding sites. The vibrational frequencies observed in experiments are labeled by gray dashed lines. The values in gray near the top x axes are the average values of H or D vibrational frequencies on the corresponding sites.

contrast, the average ν(Mo−H) vibration is at 1768 cm−1. Replacement of H with D resulted in red shifts of the frequencies. Given that only a 2530 cm−1 peak was recorded by our Raman spectroscopy studies, this peak can be reasonably assigned, in light of the DFT calculations, to the stretching vibrations of nonapical S−H bonds. Note that this analysis is in agreement with the known weak binding of apical S atoms to H (Gibbs free binding energy of H (ΔGH) >1 eV), and hence apical S is expected to be HER-inactive.15,58 Mo is unlikely to be involved in the HER as well, since no discernible signals were observed at the ν(Mo−H) range. This is also consistent with the binding energy calculations: for all amorphous MoSx models in this study we find that H cannot stably bind on Mo but moves to an adjacent S atom during the relaxation (unless the Mo is unsaturated with S, in which case the H binds too strongly for catalytic turnover). We note that Li et al. recently proposed that S vacancies (Mo atoms) in the basal plane of the semiconducting 2H-phased (trigonal-prismatic D3h) MoS2 catalyst can bond to H intermediates.59 However, the study was based on a two-dimensional basal site model, which is quite 7796

DOI: 10.1021/acscatal.6b01848 ACS Catal. 2016, 6, 7790−7798

ACS Catalysis



such as the identification of ν(Mo−H) vibrational signals at 1714−1942 cm−1 is lacking.55−57 The involvement of the S atoms of amorphous MoSx as HER-active sites is putative but has never been directly proven until this study.13,15,58 This is because many previous studies typically use X-ray-based spectroscopies, which are not suitable for detecting H. Here, through a judicious choice of electrochemical CV measurements, operando Raman spectroscopy, H/D isotopic exchange, and DFT simulations, we provided direct molecular fingerprint evidence for the formation of S−H moieties on MoSx and the participation of S atoms as HER-active sites. The Raman spectroscopy data also demonstrate clearly the structural transformation of MoSx-AE to MoSx-CE-like during HER. We believe that this investigation, along with our previous work that quantified the HER activities of S atoms in amorphous MoSx,15 will contribute significantly to a deeper understanding of the fundamental catalytic behaviors of Mo sulfide films.

ASSOCIATED CONTENT

* Supporting Information S

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acscatal.6b01848. XPS data and peak assignments, schematic diagram of the Raman setup, additional electrochemical and Raman data, calculation of TOF values, and isotopic labeling experiments (PDF)



REFERENCES

(1) Lewis, N. S.; Nocera, D. G. Proc. Natl. Acad. Sci. U. S. A. 2006, 103, 15729−15735. (2) Armaroli, N.; Balzani, V. ChemSusChem 2011, 4, 21−36. (3) Morales-Guio, C. G.; Stern, L.-A.; Hu, X. Chem. Soc. Rev. 2014, 43, 6555−6569. (4) Rod, T. H.; Nørskov, J. K. J. Am. Chem. Soc. 2000, 122, 12751− 12763. (5) Hinnemann, B.; Moses, P. G.; Bonde, J.; Jørgensen, K. P.; Nielsen, J. H.; Horch, S.; Chorkendorff, I.; Nørskov, J. K. J. Am. Chem. Soc. 2005, 127, 5308−5309. (6) Reddy, S.; Du, R.; Kang, L.; Mao, N.; Zhang, J. Appl. Catal., B 2016, 194, 16−21. (7) Lee, S. C.; Benck, J. D.; Tsai, C.; Park, J.; Koh, A. L.; AbildPedersen, F.; Jaramillo, T. F.; Sinclair, R. ACS Nano 2016, 10, 624− 632. (8) Huang, Y.; Nielsen, R. J.; Goddard, W. A.; Soriaga, M. P. J. Am. Chem. Soc. 2015, 137, 6692−6698. (9) Morales-Guio, C. G.; Hu, X. Acc. Chem. Res. 2014, 47, 2671− 2681. (10) Vrubel, H.; Hu, X. ACS Catal. 2013, 3, 2002−2011. (11) Weber, T.; Muijsers, J. C.; Niemantsverdriet, J. W. J. Phys. Chem. 1995, 99, 9194−9200. (12) Merki, D.; Hu, X. L. Energy Environ. Sci. 2011, 4, 3878−3888. (13) Lassalle-Kaiser, B.; Merki, D.; Vrubel, H.; Gul, S.; Yachandra, V. K.; Hu, X.; Yano, J. J. Am. Chem. Soc. 2015, 137, 314−321. (14) Tran, P. D.; Tran, T. V.; Orio, M.; Torelli, S.; Truong, Q. D.; Nayuki, K.; Sasaki, Y.; Chiam, S. Y.; Yi, R.; Honma, I. Nat. Mater. 2016, 15, 640−646. (15) Ting, L. R. L.; Deng, Y.; Ma, L.; Zhang, Y.-J.; Peterson, A. A.; Yeo, B. S. ACS Catal. 2016, 6, 861−867. (16) Kibsgaard, J.; Jaramillo, T. F.; Besenbacher, F. Nat. Chem. 2014, 6, 248−253. (17) Huang, Z.; Luo, W.; Ma, L.; Yu, M.; Ren, X.; He, M.; Polen, S.; Click, K.; Garrett, B.; Lu, J.; Amine, K.; Hadad, C.; Chen, W.; Asthagiri, A.; Wu, Y. Angew. Chem., Int. Ed. 2015, 54, 15181−15185. (18) Merki, D.; Fierro, S.; Vrubel, H.; Hu, X. L. Chem. Sci. 2011, 2, 1262−1267. (19) Yeo, B. S.; Klaus, S. L.; Ross, P. N.; Mathies, R. A.; Bell, A. T. ChemPhysChem 2010, 11, 1854−1857. (20) Deng, Y.; Handoko, A. D.; Du, Y.; Xi, S.; Yeo, B. S. ACS Catal. 2016, 6, 2473−2481. (21) Fabbri, E.; Habereder, A.; Waltar, K.; Kotz, R.; Schmidt, T. J. Catal. Sci. Technol. 2014, 4, 3800−3821. (22) Schmidt, T. J. In Polymer Electrolyte Fuel Cell Durability; Büchi, F. N., Inaba, M., Schmidt, T. J., Eds.; Springer New York: New York, 2009; pp 199−221. (23) Tsang, J. C.; Demuth, J. E.; Sanda, P. N.; Kirtley, J. R. Chem. Phys. Lett. 1980, 76, 54−57. (24) Enkovaara, J.; Rostgaard, C.; Mortensen, J. J.; Chen, J.; Dułak, M.; Ferrighi, L.; Gavnholt, J.; Glinsvad, C.; Haikola, V.; Hansen, H. A.; Kristoffersen, H. H.; Kuisma, M.; Larsen, A. H.; Lehtovaara, L.; Ljungberg, M.; Lopez-Acevedo, O.; Moses, P. G.; Ojanen, J.; Olsen, T.; Petzold, V.; Romero, N. A.; Stausholm-Møller, J.; Strange, M.; Tritsaris, G. A.; Vanin, M.; Walter, M.; Hammer, B.; Häkkinen, H.; Madsen, G. K. H.; Nieminen, R. M.; Nørskov, J. K.; Puska, M.; Rantala, T. T.; Schiøtz, J.; Thygesen, K. S.; Jacobsen, K. W. J. Phys.: Condens. Matter 2010, 22, 253202. (25) Mortensen, J. J.; Hansen, L. B.; Jacobsen, K. W. Phys. Rev. B: Condens. Matter Mater. Phys. 2005, 71, 035109. (26) Hammer, B.; Hansen, L. B.; Nørskov, J. K. Phys. Rev. B: Condens. Matter Mater. Phys. 1999, 59, 7413−7421. (27) Wang, Z.; von dem Bussche, A.; Qiu, Y.; Valentin, T. M.; Gion, K.; Kane, A. B.; Hurt, R. H. Environ. Sci. Technol. 2016, 50, 7208− 7217. (28) Lin, Y.-C.; Dumcenco, D. O.; Huang, Y.-S.; Suenaga, K. Nat. Nanotechnol. 2014, 9, 391−396.

5. CONCLUSION A combination of cyclic voltammetry, X-ray photoelectron spectroscopy, Raman spectroscopy, and H/D isotopic exchange was used to investigate how amorphous MoSx films (both anodically and cathodically deposited) catalyzed the hydrogen evolution reaction (HER). Anodically deposited MoSx-AE transformed to cathodically deposited MoSx-CE-like structures at potentials prior to H2 production, which corresponded to elimination of its S atoms and cleavage of disulfide S−S bonds. The participation of S atoms as catalytically active sites for evolving H2 was directly demonstrated for the first time through the simultaneous detection of both H+ discharge using CV and S−H moieties using operando Raman spectroscopy. Density functional theory (DFT) simulations on ν(S−H) vibrational frequencies further revealed that only nonapical S atoms were involved in the HER.



Research Article

AUTHOR INFORMATION

Corresponding Author

*B.S.Y.: e-mail, [email protected]; fax, +65 6779 1691; tel, +65 6516 2836. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS This work was supported by an academic research fund (R-143000-587-112) from the National University of Singapore. Y.D. acknowledges a Ph.D. research scholarship from the Ministry of Education of Singapore. A.A.P. and Y.-J.Z. acknowledge funding from the U.S. Office of Naval Research under Award No. N00014-15-1-2223. Electronic structure calculations were undertaken at Brown’s Center for Computation and Visualization. 7797

DOI: 10.1021/acscatal.6b01848 ACS Catal. 2016, 6, 7790−7798

Research Article

ACS Catalysis (29) Voiry, D.; Salehi, M.; Silva, R.; Fujita, T.; Chen, M.; Asefa, T.; Shenoy, V. B.; Eda, G.; Chhowalla, M. Nano Lett. 2013, 13, 6222− 6227. (30) Voiry, D.; Mohite, A.; Chhowalla, M. Chem. Soc. Rev. 2015, 44, 2702−2712. (31) Fan, X.-L.; Yang, Y.; Xiao, P.; Lau, W.-M. J. Mater. Chem. A 2014, 2, 20545−20551. (32) Eda, G.; Fujita, T.; Yamaguchi, H.; Voiry, D.; Chen, M.; Chhowalla, M. ACS Nano 2012, 6, 7311−7317. (33) Heising, J.; Kanatzidis, M. G. J. Am. Chem. Soc. 1999, 121, 638− 643. (34) Nichols, R. J.; Bewick, A. J. Electroanal. Chem. Interfacial Electrochem. 1988, 243, 445−453. (35) Kinoshita, K.; Ferrier, D. R.; Stonehart, P. Electrochim. Acta 1978, 23, 45−54. (36) Bose, R.; Balasingam, S. K.; Shin, S.; Jin, Z.; Kwon, D. H.; Jun, Y.; Min, Y.-S. Langmuir 2015, 31, 5220−5227. (37) Benck, J. D.; Chen, Z.; Kuritzky, L. Y.; Forman, A. J.; Jaramillo, T. F. ACS Catal. 2012, 2, 1916−1923. (38) Thomas, J. G. N. Trans. Faraday Soc. 1961, 57, 1603−1611. (39) Kuhn, A. T.; Byrne, M. Electrochim. Acta 1971, 16, 391−399. (40) Muijsers, J. C.; Weber, T.; van Hardeveld, R. M.; Zandbergen, H. W.; Niemantsverdriet, J. W. J. Catal. 1995, 157, 698−705. (41) Wang, T. Y.; Gao, D. L.; Zhuo, J. Q.; Zhu, Z. W.; Papakonstantinou, P.; Li, Y.; Li, M. X. Chem. - Eur. J. 2013, 19, 11939−11948. (42) Muller, A.; Wittneben, V.; Krickemeyer, E.; Bogge, H.; Lemke, M. Z. Anorg. Allg. Chem. 1991, 605, 175−188. (43) Hou, Y.; Abrams, B. L.; Vesborg, P. C. K.; Björketun, M. E.; Herbst, K.; Bech, L.; Setti, A. M.; Damsgaard, C. D.; Pedersen, T.; Hansen, O.; Rossmeisl, J.; Dahl, S.; Nørskov, J. K.; Chorkendorff, I. Nat. Mater. 2011, 10, 434−438. (44) Jaramillo, T. F.; Bonde, J.; Zhang, J.; Ooi, B.-L.; Andersson, K.; Ulstrup, J.; Chorkendorff, I. J. Phys. Chem. C 2008, 112, 17492−17498. (45) Chang, C. H.; Chan, S. S. J. Catal. 1981, 72, 139−148. (46) Fedin, V. P.; Kolesov, B. A.; Mironov, Y. V.; Fedorov, V. Y. Polyhedron 1989, 8, 2419−2423. (47) Stevens, G. C.; Edmonds, T. J. Catal. 1975, 37, 544−547. (48) Murugan, R.; Ghule, A.; Bhongale, C.; Chang, H. J. Mater. Chem. 2000, 10, 2157−2162. (49) Ratcliffe, C.; Irish, D. Can. J. Chem. 1984, 62, 1134−1144. (50) Iwasa, K.; Seino, H.; Niikura, F.; Mizobe, Y. Dalton Trans. 2009, 6134−6140. (51) Smith, S. J.; Whaley, C. M.; Rauchfuss, T. B.; Wilson, S. R. Inorg. Chem. 2006, 45, 679−687. (52) Gland, J. L.; Kollin, E. B.; Zaera, F. Langmuir 1988, 4, 118−120. (53) Dupre, N.; Hendriks, H. M. J.; Jordanov, J. J. Chem. Soc., Dalton Trans. 1984, 1463−1465. (54) Sundberg, P.; Moyes, R.; Tomkinson, J. Bull. Soc. Chim. Belg. 1991, 100, 967−976. (55) Zhang, S.; Bullock, R. M. Inorg. Chem. 2015, 54, 6397−6409. (56) Pennella, F. J. Chem. Soc. D 1971, 158a. (57) Green, M. L. H.; Silverthorn, W. E. J. Chem. Soc. D 1971, 557− 558. (58) Li, Y.; Yu, Y.; Huang, Y.; Nielsen, R. A.; Goddard, W. A.; Li, Y.; Cao, L. ACS Catal. 2015, 5, 448−455. (59) Li, H.; Tsai, C.; Koh, A. L.; Cai, L.; Contryman, A. W.; Fragapane, A. H.; Zhao, J.; Han, H. S.; Manoharan, H. C.; AbildPedersen, F.; Norskov, J. K.; Zheng, X. Nat. Mater. 2015, 15, 48−53. (60) Kornienko, N.; Resasco, J.; Becknell, N.; Jiang, C.-M.; Liu, Y.-S.; Nie, K.; Sun, X.; Guo, J.; Leone, S. R.; Yang, P. J. Am. Chem. Soc. 2015, 137, 7448−7455. (61) Nichols, R. J. In Developments in Electrochemistry; Pletcher, D., Tian, Z.-Q., Williams, D. E., Eds.; Wiley: Chichester, U.K., 2014; pp 183−200.

7798

DOI: 10.1021/acscatal.6b01848 ACS Catal. 2016, 6, 7790−7798