Electrocatalytic Performance of Titania Nanotube Arrays Coated with MoS2 by ALD toward the Hydrogen Evolution Reaction.

The electrochemical splitting of water provides an elegant way to store renewable energy, but it is limited by the cost of the noble metals used as catalysts. Among the catalysts used for the reduction of water to hydrogen, MoS2 has been identified as one of the most promising materials as it can be engineered to provide not only a large surface area but also an abundance of unsaturated and reactive coordination sites. Using Mo[NMe2]4 and H2S as precursors, a desired thickness of amorphous MoS2 can be deposited on TiO2 nanotubes by atomic layer deposition. The identity and structure of the MoS2 film are confirmed by spectroscopic ellipsometry, X-ray photoelectron spectroscopy, scanning electron microscopy, and energy dispersive X-ray spectroscopy. The electrocatalytic performance of MoS2 is quantified as it depends on the tube length and the MoS2 layer thickness through voltammetry, steady-state chronoamperometry, and electrochemical impedance spectroscopy. The best sample reaches 10 mA/cm2 current density at 189 mV overpotential in 0.5 M H2SO4. All of the various geometries of our nanostructured electrodes reach an electrocatalytic proficiency comparable with the state-of-the-art MoS2 electrodes, and the dependence of performance parameters on geometry suggests that the system can even be improved further.

Introduction

The increasing environmental issues caused by the release of CO2 from the combustion of fossil fuels creates an urgent need to develop a renewable, sustainable, and clean energy source to replace the current energy system dominated by fossil fuels.[1] Compared with carbon-based fossil fuels, hydrogen can theoretically be a carbon-free energy carrier as its only combustion product is water. However, hydrogen is mainly produced by steam reforming of natural gas, which does produce CO2.[2] A promising alternative method of producing hydrogen in a carbon-neutral manner is the electrochemical and photoelectrochemical water splitting.[3−5]

Losses during energy storage in the fuel H2 (and its counterpart O2) are caused primarily by the electrodes’ overpotentials required to perform each half-reaction at a practically useful rate. The minimization of overpotentials is achieved by using catalysts, the best of which currently consist of noble metals, specifically platinum and its alloys for the hydrogen evolution reaction (HER).[6−9] However, their wide-scale application is limited by their scarcity and high cost. Alternatives based on non-noble metals, usually in compounds such as oxides, heavier chalcogenides or even pnictides, have not achieved catalytic performance and durability at high potential on par with their metallic counterparts so far.[10−12] Therefore, efficient and stable catalysts based on earth-abundant, inexpensive transition metals remain to be found.[4] Recently, molybdenum sulfide has attracted significant attention as a promising alternative to the noble metal catalysts due to its low cost and long-term stability as well as the abundance of the elements molybdenum and sulfur in the earth crust.[13−16]

Significant work has been dedicated to the HER electrocatalytic activity of crystalline and amorphous MoS2 and has highlighted their respective advantages and drawbacks. For crystalline MoS2, both experimental data and theoretical calculations show that the edge sites in the layered structure mostly determine the catalytic performance.[17−20] One approach to increasing the number of active edge sites has consisted in the engineering of nanostructured MoS2, for example, as MoS2 nanosheets,[21−23] MoS2 nanoparticles,[24−26] mesoporous MoS2,[27] and MoS2 thin films.[28,29] In comparison to this, amorphous MoS2 features an enhanced overall electrode activity toward HER owing to its higher density of defect sites.[30] In the case of amorphous MoS2, as well, a number of studies have been devoted to nanostructuring aimed at increasing the surface area while maintaining sufficient electrical conductivity.[31−34] A highly porous structure typically features long electron transport distances, which can cause resistive (Ohmic) losses.[35,36]

Titanium oxide nanotube arrays (TNTAs) are a type of vertically aligned, highly ordered, high-surface area nanostructure that can be economically fabricated by anodic anodization of titanium metal.[37−40] Recently, TNTAs have been used in combination with MoS2 materials fabricated by a variety of methods, such as hydrothermal synthesis, Mo sulfurization, and electrochemical deposition, to build heterostructures for water splitting.[41−43] The highly ordered tubular structure not only increases the number of active sites exposed to the electrolyte, it can also facilitate reactant supply and product removal along straight pores with respect to typical powder-based catalyst beds featuring long diffusion paths along disordered pores. However, those MoS2 synthesis methods are not able to coat the highly tubular TNTAs with an accurate control of the MoS2 loading.

Thus, the present work focuses on atomic layer deposition (ALD) as a method to produce molybdenum sulfide. We use TNTAs as a substrate, tune the length of the tubes via the anodization duration, coat the substrate with amorphous MoS2 through atomic layer deposition, and investigate the geometric effect and catalyst thickness influence on the electrocatalytic HER performance of amorphous MoS2.

Results and Discussion

Molybdenum Sulfide Atomic Layer Deposition (ALD) on Planar and Nanostructured Substrates

Atomic layer deposition (ALD) is a thin-film deposition technique based on sequential reactions of volatile precursors with a solid surface, each of which behaves in a self-limiting manner.[44] This self-limiting surface chemistry renders ALD uniquely suited to depositing a precise thickness of continuous and pinhole-free coating in deep pores.[45]

Here, we first characterize the ALD of molybdenum sulfide from Mo[NMe2]4 and H2S (3% in N2) on the piezoelectric GaPO4 crystal microbalance of a small reactor used for method development and on planar Si wafers.[46,47]Figure a shows that the thickness t of the molybdenum sulfide film deposited increases in a linear relationship with the number N of cycles performed at 95 °C. According to the thickness measured by spectroscopic ellipsometry, the growth per cycle is approximately 2.5 Å. This rather large value corresponds to nearly half of a MoS2 monolayer. The self-limiting behavior of the surface reactions is demonstrated by the piezoelectric microbalance data of Figure b, which evidences saturation with dose. During each ALD cycle (except the first), the Mo[NMe2]4 precursor reacts in a self-limiting way with the −SH reactive groups terminating the substrate’s surface and leaves a dimethylamide termination. In the subsequent step, these amide groups are released by H2S and the surface reverts to the original −SH termination. A closer look at the piezoelectric microbalance curve shape within one cycle (Figure c) evidences a nonideal behavior, in which the frequency does not revert to a stable value after each pulse. However, we have checked that varying the purge duration does not affect the growth rate. Thus, we interpret this behavior as due to factors that affect the crystal’s eigenfrequency beyond mass, such as thermal and strain effects.

Figure 1

(a) Growth curve of MoS2 on silicon wafer pieces (thickness measured by spectroscopic ellipsometry). Three independently grown batches of samples are represented in three distinct colors. (b) Saturation curve of the MoS2 growth dependent on the molybdenum precursor pulse duration, and (inset) in situ piezoelectric microbalance growth curve when the Mo[NMe2]4 and H2S pulse durations are set to 20 and 2 ms, respectively. (c) A zoom-in view of three ALD cycles (one cycle being highlighted in light blue color in (b) and (c)). The timing of both precursor pulses within this cycle is shown by arrows.

The ALD recipe is then applied to coating annealed TNTA substrates with molybdenum sulfide in a larger reactor. The pulse and purge durations are adapted to the larger chamber volume, whereas the temperatures are maintained. The TNTAs are obtained by the anodization of Ti foil, whereby various tube lengths are achieved by various anodization durations, as shown in Figure . More specifically, tubes of 7.8, 13.6, and 17.9 μm are obtained after 1, 2, and 3 h of anodization at 60 V, respectively. The tubes’ inner diameter remains constant at approximately 100 nm near the surface. Figure shows top-view scanning electron micrographs of a TNTA sample before and after molybdenum sulfide coating. Evidence of the material coating on the top surface is clear from the fact that tube walls appear thicker after ALD. Coating into the depths of the pores is covered later.

Figure 2

Cross-sectional scanning electron microscopy images of different lengths of titanium dioxide (TiO2) nanotubes fabricated by varying anodization time: (a) 1 h, (b) 2 h, and (c) 3 h at 60 V.

Figure 3

Top-view scanning electron microscopy images of 2 h annealed TiO2 nanotubes before and after MoS2 coating: (a) TNTAs without MoS2, (b) TNTAs coated with MoS2 by 80 ALD cycles.

Characterization of the Molybdenum Sulfide Deposit

The chemical identity of the deposit is determined by energy dispersive X-ray spectroscopy (EDX) analysis recorded on the top of a sample as shown in Figure : Mo and S signals are present in an atomic ratio that integrates to around 1 to 2 despite the limitation inherent to their spectral overlap (Figure ). When an EDX profile is recorded along the tubes’ length, the 1:2 ratio is maintained (Figure ). The absolute signal intensity of these two elements, however, decays from the surface down to the depth of 7.8 μm (1 h of anodization). This signal decline of Mo and S from the top to the bottom could be due to two effects. First, the “V” shape of the TiO2 nanotubes is known to generate a decrease of surface area with depth.[48] The inner tube diameter becoming smaller and smaller from the top to the bottom causes the intensity increase of the Ti signal observed in Figure . A second effect could be due to the limited vapor pressure of Mo[NMe2]4 available and transport limitation along the deep tubes’ length.

Figure 4

Energy dispersive X-ray analysis of TNTAs anodized for 1 h and coated with MoS2 (80 ALD cycles). Oxygen signal in blue, titanium signal in green, and sulfur and molybdenum signals (overlapped) in red. The accelerating voltage is 20 kV.

Figure 5

Energy dispersive X-ray analysis profile of a TiO2 nanotube sample coated with MoS2 (80 ALD cycles), investigated in cross section. The carbon signal (from the conductive tape) serves to identify the total pore length (7.8 μm).

Regular X-ray diffraction (XRD) investigation of the MoS2/TNTA samples exhibits large, sharp anatase peaks and no MoS2 signals (Figure S1). To look for potentially broad and weak MoS2 signals, we turn to grazing-incidence XRD measurements of approximately 20 nm MoS2 on a planar glass (Figure ). The patterns of the glass substrate both before and after MoS2 coating glass show no peak. This establishes that the ALD material is amorphous, at least when deposited at 95 °C.

Figure 6

Grazing-incidence X-ray diffraction patterns of a glass slide coated with 20 nm of MoS2 (black line) and uncoated glass (red line).

Further insight into the surface chemical composition and valence state of the MoS2 deposit on TNTAs is provided by X-ray photoelectron spectroscopy (XPS) analysis. Figure a shows the high-resolution XPS data in the Mo 3d region. It can be fitted with two major contributions attributed to MoS2 (with Mo 3d5/2 and Mo 3d3/2 peaks at 229.7 and 232.9 eV, respectively) and to Mo5+ (as already described for the edge states in nanocrystalline molybdenum sulfides, with Mo 3d5/2 and Mo 3d3/2 at 230.8 and 234.0 eV where such Mo5+ states are mainly found at the edge of MoS2 crystals under a sulfur-rich environment).[49] A doublet at still higher energy is also present and characteristic of oxygen-bonded Mo (“MoO3” with Mo 3d5/2 and Mo 3d3/2 at 233.0 and 236.1 eV), probably formed by superficial oxidation of the MoS2 layer in contact with air. Correspondingly, the S 2p shown in Figure b indicates that sulfur is mainly present as molybdenum sulfide with the S 2p3/2 + S 2p1/2 doublet at 162.5 and 163.6 eV. A shoulder at higher energy (S 2p3/2 + S 2p1/2 at 163.8 and 164.9 eV) can be attributed to the presence of S–S bonds (in polysulfides). A trace amount of sulfates is also observed (S 2p3/2 + S 2p1/2 doublet at 169.0 and 170.1 eV). This corresponds to the superficial aerobic oxidation of MoS2 or to an artifact of sputtering. Note that sputtering does reduce the oxide signals (see the Mo 3d region XPS data recorded without sputtering for comparison, Figure S2).

Figure 7

X-ray photoelectron spectroscopy analysis of Mo 3d (a) and S 2p (b) regions for TiO2 nanotubes obtained by 3 h of anodization and coated with MoS2 (80 ALD cycles), after 30 s of sputter-etching.

Electrochemical Investigation

The electrocatalytic activity of nanostructured MoS2 toward the hydrogen evolution reaction (HER) was investigated using a three-electrode system with 0.1 M H2SO4 as electrolyte. Cyclic voltammetry (CV) was performed on MoS2-coated TNTA electrodes between −0.8 and +0.2 V versus a standard hydrogen electrode (SHE) at a scan rate of 50 mV/s (Figure ) first. From the CV, we can observe a significant cathodic current density starting at −0.23 V, corresponding to the onset of the hydrogen evolution reaction. The increasing current density also materializes in a continuous generation of bubbles from the electrolyte. The linear shape of the current density curve with increase of the applied potential beyond (below) −0.3 V deviates from the exponential function expected of the pure electrocatalytic case and is due to the Ohmic resistance of the electrode substrate. No additional signals appear down to overpotential of 800 mV, and the curve is reversible except for a slight hysteresis. These observations are indicative of the stability of our MoS2 catalyst in the acidic electrolyte and the capacitive behavior associated with its enhanced surface area.

Figure 8

Cyclic voltammogram of an electrode consisting of 7.8 μm long TNTAs coated with MoS2 (80 ALD cycles), from −0.8 V (vs SHE) to +0.2 V (vs SHE). Scan rate: 50 mV/s, step size: 2 mV, and electrode area: 0.0314 cm2.

In the next step, we further investigate the effect of the TiO2 tube length and the thickness of the MoS2 catalyst layer toward the electrocatalytic performance of the electrodes. Cyclic voltammetry and steady-state chronoamperometry were performed for a series of TNTAs featuring tubes of various lengths (between 7.8 and 17.9 μm) and decorated with different MoS2 thicknesses (from 7.5 to 20 nm). The voltammograms displayed in Figure a,b show the influence of the MoS2 catalyst’s thickness on both planar and nanostructured electrodes. From these two graphs, the activity of MoS2 as a catalyst (as opposed to the naked TiO2 substrate) is obvious in both the planar and nanotubular cases (compare colored and black curves within Figure a and within Figure b). Furthermore, the beneficial effect of increased surface area caused by the porous geometry appears clearly (upon comparing Figure a with 9b). Although increasing the MoS2 layer thickness from 7.5 to 12.5 and 20 nm does generate a slight current density increase at −0.45 V, the relationship is by far not proportional. This observation corresponds to a surface reaction and proves that the MoS2 deposit is dense (albeit amorphous), with a beneficial effect of a slight surface roughness. This further implies that the catalyst loading can be minimized without a significant loss of activity.

Figure 9

Effect of the MoS2 thickness (a, b) and tubes length (c, d) on electrocatalytic HER performance. Cyclic voltammograms performed on electrodes consisting of (a) planar TiO2; (b) TNTAs (anodized for 1 h) coated with various amounts of MoS2 (0, 30, 50, and 80 cycles); (c) planar TiO2 and TNTAs (anodized for 1, 2, and 3 h) without coating; (d) same substrates after coating with MoS2 (30 ALD cycles). Scan rate: 50 mV/s, step size: 2 mV, and electrode area: 0.0314 cm2, performed in 0.1 M H2SO4 electrolyte.

The effect of the TiO2 tube length is presented in Figure c,d. In the absence of a catalyst (Figure c), the tube length increases are associated with corresponding increases in the capacitive current (hysteretic behavior of the voltammetric curve), indicating that the full sample surface is in electrochemical contact with the electrolyte. The enhanced surface area of the nanotubular samples proves to be beneficial with respect to the planar counterparts in Figure d. However, pore length increases beyond 7.8 μm do not result in any current density improvements. This observation allows us to conclude that the EDX profile discussed in Figure is likely due to incomplete coating of long tubes.

The systematic increase in electrocatalytic performance from planar to structured samples and from inert to MoS2-coated ones can be expressed as a corresponding decrease in the charge-transfer resistance (Rct) at the electrolyte/electrode interface. This parameter is related to the inherent catalytic activity of the electrodes, and it is associated with the diameter of the semicircle determined by electrochemical impedance spectroscopy in the Nyquist plots (Figure ). In the planar case, adding MoS2 as the HER catalyst reduces electrolyte/electrode interface Rct from 85 to 3.3 kΩ approximately whereas the geometric effect of the nanotubular structure further improves Rct to around 164 Ω. Figure shows the steady-state current densities obtained for our various samples over 30 min of bulk electrolysis. Not only are all trends observed in voltammetry reproduced in the steady state but the current densities obtained here remain high (on the same order of magnitude as on the CV curves). Furthermore, these data are indicative of an excellent stability in the corrosive electrolyte. Let us now characterize their stability further.

Figure 10

Electrochemical impedance spectroscopies of (a) planar TiO2, (b) planar TiO2 with MoS2 coating, and (c) nanotubular TiO2 with MoS2 coating, recorded at −0.25 V vs SHE in 0.1 M H2SO4 electrolyte. Fitted curves are in solid lines.

Figure 11

Average steady-state current densities of planar TiO2 and TNTAs of different lengths coated by various amounts of MoS2: 0 ALD cycle (black), 30 cycles (red), 50 cycles (blue), and 80 cycles (green). Measurement conditions: 0.1 M H2SO4 at −0.5 V vs SHE.

Stability of the Electrodes

The current–time curve recorded on a nanotubular MoS2 electrode held at an overpotential of 450 mV for 36 h quantifies its high durability and long-term stability (Figure ). The current remains constant during the whole experiment apart from the noiselike variations occuring due to the bubble formation. A crystalline anatase substrate is a crucial prerequisite for this high stability, as shown in Figure S3. Not only is the anatase phase of TiO2 more conductive electrically than its amorphous counterpart but more importantly, TNTA samples that have not been annealed (and are correspondingly amorphous) are destroyed upon electrochemical treatment. Annealed TNTA samples, in contrast to this, display unaffected tube structures after tens of hours of continuous electrolysis. Figures S4 and S5 also show no change in the chemical composition and crystalline structure after electrolysis. Compared with the reported works, the stability of our MoS2-coated TNTA electrodes is among one of the best.[13−16]

Figure 12

Time-dependent current curve of the TNTA electrode obtained upon anodization for 1 h and coated with MoS2 (80 ALD cycles), recorded under η = 450 mV in 0.1 M H2SO4.

Conclusions

In conclusion, we have established a robust type of electrode based on anatase TiO2 nanotubes coated with MoS2 by ALD. Compared with planar samples, the nanotubular geometry serves to improve the electrocatalytic performance due to their larger surface area and concomitantly higher number of active sites. The presence of MoS2 allows for steady-state current densities beyond 50 mA/cm2 to be obtained for the HER (at 500 mV overpotential). A more direct comparison with literature values is performed in a 0.5 M H2SO4 electrolyte considered standard in some recent papers (Figure ).[50−55] In a 0.5 M H2SO4 electrolyte at 50 mV/s, our electrode consisting of nanotubular TiO2 coated with MoS2 (80 ALD cycles) reaches the current density of 10 mA/cm2 at an overpotential of η = 189 mV with a low catalyst loading (∼1.2 mg/cm2), which represents an improvement with respect to the state-of-the-art electrode (Table ).[50−55] In future, we can reduce the thickness of the ALD layer and thereby the loading of MoS2. Our data indicate that loadings below 0.5 mg/cm2 (30 cycles coating) will yield similar current densities. Thus, applications such as photoelectrochemical ones can be pursued with minimized catalyst cost and with an ease of adjustment (based on geometric parameters) unmatched by other systems.

Figure 13

Linear sweep voltammetry of TNTAs (anodized for 1 h) coated with MoS2 (80 cycles, black line) and without MoS2 (red line), as well as of a planar TiO2 sample in coated with MoS2 (80 cycles, green line, overlapped with the blue one) and without MoS2 (blue line), measured in a 0.5 M H2SO4 electrolyte, with a scan rate of 50 mV/s and electrode area of 0.0314 cm2.

Table 1

Comparison of MoS2 HER Activity Quantifiers

catalyst	electrolyte	overpotential@10 mA cm–2 (mV)	loading amount (mg cm–2)	references
flowerlike MoS2/GCNa	0.5 M H2SO4	200	0.196	(50)
MoS2 microflake film	0.5 M H2SO4	170	unknown	(51)
O-MoS2	0.5 M H2SO4	120 (onset)	0.29	(52)
MoS2/MoO2	0.5 M H2SO4	240	0.22	(53)
ultrathin MoS2-coated CNb	0.5 M H2SO4	200	0.32	(54)
MoS2/g-C3N4	0.5 M H2SO4	260	1.06	(55)
MoS2/TNTA	0.5 M H2SO4	189	1.2c	this work

GCN: graphitic carbon nitride.

CN: carbon nanospheres.

Calculated based on geometry.

Experimental Section

Chemicals

All chemicals are of analytical reagent grade and were used as received without any further purification. NH4F, ethylene glycol, concentrated phosphoric acid, and sulfuric acid were obtained from Acros Organics (Germany), VWR (Germany). Titanium foils were purchased from Advent Research Materials Ltd (England). The silicon wafer with approximately 200 nm silicon oxide on the top was supplied by Silicon Materials Inc. H2S (3% in N2) was ordered from Air Liquid (Germany). The molybdenum precursor (Mo[NMe2]4) was synthesized according to the published procedure.[56]

Fabrication of the TiO2 Nanotubes Arrays and Planar TiO2 Substrates

The TNTAs were fabricated by titanium anodic anodization according to ref (39). The titanium foil was sonicated successively in acetone, ethanol, and water for 15 min, respectively, and then dried under N2 flow. Afterward, the clean foil was bound to the homemade electrolytic cell with silver mesh serving as the counter electrode. The anodization was carried out at 60 V for 1, 2, and 3 h in the electrolyte consisted of ethylene glycol, 5% (vol %) H2O and 0.5% (wt %) NH4F to obtain the vertically aligned different lengths of TiO2 tubes on the metallic Ti foil. Thereafter, the TiO2 nanotubular membrane was cleaned with distilled water and dried at 45 °C overnight. Afterward, the membrane was sonicated in 20% (vol %) acetic acid for 10–20 min to remove the disordered top layer covering the top side of the membrane. The planar TiO2 substrates were also made via Ti anodization according to ref (40). The clean Ti foil was anodized at 60 V in a fluoride-free electrolyte (5 wt % H3PO4 solution in ethylene glycol) for 20 min to obtain a thin layer of planar compact TiO2 on the Ti surface.

MoS2 Atomic Layer Deposition

The deposition of MoS2 on planar and TNTA substrates was performed on a homemade ALD reactor by using Mo[NMe2]4 as the molybdenum precursor and H2S as the sulfur source based on refs (46) and (47). The molybdenum precursor was heated up to 65 °C, and the reactor chamber was kept at 95 °C. For the deposition on planar substrates, a 0.7 s pulse of the molybdenum precursor was first inserted into the chamber and stayed in the chamber for 20 s (exposure time) to let the precursor react with the wafer and planar TiO2 surface; then, the molybdenum precursor was flushed away completely by 40 s of pumping (pumping time). In the next step, a 0.2 s H2S pulse was introduced to the chamber and reacted with the molybdenum precursor for 20 s and the remaining H2S was also pumped away (40 s). Through the repeated, nonoverlapping exposure to the two precursors, a thin film of MoS2 was slowly deposited. For the deposition on TNTA substrates, we used a different 50 s exposure time and 60 s pumping time for both precursors since a longer time is needed for the precursors to reach the extremity of the tube.

Characterization

The samples were characterized by the X-ray diffraction (Bruker D8 Advance, Germany) with a Cu Kα source and a LynxEye XE T detector, a field emission electron microscope (Zeiss Merlin, Germany) equipped with Oxford Instruments INCA A-Act EDX system, and monochromatized Al Kα X-ray photoelectron spectroscopy (PHI Quantera II, Japan). The thickness of the MoS2 deposition was measured on a spectroscopic ellipsometer (Sentech SENpro, Germany).

Electrochemical Studies

The MoS2-coated TNTAs were first cut into small pieces and covered with a laser-cut polyamide tape (Kapton) to accurately define the electrodes’ surface area; then, small pieces of samples were glued on copper plates by using a conductive double-side copper tape. Most of the electrochemical measurements were carried out in 0.1 M sulfuric acid solution in a three-electrode system with platinum mesh as the counter electrode and Ag/AgCl (3 M NaCl) as the reference electrode. All electrochemical measurements, including cyclic voltammetry, impedance spectroscopy, steady-state chronoamperometry, and linear sweep voltammetry, were carried out on Gamry Interface 1000 potentiostats at room temperature. The measured potentials (vs Ag/AgCl) were converted to standard hydrogen electrode (SHE) scale using the Nernst equation E(SHE) = E(Ag/AgCl) + 0.20 V.