Quantum-Dot-Sensitized TiO₂ Inverse Opals for Photoelectrochemical Hydrogen Generation

With the increasing concern over the global energy crisis and the greenhouse effect by carbon dioxide emissions, development of clean and sustainable energy solutions to the alternatives of traditional fossil fuels has attracted considerable interest in both the scientific and industrial community.1, 2 Photoelectrochemical (PEC) water splitting provides a promising approach to address the above issues simultaneously by capturing and storing solar energy in the chemical bond of H₂.3–6 Titanium dioxide (TiO₂) has been one of the most attractive materials in this area due to its high photoactivity coupled with low cost and excellent chemical stability.7 However, the only obstacle is its large bandgap (≈3.2 eV), which results in limited solar-light harvesting. In the past few decades, tremendous effort has been made to improve the water splitting efficiency. In general, two main approaches have been developed to enhance the visible-light absorption. One is narrowing of the bandgap of TiO₂ by doping with either transition-metal ions8 or nonmetal elements such as N9 and C;10 the other is to sensitize with narrow-bandgap semiconductor quantum dots (QDs) such as CdS,11, 12 CdSe,13 CdTe,14 and so on.

Furthermore, a purpose-designed nanoarchitecture of the photoelectrode is equally essential to the performance, which leads to increased surface area, improved electron transportation, and reduced minority-carrier diffusion length and electron–hole recombination loss.15, 16 In particular, periodically ordered nanostructures like nanowire and nanotube arrays17–20 and three-dimensional (3D) inverse opals21–27 have been demonstrated as beneficial to the enhancement of electron transport and light trapping arising from the increased optical path length by multiscattering. Among them, the inverse opal, which is a replicated shell structure of a face-centered-cubic (fcc) opal, offers very high specific surface area and porosity (74% void volume). Besides, the periodical 3D inverse opal can provide an additional photonic bandgap effect to enhance the light–matter interactions by controlling the propagation of light via back reflections, slow photons, and surface resonant modes.28–30 As such, the 3D inverse opal structure is expected to be an ideal electrode design for energy conversion applications. Recent studies have shown promising results on the applications of TiO₂ inverse opals in dye-sensitized solar cells21–26 and photocatalysis.31–33 However, to our knowledge, there has so far been no study on the TiO₂ inverse opal structure as photoanode in PEC water splitting.

Herein, we report an innovative electrode design by the combination of a TiO₂ inverse opal with CdS QD sensitization for solar-light-driven hydrogen production. The 3D percolated periodical pore structure of the TiO₂ inverse opal provides a high surface area for QD loading plus a good electrical transport path and intimate contact with the electrolyte. While the QDs acting as “light antennas” can greatly improve the visible-light harvesting, the type II band alignment with TiO₂ also favors interfacial charge transfer and separation. Moreover, the photonic bandgap feature facilitates photon–QD interaction, as a result of enhanced light absorption. The TiO₂ inverse opals in this study were fabricated using self-assembled polystyrene (PS) spheres as sacrificial templates and subsequent TiO₂ infiltration using atomic layer deposition (ALD). It should be noted that ALD, as a cyclic self-limiting deposition method, provides an exceptional filling control (close to a theoretical filling fraction)34, 35 with minimization of cracks induced by volume shrinkage in comparison with the traditional sol–gel infiltration route.

The scheme in Figure1 illustrates the experimental procedure for the preparation of CdS QD sensitized TiO₂ inverse opals, which consists of four simple steps: 1) self-assembly of close-packed PS spheres with multilayer structures onto fluorine-doped tin oxide (FTO) glass substrates, 2) infiltration of the opal with TiO₂ by ALD, 3) removal of the PS sphere templates by thermal decomposition to obtain the TiO₂ inverse opal, and 4) final sensitization of the inverse opal with CdS QDs using a successive ionic layer adsorption and reaction (SILAR) route. Typical scanning electron microscopy (SEM) images of PS colloid films prepared from 288 nm PS spheres are presented in Figure2a and b, which clearly demonstrate a dense fcc arrangement of monodispersed spheres with the (111) plane oriented parallel to the underlying glass substrate. After the TiO₂ infiltration and calcination, an ordered TiO₂ inverse opal with spherical voids is obtained. As shown in Figure 2, all the inverse opals fabricated from PS sphere templates with 288, 510, and 900 nm sizes show ordered periodicity and wide pore interconnectivity. This important characteristic is beneficial to effective CdS QD sensitization as well as unhindered electrolyte infiltration throughout the entire thickness of the photoanode. The typical thickness of the inverse opal films is around 8 μm (Supporting Information, Figure S1). It is notable that the void size of the inverse opals decreased by a minimal 15% compared to that of the original PS sphere templates, owing to the shrinkage of the structure during calcination.

Figure 1. Schematic of the fabrication procedure of the 3D QD-sensitized TiO₂ inverse-opal photoanode.

Figure 2. a,b) SEM images of PS sphere opal films at different magnifications. Inset in (a): photograph of PS sphere films on FTO substrate. c–f) TiO₂ inverse opals obtained from templates of different sized PS spheres: c) 288, d,e) 510, and f) 900 nm.

The SEM image in Figure3a of the QD-sensitized 288 nm TiO₂ inverse opal shows that the well-ordered inverse structure is preserved and a uniform coverage of the QD layer can be seen without evident aggregation and pore clogging. More SEM images of QD-coated 510- and 900-nm-diameter TiO₂ inverse opals are shown in the Supporting Information (Figure S2). The transmission electron microscopy (TEM) image in Figure 3b gives a more detailed view of the TiO₂ opal surface, and shows clearly the coverage of CdS on both inner and outer surfaces of the TiO₂ opal. Both the TiO₂ and CdS have good crystallinity, as seen from the high-resolution TEM (HRTEM) image in Figure 3c. The thickness of the CdS layer is around 7 nm. The 0.35 and 0.26 nm fringes of the TiO₂ and CdS QDs correspond to the d spacing of (101) planes of anatase TiO₂ and the (102) planes of hexagonal CdS, respectively. More characterizations by energy-dispersive X-ray spectroscopy (Supporting Information, Figure S3) and X-ray diffraction (XRD; Figure S4) further confirm that the composites consist of anatase TiO₂ and hexagons of CdS. Through a profile refinement of the XRD pattern (Figure S4), the size of the CdS nanoparticles was determined to be 7.3 ± 0.4 nm, which is consistent with the observed sizes under TEM.

Figure 3. SEM and TEM images of CdS QD sensitized 288-nm-diameter TiO₂ inverse opals. a) 20° tilted view at the crack area. b) TEM image of several QD-coated inverse TiO₂ spheres. c) HRTEM image from the CdS/TiO₂ interface. d) Magnified HRTEM image of the CdS QD in (c). The inset is the corresponding Fourier-transform electron diffraction (ED) pattern of the QDs observed in zone axis [0001].

The formation of high-quality TiO₂ periodic structures would prohibit certain wavelengths of light from transmitting, by correlating to the cavity sizes and the refractive index of the surrounding medium. According to the modified Bragg's law36 as shown in Equation 1, the stop-band reflection peak position can be estimated:

• ((1))

where λ_max is the maximum wavelength of the reflected peak (the position of the photonic bandgap) and d_hkl is the interplanar spacing between hkl planes. In first-order Bragg diffraction from fcc (111) planes, d_hkl = 0.8165D, where D refers to the sphere's diameter of the opal or inverse opal. The n_s and n_f are the refractive indices of the spheres and their surrounding medium, respectively, while f is the volume of fraction of the spheres. The θ represents the angle of incident beam to the planes (70° in our case).

Figure4a shows the specular reflectance spectra of TiO₂ inverse opals fabricated from 288-, 510-, and 900-nm-diameter PS spheres (left to right). The corresponding stop-band reflectance peaks at 514, 981, and 1630 nm as well as their associated well-defined peaks from higher-energy bands are clearly observed, evidence of the highly ordered opal lattice in these samples. With the decreasing size of the TiO₂ inverse opal, the stop-band peak positions are blue-shifted as expected. After the CdS QD sensitization, the stop-band peak positions are red-shifted accordingly due to the increased refractive index (n_CdS ≈ 2.5, n_TiO2 = 2.2–2.4, n_air = 1). Notably, the reflectance peaks are still very sharp in intensity and highly symmetrical with almost unchanged bandwidths, thereby implying the uniform coverage of CdS without causing any damage to the structural order of the TiO₂ inverse opal. Further, Bragg peak calculations using Equation 1 after the sensitization indicate that inverse opals of all sizes are infiltrated with ≈7−8 nm thickness of CdS QDs, which is in good agreement with the SEM and TEM observations. By infiltrating with sulfide electrolyte, the stop-band peaks further red-shifted to a higher wavelength, again indicating the complete filling of electrolyte within the air voids of TiO₂.

Figure 4. Specular reflectance spectra at 20° incidence for a) TiO₂ inverse opals, b) CdS QD sensitized TiO₂ inverse opals, and c) electrolyte-infiltrated TiO₂/CdS inverse opals fabricated from 288, 510, and 900 nm sized PS sphere templates.

To evaluate the PEC performance of TiO₂ inverse opals with different pore sizes as photoanodes, current density versus potential measurements under dark and simulated sunlight illumination (AM 1.5, 100 mW cm⁻²) in a three-electrode cell were conducted. As shown in Figure5a, under dark conditions, all three photoanodes show a very small photocurrent density in the range of 10⁻² mA cm⁻², whereas under light illumination, pronounced photocurrent densities are observed, which implies efficient light harvesting and charge separation at the photoanode/electrolyte interface. Typically, the photocurrent increases from the onset potential around –1.0 V versus Ag/AgCl and approaches a plateau at –0.5 V versus Ag/AgCl. It is noteworthy that the photocurrent density is enhanced by ≈200 times through CdS QD sensitization in comparison to that of a pristine TiO₂ inverse opal photoanode (Supporting Information, Figure S5). This confirms that efficient CdS sensitization can enhance the light harvesting greatly, enabled by the effective surface decoration on the accessible voids of the inverse structure. It can be seen that the photocurrent density at 0 V versus Ag/AgCl increases from 3.02 mA cm⁻² for the 900 nm PS photoanode to 4.84 mA cm⁻² for the 288-nm-diameter one. The maximum photon-to-hydrogen conversion efficiency calculated from the equation37 for the 288-nm-diameter photoanode was ≈2.62%. The efficiency enhancement with the decrease of the void size is mainly due to the higher total surface area for smaller-pore photoelectrodes, which leads to higher QD loading. The transient photocurrent curves (Figure 5b) demonstrate that all three electrodes have a fast photoresponse speed and good photostability.

Figure 5. a) Linear sweep voltammetry measurements of the TiO₂ and QD-sensitized TiO₂ inverse-opal photoanodes, collected both in the dark and under 100 mW cm⁻² simulated sunlight illumination. b) Amperometric I–t curves at an applied potential of 0 V versus Ag/AgCl at 100 mW cm⁻² illumination with 50 s light on/off cycles. c,d) Measurements of the incident photon-to-electron conversion efficiency (IPCE) and e,f) electrochemical impedance spectra of the CdS-sensitized TiO₂ inverse-opal photoanodes with three opal sizes. c) Original IPCE curves; d) normalized IPCE spectra at 440 nm where none of the three electrodes (288, 510, and 900 nm sizes) possesses any reflective characteristics; e) Nyquist plots; f) Bode phase plots.

Incident photon-to-electron conversion efficiency (IPCE) tests of the three photoanodes at wavelengths from 300 to 650 nm were performed. The IPCE spectra in Figure 5c exhibit a strong photoresponse for all the photoanodes from 350 to 550 nm, which is in accordance with the absorption edge characteristics of TiO₂ and CdS. Maximum IPCE values of 30, ≈20, and 11% were achieved for 288-, 510-, and 900-nm-diameter sensitized TiO₂ inverse opal photoanodes, respectively, with the 288 nm one having superior quantum efficiency in most of the wavelength range. This result is consistent with the corresponding J–V characteristics. However, upon a closer inspection of the IPCE curves, the photoanodes with different pore diameters showed a slightly different photoresponse behavior relative to the incident light wavelength, which might be correlated with their respective reflectance properties (see below for a discussion based on normalized IPCE spectra).

To ensure identical light exposure conditions at the position of normalization, the spectra were normalized at the wavelength position of 440 nm where all three electrodes (288, 510, and 900 nm) do not possess any reflectance characteristics. As can be seen from Figure 5d, the 900 nm photoanode exhibits the highest IPCE value in the range of 440–525 nm, and reaches a maximum at around 490 nm corresponding to the absorption edge of CdS QDs. Referring to the reflectance spectra in Figure 4c, the 288 and 510 nm sized photoanodes possess higher-order photonic bands in the wavelength ranges of 360–430 and 450–550 nm, respectively, which can be accounted for by the lower efficiency in these wavelength regions due to the loss of photons from reflections. On the other hand, it is noted that the 900-nm-pore photoanode does not show reflectance peaks below 400 nm, but presents the lowest efficiency in this region. This may be due to its lowest surface area, which results in very low UV light harvesting by TiO₂ and thus near zero absolute IPCE values. One step further, we expect that the spectrally selective reflectance characteristics of TiO₂ inverse-opal electrodes might also be useful for the design of photonic crystal bilayer photoanodes by coupling TiO₂ inverse opals with mesoporous nanoparticle films or nanowire structures, in which the photonic crystals act as “dielectric mirrors” to enhance light harvesting.

Electrochemical impedance spectroscopy (EIS) was used to investigate the charge-transport and -transfer processes. The Nyquist plots of the obtained EIS data measured at open-circuit conditions under simulated solar-light illumination are shown in Figure 5e. The high-frequency (>10⁴ Hz) semicircular portion in the Nyquist plots corresponds to the reduction reaction at the Pt counter electrode, and the second semicircle in the middle frequency range (1–10³ Hz) represents the charge transfer at the TiO₂/CdS/electrolyte interface. The measured impedance spectroscopic data fitted well with the diffusion–recombination mode for a porous electrode38 using an equivalent circuit with Zsimpwin software. With increasing pore size, the TiO₂ inverse-opal electrode shows decreased charge-transfer resistance, thus indicating higher recombination losses in the cells with larger-pore-sized electrodes. The electron lifetime (τ_n) is correlated with the characteristic frequency peaks in Bode phase plots, that is, . As presented in Figure 5f, the frequency value slightly decreases with the reduction of pore size, from ≈2.6 to ≈1 Hz, which suggests a slower recombination occurring in the smaller-pore-sized electrode. As a result, the electrons have more chances to be collected by FTO. These EIS results confirm that the TiO₂ opal films with smaller pore diameters are more beneficial to the reduction of recombination losses, in accordance with the J–V and IPCE results.

In conclusion, we have demonstrated a novel nanoarchitectured electrode design through the use of 3D CdS QD sensitized, optically and electrically active TiO₂ inverse opals for PEC hydrogen production. A promising photocurrent density of 4.84 mA cm⁻² has been achieved for the 288 nm inverse opal as photoanode at 0 V versus Ag/AgCl bias under simulated solar-light illumination. The highly ordered and percolated 3D pore structures provide efficient and fast electron-transport pathways. Several new possibilities are expected for further efficiency enhancement by photonic-bandgap engineering to realize back reflection, surface resonant modes, and slow photons via modulation of the cavity size of the TiO₂ inverse opal, coupling of an additional semiconductor photoanode structure, and the use of other types of semiconductor QD sensitizers.

Experimental Section

Preparation of TiO₂ Inverse Opals: Fluorine-doped SnO₂ (FTO)-coated glass substrates (15 Ω □⁻¹) were cleaned by immersion in acetone, ethanol, and deionized water for 10 min each with sonication and dried under a N₂ stream. After that, commercial monodispersed carboxylate-modified PS spheres with different diameters (Thermo Scientific, d = 288, 510, and 900 nm) were assembled onto the FTO-coated glass substrates via a vertical deposition process at 90 °C.25 Subsequently, the self-assembled PS opals were infiltrated with TiO₂ by a custom-designed stopped-flow reactor type of ALD system at 70 °C, thereby providing conformal layers around the PS spheres of controllable thickness.35 Titanium tetrachloride (99.99%, Sigma–Aldrich) and H₂O were used as the Ti and O precursors, respectively. Finally, the original PS spheres were removed by calcination in air at 450 °C for 2 h, to leave spherical voids in the TiO₂ structure.

CdS QD Sensitization: The TiO₂ inverse opals were sensitized with CdS QDs using a modified previously reported SILAR route.12 In a typical procedure, the TiO₂ inverse opal films were immersed in a solution containing 50 mM cadmium acetate dihydrate (Cd(Ac)₂·2H₂O, Alfa Aesar, 98%) in ethanol for 1 min, to allow Cd²⁺ to adsorb onto the TiO₂. The electrodes were then dried with a N₂ stream. The dried electrodes were dipped into a solution containing 50 mM sodium sulfide nonahydrate (Na₂S, Alfa Aesar, 98%) in methanol for 1 min, where the preadsorbed Cd²⁺ reacted with S²⁻ to form the desired CdS. The electrodes were then rinsed in water for 1 min to remove the excess ions and dried again with N₂. This procedure was repeated ten times to obtain the desired thickness of CdS. To further increase the crystallinity of CdS, the as-prepared sensitized samples were annealed at 400 °C for 30 min in an Ar protective atmosphere.

Characterization: The morphology and microstructures of the synthetic TiO₂ and CdS QD sensitized TiO₂ inverse opal films were examined using a JEOL JSM-6700F field-emission scanning electron microscope and a JEM 2010F transmission electron microscope. Samples for TEM investigation were prepared by detaching the inverse opal from the FTO glass substrate and dispersing in ethanol, followed by transferring one drop of the suspension onto a copper grid. The XRD patterns were recorded on a Bruker D8 Advanced diffractometer using Cu_Kα radiation. The specular reflectance spectra were collected at 20° with respect to the normal incidence of light through a UV–vis–near-IR spectrophotometer (Varian, Cary 5000).

Photoelectrochemical Measurements: The PEC performance of the electrodes was evaluated in a three-electrode configuration under back-side simulated AM 1.5G illumination using a potentiostat (CHI760D, CH Instruments), similar to our previous approaches.39 Ag/AgCl in saturated KCl (+0.197 V vs. normal hydrogen electrode, NHE) and a Pt foil were used as reference and counter electrodes, respectively. The electrolyte was a mixed solution of 0.25 M Na₂S and 0.35 M Na₂SO₃ with the pH adjusted to ≈9.5. The photoresponse was measured under chopped irradiation from a 150 W Xe lamp (Sciencetech SS150) equipped with an AM 1.5G filter, calibrated with a standard Si solar cell to simulate AM 1.5 illumination (100 mW cm⁻²). Photocurrent stability tests were carried out by measuring the photocurrent produced under chopped light irradiation (light/dark cycles of 50 s) at a fixed bias of 0 V versus Ag/AgCl. The IPCE was measured as a function of wavelength from 300 to 700 nm using a specially designed IPCE system for solar cells (Zolix Solar cell Scan100), with two-electrode configuration under 0 V bias. A 150 W Xe lamp and a monochromator equipped with gratings were used to generate a monochromatic beam. The incident light intensity was calibrated by a standard silicon photodiode. Electrochemical impedance spectra were measured on a potentiostat (CHI760 D, CH Instruments) by applying a bias to the open-circuit voltage over the frequency range of 10⁻¹ to 10⁵ Hz with a 5 mV amplitude under 100 mW cm⁻² illumination.

Supporting Information

Supporting Information is available from the Wiley Online Library or from the author.

Acknowledgements

A.I.Y. Tok would like to acknowledge the financial support by the Ministry of Education, Singapore, by the Tier 2 Academic Research Fund (Grant No. T208A1225, ARC 5/08).