Introduction

Ever since the seminal paper of photoelectrolysis of water by Fujishima and Honda1, TiO2 has received wide attentions in photocatalysts, water splitting and solar cells due to its high photoactivity, low cost and excellent chemical stability2,3,4,5. The limiting factor for TiO2 is the large band gap (~3.2 eV), which defines its light absorption only in the UV range. During the past three decades, tremendous efforts have been put to enhance the visible light harvesting ability of TiO26. Heterogeneous structures have been proposed to couple TiO2 with materials exhibiting visible light harvesting ability and the first trial was done by Serpone et al. to couple TiO2 with CdS which showed a significant improvement7. Later on Graetzel made a significant breakthrough in sensitizing TiO2 with dye molecules, viz., the dye-sensitized TiO2 photoanode4. Following the invention of Graetzel cell, quantum dot sensitized solar cells (QDSSC) quickly catch up following the mature quantum dot synthesis protocol developed by Peng and Alivisatos8,9. The key development of QDSSC was made by Kamat in 2005, with the pre-synthesized CdSe nanocrystals linked to TiO2 thin films by organic molecules10. Since then various methods of sensitization have been developed and they can be summarized into two main categories: assembly of pre-synthesized QDs and direct growth11,12. Pre-synthesis provides the feasibility of facile control in the size, size distribution and morphology. However, the charge transfer would be retarded by the surface functional molecules. Also the loading of the sensitizer prepared by this method is usually low. Direct growth allows both a compact contact of the sensitizer with TiO2 and the ease of increasing the loading of the sensitizer. A diverse range of methods are reported for the direct coating of the sensitizing materials, such as chemical bath deposition13,14, successive ionic layer adsorption and reaction (SILAR)15,16,17, electrochemical deposition18, chemical vapour deposition19 and electrophoretic deposition20.

Despite the development of various sensitization methods, the sensitizers still suffer from poor thickness and uniformity control especially for deposition on high aspect-ratio TiO2 nanostructures. As the size of QDs is much larger than dye molecules, penetration of QDs into TiO2 nanoarchitectures with a depth >10 µm is more difficult than the case in dye-sensitized solar cells11. Due to the quantum confinement effect and the limited charge diffusion length, the size of the QDs plays an important role in charge transfer process. The poor control in conventional deposition techniques usually leads to aggregation of QDs into large particles, thus causing high internal recombination loss.

Atomic layer deposition (ALD) is a thin film deposition technique that is based on self-limiting surface reactions by sequential exposure of the substrate to different gas phase precursors21. ALD provides precise thickness control at the angstrom or monolayer level and deposition on high aspect ratio nanostructures with excellent step coverage. By employing ALD for QD sensitization, excellent infiltration and conformity could be achieved and the size of QDs could be varied simply by tuning the number of ALD cycles. Recently Stacey et al showed the ALD CdS for solid state QDSSC22. However, there is a serious safety issue for such ALD CdS process, as the precursors of dimethyl cadmium (DMCd) and hydrogen sulfide (H2S) are highly toxic.

Liquid-phase reaction via ion exchange has been developed as a method to fabricate semiconductor heteronanostructures23,24. The principle of ion exchange is based on the solubility of the material which allows only a critical selection of the target materials. Very recently ion exchange reaction has also been employed to fabricate nanowire p-n junctions for photovoltaics25, as well as photosensitizer layers on ZnO nanowires for semiconductor-sensitized solar cells26,27. Due to its amphoteric property, ZnO is unstable for PEC in acidic or alkaline electrolyte solution, which is the bottleneck for practical applications. On the other hand, the amphipathicity of ZnO makes it an ideal template for nanofabrications28.

In order to have the merit of ALD without the highly toxic source for the QD sensitization, in this work, ALD ZnO thin films were deposited on various TiO2 nanostructures as the sacrificial templates to convert to short bandgap semiconductor sensitizers by ion exchange reaction (we call the whole process ALDIER). The thickness and the size of the sensitizer could be controlled by the thickness of the ZnO layer and the condition of ion exchange reaction. The reproducible photocurrent levels >15 mA/cm2 is obtained using the TiO2 inverse opals (TiIO), which is so far the highest among all nanostructure TiO2-based PEC cell for hydrogen generation.

Results

Figure 1(a) illustrates the flowchart of the ALDIER based on the example of TiIO obtained by replicating a self-assembled multilayer polystyrene spheres29,30. However, the TiO2 host structures for the 3-D homogeneous photosensitization can include a wide range of common types, for example, hydrothermal-grown nanorods, anodized nanotubes and commercial P25 nanoparticles. In the first step, the host is coated with a layer of ZnO of tunable thicknesses using ALD. The second step is the ion exchange reaction. In this process the ALD ZnO layer serves as the sacrificial reactant which transforms first into ZnSe through anion exchange with the Se2- precursor and then to the final CdSe via cation exchange with Cd2+ source27. The ion exchange reaction is based on the solubility product constant (Ksp) of the material. As the constant of ZnO (Ksp = 6.8 × 10-17) is much larger than ZnSe (3.6 × 10-26), ZnO can be converted into ZnSe by anion exchange reaction with Se2- anions. With the same principle, ZnSe can be further exchanged into CdSe (6.31 × 10-36).

Figure 1
figure 1

Sensitization of TiO2 inverse opals photoelectrochemical anode by ALDIER.

(a) Schematics of the ALDIER process for uniform QD photosensitization of TiO2 inverse opals (TiIO). Step 1: coating the TiIO with ALD ZnO layer. Step 2: anion exchange reaction converting the initial ZnO layer to ZnSe. Step 3: cation exchange reactions converting the intermediate ZnSe to CdSe. (b) Schematics of the photoelectrochemical cell: The as-prepared nanostructure serves as the anode, saturated Ag/AgCl as the reference electrode and the Pt foil as counter electrode for hydrogen evolution. (c) Electron energy levels of TiO2 and CdSe. The photogenerated electrons within CdSe will be transferred to TiO2, while the holes (not drawn) will be scavenged by the Na2SO3+ Na2S electrolyte solution. CB: conduction band. VB: valence band.

The sensitized TiIO is tested as the PEC photoanode in a three-electrode system for hydrogen generation. Figure 1(b) shows the diagram of the PEC cell where a piece of Pt foil was used as the cathode. H2 gas bubbles are generated on the cathode through the water reduction reaction 2H+ + 2e → H2, whereas the holes are scavenged by the sulfide electrolyte. As CdSe has a wider light absorption range and better conduction band edge alignment with TiO2 than ZnSe (see Fig. 1c), the photocurrent level of TiO2/CdSe is much higher than TiO2/ZnSe, as also reported in previous work27. Thus in this experiment we focus on CdSe photosensitizer on TiIO.

The composition transformation from ZnO to CdSe is confirmed using X-ray diffractometer (see Fig. 2a). The blue solid lines are the calculated XRD peaks for Anatase TiO2 (PDF#21-1272) and the green dashed lines are the calculated XRD peaks for SnO2 (PDF#41-1445). All the diffraction peaks of TiIO structure on FTO correlate well with the calculated positions. The diffraction peaks of ZnO (PDF#36-1451) appear only for the TiIO samples after the deposition of ZnO. After the first step anion exchange reaction, the ZnSe peaks (PDF#37-1463) can be detected while the ZnO peaks disappear, implying that the ALD ZnO layer is totally converted to ZnSe. After the complete exchange reaction, the peaks can be indexed to hexagonal CdSe (PDF#08-0459). The peak intensities of the CdSe increase from the 60-cycle sample to 120-cycle one, in accordance with the increased loading of CdSe.

Figure 2
figure 2

Composition transformation.

(a) XRD patterns of the pristine TiIO and the derived structures after ALD and ion exchange reactions. The green dashed lines indicate the peaks of FTO and the blue solid lines indicate peaks of anatase TiO2. (b) UV-vis diffuse reflection spectra of three photoanode samples. TiIO: pure TiO2 inverse opal. TiIO/ZnSe: ZnSe-coated TiIO after the anion exchange reaction. TiIO/CdSe: CdSe-coated TiIO after the cation exchange reaction. Insets are the photographs of the samples on FTO-coated glass.

UV-visible diffused reflection spectra are recorded to reveal the light harvesting capability of the pristine TiIO, ZnSe and CdSe sensitized TiIO photoanodes (results shown in Fig. 2b). The pristine TiIO can only absorb light with a wavelength below 400 nm. The absorption range broadens up to 550 nm after the exchange to ZnSe and 700 nm after the final exchange into CdSe, consistent with their bandgaps. More diffuse reflection spectra for the samples at intermediate conversion stage are shown in Supplementary Fig. S1. The composition change after anion and cation exchange reactions and the light harvesting ability can also be revealed from the photographs of the samples in Fig. 2b inset. The samples with CdSe coating appear dark brown compared with the bright yellow coloured ZnSe and white pristine TiIOs.

Morphologies of the photoanodes prior to and after the ion exchange reactions with different ALD ZnO cycles are shown by SEM images in Fig. 3. The original TiIO surfaces are smooth. After ALDIER, the open surfaces are coated with a granular layer of CdSe. The SEM image in Fig. 3e provides an entire cross-sectional view of the structure with 60 ALD cycles (larger scale SEM images are provided in Supplementary Fig. S2). Clearly the ion exchange reactions proceed along the whole depth of 10 μm, owing to the effectiveness of ALD in creating conformal coatings on the surfaces of high aspect-ratio nanostructures. The morphology is further characterized using transmission electron microscopy (TEM). Figure 4a provides clear evidence of the attachment of nanoparticles on the available surfaces of TiO2 pores. High-resolution TEM image in Fig. 4b shows that the particles are crystalline and the lattice spacing can be indexed into (0-11) and (1-1-1) planes of hexagonal CdSe. Pure elements of Cd and Se are confirmed by the X-ray energy dispersion spectrum shown in Fig. 4c.

Figure 3
figure 3

SEM characterization.

(a) Top view of the pristine TiO2 inverse opal. (b-d) Top view images of the ALDIER samples, where the CdSe layers were converted from the ZnO starting layers obtained by using 60, 90 and 120 ALD cycles, respectively. Scale bars: 100 nm. (e) Side view of the entire cross section of the photoanode in (b) showing the uniform sensitization from top to the bottom. Inset: enlarged view of part of the cross section.

Figure 4
figure 4

TEM characterization.

(a) Low-magnification TEM image of the CdSe nanoparticle-sensitized TiO2 inverse opal. (b) Atomic-scale TEM image of one CdSe nanoparticle. Inset is the corresponding fast Fourier transformation pattern. (c) X-ray energy dispersion spectrum (EDS) recorded from the CdSe nanoparticle in (b).

The PEC performance of the ALDIER photoanodes is investigated by conducting the current density vs potential (J-V) measurements under the dark and simulated sunlight illumination (AM 1.5) in the three-electrode cell configuration (Fig. 1b). First of all, all the electrodes are fabricated from 500 nm polystyrene spheres and have the same height of 10 µm. For ion exchange reaction, temperature is a critical parameter that affects the size and morphology of the resulting sensitizer layer. Hence the first step is to optimize the reaction temperature for CdSe exchange reaction while keeping ALD ZnO thickness the same (~10 nm by 60 ALD cycles). As shown in Supplementary Fig. S3, the electrode obtained from the reaction at 120°C exhibits the highest photocurrent level (15.7 mA/cm2 at zero bias vs Ag/AgCl) compared to those from reactions at 90°C (4.6 mA/cm2) and 140°C (10.5 mA/cm2). Therefore, for the rest experiments, the reaction temperature is fixed at 120°C. Figure 5a presents the J-V curves of TiIO/CdSe anodes obtained from different ZnO ALD cycles. All the photoanodes present negligible dark currents while showing current densities above 10 mA/cm2, upon illumination. This implies an efficient light harvesting by the sensitizers (see also IPCE below) and charge separation at the TiO2 and CdSe interface rendered by the type-II band alignment (see Fig. 1c). All the samples show similar onset potentials (−1.0 V vs Ag/AgCl), as a result of similar surface flat bands. The photocurrents saturate with increasing bias voltage, indicating the good electrical conductivity of the TiIO/CdSe and the good contacts with the FTO.

Figure 5
figure 5

Photoelectrochemical properties of the ALDIER TiO2 inverse opal photoanodes.

(a) Linear sweep voltammagrams (J-V curves) under dark condition and AM1.5 light illumination for samples by ALDIER technique with different ALD cycles and one by 6 SILAR cycles. (b) Plot of the photocurrent density at zero bias in (a) versus the ALD ZnO cycle. (c) Photocurrent versus time tests (J−t curves) under chopped light illumination (light/dark cycles of 50 s) at a fixed bias of 0 V vs Ag/AgCl. (d) IPCE profile of the ALDIER photoanode with 60 ALD cycles. For comparison, the corresponding data of the optimized SILAR anode (6 cycles CdSe) are also shown.

We now discuss the effect of ALD ZnO thickness. Figure 5a-c show that the sample from 60 ALD cycles gives the highest photocurrents; further increment of the number of ALD cycles to 90 and 120 slightly bring the photocurrent down to ~13 mA/cm2. The lowering of the current with higher ALD cycles might originate from adverse effects such as increased carrier recombination within the thicker CdSe layers. It is noted that a thicker ALD ZnO layer corresponds to a larger CdSe nanoparticle size. The photocurrent versus time (J−t) curves in Fig. 5c show that all the electrodes have good photoresponse and relatively good stability. For a comparison, a series of control samples of the same height are sensitized with CdSe via the SILAR method and measured under the same condition (see SEM images in Supplementary Fig. S4). Among samples with 3, 6 and 9 SILAR cycles, the highest photocurrent is ~6.5 mA/cm2 obtained from the 6-cycle SILAR sample (also presented in Fig. 5a). The CdSe nanoparticles appear to have a lower coverage than that by ALDIER based on SEM inspection, which is consistent with the diffuse reflectance data, that is, the SILAR samples have a higher diffuse reflectance compared to that of ALDIER samples (see Supplementary Fig. S5 and Fig. S1). It is noteworthy that the photocurrent level obtained from our ALDIER electrodes is very high among nano TiO2 photoanodes, even higher than the previously reported TiO2 nanoparticle electrode with CdS and CdSe co-sensitization (14.9 mA/cm2)31.

While the voltammagram J-V curves show the overall PEC performance of the photoanode, the insight of the wavelength-dependent photocarrier generation can be obtained by studying the incident-photon-to-current conversion efficiency (IPCE), which allows the evaluation of the wavelength-dependent light harvesting efficiency. The IPCE profile from the 60-ALD cycle sample is shown in Fig. 5d. One can see that, within the test range of 300−800 nm wavelength, a strong and nearly constant photoresponse is observed in the visible region from 305 to 750 nm; the efficiency lies between 60−70% with a nearly flat profile. In contrast, the IPCE of the photoanode by SILAR sensitization is significantly lower, probably due to the aforementioned lower coverage of CdSe and the non-uniform sensitization, in correlation with its higher light reflectivity.

Discussion

The significant performance enhancement of ALDIER anodes compared with SILAR ones implies the superior quality of the sensitizer coating by this simple yet powerful ALD plus ion exchange technique, which results in an excellent penetration depth, uniform coverage and increased amount of loading of the sensitizers. As a result, both the light harvesting ability and the charge transfer would be enhanced. It is envisaged that the performance could be further improved by ZnS passivation as reported previously32 or by doping of CdSe which is recently developed by the Kamat group33.

To generalize this ALDIER technique to various nanostructures, same experiments are also conducted to TiO2 nanorod arrays and nanoparticle films (see Supplementary Fig. S6 and S7). Shown in Fig. S6 are the results from the commercial P25 nanoparticle film. Clearly, one can also see the very low diffused reflectance and a high photocurrent level up to 12 mA/cm2. The detailed study of the interface between CdSe and TiO2 by the HRTEM and the inner charge transfer process by ultrafast optical spectroscopy is currently underway to provide a comprehensive physical understanding.

The demonstrated fabrication process can also readily result in CdS and PbSe if the anion exchange precursor is S2- and the cation exchange precursor is Pb2+, respectively. Due to the high conformity of the ALD ZnO layer34, this simple two-step ALDIER process can be utilized for homogeneous coating of photoactive CdS, CdSe, PbS, or PbSe nanoparticles onto a wide range of complex and high-aspect-ratio substrates. Furthermore, as the thickness of the ZnO layer is precisely tunable by ALD cycles, the self-limiting ion exchange reaction will thus lead to different size and coverage of the final sensitizer nanoparticles.

To summarize, a new strategy based on ion exchange reaction using ALD ZnO layer as sacrificial template has been developed to coat TiO2 nanostructures with a homogeneous layer of CdSe photosensitizer. With the optimized ion exchange reaction temperature and ALD ZnO thickness, the highest photocurrent for the TiIO/CdSe electrode reaches 15.7 mA/cm2 at zero bias (versus Ag/AgCl), which is the highest value among TiO2-based photoelectrochemical cells for hydrogen generation. Further in-depth insights of the enhancement mechanism will be studied in more details. This sensitization method can be generalized to other sensitizers like CdS, PbS and PbSe. It is optimistic that such ALDIER method is ready to be extended to other TiO2 nanostructures including anodized nanotubes and to other electron transporting materials like SnO2.

Methods

Fabrication of TiO2 Inverse Opals

Carboxylate-modified, monodispersed polystyrene spheres of 500 nm diameter (Duke scientific corporation) were assembled onto the Fluorine-doped SnO2(FTO) coated glass substrates via a vertical deposition process at 90°C29,30. The self-assembled polystyrene spheres opals were then infiltrated with TiO2 using a stop-flow-reactor ALD system at 70°C, for which titanium tetrachloride (99.99%, Sigma Aldrich) and H2O were used as the Ti and O precursors, respectively. Finally, TiO2 inverse opal structures were developed by burning the original polystyrene spheres in air at 450°C for 2 h, which also improved the crystallinity of the TiO2. Last, reactive ion etching (RIE, NSC ES371) was used to cleave the top surface and open up the pores.

Preparation of TiO2 particle film

TiO2 nanoparticle films were deposited on the FTO by successive screen-printing using a TiO2 paste consisting of Degussa P25 TiO2 powder and an ethyl cellulose binder in α-terpinol35. The projected area of the TiO2 layers was approximately 0.28 cm2 (circles with 0.6 cm diameter). Then, the TiO2 electrodes were gradually heated to 450°C where they were held for 15 min before being heated to 500°C for a further 30 min.

ALD ZnO layer and Ion Exchange Reactions

ZnO layers with different thickness were conformably deposited onto the TiIO as the sacrificial layers for the ion exchange reaction by ALD with the Diethyl zinc (DEZ, 99.99%, Sigma Aldrich) and H2O as the Zn and O precursors, respectively. The ZnO coated TiIOs were then annealed in air at 450°C for 30 minutes to improve the crystallinity. TiIO/ZnSe core/shell structure was prepared by immersing the ZnO coated TiIO in a Se2- ion solution (0.05 M, prepared by reacting 0.79 g Se powder with 0.8 g NaBH4 in 200 ml deionized water) and kept at 60°C for 5 hours in order to fully exchange the ZnO layer into ZnSe27. The samples were then washed with deionized water and absolute ethanol and fiwater dried in air. Due to the fact that Se2- is vulnerable to oxygen, all the experiments containing Se2- ions were done in glove box. The TiIO/ZnSe core/shell structures were then reacted with the 0.1 M CdCl2· 2.5H2O aqueous solution at 90−140°C for 10 hours to replace Zn2+ by Cd2+ in the ZnSe shell. By controlling the reaction temperature, the composition of the TiIO/ZnxCd1-xSe core/shell structure could be tuned. Finally, TIO/ZnxCd1−xSe core/shell nanostructure was annealed in argon ambient at 400°C for 30 minutes to improve crystallinity.

Successive Ionic Layer Adsorption and Reaction (SILAR) of CdSe on TiIO

The SILAR process was modified from the previous reports15,17. In a typical procedure, the TiIO electrodes were immersed in a solution containing 0.05 M cadmium acetate dihydrate (Cd(Ac)2·2H2O, Alfa Aesar, 98%) in ethanol for 1 min, to allow Cd2+ to adsorb onto the TiO2 surface and then rinsed with ethanol for 1 min to remove the excess Cd2+. The electrodes were then dried for 2 min in an argon atmosphere. Subsequently, the dried electrodes were dipped in to a solution containing 0.05 M Se2- for 1 min. The Se2- solution was prepared by mixing selenium (Se, Sigma-Aldrich, 99.8%) and sodium borohydride (NaBH4, Sigma Aldrich, 99.8%) in water. The electrodes were then rinsed in ethanol for 1 min and dried again in an argon atmosphere for another 2 min. This procedure was repeated several times to get desired CdSe loading.

Materials Characterizations

The morphology and microstructure of the nanostructured films were examined using a JEOL JSM-7600F field emission scanning electron microscopy (FE-SEM) and a JEM 2100F transmission electron microscope (TEM). The X-ray diffraction (XRD) patterns were recorded by Shimadzu thin film XRD equipment using Cu Kα radiation. The diffuse reflection spectra were taken using Zolix Solar Cell QE/IPCE Measurement System equipped with an integrating sphere and a silicon diode detector.

Photoelectrochemical Characterizations

The PEC performance measurements were conducted in three electrodes configuration with the as prepared nanostructured photoanodes as working electrodes, Ag/AgCl in saturated KCl as a reference electrode and Pt foil as the counter electrode. 0.24 M Na2S and 0.35 M Na2SO3 mixed aqueous solution was used as the electrolyte. The current density vs potential (JV) measurements were measured in both dark and illumination with a 150W Xe lamp (Science tech SS150) equipped with an AM1.5 G filter, calibrated with a standard Si solar cell to simulate AM1.5 illumination (100 mW/cm2). Photocurrent versus time (J−t) tests were carried out by measuring the currents under chopped light illumination (light/dark cycles of 50 s) at a fixed bias of 0 V versus Ag/AgCl. The incident-photon-to-current conversion efficiency (IPCE) measurements were taken as a function of wavelength from 300 to 800 nm using a specially designed IPCE system for solar cells (Zolix Solar cell Scan100), with three electrodes configuration under zero bias versus Ag/AgCl. A 300 W Xe lamp equipped with gratings was used to generate a monochromatic beam. The incident light intensity was calibrated by a standard silicon photodiode.