Nanostructured Materials for Photolytic Hydrogen Production

Abstract

A hydrogen economy is often considered an attractive alternative to our current fossil fuel-based energy system. In order for such a hydrogen economy to become reality, several challenges associated with the production, storage, transportation and use of hydrogen must be solved. This chapter addresses the issue of hydrogen production. While the currently most widely used method to produce hydrogen is based on the conversion of fossil fuel resources and does not therefore fulfill the requirement of CO₂ neutrality, we discuss here the photolytic production of hydrogen via water splitting. This scheme is based on energy input from the most powerful and ultimately sustainable energy source mankind has at its disposal: the sun. Moreover, no carbon dioxide is released into the atmosphere, and the method has potential for cost-effective large-scale production.

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

1 Background and Introduction

Mankind is facing two challenges of hitherto unsurpassed importance and complexity, namely the “tera-Watt” and “tera-ton” challenges. The former refers to the issue of supplying a rapidly growing world population with larger and larger amounts of energy [1], while the latter encompasses finding ways to deal with the huge amounts of carbon dioxide being emitted into the atmosphere as a consequence of our fossil fuel-based energy system [2]. The formidable task is to identify an alternative energy system, which at the same time is sustainable, virtually unlimited, safe and clean. The task becomes yet more challenging when also considering economy and security of supply aspects.

The hydrogen economy is one of many suggested scenarios that may be able to address these tera-challenges. The basic idea of the hydrogen economy is that hydrogen—a light, diatomic gas with a vast energy content by weight, but a small energy content per volume—would replace the fossil fuels which form the backbone of our current energy system. Hydrogen reacts cleanly with oxygen in a highly exothermic reaction, with pure water being the only exhaust product. The energy stored in the chemical bond of hydrogen can thus be released by burning it in a combustion engine, or more efficiently by oxidizing it in a fuel cell. Apart from zero-emissions, hydrogen has further attractive properties, including that it can be stored and transported, which are crucial requirements addressing the typically intermittent nature of renewable energy sources. In order for hydrogen to become an economically viable option, several hurdles must be surpassed. Besides providing efficient, inexpensive and durable fuel cells, the production of hydrogen and its storage constitute two other important cornerstones of the hydrogen economy. The former aspect will be discussed in this chapter, while the latter is covered separately in different chapters of this book.

1.1 Overview of Current and Potential Future Hydrogen Production Methods

While hydrogen (H) is abundant on earth, molecular hydrogen (H₂) is usually not found in large quantities in nature. Most of the hydrogen is bound to carbon, oxygen or both, meaning that H₂ must be generated by extracting it from other hydrogen sources via highly endothermic processes. The energy required to drive these extraction processes may come from essentially any energy source, including renewable energy sources, as illustrated in Fig. 1. A key aspect to bear in mind is that hydrogen is not an energy source but an energy vector.

Fig. 1

Schematic illustration of various hydrogen production pathways using different primary energy sources and energy conversion schemes. The main application areas of hydrogen and unwanted emissions from selected processes are indicated on the right and left hand side, respectively

A large fraction of today’s hydrogen production is based on fossil fuel resources. The processes involved in the conversion of fossil fuels, such as the reforming of natural gas, coal gasification/liquefaction and coal electrolysis, require large amounts of energy, either in the form of heat or electricity, and all these methods are associated with the release of vast amounts of carbon dioxide. For the hydrogen economy to become a viable option of our energy future, we therefore have to access other feedstocks from which hydrogen can be produced, and we have to make sure that the energy required to transform these feedstocks to hydrogen is provided by a sustainable and carbon–neutral source. Alternative feedstocks under consideration include wood, other biomass, organic waste and water, where the first three options are not truly carbon neutral. The last option constitutes an entirely sustainable energy system, where hydrogen is produced from water and later on recovered back into water upon conversion in a combustion engine or a fuel cell. This chapter therefore focuses on hydrogen production from water. In principle, water can be split into hydrogen and oxygen along several different pathways and utilizing various (sustainable) energy sources. The method used most commonly today is to dissociate water in an electrolysis cell, where the required electricity may be produced from nuclear, geothermal, gravitational or solar energy.¹ Solar energy is clean, quiet, sustainable, and its potential is enormous. The solar influx, i.e., the energy received by the planet from the sun $\left( { 1 7 4 {\text{ petawatts }} = { 1}. 7 { } \times { 1}0^{ 1 7} {\text{W}}} \right)$, surpasses the current world energy consumption rate (ca. $1 5 {\text{ TW }} = { 1}. 5 { } \times { 1}0^{ 1 3} {\text{W}}$) by more than four orders of magnitude. In other words, less than one hour of sunshine would, in principle, be sufficient to provide us with energy for an entire year. Given this enormous potential, various forms of solar energy, including exploiting solar energy for heating and cooling purposes, converting solar energy into electricity using photovoltaic materials or solar thermal technologies and converting solar energy into storable chemical fuels, as discussed in this chapter, have recently received considerable attention.

Since electrolysis is a two-step process, the potential of achieving a highly efficient, simple, and at the same time cost-effective conversion is limited. A more appealing idea is to directly split water into its component gases using a single device without the prior production of electricity, i.e., with no external electron flow. High-temperature thermochemical cycles, i.e., a set of coupled, thermally-driven chemical reactions that sum to the decomposition of water into H₂ and O₂, and photolysis (the direct splitting of water using light as the only energy input), represent two approaches in this direction. The primary heat sources, which are discussed in the context of high-temperature thermochemical cycles, are nuclear energy and solar energy. While thermochemical cycles potentially can be run at high efficiency, they currently either require very high temperatures (> 1500°C), where material challenges are significant, or they involve aggressive chemicals. Photolysis-based approaches, on the other hand, can run at ambient temperature because the splitting of water is not thermally induced but achieved via electron–hole pair driven redox reactions. To this end, energy-rich photons from the solar spectrum are harvested in a semiconductor material, where they may excite electron–hole pairs, which in turn may drive chemical reactions at the semiconductor surface.

The latter scheme is the subject of this chapter, which provides an overview of the CO₂ -free production of hydrogen via photon-driven redox reactions on semiconductor materials and employing entirely renewable and abundant raw materials and energy sources, namely water and sunlight. Particular attention will be paid to the role of nanoscience and nanotechnology, which have attracted a continuously growing interest in this context, and which have shown a strong potential to improve the efficiency of various hydrogen production methods, not the least approaches based on photolysis.

Before diving into the details, we would like to note that the idea of using water as a raw material for energy production is not an entirely new concept. In 1874, Jules Verne, in his novel The Mysterious Island, painted a scenario where water one day will be employed as fuel, thereby replacing dwindling coal reserves and driving the machinery without which there would be no railways, no steamers, no manufactories, nothing of that which is indispensable to modern civilization. Today, almost 140 years later, Jules Verne’s visionary ideas have neither been realized nor abandoned. In fact, the concept of a hydrogen economy still attracts a lot of interest, although the hurdles on the way sometimes seem insurmountable. The hope is that this chapter may provide the reader with an understanding of the state-of-the-art in the field of photolytic hydrogen production and identify the most burning scientific and technological issues that have prevented hydrogen from becoming our prime energy carrier.

The remainder of this chapter is organized as follows: after a general introduction, we briefly review the mechanism and challenges associated with photolytic water-splitting (Sect. 2). In Sect. 3, we present an overview of various semiconductor materials, which have been scrutinized for their ability to split water into hydrogen and oxygen under sunlight illumination. The advantages and shortcomings of these materials will be described. Next, we discuss various ways in which nanoscience and nanotechnology (N&N) can improve materials and their integration into water splitting devices (Sect. 4). Our emphasis is on N&N approaches: (i) to tune the optical and electronic properties of semiconductors, (ii) to manage and optimally utilize the incoming light, (iii) to improve the separation and extraction of photogenerated charge carriers via control of the photocatalyst structure and morphology, (iv) to tailor the transport of reactants and products, and (v) to fabricate well-defined model systems. We conclude by outlining possible pathways toward sustainable hydrogen production schemes, which challenges need to be addressed on the way, and how N&N may contribute.

2 Mechanism of Photolytic Hydrogen Production

Photolytic water-splitting schemes integrate solar energy collection and water electrolysis into a single device, thereby transforming solar energy into chemical energy in the form of hydrogen. Several steps are required to achieve this transformation, which is initiated by the harvesting of solar photons (see Fig. 2). Since water is essentially transparent to the wavelengths constituting the solar spectrum, photolytic splitting of water requires the presence of a light harvesting structure, typically in the form of a semiconductor. Owing to their particular electronic band structure, semiconductors are able to interact with photons whose energy exceeds a threshold corresponding to the bandgap energy of the material, i.e., the energy separation between the valence and conduction band edges. If this condition is met, an electron may be promoted from the valence band to the conduction band, thereby leaving an empty state, normally referred to as a hole, behind. In order for such an excited electron–hole pair to drive water-splitting, it must have an energy that exceeds the standard potential, ΔE^o, for the decomposition of liquid water to form gaseous hydrogen and oxygen (−1.23 V, where the negative sign indicates that the water-splitting process is thermodynamically non-spontaneous). The excited electron–hole pairs are short lived, and the photogeneration of charge carriers must therefore be followed by efficient charge separation schemes in order for a significant fraction (defined by the quantum yield) of these electron–hole pairs to escape recombination and make their way to the semiconductor surface. If the photogenerated electrons and holes have sufficient energy to reduce protons and to oxidize water, respectively, they may be extracted at the solid/liquid interface and drive the photosplitting of water. Finally, hydrogen and oxygen gas need to be removed from solution and separated from one another in the case of photocatalytic water splitting.

Fig. 2

Schematic illustration of the elementary reaction steps of photocatalytic hydrogen production via water splitting on semiconductor nanoparticles with additional co-catalysts for oxygen and hydrogen evolution. Reprinted with permission from [3]. Copyright 1995 American Chemical Society

Note that we distinguish here photocatalytic and photoelectrochemical water splitting. In the former case, the reaction takes place in a suspension or colloidal solution of semiconductor nanoparticles. Oxidation and reduction reactions occur in close proximity to one another, and a mixture of hydrogen and oxygen is thus evolved. In the latter case, the semiconductor is made into a photoanode, from which oxygen is evolved. Photogenerated electrons are extracted from the semiconductor by a charge collection layer and, through an external circuit, shuttled to a counter-electrode, where they recombine with protons to form hydrogen. Oxygen and hydrogen production sites are thus spatially separated.

There are several potential bottlenecks, which may hamper or even prevent photolytic water splitting. The most burning issues are summarized as follows: (i) mismatch between the semiconductor’s absorption characteristics and the solar spectrum (and other light sources), (ii) inappropriate semiconductor bandgap and/or band edge positions, (iii) fast (bulk and/or surface) electron–hole recombination, (iv) short lifetime and thus limited diffusion length of photogenerated electrons and/or holes, (v) poor electron and/or hole transfer kinetics at the semiconductor-liquid interface (large over-potentials required), (vi) competing reactions, including photocorrosion and the back-reaction of hydrogen and oxygen to water in the case of photocatalytic water splitting

In the next section, different classes of materials and their properties will be discussed in light of these requirements and issues, followed by Sect. 4, which presents various nanoscience- and nanotechnology-based approaches towards improved photolytic water splitting.

3 Materials for Photolysis of Water

Materials play a key role in photolytic H₂ production, and their structure and electronic properties largely determine the efficiency of each step in the photolysis of water. The majority of photocatalysts and photoanodes are composed of metal oxides (Sects. 3.1–3.3), metal sulfides (Sect. 3.4), metal nitrides (Sect. 3.5), oxysulfides (Sect. 3.6), oxynitrides (Sect. 3.6) and composites thereof (Sect. 3.7). In these cases, metal cations with the highest oxidative states have d⁰ (red area in Fig. 3) or d¹⁰ (green area) electronic configuration, while O, S and N (blue area) show their most negative states. The bottom of the conduction band consists of the d and sp orbitals of the metal cations, while the top of the valence band in metal oxides is composed of O 2p orbitals, which are normally located at ca. +3 V (vs NHE) or higher [4, 5]. The valence bands of metal oxysulfides and oxynitrides are formed by S 3p (and O 2p) and N 2p (and O 2p), respectively. Some alkali (Li, Na, K, Rb and Cs), alkaline earth (Mg, Ca, Sr and Ba) and transition metal (Y, La and Gd) ions can construct a crystal structure of layered perovskite and cubic pyrochlore compounds, but do not contribute to the energy band structure of these compounds [5]. In this section, the most widely used materials for photolytic H₂ production will be introduced. The focus is on material properties, which influence the photoconversion efficiency.

Fig. 3

Elements constructing photocatalysts for water splitting

3.1 Titanium Dioxide (TiO₂ )

Compared to other photoactive materials for H₂ production, TiO₂ has received much attention since the initial work of Fujishima and Honda [6] due to its low cost, abundance, stability and performance.

Despite a high rate of consumption, a shortage of TiO₂ seems impossible in the near future. Titanium is the ninth most abundant element in the world and constitutes about 0.63% in weight of the Earth’s crust. Minerals, like rutile (TiO₂) and ilmenite (FeTiO₃), are the main ores of this element, and they are found in large deposits in Norway, Australia, China, Canada and many other countries [7]. The toxicity of TiO₂ is low, and it has been approved as a food colorant (E-171 in EU legislation). Actually, TiO₂ is widely used in many daily products such as toothpaste, pill coatings and chewing gum [8].

TiO₂ has four main crystal phases: anatase (tetragonal), rutile (tetragonal), brookite (orthorhombic) and TiO₂ (B) (monoclinic). All TiO₂ forms can be described as different arrangements of elongated TiO₆ octahedra connected by sharing corners and edges. These differences in lattice structure determine different mass densities and electronic band structures between different forms of TiO₂. Loose zigzag chains in octahedra coordination exist in anatase and TiO₂ (B) structures, compared to linear chains of octahedra in rutile. Therefore, the interstitial spaces between octahedra in TiO₂ (B) and anatase are larger, making them less compact than rutile and brookite (densities of TiO₂ (B) and anatase are 3.64 and 3.84 g cm⁻³, respectively, compared to 4.26 g cm⁻³ for rutile and 4.17 g cm⁻³ for brookite) [9, 10]. Compared to anatase and rutile, brookite and TiO₂ (B) are uncommon, unstable and not easily synthesized in their pure phases. Therefore, they are seldom used as independent photocatalysts or photoanodes. Although rutile is the most stable crystal phase of TiO₂, the difference in Gibbs free energies of formation for rutile and anatase is small (less than 15 kJ mol⁻¹) [11]. Generally speaking, anatase will transform to rutile at high temperature, big size and/or small surface area. However, the critical values of this transformation depend on the pH conditions, adsorbates, pressure and preparation methods, which explains some of the conflicts in the literature. It has been shown that bicrystalline TiO₂ (mixture of anatase–rutile, anatase-brookite or anatase-TiO₂ (B)) normally has better photocatalytic and photoelectrochemical performance than its single-phase constituents. This synergistic effect is attributed to the formation of n–p junctions at the contact of the two crystal phases, which improves the efficiency of charge separation.

Due to the presence of a small amount of oxygen vacancies, which are compensated by the presence of Ti³⁺ centers, pure TiO₂ is an n-type semiconductor and is transparent to visible light. The top of the TiO₂ valence band is mainly formed by overlapping oxygen 2p orbitals (O pπ states), whereas the lower part of its conduction band is mainly constituted by the 3d orbitals (t_2g bands) of Ti⁴⁺ cations [12]. The band gaps are 3.2 eV (corresponding to an absorption edge of 380 nm) and 3.0 eV (400 nm) for anatase and rutile, respectively, which limits their ability to convert visible light. Another drawback of TiO₂ is its low photon efficiency. The band transitions of TiO₂ are indirect, and indirect-band semiconductors normally present less photon absorption compared to direct-band semiconductors. In addition, 90% of the photogenerated electron–hole pairs have short lifetimes and recombine in less than 10 ns; the photogenerated carriers available for surface reactions are very limited. Quantum yields for TiO₂ reactions in solution are typically below 1%, while they can exceed 25% for some gas-phase reactions. The values depend largely on the electronic transport at different interfaces and surfaces [13].

The conduction band energy E_CB of rutile coincides with the reversible hydrogen potential $({{\text{E}}^{{{\text{H}} ^{+}} /{{\text{H}}_{ 2} }}} = \, 0{\text{ V vs}} . {\text{ NHE at pH }}0)$, whereas that for anatase is more negative by 0.2 eV. The valence band energy E_VB of both anatase and rutile is more than positive enough to oxidize water $( {\text{E}}_{{{\text{O}}_{ 2} /{\text{H}}_{ 2} {\text{O}}}} = { 1}. 2 {\text{ V vs}}.{\text{ NHE at pH }}0)$, by about 1.8 eV. Note that these relationships are relatively fixed, since both E_CB, E_VB, and reduction and oxidation potentials simultaneously shift with solution pH (–59 mV per pH unit). Therefore, one of the main advantages of TiO₂ over other semiconductors is that its electronic structure allows both the reduction of protons and the oxidation of water, which are key processes for water splitting. TiO₂ is photo-corrosion resistant because water oxidation is thermodynamically more favored than the oxidation of oxide anions by photogenerated holes. This stability, which is maintained over a large pH range, is a crucial feature and explains the widespread use of TiO₂ in water splitting [13].

3.2 Other Binary Metal Oxides

In addition to TiO₂ , there are some other traditional metal oxides, which have also been investigated extensively due to their specific advantages. Among them, ZnO, α-Fe₂O₃ and WO₃ are representative. However, they all have inherent drawbacks in photolytic H₂ production.

ZnO has a large application potential in optoelectronic devices, due to its diverse physical properties and finely tunable preparation. ZnO normally has a hexagonal (wurtzite) structure with the lattice parameters a = b = 3.25 Å and c = 5.12 Å. ZnO is intrinsically doped via oxygen and/or zinc interstitials, which act as n-type donors. However, it can be designed and doped to be either n- or p-type. ZnO is a direct-band gap semiconductor, and has a wide band gap of 3.2 eV with good charge carrier mobility. ZnO has similar conduction and valence positions as anatase TiO₂ , as shown in Fig. 4, and therefore has been frequently considered an alternative to TiO₂ for photocatalytic applications. Although ZnO and TiO₂ have similar conduction band positions, the density of states is about one order of magnitude lower in ZnO. Also, the structure of the conduction bands differs. The conduction band of ZnO is mainly composed of Zn 4s and 4p orbitals, while that of TiO₂ consists of Ti 3d orbitals [14]. This may affect the transfer rate of photogenerated electrons from the valence band to the conduction band under irradiation. The electron mobility in single-crystalline ZnO is much higher than that of anatase TiO₂. It tends to decrease with doping due to electron scattering at impurities, and decreases substantially due to scattering and energy barriers at grain boundaries [14, 15].

Fig. 4

Valence and conduction band positions of various semiconductors with respect to the standard hydrogen electrode (SHE) and the vacuum reference energy level E_AVS. All values are tested at pH 1, unless otherwise noted

The main drawback of ZnO in solution is its chemical instability. It is soluble in strong acids (Eq. 1) and its surface converts to Zn(OH)₂ in alkalis (Eq. 2), which limits its application at extreme pH settings as well as in the presence of Zn²⁺ chelating agents.

$${\text{ZnO}} + {{\text{2H}}^+}_{{({\text{aq}})}} \to {{\text{Zn}}^{2+}}_{{({\text{aq}})}} + {\text{ H}}_{ 2} {\text{O}}$$

(1)

$$ {\text{ZnO}} + {\text{ H}}_{ 2} {\text{O}} \to {\text{Zn}}\left( {\text{OH}} \right)_{ 2} $$

(2)

Besides its chemical instability, ZnO is easily photo-corroded, especially at low pH values. Under irradiation for a long time, ZnO suffers from photodecomposition due to the oxidation of O²⁻ in ZnO by photogenerated holes (Eq. 3).

$$2 {\text{ZnO}} + {{\text{4h}}_{\text{VB}}}^{+} \to 2{{\text{Zn}}^{2+ }}_{{({\text{aq}})}} + {\text{ O}}_{{2({\text{g}})}}$$

(3)

WO₃ is an n-type semiconductor with a relatively narrow band gap of 2.7 eV, which enables it to convert part of the visible solar light. Using WO₃ for solar energy conversion has been extensively investigated. WO₃, like TiO₂ , is photo-corrosion resistant. The VB of WO₃ is close to that of TiO₂, exceeding the O₂/H₂O potential, which makes it a successful photocatalyst for O₂ evolution from water. However, the bottom of its CB lies below the redox potential of H⁺/H₂, which means that the reduction of water to hydrogen is thermodynamically unfavorable for WO₃. Applying a bias can overcome the energy barrier for the ejection of photogenerated electrons from the conduction band to water. Coupling with other semiconductors (like TiO₂) and doping are alternative ways to achieve overall water splitting with WO₃ [16].

α-Fe₂O₃ is another attractive candidate for water splitting due to its narrow band gap (2.2 eV), good photo stability, chemical inertness and the abundance of iron mineral resources in the world. The band structure of α-Fe₂O₃ is similar to that of WO₃ , with the VB top more negative, but still exceeding the O₂/H₂O potential, and the CB bottom lower than the redox potential of H⁺/H₂. A variety of dopants have been employed to modify α-Fe₂O₃ and p-type semiconductor behavior has been obtained, in addition to its common n-type characteristics. The main bottlenecks of α-Fe₂O₃ are its fast e⁻-h⁺ recombination rate, poor charge transport characteristics due to the trapping of electrons at defect sites and the poor mobility of holes. Attempts to circumvent these problems by using a bias, scavengers, coupling to other semiconductors, structure configurations and compositional tuning will contribute to further applications of this material.

There are many other binary metal oxides that have been used in photoelectrolysis of water, such as ${\text{SnO}}_{ 2} ,{\text{ CeO}}_{ 2} ,{\text{ NiO}},{\text{ CdO}},{\text{ PdO}},{\text{ V}}_{ 2} {\text{O}}_{ 5} ,{\text{ MoO}}_{ 3} ,{\text{ Cr}}_{ 2} {\text{O}}_{ 3} ,{\text{ In}}_{ 2} {\text{O}}_{ 3} ,{\text{ Ga}}_{ 2} {\text{O}}_{ 3} ,{\text{ Cu}}_{ 2} {\text{O}},{\text{ CuO}},{\text{ Bi}}_{ 2} {\text{O}}_{ 3} ,{\text{ ZrO}}_{ 2} ,{\text{ Ta}}_{ 2} {\text{O}}_{ 5}$ and so on. Although all of them have inherent issues, some of them are very good coupling partners to be used in conjunction with another semiconductor.

3.3 Ternary and Quaternary Metal Oxides

When TiO₂ is fused with other metal oxides (SrO, BaO, Ln₂O₃ (Ln = lanthanide)), metal titanates with the perovskite structure are formed. Perovskites have the general formula, ABX₃, with SrTiO₃ as an example. Their framework structure contains corner-sharing TiO₆ octahedral with the A cation in twelve-coordinate interstices. Several hundred oxides own this structure. In addition to the cubic structure exhibited by SrTiO₃, a series of distorted, non-cubic structures occurs, with the framework of TiO₆ octahedra being twisted. For example, BaTiO₃ is tetragonal. Both SrTiO₃ and BaTiO₃ with band gaps of 3.3 eV are interesting semiconductors for photoelectrolysis of water. Compared to BaTiO₃, SrTiO₃ has been more intensively studied.

Titanates with tunnel structures have been investigated for water splitting [17]. BaTiO₉ has a twin-type tunnel structure in which the TiO₆ octahedra are not parallel to one other, forming a pentagonal prism space. Alkaline metal hexatitanates $\left( {{\text{M}}_{ 2} {\text{Ti}}_{ 6} {\text{O}}_{ 1 3} ;{\text{ M }} = {\text{ Na}},{\text{ K}},{\text{ Rb}}} \right)$ have Wadsley-Andersson type structures in which TiO₆ octahedra share one edge at one level in linear groups of three, creating a tunnel structure with rectangular space. These materials have been used in powder form in suspensions, normally together with a co-catalyst.

There are more complex perovskites containing two different cations, which occupy either the A or B sites, and many of these have a layered structure. Two main classes of such oxides have been studied in water photolysis: the Dion-Jacobson series $({\text{AM}}_{n - 1} {\text{B}}_{n} {\text{O}}_{ 3n + 1} ,{\text{ e}}.{\text{g}}.,{\text{KCa}}_{ 2} {\text{Ti}}_{ 3} {\text{O}}_{ 10} )$ and the Ruddlesden-Popper series $({\text{A}}_{ 2} {\text{M}}_{n - 1} {\text{B}}_{n} {\text{O}}_{ 3n + 1} ,{\text{ e}}.{\text{g}}.,{\text{K}}_{ 2} {\text{La}}_{ 2} {\text{Ti}}_{ 3} {\text{O}}_{ 10} )$. Noble metal co-catalysts can be loaded onto these materials by photo deposition. Due to a negative charge (which is balanced by the alkali cations) of the oxide sheets, the noble metal salt anions will not be intercalated in the host lattice. Instead, the noble metal particles are formed on the external surfaces of the layered perovskites. In many cases, the H⁺-exchanged layered oxides show higher activity for hydrogen production, due to the easy accessibility of the interlayer space to electron donor species [18, 19].

Another type of layered perovskite has the generic composition ${\text{A}}_{n} {\text{B}}_{n} {\text{O}}_{ 3n + 2} \left( {n \, = { 4},{ 5};{\text{ A }} = {\text{ Ca}},{\text{ Sr}},{\text{ La}};{\text{ B }} = {\text{ Nb}},{\text{ Ti}}} \right)$. Unlike the (100)-oriented structures mentioned above, these perovskites have slabs parallel to (110) and are highly donor-doped. These structural and electronic characteristics are thought to be responsible for the observed improved quantum yields for overall water splitting under UV irradiation. Among these compounds, La₂Ti₂O₇ (deduced from La₄Ti₄O₁₄) and La₄CaTi₅O₁₇ are representative titanates. Their band gaps are 3.2 and 3.8 eV, respectively, and they show quantum yields of 12% (< 360 nm) and 20% (< 320 nm), respectively [19, 20].

Tantalates and niobate oxides with corner-sharing octahedral MO₆ (M = Ta, Nb) structures have shown high photocatalytic activity for the cleavage of water. The high activity of these layered compounds can be assigned to the ease of migration and separation of photogenerated electron–hole pairs through the corner-shared framework of MO₆ units [21]. Tantalates, MTaO₃ (M = Li, Na, K) have been reported as effective photocatalysts for water splitting under UV irradiation. These oxides crystallize in perovskite structure, and their band gaps are 4.7 eV (Li), 4.0 eV (Na), and 3.7 eV (K) [22]. NaTaO₃ shows the highest photocatalytic activity among these ATaO₃ type photocatalysts when a NiO co-catalyst is loaded, which is due to the suitable conduction band level consisting of Ta 5d orbitals and energy delocalization caused by the small distortion of TaO₆ connections.

The layered oxides with ion-exchange characteristics have a net negative charge on the layered sheets. Thus, they can attract positively charged ions (such as K⁺) in the interlamellar spaces. Interestingly, some of these materials (e.g., K₄Nb₆O₁₇) have two kinds of interlayers (I and II) that alternate, as shown in Fig. 5. H₂ is evolved from one interlayer, in which co-catalysts are introduced by ion-exchange or intercalation reactions, while O₂ production occurs in the other interlayer. In this case, the sites for H₂ and O₂ evolution are thus separated by the photocatalytic niobate sheet [5]. Furthermore, an electric field gradient originating from the uneven K⁺ distribution on opposite sides of the niobate sheets assists electron–hole separation.

Fig. 5

Water splitting over K₄Nb₆O₁₇ photocatalyst with layered structure. From [5]. Reproduced by permission of The Royal Society of Chemistry (http://dx.doi.org/10.1039/b800489g)

Generally speaking, oxides containing transition metal cations with d⁰ electronic configuration like Ti⁴⁺, Nb⁵⁺ or Ta⁵⁺ have wide band gaps (> 3.0 eV). Thus, these materials do not perform well under visible-light irradiation. The ternary and quaternary titanates, tantalates and niobates suffer from the same handicap. However, a series of H₂ production photocatalysts based on K₄Ce₂M₁₀O₃₀ (M = Ta, Nb) and their solid solution ${\text{K}}_{ 4} {\text{Ce}}_{ 2} {\text{Ta}}_{{ 10 - {\text{x}}}} {\text{Nb}}_{\text{x}} {\text{O}}_{ 30} \left( {{\text{x }} = \, 0- 10} \right)$ have appropriate band gaps of ca. 1.8–2.3 eV (corresponding to absorption edges of 540–690 nm) [23, 24]. Density functional theory (DFT) calculations indicate that the valence bands of these photocatalysts are composed of hybridization with O 2p + Ta 5d (or Nb 4d) and occupied Ce 4f orbitals, while the conduction bands are mainly comprised of the Ta 5d (or Nb 4d) orbitals (Fig. 6). ${\text{K}}_{ 4} {\text{Ce}}_{ 2} {\text{M}}_{ 10} {\text{O}}_{ 30} ({\text{M}} = {\text{Ta}},{\text{ Nb}})$ has a parallelepiped (tunnel) surface structure. This is beneficial to the formation of “nano-nests”, to which the co-catalysts, nanoparticles of Pt, RuO₂ and NiO_x, can be strongly associated, thus avoiding aggregation and improving photocatalytic H₂ generation greatly. This is an example of an effect of surface nanostructures on photocatalytic performance.

Fig. 6

Band structure of K₄Ce₂M₁₀O₃₀ (M = Ta, Nb) and comparison with redox couples for photocatalytic production of H₂ and O₂ from water. Reprinted from [23], Copyright 2007, with permission from, Elsevier (http://www.sciencedirect.com/science/journal/14686996)

3.4 Metal Sulfides

Metal sulfides are considered attractive candidates for visible-light responsive photocatalysts. The valence bands of most metal sulfides consist of 3p orbitals of S, which results in a higher (more negative) valence band and narrower band gap compared to metal oxides.

Among the available metal sulfides, CdS with wurtzite structure is probably the most studied metal sulfide photocatalyst. Its narrow band gap (2.4 eV) means that it can absorb visible light below a wavelength of 510 nm. Compared to the VB of metal oxides being formed by O 2p orbitals, the VB of CdS is constituted by S 2p orbitals. The electronegativity of S is smaller than that of O, resulting in different VB positions. The CB potential of CdS is around -0.52 V (vs. NHE), high enough to reduce H₂O, and the top of its valence band is about 1.5 V (vs. NHE), suitable for the oxidation of H₂O. However, CdS is prone to photocorrosion (Eq. 4), which is a problem common to most metal sulfide photocatalysts.

$$2 {{\text{h}}_{\text{VB}}}^{ + } + {\text{ CdS }} = {{\text{ Cd}}^{ 2+}}_{{\left( {\text{aq}} \right)}} + {\text{ S}}_{{({\text{s}})}}$$

(4)

Many methods have tried to overcome this photocorrosion, such as combining with other semiconductors or using scavengers (electron donors, such as cysteines, EDTA, sulfide or sulfite species) to attract or consume photogenerated holes. In particular, H₂ generation from an aqueous medium containing S²⁻/SO₃ ²⁻ is appealing. H₂S is produced in large quantities as an undesirable by-product in coal- and petroleum-related industries, and similarly, sulfite is a pollutant. Thus, the combination of photo-oxidation of H₂S and/or sulfite (to sulfate), and co-generation of H₂ using CdS has a value-added benefit.

ZnS (zinc blende) is another important metal sulfide for photochemical water splitting. It shows high activity without any assistance of co-catalysts. It has a wide band gap of 3.6 eV, which restricts light absorption to the UV (< 340 nm). Studies have been carried out to sensitize ZnS. Doping and the formation of solid solutions are promising ways to improve the optical absorption of ZnS in the visible. Similar to CdS, it undergoes photochemical decomposition into the constituents when irradiated in the absence of sacrificial electron donors [25] (Eq. 5).

$$ 2 {\text{h}}_{\text{VB}}^{ + } + {\text{ ZnS }} = {\text{ Zn}}^{ 2+ }_{{\left( {\text{aq}} \right)}} + {\text{ S}}_{{({\text{s}})}} $$

(5)

3.5 Metal Nitrides

We have seen that doping of nitrogen into the TiO₂ lattice has a favorable effect in terms of sensitizing it to the visible light range [26]. Reasonably, the band gap of Ta₂O₅ shrinks from ~4.0 to ~2.1 eV by nitriding it in a NH₃ atmosphere to yield Ta₃N₅. The shrinking of the band gap was attributed to a conduction band derived from Ta 5d orbitals and a higher-lying valence band derived from N 2p orbitals other than O 2p orbitals of Ta₂O₅. This material evolves H₂ and O₂ under visible irradiation (< 600 nm) in the presence of a sacrificial electron acceptor such as Ag⁺ and a co-catalyst such as Pt [27].

Nitrides with d¹⁰ electronic configuration, such as Ge₃N₄ and GaN, are a new series of photocatalysts for the cleavage of water under UV-light irradiation. β-Ge₃N₄ exhibits the highest activity for the reaction among these nitrides. It is the first example of a non-oxide powdered photocatalyst for water splitting. The band gap of Ge₃N₄ is estimated to be 3.6–3.8 eV. The electronic structure of β-Ge₃N₄ was investigated using plane wave density functional theory (DFT) calculations. The density of states (DOS) indicate that the top of the valence band consists of N 2p orbitals, whereas the bottom of the conduction band is composed of hybridized Ge 4s4p orbitals. The band structure indicates that charge transfer under photoexcitation occurs from the N 2p orbitals to the hybridized Ge 4s4p orbitals. Because hybridized Ge 4s4p orbitals broadly expand, resulting from d¹⁰ electronic configuration, photogenerated electrons in such broad hybridized orbitals with large dispersion can be readily transferred to the co-catalyst, like RuO₂, on the surface without recombination, carrying out efficient overall water splitting [28]. The conduction band of GaN is 0.5 eV higher than the reduction potential of water, which is sufficient to overcome the over-potential for the release of hydrogen from a solution of methanol and Na₂S/Na₂SO₃ during the action of visible light in the absence of additional co-catalysts [29]. The results for Ge₃N₄ and GaN indicate that broad hybridized sp orbitals with large dispersion in nitrides with d¹⁰ electronic configuration have suitable band structures for overall water splitting.

3.6 Oxynitrides and Oxysulfides

Oxynitride photocatalysts consisting of metal cations of Ti⁴⁺, Nb⁵⁺ and Ta⁵⁺ with d⁰ configuration are active for H₂ or O₂ evolution in the presence of sacrificial reagents. TaON has been found to be active for water oxidation and reduction under irradiation between 420 and 500 nm [30, 31]. Two other d⁰ metal oxynitrides, LaTiO₂N and Y₂Ta₂O₅N₂, were reported effective for evolving H₂ and O₂ from H₂O under visible-light irradiation [32, 33].

In the development of oxynitrides with d¹⁰ electronic configuration, a solid solution of GaN and ZnO, (Ga_1-xZn_x)(N_1-xO_x) was first tested for decomposing water under visible-light illumination [34]. The solid solution of ZnO and ZnGeN₂, (Zn_1+xGe)(N₂O_x) is another active d¹⁰ metal oxynitride photocatalyst for pure water splitting under visible light [35].

In addition to N, sulfur is another non-metallic element that has sensitized TiO₂ to visible light. Thus, oxysulfides have also been studied as potential photocatalysts for water splitting. Ln₂Ti₂S₂O₅ (Ln = Pr, Nd, Sm, Gd, Tb, Dy, Ho and Er) with layered perovskite structure was reported for photocatalytic water splitting. The band gap of this series of oxysulfides varies from 1.94 eV (Er) to 2.13 eV (Sm). Sm₂Ti₂S₂O₅ was found to have the highest activity among all the homologues tested [21, 36].

3.7 Other Materials for Photolytic H₂ Production

Group III–V semiconductors have several positive features that make them attractive for water splitting. The combination of high charge carrier mobility, an optimal band gap (particularly for many of the alloys) and reasonable photoelectrochemical stability of the p-type materials for cathodic H₂ production, should inspire continuing scrutiny of Group III–V semiconductors. As with their chalcogenide semiconductor counterparts, Group III–V semiconductors, in n-type form, undergo photoanodic corrosion instead of evolving O₂ under illumination in aqueous media. Fortunately, their p-type form is relatively stable against cathodic photocorrosion, and the photogenerated electrons on these p-type semiconductor cathodes can be used to reduce water to H₂, particularly in the presence of a co-catalyst such as Pt or Ru [37]. p-InP photocathodes are capable of evolving H₂ from HCl or HClO₄ electrolytes with very high efficiency [38]. Photocathodes made from p-GaInP₂ (a solid solution of GaP and InP) biased with a GaAs p-n junction have also evolved H₂ with high efficiency [39].

Silicon has been studied for H₂ evolution under irradiation. One fundamental problem with Si from a water-splitting perspective is its low band gap (1.1 eV), which is too small for complete water splitting (1.23 eV). One way around this problem is to seek an alternative electron donor (e.g., ethanol, formic acid, HBr or HI), instead of water. For example, illumination of a Si powder photocatalyst whose anodic and cathodic surface were coated with polypyrrole and platinized Ag, respectively, caused hydrogen evolution from aqueous ethanol for longer than 150 h with a quantum efficiency of 2.1% at 550 nm [40]. Photocatalytic H₂ evolution from aqueous formic acid was successfully achieved by using platinized n-type silicon powder as a photocatalyst [41]. A tandem-type hydrogenated amorphous Si (a-Si) electrode having a [n-i-p-n-i-p] structure and a similar tandem a-Si electrode having [n-i-p-n-i-p] layers deposited on p-type crystalline Si showed cathodic photocurrents accompanied by H₂ evolution. These two electrodes, when connected to a RuO₂ counter-electrode (for O₂ evolution), caused sustained water splitting without external bias with solar-to-chemical conversion efficiencies of 1.98 and 2.93%, respectively [42].

CdSe nanoribbons show catalytic activity for photochemical hydrogen evolution from aqueous Na₂S/Na₂SO₃ solution under irradiation with ultraviolet and visible light [43]. This is one of few instances that displays the photocatalytic characteristics of CdSe, and the first evidence of the ability of CdSe to induce the reduction of water.

AgCl photoanodes combined with either p-GaInP₂ [44], or a single junction amorphous silicon solar cell connected to a platinum electrode [45] as the cathodic part of an electrochemical cell, split water. AgCl does not absorb light below the indirect-band gap transition, which is in the near-UV at about 3.3 eV (380 nm). The light sensitivity in the visible part of the spectrum is due to self-sensitization caused by reduced Ag species. The formation of these Ag clusters introduces new levels within the forbidden gap.

SiC was recently found to be able to reduce water upon exposure to visible light, even in the absence of electron-donating compounds [46].

4 Nanoscience and Nanotechnology Approaches Toward Improved Photolytic Water Splitting

Nature is a great source of inspiration, and the structure and function of both enzymes and the photosynthetic apparatus provide valuable guidelines for the design of optimized (photo) catalysts for water splitting. Key features of the biological analogues include: (i) judicious nanoscale design of reactant binding sites, (ii) ingenious arrangement of several components to form an electron transfer system and (iii) ultimate control of the reaction conditions by selectively operating transmembrane channels that connect the inside of a cell to the outside environment. Researchers have tried to mimic some of these concepts in artificial catalysts with varying degrees of success. We will review some selected approaches, including methods to control charge separation and transport, attempts to steer the flow of reactants and products and means to confine the reaction volume. The potential benefits of these approaches will be discussed from the viewpoint of the issues associated with photochemical water splitting outlined in Sect. 2. Additionally, we will discuss nano approaches aiming to tune the light absorption properties.

4.1 Tuning of Electronic and Optical Properties

As outlined in Sect. 2 above, a mismatch between the solar spectrum and a semiconductor’s absorption properties may severely limit the photoconversion efficiency under solar illumination. Approaches to modify the semiconductor’s electronic band structure and thus optical absorption are therefore urgently needed and discussed below. Another factor limiting conversion efficiency is the fast recombination of photoexcited charge carriers. Various ways to minimize such recombination losses are therefore covered in this section.

4.1.1 Quantum Size Effects

The most interesting feature of semiconductor nanoparticles is the remarkable blue shift in their optical absorption spectra due to size reduction, compared to the corresponding bulk material. Molecular orbital (MO) and the linear combination of atomic orbitals coupled with molecular orbitals (LCAO-MO) procedures can provide information about energy level diagrams for clusters of several molecules (left part of Fig. 7) up to bulk semiconductors (right part of Fig. 7). Increasing the number of molecules in a cluster requires the addition of filled and empty orbitals to the energy manifold. This decreases energy differences between the filled orbitals and the empty orbitals. The energy gap between the highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO) is also decreased. For a bulk semiconductor, the filled and empty states form separated continuums, i.e., the valence and conduction bands, respectively. In the Q-size regime, the energy levels within the filled and empty states remain discrete, and the gap between the HOMO and the LUMO states increases, compared to the bulk material, as shown in Fig. 7 [47]. This leads to a blue shift of the absorption edge in nanoparticles.

Fig. 7

Schematic diagram of the molecular orbital model for band structure. The number of energy levels increases with increasing number of atoms in the cluster. This results in a decrease in spacing within the occupied and unoccupied levels, and between the HOMO and LUMO levels. For bulk crystals the states within the occupied and unoccupied levels are so close that it is convenient to represent them as separated bands, i.e., the valence and conduction bands. The energy gap between these bands is the band gap, E_g, which increases in magnitude as the semiconductor particle radius decreases to the point where it becomes comparable to or smaller than the exciton radius. Reprinted from [47]. With kind permission from Springer Science + Business Media: Fig. 8

Accompanying the bandgap broadening due to reduced particle size, electrons at the lower edge of the conduction band and holes at the upper edge of the valence band then possess more negative and positive potentials, respectively. Therefore, electrons and holes have stronger reduction and oxidation powers, respectively, in such nanoparticles.

4.1.2 Modification by Doping

As mentioned above, semiconductors with a wide bandgap have stronger reduction and oxidation power. However, the main drawback of a large bandgap is the low utilization of solar light. For single semiconductor systems, dye sensitization has shown some encouraging results. It is, however, not easy to find a suitable and stable dye that can be illuminated over extended periods of time. Another way to sensitize semiconductors is by doping with ionic species. Although in many cases a red shift of the absorption edge into the visible region is observed in doped semiconductors, it should be noted that the photocatalytic activity may not necessarily increase under visible-light irradiation, since the measured absorption spectra of doped semiconductors result from several absorption transitions of different origins. Since the effects of doping depend on many factors, including the doping method, the dopant species, the amount of dopant and the distribution of dopant in the semiconductor matrix, the role and behavior of dopants are still not completely clear and need to be investigated further.

The doping of wide-band gap semiconductors with transition metal ions creates local states (new energy levels) within the forbidden band. Visible-light absorption and photoactivity are induced by the interband, as shown in Fig. 8. Here, water is reduced by the photogenerated electrons, while the dopant metal ions get back electrons from a donor. This induced visible-light photoactivity is normally low, due to the small amount of visible-light absorption, as indicated by a small shoulder in the visible-light region, instead of a total red shift of the absorption edge. Increasing the dopant concentration in the semiconductor matrix can improve visible-light absorption, but excessive doping can easily disturb the original structure. Sometimes, doping can extend the lifetime of photogenerated charge carriers by trapping them transitorily and shallowly. On the other hand, dopants can also act as recombination centers for photogenerated electrons and holes and decrease photocatalytic activity.

Fig. 8

Energy diagram of the operation of a photocatalytic system for the release of hydrogen from an aqueous solution containing an electron donor D for undoped titanium dioxide (LHS) and titanium dioxide doped with Ni²⁺ (TiO₂ :Ni²⁺, RHS). From [48]. With kind permission from Springer Science + Business Media, Fig. 3

Doping semiconductors with non-metal ions can also form intra-bandgap energy levels, as in the case of transition metal doping. For example, doping TiO₂ with F⁻ forms new energy states (color centers) close to the conduction band of TiO₂ (0.53 eV below the conduction band edge of TiO₂ for F center and 0.84 eV for F⁺ center), resulting in an enhanced photocatalytic activity of TiO₂ under both UV and visible-light irradiation [49]. N-doping can locate N 2p states near the top of the valence band of TiO₂, which was predicted by theoretical calculations [50] and confirmed by photocatalytic experiments [51]. Interestingly, anionic doping of semiconductors with suitable N, C or S concentration can also narrow the bandgap due to mixing of the p orbitals of the dopant and oxygen, as shown in Fig. 9. The concentration should be higher than that required for the formation of discrete p states of dopants, which are close to the valence band. The dopant p orbitals then contribute to the top of the valence band, while the bottom of the conduction band is still formed by transition metal d orbitals without any change. Although some fruitful results have been achieved by this approach, it is difficult for a semiconductor to maintain its electronic balance and structural stability after incorporating large amounts of dopants. Heavily-doped oxides are non-stoichiometric and contain too many oxygen defects, due to different oxidation numbers of oxygen and dopants. These defects will reduce photocatalytic activity. Furthermore, excessive N (or S) doping can also change the structure of doped samples to stoichiometric oxynitrides (oxysulfides) or even nitrides (or sulfides), which, in many cases, are not very photoactive (there are also some exceptions, as mentioned in previous Sects. 3.4–3.6). The stability of dopants during long-term photocatalytic H₂ production should also be considered, since the bonds between transition metals and dopants are normally weaker than those between transition metals and oxygen, due to the smaller electronegativity of dopants as compared to oxygen.

Fig. 9

Energy diagram of the operation of a photocatalytic system for the release of hydrogen from an aqueous solution containing an electron donor D for undoped titanium dioxide (LHS) and nitrogen-doped titanium dioxide (TiO₂ :N, RHS). From [48]. With kind permission from Springer Science + Business Media, Fig. 4

4.1.3 Modification by Nanostructured Co-catalysts

Nanostructured co-catalysts, traditionally based on noble metals, have been widely investigated and used in photocatalytic systems for H₂ production. The function of these co-catalysts is to shuttle photogenerated electrons from the semiconductor to an acceptor (H⁺ for H₂ production) in a photocatalytic process. Besides noble metals, co-catalyst chemistries also include nickel and its oxide, the mixed oxides of rhodium and chromium, tungsten carbide, MoS₂ and so on. The promoting mechanism of co-catalysts is illustrated in Fig. 10. Under irradiation by photons with enough energy (equal to or larger than the semiconductor bandgap), photogenerated electrons are promoted from the valence band to the conduction band of the semiconductor, which raises the Fermi level of the semiconductor. Then, the energy difference at the semiconductor/metal interface drives electrons from the conduction band of the semiconductor to the noble-metal nanoparticle. The Fermi level of the noble-metal is thereby negatively shifted, which drives another electron transfer to an electron acceptor (H⁺ in the case of H₂ production). This concept is also applicable to photoelectrochemical cells for H₂ production. Instead of a direct contact as in particulate photocatalysis, in photoelectrochemical water splitting, a noble metal counter-electrode (cathode) is connected to the semiconductor photoanode by an external circuit, which leads to spatial separation of H₂ and O₂ evolution sites and thus has the additional benefit of restraining their back-reaction to water (as shown on the RHS of Fig. 10).

Fig. 10

Splitting of water by photocatalyst nanoparticles modified with Pt co-catalyst nanoparticles (LHS) and in a photoelectrochemical cell based on a nanostructured TiO₂ film anode and a nanostructured Pt cathode (RHS). Reprinted with permission from [52], Copyright 2007 American Chemical Society

One of the early studies indicated how different noble metals influence semiconductor-based photocatalytic H₂ production [53]. A direct correlation between the work function of the noble metals and the photocatalytic activity was also studied for metalized TiO₂ photocatalysis systems [54]. Besides the species of noble metal, the size of the noble-metal nanoparticle also largely affects the energetics and the electron transfer between the semiconductor and noble metal by shifting the Fermi level to more negative potential for smaller sizes. In the TiO₂-Au system, higher photocatalytic reduction efficiency was obtained with smaller Au nanoparticles [55]. Identifying and developing less expensive (noble) metals (like Ag) as co-catalysts for solar H₂ production is an important research topic.

4.1.4 Coupling to Another Nanostructured Semiconductor

Photoinduced charge separation can be realized by coupling one semiconductor with another. Matching their conduction and valence band levels can drive a vectorial transfer of photogenerated charge carriers from one semiconductor to the other, thus minimizing losses via electron–hole recombination. A case of special interest is to select a narrow bandgap semiconductor as a coupling semiconductor in order to realize harvesting of visible light. Semiconductors, such as CdS, PbS, Bi₂S₃, CdSe and InP that can absorb visible light serve as sensitizers in semi-conductor/semiconductor nanocomposites because they are able to transfer photogenerated electrons to wide-bandgap semiconductors, such as TiO₂ , ZnO and SnO₂, as illustrated in Fig. 11. The band structure of two coupled semiconductors and the photogenerated charge transfer between them are determined by many factors, such as particle size, crystal structure, crystallinity, defect type and density and surface area; by tuning these factors, the flow of charge carriers can be adjusted and lead to enhanced photocatalytic and photoelectrochemical performance of such coupled-semiconductor systems. For example, the energetics can be controlled by varying the size of the semiconductor quantum dots. Increasing the band energies of these quantum dots was utilized to promote, suppress or rectify the electron transfer between two semiconductors [52].

Fig. 11

Diagram illustrating spatial separation of photogenerated charge carriers in a CdS/TiO₂ heterostructure and the formation of hydrogen during the action of visible light. From [48]. With kind permission from Springer Science + Business Media, Fig. 2

In a traditional photocatalytic water-splitting system, semiconductor photocatalysts should possess both conduction band and valence band potentials that are suitable for H₂ and O₂ evolution, respectively, as shown in Fig. 12a. This requirement seriously limits photocatalyst selection and visible-light utilization. A Z-scheme photocatalytic water-splitting system, which involved two-step photoexcitation under visible-light irradiation, was developed by mimicking the natural photosynthesis of green plants [56]. The Z-scheme system consisted of a H₂-evolution photocatalyst (PS1 in Fig. 12b) an O₂–evolution photocatalyst (PS2 in Fig. 12b) and a reversible redox mediator (Ox/Red) that acted separately as the electron donor (for PS1) and acceptor (for PS2) for the respective half reactions.

Fig. 12

Reaction mechanism of photocatalytic water splitting into H₂ and O₂ for: (a) a conventional one-step photo-excitation system, and (b) a system mimicking the Z-scheme of photosynthesis (two-step photo-excitation system). Reprinted from [56], Copyright 2002, with permission from Elsevier (http://www.sciencedirect.com/science/journal/10106030)

Photocatalysts, which alone are only effective in one of the two half reactions in water splitting, are capable of running both half reactions when arranged in a Z-scheme. SrTiO₃, TaON, CaTaO₂N and BaTaO₂N can work as H₂ evolution photocatalysts, while WO₃ , BiVO₄ and Bi₂MoO₆ can act as O₂ evolution photocatalysts. The IO₃ ⁻/I⁻ and Fe³⁺/Fe²⁺ redox couples normally serve as reversible electron mediators. An all-solid-state Z-scheme based on a CdS–Au–TiO₂ three-component nanojunction was recently reported, where PS1(CdS), PS2(TiO₂) and the electron-transfer system (Au) were spatially fixed [57]. The vectorial electron transfer of TiO₂ → Au → CdS occurs as a result of excitation of both TiO₂ and CdS. The electron supply from TiO₂ to CdS via Au restricts the self-decomposition of CdS due to the oxidation of surface S²⁻ ions by the photogenerated holes in CdS. In a Z-scheme system, H₂ and O₂ are evolved separately by two different photocatalysts, which, to some extent, restrains their back-reaction to water. The key factors in the design of a Z-scheme system are to find a pair of photocatalysts for separate H₂ and O₂ production with high efficiency, and a reversible electron mediator, the redox potential of which can meet the requirements of being electron donor and acceptor in the respective half reactions. From what we have mentioned above, the energy levels of the components and charge transfer between them should be considered when designing both semiconductor–semiconductor nanocomposites and Z-scheme systems.

4.2 Micro- and Nanostructures for Light Management

An important issue in photovoltaic and photolytic energy conversion is a large mismatch between optical and electronic length scales for photon energies close to the semiconductor bandgap; while the absorption depth of light is on the order of (hundreds of) micrometers, the electronic diffusion length is at least an order of magnitude smaller. The optimal device thickness is thus a compromise between absorbing most of the incoming photons and reducing volume recombination of excited charge carriers. Ideally, one would like to construct optically thick but physically thin absorbers. This concept provides several advantages, including resource-efficient use of (in certain cases expensive and/or scarce) materials, the opportunity to use relatively impure materials and/or the potential to achieve higher device efficiency.

While various approaches to address this issue have been presented in the context of photovoltaic (PV) solar cells, our opinion is that their usefulness for photochemical energy conversion, and in particular photochemical water splitting, has not yet been exploited fully. Much of the work reviewed in this section is therefore taken from the area of photovoltaics, with a strong belief that the presented schemes are of high relevance also for photochemical water splitting.

Traditional light-trapping approaches in PV solar cells include the use of wavelength-scale textured substrates [58] and diffractive optical structures [59], which can increase the optically effective cell thickness by a factor of 4–5. More recently, light-scattering layers [60] and optical microcavities [61], including photonic crystals [62, 63] and whispering gallery modes, have attracted significant interest. The latter is named after the whispering gallery at St. Paul’s Cathedral in London and occurs at particular resonant wavelengths of light for a given cavity size and shape when the light undergoes total internal reflection at the inner surface and becomes trapped within the void (or guide) for timescales of the order of nanoseconds [64]. Obviously, the increased photon path length increases the total absorption. An illustrative example of this effect is the observation of photolysis of water-soluble components inside cloud droplets by ultraviolet/visible radiation [65]. Most photonic dielectric cavities have traditionally been limited to sizes that are in the order of the wavelength of light. The lower bound on the effective mode volume (V_eff) arises from a mechanism of confinement based on interference effects and is therefore wavelength dependent. However, it has been shown [66] that by introducing dielectric discontinuities with sub-wavelength dimensions as a means of local field enhancement, the effective mode volume becomes wavelength independent. In this way cavities with large V_eff can be achieved, with a corresponding increase in the Purcell factor (a measure of the spontaneous emission rate enhancement for an emitter in a resonant cavity) of nearly two orders of magnitude relative to previously demonstrated high index photonic crystal cavities.

The most recent development in the area of light management for solar energy applications involves the utilization of plasmonically-active nanostructures. It is well-known that nanoparticles and perforated films of certain materials (e.g., silver, gold, platinum) show strongly enhanced optical absorption (i.e., increased cross-section for initial photon capture) due to the excitation of localized surface plasmon modes [67, 68]. Localized surface plasmon resonances (LSPRs) are collective oscillations of the conduction electrons, which may result in optical absorption cross-sections that exceed the geometric cross-section by several orders of magnitude. These plasmons are accompanied by a strongly enhanced electromagnetic near field. They may decay either radiatively (i.e., by re-emitting a photon), or non-radiatively into (quasi) particles such as electron–hole (e–h) pairs. For many metals, the resonance wavelength falls into the near-ultraviolet, visible or near-infrared regime for nanostructure sizes covering the range 20–200 nm. What is typical for the mentioned spectral range is that it covers (most of) the energetics of important chemical transformations, e.g., bond breaking and bond formation (0.5–6.5 eV). This should allow one to create conditions and propose schemes for enhanced (solar) light absorption in nanostructured materials and utilization of the deposited energy to run chemical transformations such as water splitting.

There are various pathways along which plasmonics may enhance solar-cell devices, as summarized in a recent review by Atwater and Polman [69]. In short, plasmonic structures can function as: (i) scattering structures that couple freely propagating waves from the sun into (waveguiding modes in) a thin semiconductor absorber layer (see [70, 71]), (ii) light-trapping structures, which effectively bend the incoming solar flux by 90° via the excitation of propagating surface plasmon polaritons (SPPs) [72, 73] and (iii) nanoantennas that increase the absorption in the adjacent semiconductor via their strongly enhanced plasmonic near-field [74, 75]. In the first two cases, the effective path length of the incident light in the semiconductor is increased, whereas in the final case, the incident light is concentrated into nanoscale volumes around the plasmonic nanoparticles. Experimental and theoretical work indicate that the active layer thickness of various types of solar cells, including organic solar cells, dye-sensitized solar cells and inorganic solar cells, may be decreased by 10–100-fold using such schemes with no significant impact on efficiency. While the three radiative pathways described above have shown great potential for PV devices, a potential fourth pathway, namely (iv) charge carrier generation via non-radiative plasmon decay has not been investigated to the best of our knowledge, and its usefulness for photovoltaic power generation thus remains to be shown.

Finally, we note that there are a number of recent design suggestions based on nanostructures that do not involve plasmonic effects. To mention a few, Kelzenberg and colleagues [76] demonstrated the use of microrod arrays in between which light is scattered and Zhu et al. [77] recently introduced nanodome solar cells.

In the case of plasmon-enhanced photochemistry, an area pioneered by Nitzan and Brus in the early eighties [78], both radiative and non-radiative enhancement pathways have been pursued, as illustrated in an extensive review article by Watanabe et al. [79]. To give two representative examples, the photocatalytic degradation of methylene blue was found to be significantly increased on nanocomposite photocatalysts consisting of silver nanoparticles embedded in titania. Radiative energy transfer from the Ag nanoparticles to the semiconductor was claimed to be the origin of the increased efficiency [80, 81]. Tian and Tasuma, on the other hand, reported that ethanol and methanol were photocatalytically oxidized by gold nanoparticle–nanoporous TiO₂ composites (at the expense of oxygen reduction) under visible-light illumination [82]. In this case, photoexcitation of the gold nanoparticles was followed by simultaneous charge transfer of electrons from the gold to the TiO₂ conduction band and from a donor in the solution to the gold nanoparticle, as illustrated in Fig. 13.

Fig. 13

Schematic illustration of charge separation at an illuminated gold nanoparticle–titania interface, enabling visible-light induced photocatalytic oxidation of alcohols. Reprinted with permission from [82]. Copyright 2005 American Chemical Society

While the traditional light-trapping schemes presented at the beginning of this section seem to be primarily interesting for photoelectrochemical water splitting, the discussed plasmonic amplification schemes are applicable to both PEC and photocatalytic water splitting. Despite a large potential, we are not aware of any actual realizations of plasmon-enhanced water splitting.

4.3 Control of Photocatalyst Structure and Morphology

Structure and morphology are important and hot issues in photocatalysis, since both affect the performance of a photocatalytic material. Small size and large surface area normally correspond to high photocatalytic activity. Low-dimensional nanostructures with high aspect ratio and porous nanostructures have attracted intense attention, since they favorably affect charge and mass transport, respectively.

4.3.1 Small Size and Big Surface Area: Catalytic and Photocatalytic Considerations

Heterogeneous catalysis is dependent on processes taking place at surfaces and interfaces. Since catalytic materials are often expensive, the goal has always been to fabricate catalysts with the highest possible surface-to-volume (S/V) ratio. In real-life catalysts, this is usually achieved by depositing small (< 5 nm), catalytically active nanoparticles onto a highly porous support material with very high surface area [83]. In parallel, it is known that nanostructures may possess catalytic power not found in their bulk counterparts. As an example, it was shown that tiny dots of gold on certain metal oxide supports exhibit extraordinary catalytic properties, despite the fact that bulk gold is chemically inert [84].

Similar to “normal” (dark) heterogeneous catalysis, the size of photocatalyst particles determines the surface area available for the adsorption and decomposition of reactants (e.g., water or H₂S in H₂ production), and for the harvesting of light. Besides providing more active sites due to large surface area compared to their bulk counterparts, nanosized photocatalysts can restrain bulk electron–hole recombination via charge-carrier trapping on their way to the surface. Photogenerated electrons and holes need to diffuse to the photocatalyst surface to react with electron and hole acceptors (reactants). If the dimensions of nanosized photocatalysts are small enough so that the transfer of photogenerated electrons and holes to the surface is faster than the recombination process, the photocatalytic efficiency will be largely enhanced. It should be noted, however, that the photocatalytic activity does not always increase with decreasing photocatalyst size. There are several potential disadvantages if the size becomes too small. First, too small a size cannot guarantee the best crystal structure for photocatalysis. Normally, very small particles (or quantum dots) are more prone to have amorphous structure. Secondly, very small particles easily aggregate. The size and morphology of these aggregates, or secondary particles, can affect the light-scattering properties of the photocatalyst, as well as the degree of photon penetration. The slow transport of reactants and products within the aggregates can also decrease the photocatalytic efficiency [85]. Finally, particles that are too small may result in increased surface electron–hole recombination, which offsets the benefits of high surface area. It is believed that there are respective optimal sizes at the nanoscale for different photocatalytic materials and reactions.

4.3.2 1-D Nanostructures: Nanowires, Nanorods, Nanotubes and Nanofibers

One-dimensional (1-D) nanostructured materials (nanowires, nanorods, nanotubes and nanofibers) have attracted more and more attention due to their specific properties, which differ from the bulk and other nanostructure (nanoparticle) counterparts. Compared to a nanoparticle with the same volume or weight, a 1-D nanostructure can provide higher surface area and faster interfacial charge transfer rate. As shown in Fig. 14, photoelectrodes composed of nanorods oriented perpendicular to the conductive substrate (see Fig. 14b) can shorten the transport distance for electrons to the back contact (the “electron expressway” concept [86]) and avoid recombination losses at grain boundaries between nanoparticles (see Fig. 14a) [87].

Fig. 14

Schematic representation of the electron transport through (a) spherical particles and (b) nanorods with preferential orientation. From [87]. Reproduced by permission of The Electrochemical Society

Furthermore, poor hole transport has been known to be one of the main factors that limit the conversion efficiency of Fe₂O₃ photo anodes. An elegant solution for this is to use high aspect-ratio nanowire electrodes. It was reported that the transport distance for photogenerated holes to the Fe₂O₃–electrolyte interface was reduced in Fe₂O₃ nanowires with a small diameter, as illustrated in Fig. 15 [88]. One can imagine that the limitation with regard to hole transport can be overcome when the radius of the Fe₂O₃ nanowires is smaller than the hole diffusion length.

Fig. 15

Optimized morphology for an α-Fe₂O₃ photo-anode for water splitting. The small diameter of the nanowires ensures short hole diffusion path lengths. From [88]. Reproduced by permission of The Royal Society of Chemistry, http://dx.doi.org/10.1039/B718969A

It should be noted that even with the same 1-D morphology of the basic building blocks, disorder, order and different orientations of the 1-D units affect the photoelectrochemical properties. As shown in Fig. 16, the photon-to-current efficiency is lower for the Fe₂O₃ electrode with nanorods parallel to the substrate. The IPCE at 360 nm is 3% compared to 5% with nanorods oriented perpendicularly. The electrons have a more straightforward pathway to the back contact with nanorods perpendicular to the substrate, which leads to a decrease in recombination losses (see Fig. 16b) and higher IPCE values [87].

Fig. 16

Schematic illustration (cross-section) and scanning electron micrograph (top view) of hematite nanorods with (a) parallel and (b) perpendicular orientation to the conducting substrate. From [87]. Reproduced by permission of The Electrochemical Society

We additionally note that the basic concepts described above also hold for more complex nanostructures, such as the Fe₂O₃ “cauliflower” structure reported by Kay et al. [89].

Although the initial idea of using 1-D nanostructured (nanorod) electrodes for PEC water splitting was first reported for Fe₂O₃ [90], it has been widely investigated using other semiconductor electrodes, including novel materials such as VO₂ [91] and Ta₃N₅ [92], and, especially, TiO₂ (see [93] for a recent review). Thin film and nanowire electrodes of n-type titanium oxide (n-TiO₂) were fabricated and their photoresponse toward water-splitting was measured; a more than twofold increase in maximum photoconversion efficiency was observed when a single-layer thin film of n-TiO₂ was replaced by nanowires [94]. Photoelectrochemical water splitting using dense and aligned TiO₂ nanorod arrays with well-defined length was also reported; overall water splitting was observed with an applied overpotential of 1.0 V (versus Ag/AgCl) with a photon-to-hydrogen efficiency of 0.1%. The results suggest that these dense and aligned one-dimensional TiO₂ nanostructures are promising for hydrogen generation from water by PEC cells [95]. The concept of using 1-D nanostructures for H₂ production via PEC water splitting has also been extended to nanotubes. Compared to nanorods and nanowires, nanotubes have higher surface area for redox reactions. Highly ordered titania nanotube arrays of variable wall thickness were used to photocleave water under ultraviolet irradiation. It was found that the nanotube wall thickness is a key parameter influencing the magnitude of the photoanodic response and the overall efficiency of the water-splitting reaction. Using nanotubes with a 22 nm porediameter and 34 nm wall thickness, upon 320 − 400 nm illumination at an intensity of 100 mW/cm², hydrogen was generated at a rate of 960 μmol/h W (24 mL/h W) and an overall conversion efficiency of 6.8%, which was the highest value reported for a titania-based photoelectrochemical cell [96].

Besides the application in PEC H₂ production, 1-D nanostructures also play a very important role in photocatalytic H₂ production. TiO₂ nanotubes modified with Pt were found to be a photocatalytic dehydrogenation catalyst in neat ethanol for producing H₂ gas [97]. Self-organized TiO₂ nanotube-layers were fabricated by electrochemical anodization of Ti in a HF electrolyte, and then Pt was deposited on the TiO₂ nanotube layer by plasma sputtering. The Pt-TiO₂ nanotube photocatalyst generated H₂ successfully from an alkaline water solution [98]. Pt-ionized TiO₂ nanotubes were prepared for the stoichiometric production of H₂ and O₂ by water-splitting under visible light with hydrogen evolution rates of 14.6 and 2.3 μmol/h in aqueous methanol and pure water, respectively [99]. Nanostructured TiO₂ films with controlled morphology and thickness were synthesized for use in water splitting photocells and dye-sensitized solar cells. Two different morphologies were compared: a granular morphology and a highly crystalline columnar morphology. The columnar morphology outperformed the granular morphology for both applications, achieving a UV-light to hydrogen conversion efficiency of 11% for water splitting, and a visible light to electricity conversion efficiency of 6.0% for the dye-sensitized solar cell [100]. TiO₂ nanowires (TiO₂ NWs) were synthesized through a one-step hydrothermal process followed by post heat treatment. In this study, anatase TiO₂ NWs exhibited the highest photocatalytic H₂ evolution, which was also higher than that of the starting TiO₂ powder (Degussa P25) [101]. However, Lin et al. suggested that a bi-crystalline structure consisting of TiO₂ (B) nanotubes (or nanofibers) and anatase nanoparticles could act as an active, H₂-producing photocatalyst [102, 103]. A solvothermal method was applied to synthesize CdS nanorods [104] and nanowires [105], which have high photocatalytic activity for H₂ production.

We finally note that the concepts of 1-D nanostructured materials and semiconductor–semiconductor nano-composites, as discussed in Sect. 4.1.4, can be combined and lead to synergistic effects [106]. The same holds for the deposition of co-catalyst nanoparticles (Sect. 4.1.3) onto 1-D nanostructured materials, which leads to improved charge separation properties [107, 108]. Qu et al. demonstrated the fabrication of rationally designed 1-D nanostructures, which integrate several concepts in a single nanosystem. The combination of a charge-separating nanodiode structure, which was encased in a protective insulating shell and functionalized with two exposed metal co-catalysts, was both highly efficient and stable throughout the entire solar spectrum [109].

4.3.3 2-D Nanostructures: Nanosheets, Nanoscrolls and Layered Materials

From the microstructure point of view, nanosheets, nanoscrolls and layered materials belong to two-dimensionally (2-D) structured materials. For photocatalytic H₂ production, 2-D structured materials can maintain a high surface area, selectively expose certain photocatalyst facets with high photocatalytic activity, provide fast charge transfer and stack multicomponent structures by layer assembly.

Photocatalytic H₂ production over Pt/TiO₂ nanosheets with exposed (001) facets was reported recently [110]. The authors concluded that the exposed (001) facets contributed largely to the high photocatalytic activity in H₂ production, which was also proven by Lu et al. [111, 112]. In a dye-sensitized semiconductor photosystem, the semiconductor particle mediates electron transfer between the dye and a particle catalyst for hydrogen evolution (typically Pt or Rh). For the reaction to be efficient, the rate of electron transfer to this catalyst must be faster than that of back electron transfer to the oxidized dye or electron donor. Oxide nanosheets make particularly good electron transfer mediators. An external quantum yield of 20–25% was observed for hydrogen evolution from EDTA²⁻ solutions with niobate nanoscrolls and nanosheets catalyzed by Pt and sensitized by a phosphonated [Ru(bpy)₃]²⁺ derivative (Fig. 17). Taking into account light scattering, the low extinction coefficient of the dye, and the charge injection efficiency, this system has > 50% internal quantum yield for hydrogen evolution from photoinjected electrons. This high internal quantum yield shows that the single-crystal oxide nanoscrolls and nanosheets are good mediators of electron transfer between the dye molecules and Pt particles. Since the electron transfer from EDTA²⁻ to Ru³⁺ was the slowest (limiting) step in the overall reaction, these results demonstrate that the nanosheet morphology was largely responsible for the enhanced overall photon conversion efficiency [113].

Fig. 17

(Top) Schematic representation of photo-induced electron transfer from a phosphonated [Ru(bpy)₃]²⁺ sensitizer to Pt catalyst particles, mediated by H₄Nb₆O₁₇ nanoscrolls. (Middle) TEM images of individual nanosheets and of nanoscrolls precipitated from a suspension of exfoliated H₄Nb₆O₁₇. (Bottom) The dependence of hydrogen evolution rate on Pt loading and the observed high quantum yield establish that electron transfer from EDTA²⁻ to Ru³⁺ is the slowest step in the overall process. Reprinted with permission from [113]. Copyright 2009 American Chemical Society

Layered structures have been widely used in photocatalytic H₂ production. K₄Nb₆O₁₇ has a layered structure with two kinds of interlayers. H₂ is evolved from one interlayer, in which co-catalysts are introduced by ion-exchange or interlayer reaction, while O₂ is produced in the other interlayer, as shown in Fig. 18. In this way, the sites for H₂ and O₂ evolution are separated by the photocatalytic niobate sheet [5]. Furthermore, an electric field gradient resulting from the uneven K⁺ distribution on opposite sides of the niobate sheets assists electron–hole separation.

Fig. 18

Water splitting over K₄Nb₆O₁₇ photocatalyst with layered structure. From [5]. Reproduced with permission from The Royal Society of Chemistry (http://dx.doi.org/10.1039/b800489g)

4.3.4 Porous Structures

Porous materials, behaving as spatially confined micro- and nanoreactors (see Sect. 4.4.2 below for further details), have attracted more and more attention in photocatalysis. They have high surface areas, and they can concentrate reactants with low concentrations, control the reaction environment and provide host–guest effects. Selective photocatalysis can be realized by adjusting the pore size. Co-catalysts can be easily deposited and dispersed onto these porous photocatalysts. Peng et al. reported that hydrothermally synthesized TiO₂ nanoparticles without calcination had a large specific surface area (438 m²/g) with small crystallites (2.3 nm) dispersed among amorphous mesoporous domains, and exhibited much better photocatalytic activity for H₂ production compared to samples calcined at different temperatures and also the commercial photocatalyst P25 [114]. A novel synthesis was carried out using KCl electrolyte to control the electrostatic repulsive force between TiO₂ nanoparticles toward the formation of a mesoporous structure, which owned the highest photocatalytic activity for H₂ production, compared to nonporous colloidal-TiO₂, and commercial Degussa P25 and Hombikat UV-100 (HBK) samples [115]. The photocatalytic reduction of metal cations (M = Ni²⁺, Co²⁺, Cu²⁺, Cd²⁺, Zn²⁺, Fe²⁺, Ag⁺, Pb²⁺) on the surface of mesoporous TiO₂ (specific surface area 130–140 m²/g, pore diameter 5–9 nm, and anatase content 70–90%) resulted in the formation of nanostructured metal–semiconductor composites (TiO₂/M). These metal–TiO₂ nanostructures showed a remarkable photocatalytic activity for hydrogen production from water–alcohol mixtures, and the efficiency was 50–60% greater than that of the metal-containing nanocomposites based on Degussa P25. The anatase content and pore size proved to be the basic parameters determining the photoreaction rate [116].

Porous materials were also used as active supports for catalysts and photocatalysts. Tetrahedrally coordinated metal oxide (titanium, vanadium, chromium, and molybdenum oxides) moieties can be implanted and isolated in the silica matrices of microporous zeolite and mesoporous silica materials, referred to as “single-site photocatalysts”. The single-site titanium oxide photocatalyst demonstrates a high reactivity and selectivity under UV-light irradiation, while the single-site chromium oxide operates as a visible-light sensitive photocatalyst [117]. The photocatalytic H₂ production by nanosized CdS was enhanced by immobilization of CdS on porous supports, such as aluminum-substituted mesoporous silica molecular sieve (Al-HMS) [118], microporous and mesoporous silicas [119], porous polyethylene terephthalate fibers (PET) [120], and ETS-4 zeolite [121].

4.4 Spatial Control of the Reaction Environment

Judicious control of the reaction environment, which includes tailoring the transport of reactants and products to and away from the reaction sites, respectively, and confining the reaction to nanoscale volumes, is an interesting option to fine-tune the kinetics of photolytic water splitting. It allows one to address issues related to the fast back-reaction of hydrogen and oxygen, photocorrosion and reaction selectivity.

4.4.1 Tailored Transport of Reactants and Products

Experience shows that noble metal co-catalysts, which are oftentimes used to assist water reduction, at the same time tend to act as good catalysts for water formation, i.e., the unwanted back-reaction of hydrogen and oxygen to water. While this is not an issue in photoelectrochemical water splitting due to hydrogen and oxygen being evolved in separate compartments, it constitutes a major limiting factor for photocatalytic water splitting, where the two species are evolved in close proximity to one another. To alleviate this problem, Domen and co-workers pioneered a concept, where the transport of reactants to the co-catalyst sites is controlled judiciously by coating the noble metal co-catalyst nanoparticles with a very thin chromium oxide shell using a photodeposition method [122]. An oxynitride photocatalyst loaded with such core/shell co-catalyst nanoparticles showed a much higher water-splitting activity than the same photocatalyst with unmodified noble-metal co-catalyst nanoparticles. Recently, extensive work by the same group [123] demonstrated that the beneficial effect of the chromia shell is related to its relative permeability for protons, hydrogen and oxygen; while the Cr₂O₃ layer is permeable to protons and evolved hydrogen molecules and does not interfere with proton reduction and hydrogen evolution, it prevents oxygen from accessing the underlying noble-metal core and therefore blocks the back reaction (Fig. 19).

Fig. 19

Schematic illustrating the function of noble-metal/Cr₂O₃ core/shell nanoparticles as a co-catalyst for photocatalytic water splitting. Selective permeability of the chromia layer enables proton reduction and hydrogen evolution, while the back-reaction of hydrogen and oxygen to water is blocked. Reprinted with permission from [123]. Copyright 2009 American Chemical Society

This modification method has also been shown to work in combination with an oxygen evolution co-catalyst on the same photocatalyst; the addition of a Mn₃O₄ oxygen evolution co-catalyst to a Rh/Cr₂O₃ core/shell loaded GaN:ZnO photocatalyst resulted roughly in a doubling of hydrogen and oxygen evolution rates [124].

Finally, we note that the approach of blocking the back reaction via the construction of core–shell nanoarchitectures is versatile and has been applied to different semiconductors and using other shell chemistries, such as SiO₂ [125].

Kale et al. [126] reported a strategy to protect CdS quantum dots from photocorrosion by carefully controlling the interaction between reactant and photocatalyst; by partially embedding the otherwise un-photostable CdS QDs in a glass matrix, photocatalytic decomposition of H₂S to hydrogen was possible with high activity over extended time periods, thus demonstrating an increase in photostability without compromising the catalytic function.

4.4.2 Nanoreactors to Confine the Reaction Volume

Traditionally, catalytic systems are described in terms of interactions between the reacting species and the (open) catalytic surface. Architectures in which the interaction between the catalyst and the reactant(s) is confined to nano-liter volumes, i.e., nanoreactors, have recently attracted a great deal of attention. Various methods of confining the reaction volume have been identified and include diverse porous inorganic structures (mesoporous materials, zeolites, etc.), approaches based on molecular self-assembly (oil-in-water emulsions, liquid foams, polyelectrolyte capsules [127], co-polymer vesicles [128], lipid vesicles [129] and lipid nanotube-vesicle networks [130], etc.,) and lithographically-defined nanocavities [131].

While the most evident way in which confinement can affect a reactive system is to alter the thermodynamic properties, there are a number of other effects that can influence chemical reactions in nanospace, such as geometrical constraints in pores with a size comparable to the molecular sizes, selective adsorption of reacting molecules, and changes to the potential energy surface.

A nanoreactor may stabilize a transition state of a chemical reaction and thereby change the activation parameters of the reaction. Both enthalpic stabilization, i.e., non-covalent interactions between the transition state and the surrounding, and entropic stabilization must be considered. The latter contribution may be appreciable since the translational and rotational degrees of freedom of reactants are reduced inside a confined nanospace, leading to a pre-organization of reactants toward the transition state. Even in cases where the nanoreactor does not change the activation parameters, the kinetics of the reaction may be modified due to a locally increased concentration of reactants. This mechanism is important for low (bulk) reactant concentrations.

Stabilization of reaction intermediates, or reaction products for that matter, has the added benefit of providing new pieces of information regarding the reaction mechanism, which would not have been accessible otherwise owing to the reactive and/or labile nature of the intermediates and products; encapsulation traps intermediates and/or final products for spectroscopic analysis. Such confinement also means that several successive reactive events can be induced in a very small confined space of molecular dimensions. For example, a first photon can generate an electron that causes one bond to break and thereby creates some reaction intermediates that are trapped at the interface. A second photon can then later generate another electron that induces a second type of reactive event where the intermediates from the first event participate, and so on. This creates significant probabilities for reaction paths and products that would be vanishingly small if the reactants and intermediates were not confined. We would like to point out that this reaction mechanism, as presented, is a batch process. In order to continue, it is necessary to take the products out and to regenerate the active sites to make this a true catalytic process. There are several possible ways to achieve this, such as photo-excitation, or further reaction with other molecules. If such a process could be found, the system would thereafter work as a molecular energy conversion cycle: a true nanoscale chemical reactor.

The combined result of these effects is that confinement may increase the activity and/or selectivity of a photocatalytic reaction. Additionally, the large surface-area-to-volume ratio, associated with small dimensions and small volumes, short material transport and high attempt frequency for the reaction in nanoreactors make it possible to conduct unique processes not possible for ordinary heterogeneous schemes [132–135].

While the nanoreactor concept has been applied to a diversity of chemical (photo) processes [136–139], it has not yet found widespread attention in the photocatalytic water-splitting community. Early experimental efforts report successful incorporation of photocatalyst nanoparticles into the cavities of lipid and surfactant vesicles and the ability of these to generate hydrogen under visible-light illumination. The focus of these studies was, however, mainly on electron relays for charge separation across the membrane rather than confinement effects [140–143]. Some more recent and encouraging results for photocatalytic conversion in porous media are presented in Sect. 4.3.4 above.

A recent example from our own research [144] further illustrates the applicability of the nanoreactor concept to photocatalytic water splitting. Our experimental system consists of a graphite substrate onto which water and potassium are co-adsorbed. Upon photoexcitation of the graphite substrate, energetic charge carriers are generated and drive photo-dissociation reactions of water molecules at the graphite/K⁺/water interface, leading to the formation of H₂, CO and CO₂ (Fig. 20). Covering the graphite substrate with ice layers of varying thickness and with different morphologies drastically changes the product composition due to (partial) confinement effects imposed by the ice layer, the product permeability of which depends on thickness and morphology.

Fig. 20

Schematic illustrating the mechanism of photodissociation of water on potassium-functionalized graphite. Reprinted from [144]

Although the limited body of results generated to date is perhaps mainly of academic interest, we believe that further research efforts will demonstrate the usefulness of the concept for large-scale industrial schemes.

4.5 Fabrication of Supported Model Systems

Enormous progress in the area of nanoscience and nanotechnology has resulted in a plethora of nanofabrication and characterization methods becoming available today. While the nano-architectures in most of the examples given in the previous sections were realized via bottom-up (often wet-chemical) approaches, the role of lithographic, top-down methods will be discussed in this section. Lithographic nanofabrication schemes such as electron beam lithography, nanosphere lithography [145] and (hole-mask) colloidal lithography [146] allow one to manufacture photocatalysts and photoelectrodes with an extremely high degree of control over structural parameters such as the size, shape and mutual arrangement of nanostructures. Given the high cost, limited speed and restrictions regarding the maximum area that can be patterned typically associated with lithographic nanofabrication approaches, we note that these methods are perhaps not expected to be of great importance for large-scale applications. They are, however, an ideal playground to fabricate supported model systems, the nano-architecture of which can be controlled and tuned accurately and over a large parameter space. By systematically varying certain key parameters, such model systems allow one to construct structure–activity maps, providing a detailed understanding of the physical and chemical processes underlying water splitting, and predictions for improved photocatalyst architectures may also be derived. An additional attractive feature of lithographically fabricated, well-defined model systems is that they are easily modeled, which is often not so for wet-chemically synthesized systems.

The concept of supported model systems has entered several fields related to energy and environment, including catalysis [147], solar cells [75], hydrogen storage materials [148] and fuel cells [149–151]. It has, to our surprise, not seen a breakthrough in the photolytic water-splitting area however. Considering the enormous, demonstrated power of this approach, we believe that it is only a matter of time until suitable model systems will be utilized to study the intriguing details of photolytic water splitting.

5 Conclusions and Outlook

There is no doubt that the world energy consumption will continue to increase rapidly, and man-made global warming is a fact, which is closely associated with our current, fossil fuel-based energy system. In order to avoid serious political unrest, energy shortages and catastrophic climate change, mankind must tap into renewable, CO₂ -neutral energy sources as soon as possible. Hydrogen might be part of the answer to this immense challenge. Yet, for hydrogen to become a viable option numerous fundamental problems associated with both the production, the storage, the distribution and the use of hydrogen must still be solved. Without cost-competitive production schemes, which neither emit greenhouse gases nor have any other negative environmental impacts, the production of hydrogen will be the bottleneck of a future hydrogen economy.

Currently, commercially available hydrogen production methods are limited to fossil fuel-based procedures and electrolysis (as driven by a variety of electricity sources). Initial efforts in the transition to the hydrogen economy may be based on these methods, with improved efficiency, lower costs and minimized carbon dioxide footprint being important prerequisites. On a longer time scale, only hydrogen which is derived from carbon-neutral sources and produced using renewable energy sources fulfills all demands posed on a sustainable energy carrier. Research and development in all fields of hydrogen production conforming to these basic criteria should be intensified, so that a wide variety of processes will become available and constitute a flexible energy system, which can be adapted to varying local conditions and requirements.

Among the renewable energy sources, we believe that solar energy will be the dominant resource for hydrogen production. Of the various pathways to harvest solar energy and use it to produce hydrogen, photolytic schemes seem to be particularly interesting, owing to their potential to achieve an attractive trade-off between high efficiency and low costs. Considering that “traditional” metal oxides, such as TiO₂ , WO₃ and Fe₂O₃, have, despite several decades of investigation, not reached the target of 10% solar-to-hydrogen conversion efficiency, it is clear that further research into materials that are inexpensive, durable, abundant and environmentally-benign is needed; materials with good absorption and carrier-transport properties need to be developed, and novel interfaces need to be designed with energetic and kinetic properties favoring the water-splitting reactions while inhibiting corrosion reactions. In the case of PEC hydrogen production, the materials challenge is not only limited to the photoactive material, but also concerns the materials found in the transparent conductive layer used for charge collection. Experimental efforts in the area will be complemented by computational catalysis.

Further down the road, we foresee a significant engineering challenge to integrate improved materials and interfaces into integrated devices that can be manufactured on a commercial and large-scale basis at low cost.

But what is the role of Nanoscience and Nanotechnology (N&N) in this process? We believe that N&N will have several roles to play and that they will continue to push the field forward. Many of the improvements we have seen over the years have, in fact, been made possible thanks to N&N, albeit with a lack of consciousness and direction. This has changed over time, however, due to analytical tools to analyze very small structures becoming widely available as a consequence of developments in N&N. Today N&N are being used to systematically investigate and improve water-splitting schemes. The design of nanocomposites, which feature a genuine combination of light-harvesting and charge-separation units, is perhaps the most prominent research topic at present.

Considering the rapidly growing number of important N&N contributions to other related fields, there seems to be an immense backlog demand for such schemes to be implemented in the area of photolysis. In particular, light-harvesting and light-management schemes, which have been investigated in conjunction with solar photovoltaics, are expected to make their way into the water-splitting community in the near future. On a somewhat longer term, we expect nanoreactor concepts to gain momentum and become an area of intense research. Throughout this development, N&N will enable scientists to achieve a better understanding of the basic physics and chemistry involved in photolytic water splitting, owing to the ability to manufacture and characterize well-controlled model systems. Based on such an improved understanding, more efficient and durable systems can be constructed.

We conclude by noting that alternatives to the hydrogen economy are being and should be considered. One such example is the concept of the methanol economy [152], where methanol rather than hydrogen is used as an energy storage medium. The potential advantages of such a scheme include convenient and safe storage and handling options, and readily available transport and distribution infrastructure. Additionally, methanol may replace oil and gas resources to serve as a feedstock for the synthesis of hydrocarbons.

The most attractive scheme of producing methanol is via the photocatalytic conversion of carbon dioxide and water, i.e., using abundant and clean energy from the sun and offering ways to chemically recycle CO₂ . This process has many similarities to photocatalytic hydrogen production as described in this chapter. The oxidation of water is in fact a prerequisite for both reactions, and there are thus obvious synergies. Many of the N&N concepts described here for water splitting are thus expected to be readily applicable also to photocatalytic CO₂ conversion.