Nanomaterials for Environmental Solar Energy Technologies : Applications & Limitations †

Environmental remediation and energy production are currently listed among the priority tasks by administration bodies, stakeholders and market competition. In this context, nanomaterials present competitive advantages in terms of performance and production costs. In the present critical review, the current state of the art of nanomaterials used in environmentally benign technologies exploiting solar irradiation such as H2 production, water-splitting and photocatalysis are discussed. Factors determining the overall efficiency are articulated in a single “photophysical efficiency” equation. The physical meaning, limits and constraints of each factor are analyzed, updated and examples are discussed. The article highlights the main structure-function relationships in tandem with the limitations posed by the particle’s physicochemical properties, production method, and the prerequisites posed by regulatory bodies and market needs. Several misconceptions are highlighted with regard to performance and yields that ultimately impact the end use of the nanomaterials. Current targets/limitations posed by the US Department of Energy are discussed as case studies. For context-coherence reasons, the present review focuses on metal and metal-oxide nanoparticles, i.e. carbon nanomaterials are not covered herein.


Introduction
The interconversion of different forms of energy ultimately determines environmentally benign technologies that are quantifiable in a product/byproduct/cost balance (Chen X. et al., 2010;Froschl T. et al., 2012;Holdren J.P., 2007;Kamat P.V, 2010).Research on interactions of solar photons at particle interfaces or electrode surfaces boosts technology, the economy, as well as environmental policy interests.The energy content of the earth's sunlight is 0.9 to 3.2 electron volts (eV) photon, i.e. 87 to 308 kilojoules per einstein, i.e. "mole" of photons.On average, this makes a total maximum ~120.000TW of solar radiation arriving on the earth's surface (Lewis N.S. and Nocera D.G., 2006).Today, the world's energy consumption is 15 TW, which is estimated to rise to 30 TW by 2050 (Lewis N.S. and Nocera D.G., 2006).In this context, if properly exploited, solar energy input could provide a significant contribution to our energy needs.This requires efficient/ low-cost/environmentally sustainable technologies utilizing readily available materials for the solar-conversion process.Hydrogen (H 2 ) combustion emits only water vapor without carbon emissions, thus it is an ideal fuel.H 2 has the highest energy density of all fuels (143 kJ kg -1 ), i.e. ideally 1 kg of H 2 can replace 2.6 kg of gasoline (Holdren J.P., 2007;Lewis N.S. and Nocera D.G., 2006).Today, however, 96 % of the global demand for H 2 is produced from fossil fuels (gasoline, coal).In this context, nanophotocatalytic materials emerge as being among the most promising solutions for photonic-energy storage in the form of H 2 .As we discuss herein, the same photophysical machinery can be used to catalyze the degradation/transformation of environmental pollutants such as toxic organics of high-valency metals (Dong S. et al., 2015;Fagan R. et al., 2016).
In the present critical review, the current state of the art of nanomaterials used in environmentally benign technologies exploiting solar irradiation via H 2 production, water-splitting and photocatalysis are discussed, using a unified nanomaterials-technology view as originally envisaged by A. Bard (Bard A.J., 1980).Solar photoelectrochemical (PEC) hydrogen production, since its original conception in 1972 (Fujishima and Honda, 1972), surpassed the limit of 12 % efficiency in 1998 (Khaselev O. and Turner J.A., 1998).Today, photocatalysis is envisaged as a technology that can potentially provide a clean, costeffective, domestically produced energy vector using solar energy (Lewis N.S and Nocera D.G., 2006;ENERGY. GOV, 2015).It should be underlined, however, that currently (2016), low-temperature (< 80 °C) H 2 production experiences strong research interest aimed at setting up standards and parameters in parallel with cost reduction and factory-scale economy.In the 2015 report of the US Department of Energy on hydrogen production technologies (ENERGY.GOV, 2015), it is stated that photoelectrochemical H 2 production based on semiconductor photoelectrodes or photocatalysts is at an early stage of development.Thus it requires advances in materials development and reactor concept development.The progress on these fronts is envisaged through priority-ranked tasks: (a) in an initial stage, the study/development of high-efficiency, probably high-cost, materials to establish performance parameters, (b) attain a fundamental understanding of PEC hydrogen production vs. corrosion mechanisms, (c) then, in a later stage, the study of durable, low-cost, probably lowerperformance materials to improve efficiency by mitigating loss mechanisms, (d) finally, development of complex multicomponent devices capable of achieving water splitting via optimization [light absorption]-[electron-hole transport]-[interfacial catalysis].In this context, the fabrication of artificial photosynthetic systems can be directed towards two targets: [a] direct conversion of abundant materials, e.g.H 2 O and CO 2 to fuels (H 2 , HCOOH, CH 3 OH), (Chen Z. et al., 2010;Lianos P., 2011;Miller E.L. et. al., 2010), or [b] production of electricity.Moreover, photochemical reactions could be employed to replace other energy-consuming chemical processes, for example, for pollution abatement or in chemical synthesis.
The present article is a review on the current status of photoactive nanomaterials used in such environmental technologies.The aim is to highlight the common (nano) photophysical properties that determine their performance, limitations and potential in technological applications.Emphasis is focused on the materials' physicochemical properties in an effort to highlight the main factors that determine nanomaterials' production targets and limitations.A detailed tabulation of the literature data per material is beyond the scope of the present review.The interested reader is encouraged to consult pertinent reviews on photocatalysis (Teo W.Y. et al., 2012), heterogeneous photocatalytic water treatment using visible light-responsive photocatalysts (Dong S. et al., 2015), solar and visiblelight-active TiO 2 photocatalysis for treating biocontaminants (Fagan R. et al., 2016), N-TiO 2 for green energy applications under UV/visible light (Devi L.G. and Kavitha R., 2014), as well as alternative photocatalysts to TiO 2 (Hernandez-Alonso et al., 2012).Moreover, the interested reader is encouraged to visit a virtual issue published by the American Chemical Society (JPCCSVI, 2016) where a comprehensive set of review articles-up to 2013-is available.

The general photo-nano-technology concept
Solar electromagnetic radiation is consistent with that of a black body at T = 5800 K with a radiation spectrum peak at 0.8 eV.A significant part of this spectrum is in the visible range of the spectrum (400-700 nm).The fraction of solar light which reaches the earth's surface has a maximum power density of ~100 mW/cm 2 (Holdren J.P., 2007;Lewis N.S. and Nocera D.G., 2006).However, only photons with hν exceeding the semiconductors band gap (E g ), generate electron-hole (e -/h + ) pairs.The overall powerconversion efficiency of single crystalline p-n solar cells ranges from 10 to 30 %, yielding 10 to 30 mW/cm 2 .On the other hand, when pursuing visible-light photocatalysts, Maeda and Domen (2007) have proposed that, before commercial development, photocatalysts must reach a quantum yield of 30 % at 600 nm.This seems a bold challenge at present, however, as we discuss herein, this is exactly the driving-force for science and technology in this field (ENERGY.GOV, 2015).
There are three main technology/research approaches for the exploitation of solar photons.The common basis of these was originally conceptualized by Bard A. (1980), as depicted schematically in Fig. 1.

Photoelectrochemical cells (PEC):
In this general scheme, Fig. 1, the archetypical Fujishima-Honda photoelectrochemical cell (PEC) (Fujishima and Honda, 1972) is the combination of the n-type semiconductor as the anode and an inert electrode as the cathode immersed in the electrolyte solution, see Fig. 1(A).PEC provides both the required field for e -/h + separation as well as considerable spatial separation of the products to minimize recombination hurdles or unwanted product reaction.Until today, the nanophotoelectrodes used are of n-type.Fewer examples of p-type nanophotoelectrodes exist, i.e.CuO, Co-phosphates, IrO 2 , that will be discussed hereafter.The definitions on efficiency for PEC devices are aligned with those used for photovoltaics (PV) because the two fields share many common characteristics, and they are in agreement with previous analyses of efficiencies for PEC water splitting (Chen Z. et al., 2010;Rocheleau R. and Miller E., 1997).In energy terms, in a PEC it is light that pumps the electrons "uphill" and provides the required energy.Note that the PEC provides both the required field for e -/h + separation as well as the considerable spatial separation of the reaction products to minimize unwanted back-reactions.
The concept of PEC for H 2 O splitting for H 2 production has been discussed for decades since its first demonstration in 1972 (Fujishima and Honda, 1972), but the worldrecord solar-to-hydrogen (STH) efficiency of 12.4 % was achieved almost two decades ago in 1998 (Khaselev O. and Turner J.A., 1998).This technology combines the harvesting of solar energy and the electrolysis of water into a single device.When a semiconductor-with the right set of properties-is immersed in an aqueous electrolyte and irradiated with sunlight, the photon energy is converted to electrical energy which is directly used to split water into hydrogen and oxygen (chemical energy), Fig. 1B.Thus, intermittent energy from sunlight is converted into an inherently more storable form of energy, i.e. that of chemical bonds.An excellent account on the physicochemical, technical and economical issues of PECs is given by Chen et al. (2010).
For the direct PEC decomposition of water to occur, several key criteria must be met simultaneously: (i) the semiconductor system must generate sufficient voltage upon photo-excitation to split H 2 O; (ii) the bulk band gap, E g , must match the energy of solar photons, preferably in the visible spectrum, i.e. 2.0-3.0 eV; (iii) the band-edge potentials at the surfaces must encompass the H + /H 2 and H 2 O/O 2 redox potentials; (iv) the system must exhibit physicochemical resilience in aqueous electrolytes; (v) for H 2 O splitting, the charge transfer from the surface of the semiconductor to the solution must avoid corrosion and be energetically/kinetically facile enough to reduce energy losses due to kinetic overpotential.So far, no cost-effective material satisfies all of these technical requirements (Chen Z. et al., 2010).Here, we present a survey on some of the key factors that may influence the efficiency.
Overal quantum yield vs. efficiency in PECs: Photoinduced e-transfer reactions are now well understood.High quantum yields for photochemical e -/h + separation are very often reported.When these systems are coupled to sacrificial electron donors, such as methanol, or acceptors, such as persulfate, then H 2 or O 2 can be produced from H 2 O (Chen Z. et al., 2010;Harriman A. et al., 1988).
When practical energy conversion is considered in real systems, the use of sacrificial systems gives a simplistic "efficiency" that should be revised when taking into account electrolyte redox-couple cycling.However, such sacrificial donors are useful in mechanistic studies because [a] they minimize e -/h + recombination and [b] they permit the examination of a light-driven half-cell reaction, i.e. oxidation or reduction, separately.Thus, from the practical point of view, estimating separately the quantum yield of H 2 or O 2 evolution in sacrificial half-cycles is also an effective method for rapidly screening new catalysts is the positioning of the energy levels of the redox couples vs. the photo-induced electrons' and holes' energy levels.For simplicity, external bias voltage is not included in these schemes.External bias, depending on polarity and voltage, will shift upwards or downwards the energy levels of e -and h + .(Chen Z. et al., 2010;Goldsmith J. et al., 2005).
Achieving high quantum yields in sacrificial halfcycles implies the use of efficient key components (catalytic material/redox mediators/photosensitisers) that allow a proof-of-concept level for overall water splitting.The problem is to combine them in such a way that diminishes e -/h + and H 2 /O 2 recombination.Control of e -/h + recombination is a complex process, even in the simplest of such systems, such as a Pt 0 -loaded n-semiconductor particle coupled to an O 2 -evolving catalyst, see Fig. 1B.A high quantum yield can be achieved only if the forward electron transfer rate (solid arrows in Fig. 1B) at each branch in the chain is faster than the sum of all the reverse rates.For example, in Fig. 1B, the forward electron transfer from the semiconductor to the H 2 -evolving catalyst must compete effectively with e -/h + recombination and also with e-transfer to the H 2 O-oxidation catalyst.Here, a key hurdle is that the reverse reactions are fast because the reverse pathways have much higher driving forces than the forward ones.However, because e -/h + recombination rates are also strongly distance-dependent, it should be possible to arrange the components in space in such a way that the forward pathway becomes kinetically competitive or even dominant.Photosynthesis is living proof of this principle.On this front, the use of heterostructural nanoparticles containing all components (O 2 -evolving particle, main photoexcitable semiconductor, and e-accumulating particle) is a very attractive technology that we discuss hereafter.

Photocatalysis:
The concept of photocatalysis for environmental remediation, depicted in Fig. 1C, implies that photons are converted in a first step to chemical intermediates that are highly energetic, thus very reactive.Typically, these are hydroxyl radicals ( • OH) or superoxide radicals ( • OO) and, in more rare cases, singlet-oxygen radicals ( • O) that are generated at the particle-solution interface (Teo W.Y. et al., 2012;Linsebigler A.L. et al., 1995), see Fig. 2.
When a semiconductor surface is irradiated by light (hν ≥ E g ), there is a generation of electron/hole pairs (e − / h + ) via the promotion of an e -from the valence band (lower energy level) to the conduction band (higher energy level).The surface water molecules and/or hydroxyl ions react with the holes on the valence band to generate hydroxyl radicals ( • OH), which are a powerful oxidizing agent (+ 2.80 V).In Fig. 1(C), this shows that a strong ΔG gradient has to be overcomed for the degradation of persistent organic pollutants: notice the energy levelpositioning differences in Figs.1A, B, C.

Basic photophysical principlescharacteristics of photocatalytic nanomaterials
In the context of photocatalysis, a reference solar spectrum (atmospheric mass, AM1.5) is commonly considered.1.5 atmosphere thickness corresponds to a solar zenith angle of z = 48.2°.Therefore, AM1.5 is useful to represent the overall yearly average for mid-latitudes on Earth.The solar intensity versus air mass (AM) is given by where solar intensity external to the earth's atmosphere, I 0 = 1.353 kW/m 2 , and the factor 1.1 is derived assuming that the diffuse component of sunlight is 10 % of the direct component.For AM1.5 this results in a solar power density ~930 W/m 2 (Würfel P., 2005) In the fundamental Shockley-Queisser model, it is assumed that photons with energy below the band gap (hν< E g ) are not absorbed at all, while photons with energy far above the band gap (hν > E g ) are absorbed, but all that excess energy is wasted as thermal energy (Shockley W. and Queisser H. J., 1961).
Here, we notice that for (hν >> E g ), electrons are excited to an energy-level far above the conduction-band minimum, and holes far below the valence-band maximum.After such an excitation, e -and h + relax to the band edges very rapidly (t < ns).If this relaxation could be stopped, then the highly energetic "hot" h + /e -pair efficiency can theoretically considerably exceed the existing theoretical limits posed by Shockley and Queisser for p-n junction photovoltaics (Shockley W. and Queisser H.J., 1961).In practice, this is very difficult; thus far, even a proof-of-principle laboratory demonstration of a complete hot-electron device is lacking.

Formulation of the Photophysical Efficiency Concept
In their report, Chen et al. ( 2010) make clear that solarto-hydrogen (STH) efficiency should be used as a common reference benchmark reporting.Other efficiency parameters can also be used, providing insight into the functionality and limitations of a device; definitions are reviewed in (Chen Z. et al., 2010;ENERGY.GOV, 2015;Lianos P., 2011).In all technological applications of photocatalysts, the photophysical efficiency of a given material is the core parameter that will determine its ultimate eligibility.Then, scalability, cost and life-cycle issues have to be taken into account to permit an industrial-scale consideration.The two main parameters are: • Quantum efficiency = molecules of product produced per photon absorbed; • Solar efficiency = molecules of product produced per incident solar photon.Focusing on the h + /e -production efficiency of a given material, without reference to the subsequent interfacial or liquid-phase reactions, we may use a 'photophysical efficiency parameter'.The photophysical efficiency of a photocatalyst concerns all processes inside the particle that include the photon absorption until migration of the electron and the hole to their first interfacial acceptors.
As detailed by Bolton (1978Bolton ( , 1985)), the photophysical efficiency parameter can provide a common base of discussion for all photodevices described in Fig. 1 (Bard A.J., 1980).In this context, the overall catalytic efficiency of a given photocatalyst can be formulated as follows: where the two factors f particle and ELP are defined as follows: • f particle = efficiency of electron-hole photoproduction • ELP = energy level positioning of the e -/h + relative to their interfacial electron-acceptor and holeacceptor, respectively.The ELP factor takes into account the efficiency of energylevel alignment for the elementary-step, for example from a TiO 2 nanoparticle to Pt 0 , in the TiO 2 -Pt 0 system.Equation ( 1) can be further expanded by considering that f particle consists of two factors: where the two factors f structural and R kinetic are defined as follows with α = solar-light absorbance efficiency RR = h + /e -recombination rate M e , M h = mobilities of the electrons and the holes and SSA=specific surface area of the catlyst.

Direct vs. Indirect Band-Gap Semiconductors:
[A]-Direct band-gap semiconductor: For a direct bandgap semiconductor, α is related to the frequency of the light, ν, according to the following formula: where α is the light-absorption coefficient, v = light frequency, h = Plank's constant, E g is the semiconductor's band-gap energy, Α* is a frequency-independent constant, with formula (Würfel P., 2005) and is the reduced mass, based on the effective masses m e * and m h * of the hole and the electron, respectively.n is the real part of the refraction index of the material, χ υc is a crystal-matrix parameter usually taken equal to the lattice constant for each material.q and ε 0 are the electron charge and vacuum permittivity, respectively.This formula is valid only for light with photon energy larger, but not too much larger, than the band gap (more specifically, this formula assumes the bands are approximately parabolic), and ignores all other sources of absorption other than the band-to-band absorption in question, as well as the electrical attraction between the newly created electron and hole.
[B]-Indirect band-gap semiconductor: On the other hand, for an indirect band gap, the formula is: where Ε p the energy of the phonon i.e. a lattice vibrational state, that assists in the transition; kT is the Boltzmann thermal factor.This differentiates a so-called direct from an indirect semiconductor.
A direct band-gap semiconductor is formed when the minimum energy of the conduction band is at the same k-position in the Brillouin zone to the maximum of the valence-band, see Fig. 3.The notion of direct vs. indirect band gap has profound impact on the e -/h + lifetime and recombination dynamics: when the excited electrons come back from the conduction band to the valence band due to recombination of photogenerated e -/h + , they will release their extra energy as photons of frequency v, i.e.
Moreover, the excited electrons must meet the transition selection rule of momentum conservation, ħk e(conduction band) -ħk e(valence band) = ħq phonon (5b) (indirect band gap e -/h + recombination) where ħ is the reduced Planck constant, k e(conduction band) , k e(valence band) are the electron wave vectors at the conduction band and valence band, respectively, q is the wave vector of the assisted phonon (Zhang J. et al., 2014).Using this formulation, for a direct band-gap semiconductor: ħk e(conduction band) = ħk e(valence band) => ħq phonon = 0 (5c) (direct band gap e -/h + recombination) i.e.only a photon without phonon is involved in e -/h + recombination (Zhang J. et al., 2014).Examples are rutile and brookite: an electron only emits a photon following recombination of photogenerated e -/h + .However, the recombination of photoexcited electron and hole in anatase is assisted by a phonon, because ħk e(conduction band) ≠ ħk e(valence band) for this indirect band-gap semiconductor (Zhang J. et al., 2014).Thus, the excited electrons cannot recombine directly with holes, resulting in an increase of the photogenerated electron-hole lifetime in anatase, relative to that of rutile and brookite.As a result, the diffusion length and reaction time of the electron and hole excited in anatase also increase.
In this context, in general, direct band-gap semiconductors are expected to have higher h + /e -recombination rates (shorter e -/h + lifetimes) than indirect band-gap semiconductors.Examples of common direct band-gap semiconductors include ZnO and GaAs, while crystalline Si forms indirect band-gap semiconductors.All quantum dots are direct band-gap materials.
The direct or indirect band-gap character of BiVO 4 has been a matter of debate (Cooper J.K. et al. 2015;Walsh A. et al., 2009).In an earlier report, Walsh et al. (2009) reported BiVO 4 to be a direct-band-gap material, with a direct band-gap E g = 2.4 eV, however recent data by Cooper et al. (2015) suggest that BiVO 4 is an indirect band-gap semiconductor with a transition located at 0.2 eV below the direct one at 2.4 eV.This very small energypositioning difference dictates the difficulty for its detection (Cooper J.K. et al., 2015;Walsh A. et al., 2009).Experimentally, the direct and indirect band gaps can result in resolvable differences in UV-vis transitions.Thus direct and indirect band gaps can be estimated by the so-called Tauc analyis (Tauc et al. 1970) as follows: using the UV-vis data for a given semiconducting material, we typically plot (αhν) n versus (hv) where α is the measured absorption coefficient in the UV-vis spectrum.To do this, we have to convert the x-axis from wavelength (λ) to eV using the relation This method can determine the optical bulk band gap (E g ) and, ideally, can distinguish between allowed direct (n = 2) and allowed indirect (n = 1/2) transitions (Tauc et al 1970;Elliott R.J., 1957).

Effective mass-effect on e -/h + transfer rates
The quantum efficiency of a photocatalyst is also affected by the transfer rate of photogenerated electrons and holes (Linsebigler A.L. et al., 1995).The transfer rate of h + /e -can be indirectly assessed by the effective mass (m*) of electrons and holes (Linsebigler A.L. et al., 1995).In general, the transfer rate of photogenerated h + /e -is inversely proportional to their effective masses, i.e. the larger the effective mass of photogenerated carriers, the slower their transfer rate.Thus the small effective mass can promote the migration of charge carriers and inhibit their recombination.For example, the effective mass of photogenerated electrons and holes in CaZrTi 2 O 7 have been estimated to be m e * = 1.35m 0 and m h * = 1.23m 0 , respectively (Liu J.J. et al., 2012).In NaBiO 3 , effective masses of m e * = 0.14m 0 and m h * = 1.1m 0 have been estimated, respectively (Liu J.J. et al., 2012;Zhang J., 2014).For anatase TiO 2 , effective masses of m e * = 0.0948 and m h * = 0.1995 have been calculated (Zhang J., 2014).This correlates with the fact that anatase TiO 2 has a higher photocatalytic activity than CaZrTi 2 O 7 and NaBiO 3 (Liu J.J. et al., 2012(Liu J.J. et al., , 2013;;Zhang J., 2014).
Mobility-diffusion is intimately related with the e -/h + separation efficiency or the recombination rate RR in equation (2b).The case of Fe 2 O 3 is very instructive: This material has a great advantage in that it absorbs visible light E g = 2.2 eV (Miyauchi M. et al., 2002), but it suffers from a short hole-diffusion length (2-4 nm) (Formal F.L. et al., 2010).In summary, formulas (1-2) entail that the overall photophysical efficiency (PPE) is determined by three main terms: -[a] f structural is determined by the energetic and solid-state properties of a given photocatalyst.These are physical properties determined by the solid-state characteristics of a given particle.
-[b] R kinetic is determined by the dynamic, kinetic characteristics of photogenerated e -/h + in the photocatalyst.These are determined by the crystal structure and also by the defects, dopings, etc.
-[c] the energy level positioning (ELP) factor describes the efficiency of the electron/hole transfer from the particle to the first interfacial acceptor.These factors are often described in texts and visualized in pertinent reaction schemes such as in Fig. 4.
Thus, equations (1,2) offer a simple, compact form allowing an overview of the e -/h + photogeneration efficiency.They indicate that in a given nanoparticle type, the factors f particle = f structural * R kinetic are determined uniquely by its structure.Then, when in contact with an exogenous agent, electron level positioning will have to be determined for each type of application, since the types of acceptors will be different in different experiments.

Efficiency estimates
The energy levels of the conduction and valence bands, as determined form their ideal structures and compiled by Li and Wu (2015), are displayed in Fig. 5. Bolton et al. (1985) considered the thermodynamics of photochemical water splitting in detail and concluded that it is possible to store about 12 % of the incident solar energy in the form of hydrogen, allowing for reasonable losses (a total of 1.0 eV) in the electron transfer steps and the catalytic reactions of water splitting.In practice, there have been very few reports of photocatalytic water splitting using visible light (Hernandez-Alonso M. D. et al., 2009;Maeda K. et al., 2006;Zou Z. et.al., 2001), and the best reported quantum yield is so far about 6 % (Maeda K. and Domen K., 2008).
The solar power conversion based on semiconductor p-n junctions and semiconductor-liquid junctions can be quite efficient.In such p-n junctions, charge recombination is inhibited by the electric field that separates the lightgenerated minority and majority carriers (Hernandez-Alonso M. D., et al., 2009).
Photon-to-electron quantum yields approaching unity and power conversion efficiencies up to 18 % have been reported with single-crystal photoelectrodes (Khaselev O. et al., 1998).As noted above, if charge recombination reactions could be suppressed, then photocatalytic water splitting could become similarly efficient.On this front, significant advances have been made towards understanding photo-induced charge separation and interfacial  charge-transfer processes in semiconductor photocatalyst assemblies (Tang J.W. et al., 2011).Thermodynamic and kinetic limitations determine the efficiency of electron and hole transfer in semiconductor systems.The existence of shallow surface traps may enhance the overall efficiency by increasing the e -/h + lifetime.In any case, it is worth emphasizing that as much as 90 % of the electron-hole pairs recombine in less than 10 ns and consequently, photogenerated carriers available for surface reactions are quite limited.Values of quantum yield vary broadly with the process considered.
As an example, for TiO 2 reactions in solution, quantum yields are typically around 1 % (Wang C. et al., 2002) but they can exceed 5 % for some gas-phase reactions (Coronado J.M. et al., 2008).These values depend, as dictated by relations (1) and ( 2), on e -/h + transfer at the particle interface and the surface characteristics, but considering exclusively the photoactivation process, TiO 2 shows a limited performance.Thus in contrast to silicon, which presents an internal quantum efficiency (IQE) close to 100 % under illumination at 600 nm (Svrcek V., 2004), in TiO2, the absorbed photon-to-current efficiency is about 30 % at 360 nm (Lindgren T. et al., 2004).

Experimental Issues
Using diffuse-reflectance UV-vis spectroscopy, we typically measure the optical bulk band gap E g of the semiconducting particles.One has to discriminate between a direct or indirect band gap, allowed or forbidden transition.
As described in Section 3.2, UV-vis data are routinely analyzed using Tauc plots (Tauc J. et al., 1970), i.e. (αhv) n versus (hv) and fitting the data with the appropriate n-value, we can decide if we have allowed direct (n = 2), forbidden direct (n = 2/3), allowed indirect (n = 1/2), and forbidden indirect (n = 1/3) transitions.An example is given in Fig. 6 where absorption data for a Cu 2 O material are analyzed in the form of Tauc plots (Tauc J. et al., 1970).By plotting the (αhv) 2 versus (hv), the band gap for the allowed direct transitions is determined to be E g direct = 2.4 eV, Fig. 6.
By plotting (αhv) 1/2 versus (hv), the indirect optical band gap is determined E g indierect = 2.0 eV, Fig. 6c.Attention should be paid to the fact that in nanoparticles, a UV-vis-derived band gap for the bulk is not necessarily valid for the surface.Moreover, the slope of the tangent to the data is prone to subjective error and should be treated with caution.
Typical error-prone cases are the verification of 'visible light absorbance'.In cases where a semiconducting particle is just covered or intermixed with a colored substance, this will give a change in the UV-vis profile.If this spectrum is analyzed by a typical Tauc plot, then this will result in the erroneous conclusion that the band gap of TiO 2 has been narrowed, whereas it is just an overlap of two UV-vis spectra.
This can be a common mistake even for the common TiO 2 -Pt 0 particle characterization where Pt 0 particles are deposited on the TiO 2 surface for H 2 -generation applications.The color of Pt 0 particles causes a shift of the TiO 2 UV-vis, however this should not be interpreted as bandgap narrowing (Giannakas A. et al., 2016).Similarly, in dye-sensitized solar cells (DSCC), the dye-covered particles give a visible-light absorbance band, however this is not due to band-gap narrowing, see the example in Fig. 7.As explained by Grätzel M. (2005), the basic mechanism of DSSC involves photoexcitation of the dye, donation of electrons by the dye on the conduction band of the semiconductor, and filling of the dye-hole via the counterelectrode/electrolyte circuit.As a general guideline, the most interesting materials will exhibit a band gap, E g , of between 1.5 and 3.5 V, because it needs to be large enough to account for the thermodynamic energy requirements of water splitting (1.23V).In addition, conduction and valence-band potentials should encompass the hydrogen and oxygen evolution reactions, both the reduction of protons (E NHE(H/H2) = 0.0 eV) and the oxidation of water (E NHE(O2/H2O) 1.2 eV) (Kamat P.V., 2010).
Open-circuit potential is the maximum potential that a PEC can provide under zero current flow.This is the simplest commonly used method to measure conductivity and flat-band potential V FB.In practical terms, to obtain measurable V FB data for a photoactive material, one has to use high illumination intensities.However, depending on the type of material, high photon flux may cause photocorrosion or surface alteration of the material.This is evidenced as a non-linear change in the measured mVolts at increased photon flux.Chen et al. (2010) exemplified this effect for a p-GaAsPN material using a tungsten 150-W lamp, where for illumination intensities from 100 mW/cm 2 to 600 mW/cm 2 , the measured potential changed from ~430 mV to 600 mV (Chen X. et al., 2010).
H 2 /O 2 ratios are used in standard test protocols to prove that H 2 O splitting occurs.This requires perfect gasshielding of the reactor components and thermal stability of the reaction set-up, particularly the illumination set-up.
Stability tests often take days of repetitive use (Kudo A. and Miseki Y., 2009).In water splitting, both H 2 and O 2 should form with a stoichiometric amount, 2:1, in the absence of a sacrificial agent.Often H 2 is observed with a lack of O 2 evolution.In this case, the amount of H 2 evolution is usually small compared with the amount of a photocatalyst.It is not clear if such a reaction is photocatalytic water splitting, and it is important to clarify that it is not a sacrificial reaction (Froschl T. et al., 2012;Kudo A. and Miseki Y., 2009).Because the photocatalytic activity depends on the experimental conditions such as the light source and type of reaction cell, the activities cannot be compared with each other if the reaction conditions are different from each other.Therefore, determination of a quantum yield is important.The number of incident photons can be measured using a solid-state photodiode under non-saturating light-flux conditions.In a particle dispersion/slurry system, the real flux of photons absorbed by a photocatalyst will be lower than the incident photons because of scattering.So the obtained quantum yield is an apparent quantum yield (Kudo A. and Miseki Y., 2009).The apparent quantum yield is estimated to be smaller than the real quantum yield because the number of absorbed photons is usually smaller than that of incident light (Lewis N.S. and Nocera D.G., 2006).
In many cases, the most common test protocols in photocatalysis involve comparison versus the commercial Degussa-P25-TiO 2 for the degradation of dyes.However, extracting meaningful conclusions by such studies requires a complete data set that assesses the solid-state, interface and solution parameters.Ryu and Choi (2008) have compared the photocatalytic performance of 8 commercial types of TiO 2 for 19 different substrates.They concluded that not one of the tested TiO 2 is best for all the substrate types.As suggested (Ryu and Choi, 2008), multivariant statistical analysis, taking into account the key physicochemical properties of the photocatalyst in relation to the photocatalytic mechanistics, may provide a meaningful insight into the real performance of a photocatalyst.Useful details of experimental design for general photocatalysis can be found in Ohtani B. (2008).
Deactivation of surface sites.Deactivation of surface sites may occur in photocatalytic materials due to various reasons such as surface (photo) corrosion or surface fouling, or due to adsorption of reaction products on the surface.This phenomenon, however small it may be, has to be considered carefully (Dong S. et al., 2015).
Both these events result in deactivation of surface sites.As an example, if we have a loss of L % in each catalytic cycle, then, after k cycles, the remaining active sites are given by equation ( 7) Fig. 8 reveals that after just 100 cycles, a L = 0.01 % will result in more than 40 % loss of active surface sites.

Dispersed Powders vs. Electrodes
In nanosized materials, the ability to manipulate the particle size permits the chemical properties in nanoparticles to be tuned.Thus, size control has tremendously extended the potential of nanomaterials.It is well known that in its simplest form, water-splitting is the transformation of the energy of photons in electrons and holes and ultimately to H 2 and O 2. Thus it is easy to anticipate that if 100 % of input electrons (e -) can be successfully put to work and not lost, then a maximum amount of photons is transferred and stored in H 2 molecules.Ideally, maximizing the surface of the cathode and anode materials, a maximum conversion of photons to e -/h + pairs and ultimately to H 2 /O 2 is feasible.Thus, using nanoparticles instead of bulk materials should boost the fundamental e -/h + transfer efficiency and therefore, in water splitting, see Fig. 9 (Teo W.Y., 2012).In this context, the know-how gained from studies of PEC cells with bulk semiconductor electrodes can be applied to the design of systems in which semiconductor nanoparticles are used (Chen X. et al., 2010;Getoff N., 1990;Kudo A. and Miseki Y., 2009).For example, using powders of TiO 2 with platinum dispersed on the TiO 2 -surface, each particle can be pictured as a "short-circuited" PEC cell (Bard A.J., 1980;Chen X. et al., 2010;Teo W.Y et al., 2012), where the TiO 2electrode and Pt-counter electrode are in physical contact (Chen X., et al., 2010;Bard A.J., 1980;Teo W.Y et al., 2012).Irradiation of such nanoparticulate systems still involves the e -/h + pair formation and surface oxidation/reduction reactions, found in the cells but without external current flow.
Although nanopowders are obviously much simpler to use, the advantage of the large spatial separation between the oxidation and reduction sites that occurs in bulk electrodes used in PEC, remains a weak point of the nanopowders (Teo W.Y. et. al., 2012).However, at a molecular scale, the distance between oxidation and reduction sites on large-enough particles is still probably large compared to those found in solution-phase photochemical redox reactions (Teo W.Y. et al., 2012) where the redox products are in close proximity within the same solvent cage.Keeping this in mind, the know-how of PEC electrodes can serve as a useful guide to the design of the particlebased catalytic devices.

Choice of photocatalyst material
So far, nanocrystalline TiO 2 is the most commonly used photoactive material to build a photoanode.However, other oxides and other n-type semiconductors have been studied as well.Man I. et al. (2011) have explored trends in electrocatalytic activity of the oxygen evolution Fig. 9 Conceptualization of the equivalence of a suspended nanoparticle system as a miniaturized PEC Reprinted with permission from Ref. (Teo W.Y. et al., 2012).Copyright: (2012) American Chemical Society.Then, based on this, they have analyzed the reaction-free energy diagrams, ΔG, of many oxides and derived an "activity volcano", see Fig. 10.This plot can provide a universal descriptor for the oxygen evolution activity.Fig. 10 suggests a fundamental limitation on the maximum oxygen evolution activity of oxide catalysts.For example, Fig. 10 suggests that Mn-oxide and Co-oxide, i.e. much cheaper oxides than RuO 2 or IrO 2 , should be promising co-catalysts for the O 2 production reactions in water-splitting heterostructures.
A similar "volcano activity" plot has been derived by Morales-Guio C.G. et al. (2014), see Fig. 11, for the electron-accepting particles.This shows that the-expensive-Pt 0 , Pd 0 , Rh 0 particles are the most efficient for H 2 production technologies.Cu 0 , Ni 0 , Co 0 -as low-cost alternatives-are worth exploiting; however, avoiding surface oxidation is a great limitation at present.Non-oxide chalcogenide n-type semiconductors such as ZnS, CdS, CdSe, etc. have been extensively studied for application in the photocatalytic (Kakuta N. et al., 1985) and photoelectrochemical (Hodes G. et al., 1976) production of H 2.
The reason is that, with the exception of ZnS, they absorb visible light.However, these substances are vulnerable to oxidation (for example, CdS + 2H + → Cd 2+ + H 2 S), which dissolves the material and destroys it.In the presence of a sacrificial agent, it is possible to harvest h + , thus restricting the particle degradation in solution.Also, chalcogenide materials can be synthesized in the form of core-shell nanostructures (Amirav L. and Alivisatos A.P., 2010) which have greater stability.ZnS-in spite of its highest Eg among all halcogenides-has been a popular research target since its CB and VB levels are favorably placed, as can be seen in Fig. 5. Thus, ZnS possesses strong oxidant and reductant power.It combines with CdS, so that, by changing their proportions, the overall light-absorbance can be tuned within a wide range.In the most advanced Z-scheme concept, the combination of two semiconductors assists electron-hole separation anyway (see Section 4.1.5).

Bi-Vanadate-Based Nanostructures
Monoclinic sheelite BiVO 4 is an n-type semiconductor with a direct band gap of 2.4 eV, thus it absorbs visible light, λ = 420-530 nm, and is stable in neutral electrolyte, non-toxic and relatively cheap (Kudo A. et al., 1999).
BiVO 4 has an optical penetration depth of lp = 100-500 nm at λ = 420-530 nm.Compared to other common O 2 -evolving photocatalysts such as WO 3 or Fe 2 O 3 , BiVO 4 has a relatively high CB edge (+ 0.02 eV) (Kudo A. and Miseki Y., 2009;Park Y., et al., 2013) and, as a consequence, requires less bias potential to raise the potential of photoelectrons above the H + /H 2 reduction potential (0.0 V).BiVO 4 is thermodynamically favorable for the half-reaction of H 2 O oxidation, but requires an externally applied bias (Kudo A. and Miseki Y., 2009;Park Y. et al., 2013) for the water reduction half-reaction.The theoretical maximum STH efficiency is 9.1 % and the theoretical maximum photocurrent is 7.4 A/cm 2 (Park Y. et al., 2013).In the last decade, BiVO 4 has arisen among the main materials for visible-light photocatalysis, despite its shortcomings, i.e. low e-conductivity and poor H 2 O-oxidation kinetics.These undesired properties may be improved by modifications, including the formation of heterostructures with another semiconductor, charge mediator, or use of a co-catalyst (Kudo A. et al., 1999;Kudo A. and Miseki Y., 2009).So far, the theoretical maximum STH efficiency  9.1 % cannot be achieved due to serious charge recombination and low water oxidation kinetics.BiVO 4 has an electron-diffusion length of only ~10 nm and a holediffusion length of ~100-200 nm (Hong S.J., et al., 2011;Li Z., et al., 2013).This causes excessive consumption of electrons by charge recombination.Doping BiVO 4 with Mo and W can significantly increase its electron-diffusion length to ~300 nm (Kudo A. et al., 1999).Poor kinetics for O 2 evolution at the BiVO 4 surface constrains its efficiency.This can be improved by co-catalyst loading on BiVO 4.
In an innovative approach, Amal's and Kudo's groups (Ng Y.H., et al., 2010) demonstrated that a reduced graphene oxide RGO/BiVO 4 nanocomposite, see Fig. 12, can achieve almost an 80-fold [i-V] enhancement in the photoelectrochemical water splitting reaction.In the same work (Ng Y.H. et al., 2010), BiVO 4 was used successfully as a visible-light-photocatalyst to photocatalytically reduce GO to reduced graphene oxide.Recently, Kim et al. (2014) demonstrated that a nanoporous BiVO 4 (SSA = 31.8m 2 /g) effectively suppresses bulk carrier recombination of BiVO 4 without additional doping, manifesting an electron-hole separation yield of 0.90 at 1.23 V. Use of two different oxygen evolution catalyst (OEC) layers, FeOOH and NiOOH, resulted in a BiVO 4 /FeOOH/NiOOH photoanode that achieved a photocurrent density of 2.73 mA/cm 2 at a potential as low as 0.6 V.The use of FeOOH and NiOOH reduces interface recombination at the BiVO 4 / OEC junction, while creating a more favorable Helmholtz layer potential drop at the OEC/electrolyte junction, Kim et al. (2014).

Z-Scheme Nano Photo Structures
A "Z-scheme" (Sasakia Y. et al., 2008) can be considered a special type of PEC-PEC tandem cell that imitates the architecture of the two natural photo systems of green plants, i.e. photosystem I and photosystem II.In the nano-Z-scheme approach, see Fig. 13, two semiconductors, I and II, are bridged with a redox particle or a liquid-redox couple mediator.Energetic considerations of the photodecomposition of water by semiconductor systems using sunlight suggest that a single light-absorbing system necessitates the use of a semiconductor with a rather large band gap of 2.5 to 3.0 eV (Chen X. et al., 2010;Kubacka A. et al., 2012;Sasakia Y. et al., 2008).This stems from the fact that the requisite energy and driving forces must suffice to overcome the redox barrier of 1.23 eV for water splitting, see Fig. 13.
Alternatively, efficient utilization of solar energy for water splitting may imitate the successful strategy of green plants, that is, the use of two photo systems with the absorption of two photons of lower energy per electron transferred.So far, several Z-schemes with twoheterostructured photocatalysts have been investigated in H 2 and O 2 production from water using solar energy (Chen X. et al., 2010;Kubacka A. et al., 2012;Sasakia Y. et al., 2008).
These works have established that overall water splitting can be achieved by constructing a Z-scheme photocatalytic system using photocatalysts, each one being active only for one of the half-reactions of water splitting (H 2 or O 2 evolution reactions) in the absence of sacrificial  reagents.In general, a redox mediator is required to couple with two photosystems.In a further leap forward, to eliminate undesirable backward reactions caused by redox mediators (such as Fe 3+ /Fe 2+ ), a Z-scheme system without an electron mediator, i.e. an all-solid-state Z-scheme photocatalyst, has attracted much attention, see Fig. 13.This advanced Z-scheme retards back-reactions and increases the reaction efficiency.A number of allsolid-state Z-scheme photocatalysts (Chen X. et al., 2010;Kubacka A. et al., 2012) such as CdS-Au-TiO 2 (Tada H. et al., 2006) have been demonstrated to exhibit high photocatalytic activity, far exceeding that of the single-component systems.In all these multi-component Z-scheme photocatalyst systems without exception, the separation of e -/h + pairs is promoted within the two semiconductors and further improved by the coupled metal particle that acts as the electron-transfer mediator, Fig. 13.
TiO 2 is a benchmark semiconductor, but it has a wide band gap (3.2 eV for anatase).Hence it can only absorb UV light, which accounts for less than 5 % of total solar radiation (Grätzel M., 2001;Miller E.L. et al., 2010).The most effective approach to increase the STH efficiency is to decrease the E g in order to extend the light absorption spectral range into the visible-light region (~43 % of total solar radiation) and even the near-infrared light region (~80 % of total solar radiation).However, the band gap must be large enough to meet the thermodynamic and kinetic requirements for water splitting.Thermodynamics dictate the minimum energy required to overcome the standard Gibbs free energy change (1.23 eV) for water splitting.Thermodynamic losses (0.3-0.5 eV) should be added to this value.Thus an overpotential of 0.4-0.6 eV is required to enable a fast reaction (Asashi R. et al., 2001;Liu G. et al., 2012).Therefore, an ideal band gap is 1.9-2.3eV.
Nitrogen-doping of TiO 2 (Asashi R. et al., 2001) has boosted research in visible-light photocatalysis.Other examples of innovative visible-light structures involve the following: -[i] "Red anatase" TiO 2 , which has been prepared by heating anatase TiO 2 microspheres with a pre-doped interstitial boron shell to between 580 and 620 °C in a gaseous NH 3 atmosphere.
This material has achieved a high photoelectrochemical water splitting activity (Liu G., et al., 2012).The UV-vis absorption spectrum of red anatase TiO 2 shows an extended absorption edge up to 700 nm, covering the full visible-light spectrum due to a band-gap gradient which varies from 1.94 eV on the surface to 3.22 eV in the core by gradually elevating the VB maximum, Fig. 15.The approach for producing this red-TiO 2 involved a predoping step, i.e. introduction of an interstitial boron gradient (Liu G. et al., 2012).This in turn improved the incorporation of substitutional N in the TiO 2 bulk without introducing Ti 3+ impurity levels (Liu G. et al., 2012).Structurally, the interstitial boron dopant effectively weakened the surrounding Ti-O bonds to facilitate easier N-substitution and increased the chemical stability of B-N co-doped  TiO 2 .Moreover, the extra electrons from B-atoms compensated for the charge difference between the lattice oxygen and a substituted ion to maintain charge neutrality.
-[ii] "Black TiO 2 " can be produced by two different technologies: treatment of TiO 2 particles under H 2 at high pressures and temperatures (Chen X. et al., 2011), which results in the formation of a Ti 3+ shell by reduction of the Ti 4+ .This Ti 3+ shell has a distorted structure, which in turn creates additional interband states.This results in light absorbance in the visible wavelengths (Chen X. et al., 2011).
Recently, we have discovered that a black TiO 2 can be produced by a scalable flame spray pyrolysis process in one step (Fujiwara K. et al., 2014), Fig. 16.Reductant combustion intermediates present during flame synthesis of these materials partially induce strong TiO 2 metal support interactions (SMSI), resulting in crystalline Tisuboxides (Ti 3 O 5 , Ti 4 O 7 ).The growth of such suboxides can be controlled via control of the flame spray pyrolysis conditions, allowing the light absorption intensity in the visible spectrum to be tuned.

The notion of "renewable H 2 " production
Currently, industrial electrolysis uses clean water in order to prevent the fouling of system components.However, the need for clean water prevents the exploitation of large-body sources such as seawater, industrial waste and municipal waste as water feed.This poses the challenge that the use of cheaper, alternative sources of water instead of purified distilled water should be envisioned as the next target for truly unlimited production of "renew-able H 2 ".At the consumer scale, H 2 can be either produced centrally or on-site.The market for on-site generation systems is growing due to certain advantages associated with not needing transportation and delivery.Currently, most of industrial-use H 2 is produced by the steam reforming of natural gas or CH 4 (Conway, B.E, and Tilak, B.V. 2002;Holdren J.P., 2007) due to the availability and low prices of natural gas at present.Next, after the steam reforming of natural gas, the partial oxidation of petroleum oil is second in H 2 production capacity.
So far, with regard to biofuels, most researchers choose to study methanol, ethanol or glycerol as model fuels (Chen X. et al., 2010;Lianos P., 2011).There are three main reasons for this: (1) methanol, ethanol or glycerol give the highest yield; (2) they have been extensively studied, so a lot of data and know-how on their use exists; and (3) they can be cost-efficient products of biomass.Thus in a mature biomass-based market, methanol, ethanol or glycerol should be available and renewable.As an example, glycerol, a waste byproduct material of biodiesel production (Lianos P., 2011), is available in mass-produced quantities.In addition to small alcohols, other organics such as sugars and organic acids have also been a popular choice for photocatalytic H 2 production (Chen X. et al., 2010;Lianos P., 2011).
Thus truly "renewable H 2 " production can be achieved if the photocatalytic technology can exploit seawater or wastewater feed or via the use of small organics produced via a sustainable process.

Next-decade targets
Currently, the long-term objective aimed at by the US Department of Energy is the photocatalytic H 2 production based on a durable, semiconductor-based, solar-driven water-splitting device with the following specifications: (1) solar-to-hydrogen (STH) efficiency > 20 %, (2) that can operate under at least 10-15 times solar concentration, and generate renewable hydrogen for <$2/ kg .(ENERGY.GOV, 2015) In the same context, the US-DoE near-future objectives specify the nanomaterials' physicochemical standards as follows: development of surface modifications that will make the materials viable at high current densities.This is related to the need for highly efficient, durable materials for PEC water splitting using concentrated solar energy.This is important since in order to minimise H 2 -production costs, operation under high photon and electron flows is identified as the most significant driver.Mechanical surface durability can be improved through the crystal structure and surface engineering.To this end, surface defects should be engineered in a balanced way, since surface defects might be desirable for enhanced reactiv- ity; however, at the same time, too many surface defects might decrease the durability of a nanocrystal.In parallel, the optimization of existing materials or the discovery of novel materials should incorporate the predictive power of theory and modeling.The target of 20 % STH efficiency targeted by the US-DoE is a feasible goal.As a reference, a 10-fold solar light concentration with a 20 % STH gives approximately 160 mA/cm 2 (Conway, B.E and Tilak, B.V., 2002;Lewis N.S. and Nocera D.G., 2006).However, although the current H 2 oxidation reaction rates in fuel cells achieve this range, i.e. 100-1000 mA/cm 2 , a severe rate-limiting step is imposed by the very slow O 2 reduction reaction that typically imposes low rates, i.e. as low as 10 -6 -10 -11 A/cm 2 (Conway, B.E and Tilak, B.V., 2002;Lewis N.S. and Nocera D.G., 2006).Thus intensive research on novel or improved O 2 -reduction materials is in high demand.
The US-DoE targeted (by 2018) an increase in chemical conversion process efficiency for PEC from 4 % to 12 % (ENERGY.GOV, 2015;Miller E.L., et al., 2010).This will require significant breakthroughs in the development of new materials.The challenge can be seen through Fig. 14, which allows an estimate of the fundamental solar absorption limitations in H 2 production in PEC for a number of the tested metal-oxide semiconductors (Li J. and Wu N., 2015;Miller E.L. et al., 2010).In Fig. 14, the potentially achievable photocurrent levels and STH conversion efficiencies of a given semiconductor/ electrolyte interface are related to the semiconductor band gap.This is based on an idealized case where 100 % of the photons in the solar spectrum with energies exceeding the band gap are absorbed and converted.It must be strongly emphasized that the STH efficiencies, as shown in Fig. 14, are only defined in stand-alone material systems with sufficient photo-induced potential to thermodynamically split water.To date, only very high-band-gap materials have exhibited stand-alone water-splitting in single junction configurations, as seen in Fig. 14.It is seen that the resulting STH efficiencies of these materials is less than 1 % (Li J. and Wu N., 2015;Miller E.L., et al., 2010;Rocheleau R. and Miller E., 1997).Interestingly, today's PEC performance status of a STH = 4 % has been achieved in multi-junction configurations (Grätzel M., 2001), not in a single-particle system.Thus, it is important to emphasize that Z-scheme, multi-junction configurations are the way forward to create stand-alone watersplitting systems.This will allow the use of some of the low-E g materials under a proper "branching" of the CB/ VB levels, see the concept in Fig. 13.If this is the case with Z-scheme configurations, the photo-generated currents and therefore their STH efficiencies will not be constrained by the E g value or e -/h + transfer rates, but rather by the semiconductor/electrolyte interface limitations.Even in such Z-scheme configurations, achieving STH performance ~15 % will require development of efficient and stable semiconductors with the E g near 2.0 eV.Ultimately, after the lab work, marketing of the most competitive materials or devices will require low-cost nanomaterials production and device synthesis.On this front, scalable one-step particle production technologies such as flame spray pyrolysis (Teoh W.Y. et al 2012;Ng, Y. et al., 2010;Fujiwara K. et al., 2014), would contribute decisively.

Conclusions
Photoactive nanomaterials hold great potential for environmental technologies.These include the production of electricity, energy-fuel production and photocatalytic water remediation.Current targets posed by the US Department of Energy are discussed based on a 2015 report.In contrast to the case of p-n junction photovoltaics where the manufacturing cost of the required high-purity single crystals emerges as the main objective (mostly commercial, rather than fundamental research), a key target for the field of photoelectrochemical nanotechnology is the discovery and development of new materials to advance the field.These include: -raising the solar to hydrogen yields > 15 %, -manufacturing of all-solid-state Z-scheme photoactive materials, -optmized materials for visible-light absorbance with enhanced quantum yields.

Nomenclature
Symbols and units should be listed in alphabetical order with their definition and dimensions in SI units.

Fig. 1
Fig. 1 Schematic description of the fundamental photoactive nanocatalytic systems.A fundamental differenceis the positioning of the energy levels of the redox couples vs. the photo-induced electrons' and holes' energy levels.For simplicity, external bias voltage is not included in these schemes.External bias, depending on polarity and voltage, will shift upwards or downwards the energy levels of e -and h + .

Fig. 2
Fig. 2 Photo-induced reactions and radical transient species produced arising from the excitation of a photocatalyst particle in aqueous solution Reprinted with permission from Ref. (Teo W.Y. et al., 2012).Copyright: (2012) American Chemical Society.

Fig. 4
Fig.4Efficiency determining reactions in a photo-induced e -/h + generation on a semiconductor nanoparticle Reprinted with permission from Ref.(Kudo A. et al., 2009).Copyright: (2009) Royal Society of Chemistry.

Fig. 5
Fig. 5 Conduction-band and valence-band edge positions (in eV units) for selected semiconductors at pH 0.Left Y-axes provide energies vs. the vacuum energy level.Right Y-axes provide energies vs. the NHE energy level.Reprinted with permission from Ref. (Li and Wu, 2015).Copyright: (2015) Royal Society of Chemistry.

Fig. 7
Fig.7The UV-vis spectrum of a Ru-based dye entails strong absorbance bands at visible wave lengths, with no TiO 2 band gap narrowing Reprinted with permission from Ref.(Grätzel M., 2005).Copyright: (2005) American Chemical Society.

Fig. 8
Fig.8Total loss of surface sites, for a [loss/per cycle] factor L, after N reaction cycles, estimated from equation (7).After just 100 cycles, L = 0.01 % will result in the loss of ~ 40 % of the active surface sites.

Fig. 14
Fig. 14 Solar-to-hydrogen (STH) efficiency and maximum photocurrent trends vs. the band gap for photoactive semiconductors Reprinted with permission from Ref. (Li J. and Wu N., 2015).Copyright: (2015) Royal Society of Chemistry.