Design and Development of Photoanodes for Water-splitting Dye-sensitized

8/16/2019 Design and Development of Photoanodes for Water-splitting Dye-sensitized

1/28

This journal is c The Royal Society of Chemistry 2012 Chem. Soc. Rev.

Cite this: DOI: 10.1039/c2cs35246j

Design and development of photoanodes for water-splitting dye-sensitized

photoelectrochemical cells wJohn R. Swierk and Thomas E. Mallouk*

Received 8th July 2012DOI: 10.1039/c2cs35246j

Dye sensitized solar cells (DSSCs) use low-cost materials, feature tunable molecular sensitizers,and exhibit quantum efficiencies near unity. These advantageous features can be exploited in thecontext of solar water splitting by functionalizing DSSCs with catalysts for water oxidation andreduction. This article will cover the development of photoanodes for water splitting DSSCs fromthe perspective of water oxidation catalysts, sensitizers, electron transfer mediators, photoanode

materials, and system level design. Within each section we will endeavor to highlight criticaldesign elements and how they can affect the efficiency of the overall system.

1. Introduction

Dihydrogen, the smallest and simplest molecule, plays anincreasingly important role in the global energy economy. In2010, over 40 million tons of hydrogen were produced globally,the vast majority used in the production of petrochemicals andammonia. 1 Beyond these industrial applications, the demand forhydrogen is expected to rise as it nds greater use as an energycarrier. Unlike hydrocarbons, hydrogen gas produces onlywater as a combustion product and can readily run any system

currently using natural gas. The stored chemical energy inhydrogen can be efficiently converted to electricity by air-breathing fuel cells. 2 In applications that require carbon-basedfuels, hydrogen can be used to form methane via the Sabatierreaction, combined with carbon monoxide to make hydro-carbon fuels by the Fischer–Tropsch process, catalytically

reacted with carbon dioxide to make methanol, or added toliqueed coal to make synthetic gasoline. 3

Enabling a fuel economy based on hydrogen is however amajor challenge. Currently, hydrogen is produced directlyfrom methane or indirectly via electrolysis using electricityfrom non-renewable resources. 1 A more sustainable approachwould be to produce hydrogen from renewable energy sources. 4

Of the renewable energy sources, sunlight has enormous potential.Its power density is 1 kW m 2 on a clear day and the globalpotential of solar power is approximately 10 000 greater than the

current primary power use (16 TW) of the entire world. 4 In thiscontext, hydrogen can be liberated by splitting water intomolecular hydrogen and oxygen, either directly by photocatalysisor in a photoelectrochemical process:

H 2O - H 2(g) + 1/2O 2(g) (1)

The basic thermodynamic requirements for splitting waterare modest. It is an uphill reaction, requiring the input of energy, with a Gibbs free energy of 237.178 kJ mol 1 .

Department of Chemistry, The Pennsylvania State University,University Park, Pennsylvania 16802, USA. E-mail: [email protected] Part of the solar fuels themed issue.

John R. Swierk

John R. Swierk received his BS in Chemistry and BSE in Materials Science & Engi-neering from the Universityof Pennsylvania. In 2008, hebegan his PhD at Penn Stateunder the supervision of Pro- fessor Thomas E. Malloukstudying electron transfer inwater splitting systems.

Thomas E. Mallouk

Thomas E. Mallouk received his ScB degree from BrownUniversity (1977) and hisPhD in Chemistry from theUniversity of California,Berkeley (1983). He was a postdoctoral fellow at MIT (1983–1985). His research focuses on the synthesis,assembly, and applications of nanoscale inorganic materials.

Chem Soc Rev Dynamic Article Links

www.rsc.org/csr REVIEW ARTICLE

O

View Online / Journal Homepage

http://dx.doi.org/10.1039/c2cs35246jhttp://pubs.rsc.org/en/journals/journal/CShttp://dx.doi.org/10.1039/c2cs35246jhttp://dx.doi.org/10.1039/c2cs35246jhttp://dx.doi.org/10.1039/c2cs35246j


2/28

Chem. Soc. Rev. This journal is c The Royal Society of Chemistry 2012

Considered as an electrochemical potential, a minimumpotential of 1.229 V is required at 298 K. In principle anywavelength of light shorter than 1 mm has enough energy tosplit the water molecule, allowing for the use of the entirevisible solar spectrum and much of the near-infrared, whichtogether comprise B 80% of the total solar irradiance. 5

How well photochemical water splitting transitions to realworld systems hinges on how efficiently photons are absorbedand used to form hydrogen. In real systems not all of thesolar energy available can be converted into a chemical fuel.Detailed balance calculations can be used calculate a maximumtheoretical efficiency by considering various loss processes.Shockley and Queisser 6 rst developed this type of analysisfor single junction silicon solar cells with many other authorsextending it to multiple junction cells 7 and photochemicalprocesses. 5,8,9 All treatments begin by assuming an ideal absor-ber, a material with a single bandgap or absorption threshold(U g) that absorbs all photons with energy greater than U g andabsorbs no photons with energy less than U g. As a consequence,some fraction of the solar spectrum with energy less than U gis not used by the photochemical system. It is also assumedthat excited electrons rapidly lose energy in excess of U g bythermalization and relax back to the bandgap or absorptionedge. Thus much of the energy from photons with energygreater than U g is lost to the system as heat. Considering onlythese two loss mechanisms would mean that at most an idealsystem could convert B 50% of incident solar energy into achemical fuel. However a third loss mechanism is introducedbecause the maximum extractable work from a photovoltaic orphotochemical system is always less than U g. This is because asea of electrons in the ground state surrounds each electron inan excited state, creating signicant entropy of mixing andintroducing an unavoidable thermodynamic loss parameter.Finally, a small fraction of excited states must decay radiativelyin order to maintain a high chemical potential. When these lossmechanisms are taken together, a single absorber with U g =1.59 eV has a maximum efficiency of 30.6% under idealconditions. Multiple absorber systems can exceed this limit,though the theoretical efficiency depends on how many absorbersare used. Realistically attainable efficiencies are well below thesetheoretical limits due to electron transfer losses, catalyst over-potentials, and reection losses. Fig. 1 demonstrates how additionalprocesses can contribute to energy loss.

2. Water splitting systems

Nature provided an enduring blueprint for photochemical watersplitting with the evolution of photosynthesis 2.4–3 billion yearsago. 10 Algae and higher green plants use two coupled photo-systems, photosystem II and photosystem I, to absorb lightand split water into oxygen and NADPH. Briey, a series of light harvesting pigments absorb visible light and within onepicosecond funnel the excitation energy to a P680 chlorophylldimer located within the reaction center protein of photo-system II. The excited P680 passes an electron to a nearbypheophytin, a chlorophyll molecule lacking a Mg 2+ ion, which isthen followed by rapid electron transfer to a plastoquinone. 11

Within200ps, the electronis physically separated from the oxidizedP680 + by a distance of about 26 A ˚ giving a charge-separated

state stable for hundreds of microseconds. 12 A redox activetyrosine group is oxidized by P680 + and mediates the rapidcharge transfer steps at P680 with the slow oxidation of theoxygen-evolving complex (OEC). At the heart of the OEC liesan oxo-bridged cluster of one calcium and four manganeseions. The OEC accumulates four oxidizing equivalents beforereleasing oxygen and being regenerated to its reduced restingstate. 13 As the OEC is being oxidized, the electrons from P680leave photosystem II via another plastoquinone and cytochromeb6f molecule to make their way to photosystem I. Here anotherchlorophyll, P700, is excited and oxidized, with electrons beingfunneled down a charge transfer chain until they are used toreduce NADP + to NADPH. The electrons from photosystemII regenerate the oxidized P700. Overall, two photons arerequired for every electron transferred (Fig. 2). 14

Despite all of the exquisite complexity and sophistication of natural photosynthesis, plants are optimized for reproductivesuccess and not efficient energy conversion. For example, P680and P700 have signicant spectral overlap, meaning that a largeportion of the usable solar spectrum is wasted. Likewise, plantsabsorb more light energy than they are able to process with upto 80% of absorbed energy being discarded. 16 Additionally,photorespiration can be responsible for a loss of up to 25% of the stored energy in plants. 17 These factors limit the overallefficiency of photosynthesis to less than 10% at low light levels,and 1–3% in full sunlight. 18–21 While the natural photosyntheticapparatus provides inspiration for the design of efficientbiomimetic systems, plant photosynthesis itself is unlikely tobe bioengineered to the level where it can be competitive withother commodity fuel sources (including photovoltaicscoupled to electrolyzers). 22 Among the major impedimentsto such improvement is the mass transport limit of atmo-spheric CO 2 . Because of the low concentration of CO 2 in the

Fig. 1 Energy diagram demonstrating energy losses in a water split-ting system sensitized by a molecular dye. Reproduced from ref. 5 withpermission from The Royal Society of Chemistry.

O

View Online

http://dx.doi.org/10.1039/c2cs35246j


3/28


atmosphere (390 ppm), photosynthesis is forced to operate atlow concentrations of CO 2 . The equivalent current densitythat a photosynthetic system can provide in atmospheric CO 2can be calculated from the large-area, boundary-layer mass-transport limit for atmospheric CO 2 capture (400 tC per hayear). 23 This mass-transport limit translates to an equivalentcurrent density of 3.7 mA cm 2 , assuming that the reductionof CO 2 to carbohydrates involves four electrons per carbonatom. The free energy stored per mole of carbon is 480 kJ inglucose, a representative carbohydrate product. It follows thatthe power that can be stored in carbohydrates at the masstransport limit is 4.6 mW cm 2 , which corresponds to amaximum power conversion efficiency of 4.6% (the powerinput from the sun being B 100 mW cm 2). However, likeany other energy conversion system, photosynthesis cannotoperate efficiently at the mass transport limit because of concentration polarization of CO 2. Some of the stored energyis also needed for respiration in living plants. 24 Consideringthese loss mechanisms, the CO 2 mass-transport limit corre-sponds closely to the B 1–3% efficiency of plant photosynthesisin full sun. In contrast, efficient solar energy conversion devicessuch as crystalline silicon cells, which operate at quantumyields (electrons generated per photon absorbed) near unity,have power conversion efficiencies and current densities thatare approximately one order of magnitude higher than photo-synthesis. For example, the short-circuit current density at a24.4% efficient silicon solar cell in full sun is 42.0 mA cm 2 .2,25

It follows that articial photosynthetic systems will need to bemore like the latter kind of device, ultimately operating nearunit quantum efficiency, in order to be competitive with otherforms of commodity energy.

The development of articial photosynthetic systems canroughly be divided into two general approaches, photocatalysisand photoelectrohemical cells (PEC). Any material that canboth absorb light and perform water oxidation or reduction canbe considered a photocatalyst, though in the context of photo-chemical water splitting the term is more narrowly applied.Overall water splitting by a photocatalyst typically refers to aphotoactive colloidal suspension, which produces hydrogen andoxygen in close proximity to each other. PECs can perform avariety of electrochemical half-reactions, and in regenerative

cells (like the DSSC), the anode reaction is simply the reverseof the cathode reaction and the cell generates electricity.Water-splitting PECs are a special case where hydrogen ismade at the cathode and oxygen at the anode. These electrodesare physically well separated from each other, in contrast to thesituation in photocatalysis. Many of the materials that are usedas photocatalysts are also employed as photoelectrodes in PECs.This approach is reminiscent of the compartmentalization of oxygen and NADPH production in natural photosynthesis.

More than 130 inorganic materials have been identied aspossible photocatalysts for water splitting. 26 Early work onphotocatalysts focused on titanium dioxide and demonstratedthe production of hydrogen and oxygen using UV light and aco-catalyst. 27,28 Photocatalysts with enhanced visible lightabsorption, such as BiVO 4, produce oxygen photochemicallywith visible light, but require the use of a sacricial Ag +

electron acceptor because the conduction band potential is toopositive for hydrogen evolution. 29,30 Overall water splittingusing a (Ga 1 x Zn x )(N 1 x O x ) photocatalyst modied withrhodium and chromium mixed oxide was achieved by Domenand co-workers. 31,32 An early version of this system had aquantum yield of 2.5% with 420 nm excitation, 33 which waslater improved to 5.9%. 33 In general though, these materials haverelatively wide bandgaps and poor absorption characteristics. Forexample the onset of absorption in (Ga 1 x Zn x )(N1 x O x ) is510 nm, and the quantum yield is low due to electron–holerecombination, another problem typical of photocatalytic systems.

The rst demonstration of photoelectrochemical water splittingused an oxygen-evolving rutile TiO

2 photoanode and hydrogen

evolving platinum cathode. 34 Rutile has a bandgap (3.0 eV) thatlimits light absorption to the ultraviolet, so the quantum yield of water splitting in sunlight is typically below 2%. Many groupshave worked to extend the visible absorption of TiO 2 via dopingwith transition metal 35,36 and main group elements. 37,38 Whiledopants do increase visible light absorption by introducinglocalized color centers, or in some cases by shifting the valenceand/or conduction band edges, doping also introduces a highdensity of trap states and can decrease hole mobility leading toincreased electron–hole recombination. 39 Other metal oxidessuch as WO 3,40 BiVO 4,41 and Fe 2O342 have also been investigatedas possible visible light absorbing photoanodes, but are generally

Fig. 2 (A) Redox potentials with photosystem II reaction center. (B) Photosystem II reaction complex after light excitation with surroundingprotein structure removed. Reproduced from ref. 15 with permission from The Royal Society of Chemistry.

O

View Online



4/28


limited to wavelengths shorter than 500 nm, exhibit poor holetransport properties, and require large bias voltages. Somerecent work has focused on the development of metal oxideheterostructures as a route to improving the performance of these photoanode materials. 43

Oxynitrides photoanodes offer an alternative strategy forvisible light absorption. Introduction of nitrogen atoms into oxygensites shifts the valence band edge to more negative potentialsthrough hybridization of N 2p and O 2p orbitals. Domen andco-workers have demonstrated high quantum efficiency (IPCE =B 76% at 400 nm) with TaON based photoanodes, though theapplication of an external bias and a co-catalyst were required. 44,45

Other oxynitrides such as LaTiO 2N 46 and SrNbO 2N 47 havesimilarly been studied as photoanodes.

An alternative strategy for PECs is to decouple the lightabsorption and catalytic functions of the electrode. In thesimplest form of this concept, a semiconductor photovoltaicis coupled to oxygen- and hydrogen-evolving catalysts. Anearly demonstration of this approach was photolysis of HBrand HI on a silicon p–n junction coated with aluminum. Thissystem, developed in the 1970’s by Texas Instruments, usedconcentrically doped silicon microspheres. Because of thesmall size of these spheres, inexpensive metallurgical siliconcould be used. 48–50 Bipolar series arrays of TiO 251 and CdSe/CoS 52 photoelectrochemical cells were subsequently demon-strated to drive overall water splitting, although in both casesefficiency was low ( r 1%). The CdSe/CoS-based array wasdesigned to avoid the well-known stability problem of oxygen-evolving photoanodes based on semiconductors with visiblelight bandgaps. In the bipolar CdSe/CoS cell, the chalcogenideportion of the cell was in contact with a stabilizing polysuldesolution and prevented from contacting the aqueous side of the cell. Khaselev and Turner 53 later developed a monolithicsystem with an overall water splitting efficiency of 12.4% bycombining an oxidatively unstable GaAs p–n junction with acathodically protecting p-type GaInP 2 layer connected througha tunnel junction. Licht and co-workers 54 achieved the highestreported water splitting efficiency, 18.3%, by extending thismultijunction approach to include as many as 10 components.More recently, arrays of silicon microwires have been used todrive the hydrogen-evolving half cell of a water splittingsystem, 55,56 although as single-junction devices they lacked thephotovoltage needed for unassisted water splitting. Multijunc-tion wires 57 coupled to electrocatalysts could in principleprovide a sufficient photovoltage to drive the overall reaction.In related work, Nocera and coworkers 58 recently demonstrateda 4.7% efficient water-splitting PEC by coupling a monolithicthree-junction amorphous silicon–germanium photovoltaic cellto a cobalt phosphate oxygen evolving catalyst and aNi–Zn–Mo hydrogen evolving catalyst (Fig. 3). This system isinteresting because it uses only abundant elements and becausethe cathode catalyst, unlike the noble metals used in manywater-splitting systems, is relatively insensitive to impurities inthe water.

An ideal photoelectrochemical water splitting cell wouldcombine the low cost of terrestrially abundant materials withthe high quantum efficiency of a photovoltaic cell. In addition,tunability of the absorption characteristics and the use of multiple absorbers is desirable in order to ‘‘evolve’’ towards

a system that can use the solar spectrum efficiently. Dyesensitized solar cells (DSSCs) are made from inexpensiveTiO 2 anodes sensitized with visible light absorbing dyes.

59

The latter are tunable by molecular design to cover differentparts of the solar spectrum, and tandem arrangements withmultiple dyes 60,61 and dyes combined with semiconductorabsorbers have been demonstrated. 62–66 The quantum efficiencyof DSSCs is near unity for charge injection into the semi-conductor electrode.

When employed as photovoltaics, DSSCs use a redoxshuttle to complete the photoelectrochemical circuit. If thedye is instead coupled to a water oxidation catalyst, water canserve as an electron source for regenerating the oxidized dye. 67

In this arrangement, the photoanode of the dye cell oxidizeswater, and proton reduction to hydrogen occurs at thatcathode. So far such systems employ only one light absorberand an external bias voltage must be applied because thephotovoltage generated by the dye is not sufficient to driveoverall water splitting. In principle, these dye-sensitized photo-anodes could be coupled to complementary photocathodes toeffect water splitting in an eight-photon, four-electron processthat is mimetic of the two coupled photosystems in plantphotosynthesis. For example, Kaschak et al . demonstratedphotoinduced electron transfer from a porphyrin to poly-viologen electron acceptor using a semiconducting nanosheetto mediate the forward charge transfer. 68 Though not exploredin the paper, the potential of the polyviologen is sufficientlynegative to drive hydrogen evolution. In principle thissystem, or one similar, could be grown upon a conductivesupport to prepare a photocathode. A possible dual absor-bing photoelectrochemical cell is sketched in Fig. 4. Thisreview will focus on the recent design and developmentof the dye-sensitized photoanodes of such water splittingsystems.

Fig. 3 Monolithic wireless water splitting device developed by Noceraand co-workers. Migration of anions and cations occurs in the directionshown at the bottom of the gure.

O

View Online



5/28


3. Dye-sensitized photoanodes

Dye-sensitized water splitting cells are comprised of four majorcomponents: visible light-absorbing sensitizer, water oxidationcatalyst (WOC), water reduction catalyst, and semiconductoranode. Electron transfer mediators can be added and choice of pH and buffer must be considered. Here, we will focus on thedesign of WOCs, sensitizers, mediators, and electrodes andreview examples that have been shown to participate in wateroxidation reactions either with sacricial reagents or underelectrochemical/photoelectrochemical conditions. Because themajor bottleneck in the design of efficient and durable watersplitting cells is the catalytic four-electron oxidation of water,we do not discuss here the rapidly developing science of catalysts and sensitizers for water reduction and refer theinterested reader to several recent reviews of that topic 69,70

3.1 Water oxidation catalysts

An ideal WOC must collect four oxidizing equivalents peroxygen molecule generated, facilitate the formation of dioxygen,and be chemically stable; for practical use over a 20–30 yearsystem life, the catalyst should be active for approximately 10 9

cycles or self-repairing. Catalysts are classied as either mole-cular or particle-based. Molecular WOCs, which are oftenstudied in solution as homogeneous catalysts, can be morereadily characterized structurally and are amenable to detailedkinetic/mechanistic studies. These benets of molecularcatalysts come with a price, namely synthetic complexity andoften a lack of stability. Conversely, heterogeneous, nanoparticle-based catalysts are simpler to prepare and are more chemically

robust, but their structural and mechanistic characterization ismore challenging.

The utility of WOCs can be quantied by three parameters:the turnover number (TON), turnover frequency (TOF), andoverpotential at the desired TOF. The TON of a catalystdescribes how many cycles it is able to cycle through beforebecoming inactive. TOF is the number of cycles a catalystundergoes per unit time. Ideally the TOF of the catalystshould substantially exceed the ux of photons in full sunin order to obtain a high quantum yield. Following theanalysis of Frei and Jiao, 71 a lower limit for TOF can becalculated. Integrating the solar photon ux over wavelengthsshorter than 1000 nm gives approximately 4290 mE m 2 s 1 or2600 photons nm 2 s 1.5 Assuming a dual absorber system,eight photons must be absorbed for each molecule of oxygenproduced. At a catalyst coverage of one site per nanometersquared, this gives a minimum TOF of 325 s 1. Complete lightabsorption by dyes, which usually have extinction coefficientsin the range of 10 4 –105 M 1cm 1 (B 10 3 –10 2 cm 2 at acoverage of 1 10 10 mol cm 2) typically requires porouselectrodes with surface areas that are hundreds of times largerthan their geometrical area, relaxing the TOF requirement forthe dye–catalyst dyad to approximately 3 s 1. However, asdetailed below, articial photosynthetic systems generallyrequire very fast electron transfer from the WOC in order tocompete with faster back electron transfer processes, so inpractice much higher TOFs are needed.

The overpotential is the excess energy per unit charge thatmust be added to drive the process as the catalyst cycles. Inpractice this is often measured as the difference between theonset of water oxidation and the formal potential for water

Fig. 4 Scheme of an 8-photon, 4-electron water splitting system.

O

View Online



6/28


oxidation at the pH where the cell is operated. Commonoverpotentials for WOCs are 0.3 to 0.7 V, though WOCs withlower overpotentials exist. 72 The high overpotential of thefour-electron water oxidation reaction is one of the majorreasons why real system efficiencies cannot meet theoreticalefficiency and much work is being devoted to developingcatalysts with lower overpotentials at adequate TOFs.

Rationally designing catalysts with lower overpotentials andTOFs requires understanding the catalytic mechanism andidentifying rate limiting steps. Much of the detailed mechanisticwork on water oxidation has used homogeneous catalysts. Inthese systems two mechanisms of oxygen bond formation havebeen identied: nucleophilic attack of water on a high oxidationstate M–O group and O–O bond formation by two catalyticM–O units. In the rst mechanism, a water molecule attacks abound oxygen atom to form a hydroperoxide intermediate. Thisis the likely mechanism of the OEC in photosystem II. In thesecond mechanism, two adjacent M–O bonds form a metalperoxo intermediate that then breaks down into M–OHand M–OOH groups by addition of water. 73 Studies of TiO 2photocatalysts have identied a mechanism with both features.Water attacks a bridging oxo group to give a surface hydroxyland an oxygen radical, which then couple to form a peroxo bondbetween two adjacent Ti atoms. Subsequent water attack formshydroperoxo and surface hydroxyl groups that then collapse toform dioxygen. 74 A hydroperoxide surface intermediate has alsobeen observed on iridium oxide nanoparticles, 75 suggesting thatthis species is common to heterogeneous WOCs (Fig. 5).

In both natural and articial photosynthetic systems, protoncoupled electron transfer (PCET) has been identied as part of the multi-electron catalysis cycle. In PCET, the movement of electrons and protons is concerted in a process that has a lowerenergy barrier than sequential transfer. This process can beunderstood within the conceptual framework of semi-classicalMarcus theory in the limit of weak coupling between thereactant and product potential surfaces. Solvent uctuationsoccasionally make the energy of the donor and acceptor statesequal, at which point protons and electrons can cross iso-energetically between the reactant and the product potentialsurfaces. Quantum mechanical tunneling of protons can contributeto the kinetics of PCET. 76 In photosystem II, the redox mediatortyrosine in the OEC uses PCET to simultaneously transfer anelectron to P680 + and a proton to a nearby histidine residue. Thisavoids the formation of a charged tyrosine, which is unfavorable inthe low dielectric medium of the protein, and thus proton andelectron transfer occur together in a single, lower energystep. 77 In mutants that lack the histidine proton acceptor,

the quantum yield of water oxidation drops dramatically to zero,illustrating the importance of the PCET mechanism. 78,79 PCEThas also been identied in many articial photosynthetic systems.Because water oxidation catalysts must manage the transfer of four protons and electrons for each molecule of oxygen generated,efficient proton management is a key for a good WOC. 80

Four types of WOCs have so far been integrated into dye-sensitized photoanodes: molecular catalysts, cubanes, poly-oxometalates, and heterogeneous catalysts. As a starting pointfor comparison across different classes of catalysts, we reviewTOF, TON, and overpotential data for representative catalysts.Wherever possible we have presented numbers given by theauthors but in some instances we have calculated the valuesfrom published data. In general, TOF and TON are calculatedassuming every metal center is catalytically active. Overpotentialis taken at the initial onset of catalytic water oxidation current.

3.1.1 Molecular water oxidation catalysts. One attractivefeature of the OEC in PSII is its remarkable level of structuralorganization. Essential amino acid functional groups areprecisely arranged around the CaMn 4 cluster that is the heartof the OEC. A logical starting point for the development of WOCs is a molecular design that nely tunes atomic positionsthrough synthetic chemistry. Many examples of molecularWOCs have been developed, 81–83 of which we will highlightsome notable examples.

Meyer and co-workers reported the rst molecular WOC inthe early 1980s, cis,cis- [Ru(bpy) 2(H 2O)]2(m-O)

4+ also knownas the blue dimer (Fig. 6, # 1 Table 1). 84,85 This catalyst featuresmany of the points common to molecular catalysts. It uses twotransition metal atoms connected by a bridging ligand, andexploits polypyridyl chemistry to tune the redox properties of the catalyst as well as to satisfy ruthenium’s coordinationsphere. The blue dimer has an overpotential of B 470 mV,demonstrating the moderate overpotentials typical of rutheniumdimers. Through its easily cleaved bridging oxo group, it alsoexemplies a major pitfall of molecular catalysts, instability.With this unstable oxo group, a TON of B 13 and TOF of 0.004 s 1 were obtained. 86,87 If a rigid bridging ligand is usedinstead of oxygen the TON can surpass 10 000 (# 5 Table 1). 88

It is important to note that molecular catalysts are oftenstudied not under photochemical or electrochemical conditionsbut using a chemical oxidant, such as Ce( IV), to drive the wateroxidation reaction. Chemical oxidants avoid the complicationof competing forward and back electron transfer pathways aswell as direct water oxidation at electrode surfaces. While thesereagents simplify the kinetics of water oxidation, the concentrationof the oxidant can be signicantly higher than in experiments inwhich the reaction is driven electrochemically or photochemically,leading to articially high TOFs and TONs. Additionally,sacricial reagents can generate highly oxidizing species, insome cases in excess of 3.45 V vs. NHE. 89 These species can lead toundesirable side reactions and complicate the overall kinetics.

In parallel with dinuclear catalysts, single-site catalysts havealso been developed. Single-site catalysts avoid some of thestability issues encountered with binuclear catalysts, thoughtheir ligands are still susceptible to oxidative degradation.Thummel and co-workers developed the rst mononuclear ruthe-nium complexes with high turnover rates as WOCs (Fig. 6). 90

Fig. 5 Reaction scheme for the photochemical formation of oxygenon TiO 2. Reprinted with permission from ref. 74. Copyright 2004American Chemical Society.

O

View Online



7/28


Initially, the ruthenium center is in the +2 oxidation state witha coordination number of 6 and a total electron count of 18.Two oxidizing equivalents increase the oxidation state of ruthenium to +4, leaving the highly electrophilic metal opento attack by water. Two more oxidizing equivalents give a Ru( VI)doubly bound to an oxygen atom. This complex undergoes waternucleophilic attack to produce a peroxo intermediate that then losesa proton to oxygen. 91 An interesting feature of the complex that isreminiscent of the OEC in PSII is hydrogen bonding between abound water molecule and a free nitrogen atom in the napthyridinering, which helps stabilize the complex (Scheme 1).

Single site ruthenium catalysts 92,92 developed by the Meyergroup have likewise exploited polypyridyl chemistry to operateat lower overpotentials than dinuclear catalysts (280–377 mVvs. NHE) and with oxidant-limited TONs (# 9–10 Table 1).

A single site ruthenium( II ) complex was suspended in Naonand used as a WOC in a dye sensitized PEC (# 11 Table 1). 93 Arelatively low TON of 16 and TOF of 27 h 1 were measuredfor this system at neutral pH. Recently, a single site rutheniumcatalyst with a TOF of greater than 300 s 1 was reported. 94

This TOF was achieved at very high molar concentrations of Ce( IV) and is probably not representative of the catalystperformance under photochemical conditions.

In 2008, a new class of molecular iridium WOCs was introduced.The rst of these WOCs, a family of cyclometalated iridiumcomplexes, was synthesized by Bernhard and coworkers (# 12Table 1). 95 These catalysts are synthetically simple, easilytunable, highly soluble in water, and chemically robust dueto strong carbon–iridium bonding. An extension of this approachuses electron-donating Cp* ligands to stabilize high oxidation

Fig. 6 (Left) Blue dimer developed by Meyer and co-workers; (Right) single site ruthenium WOC developed by Thummel. Reprinted withpermission from ref. 90. Copyright 2005 American Chemical Society.

Table 1 Comparison of selected molecular water oxidation catalysts

Catalyst Oxidant TOF (s 1) TON Overpotential (mV vs. NHE) Ref.

1 cis,cis- [Ru(bpy) 2(H 2O)]2(m-O)4+ Ce( IV) 0.004 13 474 84, 85

2 in,in-[(Ru(tpy)(H 2O)]2(m-bpp) 3+, Ce( IV) 0.86 512 993 trans,trans -[Ru 2(L1)(4-CH 3O-py) 4Cl]

3+ Ce( IV) 3.8 10 5 689 1004 [Ru 2(L2)(4-CH 3-py) 6]

1+ Ce( IV) 0.24 1690 424 1015 [Ru 2(L3)(4-CH 3-py) 4Cl]

1+ Ce( IV) 1.2 10 400 330 886 [Ru(tpy-PO 3H 2)(H 2O) 2]2O

4+ 1.25–1.5 V vs. Ag/AgCl 1.8 414 1027 [Ru 2(OH)(3,6-

tBu 2qui) 2(btpyan)]2+ 1.7 V vs. Ag/AgCl 0.232 33 500 405 103

8 trans- [Ru(L4)(4-CH 3-py) 2(H 2O)]2+ Ce( IV) 0.0028 260 90

9 [Ru(tpy)(bpm)(OH 2)]2+ Ce( IV) 0.019 7.5 377 92

10 [Ru(Mebimpy)(bpy)(OH 2)]2+ Ce( IV) 0.0067 7.5 280 10411 Ru(6,6 0-dcbpy)(pic) 2 RuP 0.0069 15 330 10512 [Ir(ppy) 2(OH 2)2]

+ Ce( IV) 0.004 2490 185 9513 Cp*Ir(ppy)Cl Ce( IV) 0.167 >1500 585 9614 Cp*Ir(3 0-ppy)Cl ZnCCPP 0.001 0.066 185 9815 [Mn(dpa) 2] Oxone 8.3 10 6 0.6 10616 [Mn(tpy)] 3+ Oxone 0.00014 >50 10617 [Fe(OTf) 2(mcp)]

2+ Ce( IV)/NaIO 4 0.23 >1050 107

1. bpy = 2,2 0-bipyridine; 2. tpy = 2,2 0:60,20 0-terpyridine, bpp = bis(2-pyridyl)-3,5-pyrazolate; 3. L1 = 3,6-bis[6 0-(10 0,80 0-napthyrid-2 0 0-yl)pyrid-2 0-yl]-pyridazine, 4-CH 3O-py = 4-methoxypyridine; 4. L2 = 3,6-bis-(6 0-carboxypyrid-2 0-yl)-pyridazine, 4-CH 3-py = 4-methylpyridine, TOF calculatedusing initial rate of oxygen evolution; 5. L3 = 1,4-bis(6 0-COOH-pyrid-2 0-yl)phthalazine; 6. tpy-PO 3H 2 = 4 0-phosphonato-2,2 0:60 0,20 0-terpyridine; 7.3,6- tBu2qui = 3,6-di- tert- butyl-1,2-semiquinone, btpyan = 1,8-bis(2,2 0:60,20 0-terpyridyl)anthracene, TOF represents a lower limit; 8. L 4 =4-tert -Butyl-2,6-di([1 0,80]-naphthyrid-2 0-yl)pyridine; 9. bpm = 2,2 0-bipyrimidine, TOF calculated from initial O 2 evolution rate; 10. Mebimpy =2,6-bis(1-methylbenimidazol-2-yl)pyridine, TOF calculated from initial O 2 evolution rate; 11 . 6,60-dcbpy = 6,6 0-dicarboxylic acid-2,2 0-bipyridine,pic = 4-picoline, RuP = [Ru(bpy) 2(4,4 0-diphosphonic acid-2,2 0-bipyridine)]

3+ photoelectrochemically generated; 12. ppy = 2-phenylpyridine;13. Cp* = pentamethylcyclopentadienyl, TON is number of turnovers in 5.5 hours; 14. 30-ppy = 3 0-carboxy-2-phenylpyridine, ZnCCPP =Zinc 5-(4-carbomethoxyphenyl)-15-(4-carboxyphenyl)-10,20-bis(pentauorophenyl) photoelectrochemically generated; 15. dpa = dipicolinate; 17.OTf = triuoromethanesulfonate, mcp = ( N ,N 0-dimethyl- N ,N 0-bis(2-pyridylmethyl)-cyclohexane-1,2-diamine).

O

View Online



8/28


state Ir centers. 96,97 Of particular relevance to dye sensitizedphotoanodes, Cp*–Ir complexes can be functionalized withcarboxylic acids and other ligating groups for attachment tometal oxide surfaces. 98

Some success with molecular manganese WOCs has beenachieved. Limburg, Brudvig, and Crabtree 106 prepared Mn

complexes with dipicolinate and 2,20:6

0,2

0 0-terpyridine ligands

that generated oxygen when exposed to oxone. It wasproposed that these complexes initially formed an oxo-linkeddimer before decomposing to form permanganate (Fig. 7). Asubsequent study 108 examined the behavior of the terpyridinedimer in sodium hypochlorite and found that a Mn VQ Ospecies was the key intermediate in the formation of oxygen.Stoichiometric oxygen evolution from a Mn III porphyrin dimerwas observed by Naruta and co-workers. 109,110 Individually, theMn III porphyrins were not able to generate oxygen, but whenheld in a xed geometry by a 1,2-substituted phenyl ring,oxygen evolution was observed. They also identied aMn VQ O intermediate and postulated that the O–O bond

was formed either by nucleophilic attack on this species orvia a coupling of two oxo groups. Recently, a tetra-sulfonicacid functionalized Mn III porphyrin was incorporated into apoly(terthiophene) lm and demonstrated catalytic behaviorwith an apparently low overpotential at pH = 7 ( B 90 mV). 111

The authors postulated that catalytic behavior comes from thefraction of porphyrin molecules in the lm that are in close

enough proximity to mimic the behavior of the Naruta dimer.Detailed reviews on molecular Mn WOCs 112,113 and on theMn-based 15 OEC in PSII have recently been published.

Tetraamido iron-centered molecular WOCs were recentlyintroduced by Ellis et al .114 Using Ce( IV) as a chemicaloxidant, they observed biphasic behavior with an initial rapid

release of oxygen, followed by steady oxygen evolution overthe course of hours. Building upon those initial ndings, Fillolet al . studied a series of iron coordination complexes foroxygen evolution activity. 107 When two adjacent, labile siteswere present in the catalyst, oxygen evolution was observed.With trans sites or a single site, there was no activity. They alsoproposed a catalytic cycle that involves water nucleophilicattack on an Fe( V) oxo group to give a peroxo intermediatethat releases oxygen upon subsequent oxidation.

A molecular cobalt catalyst, [Co II (qpy)(OH 2)]2+ (qpy =2,2 0:60,2 0 0:60 0,2 0 0-quaterpyridine), was shown by Leung et al .to undergo chemical and photochemical water oxidation atpH > 8 in the presence of [Ru(bpy) 3]

2+ /S 2O82 . The complex

exhibited more than 330 turnovers and using [Ru(bpy) 3]3+

asa chemical oxidant, a TOF of 4 s 1 . The onset of wateroxidation occurred at an overpotential of B 250 mV. 115

One of the most advantageous aspects of molecular WOCsis the ability to add specic functionality to the catalyst. Wadaet al. 103 developed a ruthenium dimer bridged with an anthra-cence backbone functionalized with two terpyridines to bind

Scheme 1 Mechanism for water oxidation by single site ruthenium WOC. Reprinted with permission from ref. 91. Copyright 2008 AmericanChemical Society.

Fig. 7 (Left) Manganese terpyridine dimer introduced by Limburg et al ., (Right) Naruta’s porphyrin dimer catalyst.

O

View Online



9/28


the ruthenium atoms (Fig. 8, # 7 Table 1). Notably, this catalystused 3,6-di- tert -butyl-1,2-benzoquinone a ligand on the rutheniumto introduce quinone functionality. As the SbF 6 salt, this catalystcould be deposited onto an ITO substrate and catalyze 30000+turnovers of water electrolysis at the mildly acidic pH of 4 with amoderate overpotential ( B 400 mV). The quinones play an activerole in the catalytic cycle, switching between quinone and semi-quinone oxidation states, providing a pathway for proton coupledelectron transfer, and allowing the ruthenium centers to stay inthe low Ru( II)/Ru( III) oxidation states. Interestingly, the mono-ruthenium analog does not show activity nor does an analogin which the benzoquinone ligand has been replaced with2,20-bipyridine. This demonstrates that the ruthenium centers actin concert with the quinone ligands to split water.

Before leaving the topic of molecular WOCs, it is importantto note that in the study of these catalysts, care must be takento identify the active catalytic species. Decomposition of themolecular species to form an active heterogenous catalyst is acommon issue with molecular WOCs. 116–119 A detailed discussionof the techniques used to identify the active form of the

catalyst is beyond the scope of this review, but the interestedreader is directed to a detailed review by Widegren andFinke 120 as well as a recent article by Schley et al .121

3.1.2 Cubanes. Over the last decade a class of homogeneouscatalysts based on the cuboidal structure of the active CaMn 4core in photosystem has been developed. Generally cubanes havethe core structure [M 4O4]6+ or 4+ with six to eight bidentateligands helping to hold it together. Dismukes and coworkers rstdemonstrated gas phase photochemical oxygen evolution fromMn 4O4(PPh 2)6 (1, Fig. 9) under UV illumination.

122,123 It wasproposed that a cationic species, [Mn 4O4(PPh 2)5]

+ (2), is initiallyformed. The exibility of 2 allows the two manganese atoms

on the open face to move apart due to repulsion. As the distancebetween the two manganese atoms increases, the pair of oxygensbridging between the two are brought into contact. This allows aperoxo species ( 3) to develop as a bond between the two oxygenatoms is formed. Moving from species 2 to 3 is the rate-limitingstep in the formation of dioxygen because of the strainintroduced in exing the structure. Following the formationof 3 a superoxo species ( 4) forms, which releases O 2 to give theopen ‘‘buttery’’ ( 5) conformation. Via the uptake of two newwater molecules and proton coupled electron transfer, 5 isregenerated to the starting state.

Using Naon as a support, a PEC with Mn 4O4(( p-OMe– C 6H 4)PO 2)6 as the WOC was developed and showed sustainedwater oxidation photocurrent over a period of 10 hours underUV illumination. Based on the total photocurrent, a TON of greater than 1000 turnovers was reported, with TOFs from0.014 s 1 to 0.075 s 1 (Table 2). 125 This approach was extendedto use visible light by the addition of ruthenium sensitizer 126 andby coupling to two DSSCs in series. 127 With the Naonsupport, water oxidation by the Mn cubane was possible inthe condensed solution phase. The authors postulated that theNaon may have provided a stabilizing, protecting inuence onthe oxidized cubane. This conclusion was recently challenged byHocking et al. 128 They reported that in the Naon membrane,the tetramanganese cubane dissociates into Mn( II) compounds,which are then reoxidized to form a mixed( III /IV) disorderedheterogeneous phase similar to birnessite.

A novel cobalt cubane, Co 4O4(OCMe) 4(py) 4 (py = pyridine),was recently reported by McCool et al. 129 They observed a TONgreater than40 with a TOF of0.02 s 1 at a pH of 7. Working withthe same compound, La Ganga et al. 130 measured a quantum yield

Fig. 8 [Ru 2(OH)(3,6-tBu 2qui) 2(btpyan)](SbF 6)2 where 3,6-

tBu 2qui =3,6-di- tert- butyl-1,2-semiquinone and btpyan = 1,8-bis(2,2 0:60,20 0-terpyridyl)anthracene.

Fig. 9 Mechanism of photochemical O 2 formation and release ( l =350 nm) in the gas phase. Reproduced from ref. 124 with permissionfrom The Royal Society of Chemistry.

O

View Online



10/28


of 30%, though they noted that pH and catalyst concentrationplayed a major role in determining the quantum yield. A recentreview detailing the function of the cubane structure in enzymatic,homogeneous, and heterogeneous catalysts is available. 124

3.1.3 Polyoxometalates. Good WOCs must have oxidative,hydrolytic, and thermal stability; which are the major issues formolecular WOCs. Molecular WOCs typically fail on oxidativestability due to the presence of carbon–hydrogen bonds in theligands. Polyoxometalates (POM) are entirely inorganic and soavoid the problem of carbon-containing bonds. POMs are formedby condensing small oxometalate clusters around templatinganions such as SO 42 , SiO 44 , o r PO 43 . Most often theoxometalate clusters are vanadates, molybdates, and tung-states. Molybdate and tungstate POMs can incorporate up tohundreds of octahedrally coordinated metal centers whereasvanadate POMs are much smaller, between 4 and 30 vanadiumcenters, and are structurally more diverse. 131

Recently much effort been devoted to the development of POMsas WOCs. In 2008, [Ru 4(m-O)4(m-OH) 2(H2O)4(g-SiW 10O36)2]

10

(Ru4SiPOM) was simultaneously reported by two groups as theCs10 salt132 and Rb 8K 2 salt. 133 Ru4SiPOM is based around atetraruthenium core that is prepared by in situ decomposition of Ru 2OCl10

4 to generate [Ru 4O6(H2O)n ]4+ (Fig. 10).

Catalytic activity for water splitting was initially testedusing Ce( IV) as a chemical oxidant. The TON was limited bythe amount of oxidant added and over 500 turnovers weredemonstrated with no apparent loss of catalytic activity andan O 2 yield of 90%. A maximum TOF of >0.125 s

1 wasobserved. Photochemical experiments using [Ru(bpy) 3]

2+ andS2O82 resulted in persulfate-limited TONs up to 350 andinitial an TOF of 0.08 s 1 .134 The phosphorous-containinganalog [Ru 4(m-O)5(m-OH) (H 2O) 4(g-PW 10 O36 )2]

9 (Ru4PPOM)was prepared and the catalytic activity investigated. WhileRu4PPOM proved capable of oxidizing water, the rate of photodriven water oxidation was approximately 20% lowerthan with Ru4SiPOM. 135 In an effort to integrate Ru4SiPOMinto a PEC, Ru4SiPOM was attached to a TiO 2 nanocrystallinelm via ruthenium( II) tris(4,4 0-dicarboxcylic-2,2 0-bipyridine). 136

This assembly was investigated spectroscopically and showedthat electron transfer between Ru4SiPOM and the rutheniumdye was on the order of a few microseconds and competitivewith back electron transfer from the TiO 2 electrode. An electro-chemical oxygen-evolving electrode prepared from Ru4SiPOMattached to multiwalled carbon nanotubes by dendrimers wasdemonstrated and operated at modest overpotentials ( Z =0.35 V). 137 Single site ruthenium POMs [Ru(H 2O)SiW 11 O39 ]5

and [Ru(H 2O)GeW 11 O39 ]5 demonstrate catalytic water oxida-tion when treated with Ce( IV).138

A breakthrough in using POMs for water oxidation came withthe development of [Co 4(H 2O)2(PW 9O34)2]

10 (Co4PPOM) 139,140

and more recently [Co 4(m-OH)(H 2O) 3(SiW 19 O70 )2]11 (Co4Si-

POM), 141 [Co 2Mo 10 O 38 H 4]6 , and [CoMo 6O24 H 6]3 .142 TheseWOCs have the advantage of containing only earth abundantelements, which makes them especially attractive compared tonoble metal-containing catalysts. The proposed active WOC inCo4PPOM and possibly in all cobalt polyoxometalates hasrecently been called into question by Stracke and Finke. 119 Underelectrochemical conditions, they determined that Co4PPOMwas actually a pre-catalyst for the formation of a hetero-geneous CoO x lm. This lm was examined with a scanningelectron microscope and found to have a different morphologythan an authentic Co4PPOM lm. Furthermore, as thesolution was aged for several hours in a pH 8 sodiumphosphate buffer, the oxidation current increased indicatingthe formation of a new species different from Co4PPOM.Finally, the authors noted that cobalt dissociation constantshad been measured for other cobalt-containing POMs andthat the electrochemical activity of an aged Co4PPOM couldbe completely accounted for by authentic Co( II) at a levelequivalent to the level of leached cobalt. Stracke and Finke donote that Co4PPOM may actually function as a WOC withuse of achemical or photochemical oxidant, but their results suggest moredetailed study into the true nature of WOC by cobalt-containingPOMs under a variety of conditions is warranted.

The question of heterogeneous versus homogeneous catalysis inpolyoxometalates is an interesting one. An iridium( III) containingPOM, IrCl4PPOM, demonstrates water oxidation at a ratetwo orders of magnitude faster than IrO x nH 2O, a known andhighly efficient WOC, before hydrolytically breaking down toform IrO x nH 2O.

143 It is an interesting observation that TOFsof Co4PPOM and IrCl4PPOM are noticeably lower whencompared to other WOC POMs (Table 3). The TOF for

Table 2 Comparison of cubane catalysts


18 [Mn 4O 4L6]+ 1 V vs. Ag/AgCl 0.075–0.005 >1000 380 125

19 Co 4O4(OAc) 4(py) 4 [Ru(bpy) 3]2+ /

S2O 82

0.02 40 332 129

18. L = di-( p-methyoxyphenyl)-phosphine; 19 . OAc = acetate, py = pyridine, TON given after 60 min.

Fig. 10 Synthesis of Ru4SiPOM by metalation of SiW 10 by [Ru 4O6-(H 2O) n ]

4+ . Reprinted with permission from ref. 132. Copyright 2008American Chemical Society.

O

View Online



11/28


Co4PPOM is also comparable with other examples of hetero-geneous cobalt WOCs discussed in the next section (Table 4). Itshould be noted that while [Ru(H 2O)SiW 11 O39 ]

5 also exhibitsa low TOF, the value of 0.003 s 1 represents a lower limit sinceoxygen evolution data was not presented in the paper.

3.1.4 Heterogeneous metal oxides. Compared to molecularcatalysts, heterogeneous WOCs have the major advantage of being synthetically simple to prepare. Typically heterogenousWOCs either refer to colloidal suspensions of nanoparticles orelectrochemically deposited catalytic lms.

The most successful WOCs to date have been based on noblemetal oxides, with ruthenium and iridium oxides receiving the mostattention. Both oxides have a long history as anode materials inwater and chloride electrolyzers. 144–146 These catalysts can accesshigh oxidization states, stabilizing the highly electrophilic inter-mediates in the water oxidation process. RuO 2 received a great dealof early attention as a heterogeneous WOC for photocatalysis.Those studies 147–149 identied colloidal RuO 2 as a WOC that couldbe driven with a chemical oxidant, Ce( IV) or [Ru(bpy) 3]3+ , orphotochemically with [Ru(bpy)

3]2+ /S

2O

8

2 or [Co(NH3)5Cl]2+ .

The major attraction for RuO 2 is a low overpotential and

high TOF. Unfortunately, RuO 2 also corrodes under oxidizingconditions, which limits its utility. In recent years, work onRuO 2 as a WOC has slowed, though a recent study of rutileRuO 2 has suggested that it may be a more active WOC thanpreviously thought. 150 Although it corrodes under anodicconditions, under cathodic conditions it shows better chemicalstability and can function as a hydrogen evolution catalyst. 151

Iridium oxide (IrO x nH 2O) is unfortunately prepared fromthe least abundant stable element in the periodic table, but itis so far unparalleled as a WOC across a wide range of pHvalues. Initially used to catalyze water oxidation by Ce( IV) and[Ru(bpy) 3]

3+ by Kiwi and Gra ẗzel, 152 crystalline IrO 2 andamorphous colloidal IrO x nH 2O were soon found to be highlyactive WOCs. 153,154 IrO x nH 2O exhibits a low overpotential,B 200–300 mV, for water oxidation and is active over a widepH range. 155 TOFs for surface atoms in colloidal suspensionsof IrO x nH 2O are as high as 40 s

1 .156

Iridium oxide has played a major role in the development of dye-sensitized photoanodes for water oxidation. Our groupdeveloped chemistry for capping iridium oxide with dicarboxylateligands. 156,157 Using this chemistry, colloidal iridium oxide hasbeen covalently coupled to a variety of ruthenium poly-

Table 3 Comparison of polyoxometalates active for water oxidation


20 [Ru 4(m-O) 4(m-OH) 2(H 2O) 4(g-SiW 10 O36 )2]10 Ce( IV) 0.131 488 246 133, 13221 [Ru 4(m-O) 4(m-OH) 2(H 2O) 4(g-SiW 10 O36 )2]

10 [Ru(bpy) 3]2+ /S 2O8

2 0.08 35 13422 [Ru 4(m-O) 5(m-OH)(H 2O) 4(g-PW 10 O36 )2]

9 [Ru(bpy) 3]2+ /S 2O8

2 0.13 120 248 13523 [Co 4(H 2O) 2(PW 9O34 )2]

10 [Ru(bpy) 3]2+ /S 2O8

2 0.0013 224 441 14024 [Co 4(m-OH)(H 2O) 3(SiW 19 O 70 )2]

11 [Ru(bpy) 3]2+ /S 2O8

2 0.1 80 14125 [(IrCl 4)KP 2W 20 O72 )]

14 [Ru(bpy) 3]3+ 0.0292 5.25 215 143

26 [Ru(H 2O)SiW 11 O39 ]5 Ce( IV) 0.003 20 188 138

27 [Co 2Mo 10 O 38 H 4]6 [Ru(bpy) 3]

2+ /S 2O82 0.171 154 350 142

28 [CoMo 6O 24 H 6]3 [Ru(bpy) 3]

2+ /S 2O82 0.119 107 420 142

21. 20 mM pH 7.2 sodium phosphate buffer, turnover number persulfate limited; 22. pH 5.8 NaSiF 6 buffer; 23. 80 mM pH 8 sodium borate buffer;24. 25 mM sodium borate pH 9 buffer; 25. 20 mM sodium phosphate pH 7.2 buffer, TOF represents a lower limit TOF based on completereduction of Ru( III ) within 3 minutes; 26. 0.1 M HNO 3, TOF calculated based on number of turnovers after 20 minutes; 27. and 28. overpotentialrepresents upper limit.

Table 4 Comparison of heterogeneous WOCs


29 RuO 2 [Ru(bpy) 3]2+ /[Co(NH 3)5Cl]

2+ 0.052 68 310 14830 RuO 2 (5 nm) [Ru(bpy) 3]

2+ /S2O82 0.0045 19.2 310 149

31 RuO 2 (10 nm) [Ru(bpy) 3]2+ /S2O8

2 0.089 2.7 310 14932 rutile RuO 2 (6 2 nm) 1.48 V vs. RHE Z 0.000069 280 15033 IrO x nH 2O [Ru(bpy) 3]

2+ /S2O82 0.0004 3 310 153

34 citrate-IrO x nH 2O (20 nm) [Ru(bpy) 3]2+ /S2O8

2 0.05 80 330 15635 succinate-IrO x nH 2O [Ru(bpy) 3](PF 6)2/S2O 82 0.049 28 330 15736 IrO x nH 2O 1.4 V vs. Ag/AgCl 4.71 220 17637 IrO x nH 2O 1.3 V vs. Ag/AgCl 0.64 330 17738 Co 3O4 [Ru(bpy) 3]

2+ /S2O82 0.035 325 158

39 Co 3O4 Z 0.0025 350 15840 Co 3O4 Z 0.020 295 15841 Co 3O4 Z 0.0008 414 15842 Co 3O4 Z 0.006 235 15843 SBA-15/Co 3O4 (4%) [Ru(bpy) 3]

2+ /S2O82 0.01 350 158

44 Co 3O4 (5.9 1.1 nm) 0.534 V vs. Ag/AgCl 0.0187 328 15945 CoPi 1.29 V vs. NHE Z 0.0007 410 15846 MnO 2 Z 0.013 440 15847 Mn 2O3 [Ru(bpy) 3]

2+ /S2O82 0.055 325 158

48 CaMn 1.6IV Mn 0.4

III O4.5 (OH) 0.5 zH 2O Ce( IV) 0.002 16949 l -MnO 2 (B 20 nm) [Ru(bpy) 3]2+ /S2O82 3 10 5 370 175

O

View Online



12/28


(pyridyl) dyes that contain malonate or succinate linkers.Functionalizing the ruthenium dye with both malonate andphosphonate groups allows for adsorption of the dye– IrO xnH 2O dyad to a nanocrystalline TiO 2 electrode. In principle,direct coupling of the sensitizer to the WOC should facilitaterapid electron transfer and enhance efficiency. Actual quantumyields are low due to two factors. First, the covalently attachedIrO x nH 2O colloids can rapidly quench the excited state of the dye.Second, multiple sensitizer molecules were bound to each colloidalparticle with the result that some of the bound sensitizers wereunable to inject into the TiO 2 electrode.

Despite the success of noble metal oxides as WOCs, it wouldbe desirable to develop alternative catalysts based on terrestriallyabundant materials. To this end a great deal of recent attentionhas been paid to cobalt oxide structures as WOCs. Early workon Co 3O4 demonstrated TOFs between 0.0008 to 0.035 s

1 atoverpotentials between 235 and 414 mV. 71 Jiao and Frei grewclusters of Co 3O4 in the pores of mesoporous silica and observedTOFs of 0.5 s 1 nm 2 at an overpotential of 350 mV. 158

A subsequent study established that the smaller the Co 3O4particle, the lower the overpotential and the higher the activityof the catalyst. 159 The dramatic difference between the activityof unsupported and silica-supported Co 3O 4 may arise from abuffering effect of the support near neutral pH.

Nocera and Kanan reported the anodic deposition of acatalytically active, amorphous cobalt phosphate (CoPi) lmon ITO from a solution of Co( II) in neutral potassiumphosphate buffer. 161 Subsequent work demonstrated that useof a proton accepting buffer was critical in depositing active,stable lms 162 and that the lms were self-healing through acycle of dissolution and re-deposition. 163 Gerken et al . extendedthis work by depositing from a uoride buffer, allowing theelectrodeposited catalyst to function at mildly acidic pHs. 160

They also reported detailed mechanistic studies on water oxida-tion from pH 0 to 14 (Fig. 11). Above pH 3.5, heterogeneousCoO x is the active electrocatalyst. Below pH 3.5, homogeneouscatalysis dominates and oxygen is formed via a hydrogenperoxide intermediate. Most recently, photoelectrochemicalactivity was observed with a heterogeneous cobalt alumino-phosphate (CoAPO5) photocatalyst doped into a Naon-coatedelectrode. 164 A photocurrent was observed at bias voltages greaterthan 0.8 V vs. Ag/AgCl.

Keeping with the idea of using abundant rst row transitionelements as WOCs, heterogeneous Mn WOCs have been studied.MnO x was investigated in the 1970’s as an anode material forelectrochemical water oxidation. 165 A later study identied colloidalMnO 2 as active for water oxidation using [Ru(bpy) 3]3+ as achemical oxidant. 166 Harriman also identied Mn 2O3 as one of

Fig. 11 Scheme for the electrodeposition of CoO x lms and water oxidation in the pH range of 0–3.5 (left) and 3.5–14 (right). Reprinted withpermission from ref. 160. Copyright 2011 American Chemical Society.

O

View Online



13/28


the more active WOCs within a series of heterogenous metaloxides. 153 In the past few years there has been a resurgenceof interest in manganese oxides as WOCs. Using a strategypreviously applied to Co 3O4, Jiao and Frei deposited manganeseoxide into the pores of a mesoporous silica support and observedefficient photochemical oxygen evolution. 167 Jiao later examineddifferent polymorphs of nanostructured MnO 2 and observedTOFs on the order of 10 5 s 1 per Mn with little differencebetween crystal structure and morphology. 168 A biomimeticcalcium manganese oxide, CaMn 1.6 IV Mn 0.4 III O4.5 (OH) 0.5 zH 2O,demonstrated water oxidation activity under chemical andphotochemical oxidation. 169,170 Though an amorphous material,XAS analysis showed a layered oxide similar to birnessite withdisorted cuboidal Mn 3CaO 4 and Mn 4O4 units throughoutthe layers. 171 As noted above, manganese cubanes form aheterogeneous oxide when supported by naon. 128 Nanoscale(o 50 nm) particles of calcium manganese oxide showed arelatively high TOF ( B 0.002 s 1) using Ce( IV) as a chemicaloxidant. 172 Replacement of the Ca( II ) with Zn( II ) and Al( III )led to efficient oxygen evolution with Ce( IV).173 Inspired bystudies of electrodeposited cobalt oxide lms, Zaharieva et al .electrodeposited a manganese oxide by voltage cycling. 174 At apotential of 1.35 V vs. NHE at neutral pH, they obtained aTOF comparable to Nocera’s cobalt phosphate lms (0.01 s 1

vs. 0.017 s 1). Treatment of nanocrystalline LiMn 2O 4 withnitric acid delithiates the material while leaving the l -MnO 2spinel structure intact. 175 The delithiated spinel structure iscuboidal with open sites for water oxidation at the lithiumvacancies. Oxygen evolution was limited to surface sites whichcould be accessed by photogenerated [Ru(bpy) 3]

3+ making theTOF strongly dependent on particle size.

3.2 Dyes

Efficiently converting light energy into chemical energy requiresthat light be absorbed to create excited states that have sufficientoxidizing or reducing power to drive at least one of the half-cellreactions of water splitting. At a dye-sensitized photoanode,this role falls to a dye molecule (also called a sensitizer)adsorbed at the solid–liquid interface. The ideal dye shouldabsorb a signicant fraction of visible spectrum, convert allabsorbed photons to electron–hole pairs, bind persistently tothe surface, and have the appropriate redox potential to drivethe catalytic oxidation of water at a WOC. Most molecules failto meet these stringent requirements. New photosensitizers continueto be developed, although much of the current emphasis is on thehydrogen-evolving half-cell reaction. 178–180 To date, most studiesof dye-sensitized photoanodes have employed [Ru(bpy) 3]

2+

derivatives or high potential porphyrins, which have a longhistory in studies of light-driven electron transfer reactionsand in conventional dye-sensitized solar cells. 181–184

As noted above, functional water-splitting systems requiremore energy than the 1.23 V stored in the products (Fig. 1).Most of the dyes that have sufficiently negative excited stateredox potentials to transfer an electron to TiO 2 and aresufficiently oxidizing to accept electrons from water absorbin the blue part of the visible spectrum. In [Ru(bpy) 3]

2+

derivatives, the major visible absorption is a metal-to-ligandcharge transfer (MLCT) band with an absorbance maximum

around 450–470 nm (454 nm = 2.7 eV). Absorption of aphoton initially produces a singlet state, which undergoesintersystem crossing within 300 fs to give a predominantlytriplet MLCT state. 185 This rapid decay to the triplet state,which is several hundred mV lower in energy than the singletstate, is the rst energy loss within the cell. 186 As the return of the excited electron to the ground state is spin-forbidden, thetriplet MLCT state lifetime is relatively long ( B 600 ns) for[Ru(bpy) 3]2+ and many of its derivatives. The electrochemicalpotential of the triplet MLCT state is typically reported as theoxidation potential of the photoexcited dye, although there isevidence that ultrafast hot electron injection into TiO 2 occurswith ruthenium polypyridyl sensitizers. 187,188 In the lowesttriplet excited state of the dye, the metal center is oxidizedand one of the bipyridine ligands is reduced. 189

For a dye-sensitized photoanode to operate efficiently, theexcited electron must be transferred before the molecule canrelax back to the ground state. This is accomplished by one of two means, transferring an electron to a metal oxide electrodeor to a sacricial reagent. Model systems typically use sacricialoxidants, commonly sodium persulfate (Na 2S2O8) or Co(NH 3)5Cl,as these reagents irreversibly oxidize the dye and simplify thekinetic analysis of subsequent steps in water oxidation catalysisby eliminating back electron transfer. When the photoanode isoperated without sacricial reagents ( i.e. , as a water-splittingcell), the excited dye injects an electron into the conductionband of a metal oxide semiconductor, usually titanium dioxide.The electron then percolates through the semiconductor lmsuntil it nds its way into an external circuit before being usedto reduce protons to hydrogen at a counter-electrode. 190 If theoxidized dye is not rapidly reduced by the WOC, the electronsinjected into semiconductor can be transferred back to theoxidized dye, regenerating the reduced form of the dye in itsground state. Rapid back electron transfer is the dominantkinetic pathway in dye-sensitized photoanodes and is theprimary reason for their low quantum yield. Dye stability isalso a signicant problem. Ruthenium based sensitizers areunstable in the 3+ oxidation state and susceptible to nucleo-philic attack by water and buffer anions on a timescale of tensof seconds. 191 Fast electron transfer between the WOC and theoxidized sensitizer is thus critical to both the efficiency andstability of the water-splitting dye cell.

The chemical attachment of the dye to the high surface areaoxide semiconductor is an important and subtle issue. Carboxylicacids are the most frequently used linking groups in dye-sensitized solar cells (DSSCs), and provide strong electroniccoupling for ultra-fast electron injection from the dye excitedstate. However, photoanodes for water splitting necessarilyoperate in solutions that contain water, and are often usedwith aqueous buffer solutions, making hydrolysis of thecarboxylic acid–metal bond a signicant problem. Under theaqueous conditions, phosphonate linkers provide a morerobust linkage to the oxide surface. The performance of phosphonate-functionalized dyes is generally comparable orbetter than that of carboxylate analogues under non-aqueousconditions because of reduced dye desorption. 192–194 Recentwork by Hanson et al. 195,196 has shows that the photodesorptionof phosphonate-bound dyes in water is accelerated in thepresence of oxygen, possibly because of the generation of

O

View Online



14/28


superoxide ions by back electron transfer from TiO 2. However,dye desorption can be hindered with by having multipleligands with phosphonic acid functionalities. Unfortunately,the electron injection efficiency of dye molecules decreases withincreasing number of phosphonic acid groups, 196 suggestingthat further work on optimizing dye attachment and electroninjection efficiency is needed.

3.2.1 Ruthenium polypyridyl sensitizers. Ruthenium( II )tris(bipyridine) gures prominently in early studies of sensitizedhydrogen 197,198 and oxygen 199–201 production from water. As amodel sensitizer, [Ru(bpy) 3]

2+ has many attractive features. Itabsorbs strongly below 500 nm ( e450nm = 14400 M

1 cm 1)and has a sufficiently long-lived excited state lifetime ( B 600 ns)for a diffusional encounter with an electron donor or acceptorin solution. 202 Perhaps most importantly, the internal quantumyield for formation of the MLCT excited state is nearly unity,meaning that virtually all excited sensitizer molecules canparticipate in electron transfer reactions. 203 Thermodynami-cally, the redox potential of the excited state is sufficientlynegative ( 0.66 V vs. NHE) to form hydrogen from waterbelow pH 10. The Ru( III )/Ru( II ) reduction potential is suffi-ciently positive at 1.26 V vs. NHE to drive the water oxidationprocess, even in acidic solutions (Fig. 14). 202 However if anoverpotential of 300 mV is needed to drive electron transferfrom the WOC at a rate that can compete with back electrontransfer, then [Ru(bpy) 3]

2+ cannot be used below pH B 4.5 asa sensitizer for water splitting.

Because both its oxidized and reduced forms are relativelystable in water, the photocatalytic reactions of [Ru(bpy) 3]2+

have been well studied using sacricial reagents. Of particularinterest in the context of water oxidation is the reactionof [Ru(bpy) 3]2+ and a sacricial oxidant, most commonlypersulfate, S

2O

8

2

. The generally accepted mechanism for

oxidation of [Ru(bpy) 3]2+ by persulfate. 204,205 begins withthe initial excitation of [Ru(bpy) 3]

2+ (Scheme 2, eqn (1)).From a precursor complex, S 2O8

2 –[Ru(bpy) 3]2+ *, an electron

is rapidly transferred with a bimolecular rate constant on theorder 10 8 M 1 s 1.206 If the persulfate ion is in excess atconcentrations on the order of millimolar or higher, the reactionbetween [Ru(bpy) 3]

2+ * and persulfate is essentially diffusioncontrolled, with nearly all [Ru(bpy) 3]

2+ * oxidized as in eqn (3).At lower persulfate concentration, some excited sensitizer mole-cules decay via eqn (2). Upon receiving an electron, S 2O82

irreversibly decomposes into one molecule of sulfate, SO 42 ,and one sulfate radical anion, SO 4 . Sulfate radical anion is apowerful oxidizing agent, >3.45 V vs. NHE, and rapidly reactswith any ground state sensitizer molecules it encounters (eqn (4))to generate a second molecule of [Ru(bpy) 3]3+ .89

As noted above, derivatives of [Ru(bpy) 3]2+ functionalizedwith carboxylic acid or phosphonic acid groups can be directlycoupled to a metal oxide semiconductor electrode and fromtheir excited state can inject an electron directly into theelectrode. The kinetic theory of electron transfer from anexcited sensitizer to a nanocrystalline electrode follows fromsemi-classical Marcus theory and has been reviewed in detail. 207,208

Using the Franck–Condon principle that nuclear coordinatesdo not change during electron transfer, a distribution of donorenergy states for the excited sensitizer molecule can be

calculated. The width of this distribution depends on the freeenergy of the excited state as well as the reorganization energyrequired for the change in oxidation state of the sensitizer. Forreorganization energies between 0.3 and 0.8 eV the width of distribution is approximately equal to the reorganizationenergy, and for larger reorganization energies, the width of

the distribution is less. If the width of the donor distribution isequal to the reorganization energy then the maximum rate of electron transfer occurs when the free energy of an electron inthe semiconductor conduction band ( Gc) is lower than or equalto the free energy of the ground state ( Go ) plus the difference infree energy between the ground and excited states ( G00 ) minustwice the reorganization energy ( l ) (eqn (2)).

Gc r Go + G00 2l (2)

Eqn (2) describes the situation shown on the left in Fig. 12where the entirety of the distribution of donor states lies abovethe conduction band edge potential. When considering realsensitizers, this means that the maximum rate of electron transferwill occur when the quantity, (Go + G00 2l ) nF 1, is morenegative than the conduction band edge potential of theelectrode. For [Ru(bpy) 3]

2+ , the potential of the excited state,(Go + G00 ) nF

1, is 0.84 V vs. NHE and the reorganizationenergy for the excited state is approximately 0.35 eV. 209 Usingthese values, the maximum rate of electron injection into TiO 2will occur when the potential of the conduction band edge ismore positive than 0.14 V vs. NHE. The conduction bandedge potential of anatase TiO 2 is 0 V vs. NHE at pH = 0 butdecreases by 55 mV per pH unit. Thus it is expected that at lowpH there is a signicant overlap between unoccupied states inthe conduction band and the distribution of donor states in theexcited sensitizer. As the pH increases, the overlap decreases asthe potential of the conduction edge band becomes morenegative than 0.14 V vs. NHE and the rate of electron transferfalls off (Fig. 12). In the case of TiO 2, this model is insufficientto explain the experimental behavior. As we shall discuss in thephotoanode section, the high density of conduction band statesallows for injection from excited states higher in energy than thelowest MLCT excited state.

One attractive quality of ruthenium tris(bipyridyl) derivativesand related complexes is the possibility of including multiplefunctionalities into the three l igands. In addition to phosphonicand carboxylic acid attachment functionalities, 192–194 bipyridineligands can be modied to act as linkage units to catalysts.

Scheme 2

O

View Online



15/28


Hoertz et al . functionalized 2,2 0-bipyridine with dicarboxylicacid units in the 4,4 0 positions. 157 These dicarboxylic acidgroups were used as capping agents for the controlled hydro-lysis of IrCl 62 to form IrO x nH 2O with sensitizers directlycoupled to the WOC particle. In an extension of that work,Youngblood et al. prepared a sensitizer containing a malonicacid functionality, which was used to cap an IrO x nH 2Onanoparticle, as well as a 4,4 0-diphosphonic-2,2 0-bipyridineligand for attachment to the nanocrystalline electrode (Fig. 13). 67

Beyond attachment functionality, ligand modication canbe used to tune the redox and spectroscopic properties of thesensitizer. 210,211 The lowest excited state of a rutheniumtris(bipyridyl) sensitizer, Ru II L3 , is best described in terms of a ligand-localized state, Ru III L2(L ). Because the electronlocalizes on one bipyridine ligand, the excited state of thesensitizer mirrors the electrochemical and spectroscopicproperties of the free reduced ligand. Upon excitation, theexcited electron hops rapidly ( o 15 ps) 212 between ligands andthermally equilibrates to reside primarily on the most readilyreduced, that is the most electron-decient, ligand. The excitedstate oxidation potential can be made more positive by addingan electron-decient ligand or more negative by addingelectron-rich ligands. 210 By using a combination of s -donorand p-acceptor effects, the ground state oxidation potential

can also be shifted. Broadly considered, a more electron-decient ligand donates less electron density via s -donationgiving a higher nuclear charge on the metal center and as aconsequence, stabilizing the d p orbitals. Back-bondingbetween the d p and p* orbitals of the ligands results in furtherstabilization of the d p orbitals. The net effect is that complexeswith ligands that are more readily reduced exhibit morepositive oxidation potentials. 189,213 In the context of water-splitting, electron injection into a semiconductor electrode canbe accelerated by making the ligand bound to the electrode themost electron-withdrawing. Along the same lines, an increasedreduction potential to drive a WOC can be generated by usingmore easily reduced ligands. Of course, this strategy also leadsto more positive excited state potentials, which can affect theefficiency of electron injection.

Multinuclear dendrimer sensitizers present an interestingpathway towards enhanced light harvesting in water-splittingdye cells. By coupling multiple photoactive metal centers,dendritic sensitizers show both extended and increased absorp-tion in the visible region relative to mononuclear sensitizers.Through careful selection of the core and branching metal ionsand ligands, a directed energy cascade reminiscent of photo-synthesis can be designed. 214,215 Recently the tetrarutheniumsensitizer, [Ru{( m-dpp)Ru(bpy) 2}3](PF 6)8 (dpp = -2,3 0-bis(2 0-pyridyl)pyrazine) was shown to drive water oxidation undersacricial conditions. 216,217 Notably, this complex drove wateroxidation under illumination with light >700 nm using IrO xnH 2O as the WOC. Extended light absorption comes with aprice, namely that the excited state redox potential is roughly0.4 V vs. NHE (Fig. 14) limiting the use of this complexto sacricial systems. In principle, such a sensitizer couldfunction at a water-splitting photoanode using a semiconductorsuch as SnO 2 or In 2O3, which have more positive conductionband edge potentials than TiO

2.218

3.2.2 Porphyrin sensitizers. Poprhyrins and related macro-cyclic chromophores have been studied as sensitizers for watersplitting reactions for nearly three decades. 222,223 Porphyrins arestrongly absorbing with intense Soret bands ( e B 105 M 1 cm 1)in the blue part of the visible spectrum and Q-bands ( e B 104)in the green to red. 224 In recent years attention has returnedto high potential porphyrins as sensitizers for both DSSCsand water-splitting photoanodes. A malonate substituted

Fig. 12 Diagram of donor distribution functions, W don (G), and thedensity of states within the semiconductor electrode, D (E ) at (a) lowpH and (b) high pH. Reprinted with permission from ref. 183. Copy-right 2005 American Chemical Society.

Fig. 13 Structure of [Ru(bpy) 3]2+ (Left), [Ru(dcbpy)(bpy) 2]

2+ (Middle), and [Ru(dpbpy)] 2+ (Right).

O

View Online



16/28


5,10,15,20-tetrakis(4-bromomethylphenyl)porphyrin was usedto cap IrO x nH 2O nanoparticles. 225 The porphyrin-capped particlesdemonstrated strong electronic coupling and sustained catalyticactivity when poised at a potential of 1.4 V vs. Ag/AgCl. Mooreet al . demonstrated visible light water-splitting via the co-depositionof a high potential porphyrin zinc 5-(4-carbomethoxyphenyl)-15-(4-carboxyphenyl)-10,20-bis-(pentauorophenyl)porphyrin (ZnBPFP)and a Cp* iridium WOC onto a TiO 2 electrode. 98 A subsequentstudy considered photoanodes prepared from ZnBFPP as wellas the free base version and palladium analogue. 221 Whileno water oxidation catalyst was present, each porphyrindemonstrated sufficient oxidizing potential for water oxida-tion, however, only ZnBFPP had an excited state sufficientlynegative to inject an electron into the TiO 2 anode (Fig. 15).

3.2.3 Sensitizer–catalyst dyads. Direct coupling of the sensitizerand WOC should lead to faster electron transfer rates and inprinciple to higher overall efficiency. As noted above, Youngbloodet al .67 coupled a ruthenium sensitizer directly to an IrO x xH 2Onanoparticle. Unfortunately, rapid quenching of the excitedstate byIrO x xH

2O limited any gains from faster electrontransfer. Since the

publication of that work several groups have adopted a coupledsensitizer–catalyst dyad strategy to address this problem.

Kaveevivitchai and co-workers coupled a ruthenium poly-pyridyl sensitizer to a single-site ruthenium WOC (Fig. 16). 226

Under photochemical conditions with S 2O 82 as sacricialelectron acceptor they observed a TOF of B 0.005 s 1 and a

TON of 134 over six hours. While low, the TON for the uncoupledsystem at higher concentrations was only 6. Furthermore, the[Ru(bpy) 2(pynap)]

2+ (pynap = 2-(pyrid-2 0-yl)-1,8-napthyridine)sensitizer upon which the dyad is based was shown to be anineffective sensitizer.

A similar system built around a single site [RuL(pic) 2] (L =6,6 0-dicarboxy-2,2 0-bipyridine; pic = 4-picoline) complex wasintroduced by Li et al .227 This system coupled a modiedruthenium trisbipyridine sensitizer through each picoline ligand.As above, under photochemical conditions, a signicantly higherTON was observed for the coupled system than for theanalogous uncoupled system, 38 vs. 8 turnovers.

Ashford and coworkers introduced a general approach forcoupling chromophores and molecular WOCs. 228 Via an amidelinkage formed at elevated temperatures, they coupled an aminefunctionalized ruthenium trisbipyridyl sensitizer to a single site,ruthenium terpyridyl catalyst. Although they did not reportphotochemical water oxidation, electrochemical and chemicalwater oxidation was found with oxygen yields of B 70% relativeto initial Ce( IV) concentrations. Transient spectroscopy showedrapid intrassembly energy transfer. This followed work by thesame group on coupled a sensitizer–catalyst dyad absorbed on aTiO 2 electrode. 229 Rapid injection into TiO 2 and electrontransfer from the catalyst to the sensitizer were observed. Backelectron transfer to the oxidized catalyst was observed on themicrosecond time scale. Injection yields into TiO 2 were low,possibly due to unfavorable energetics. Steady state photolysis

Fig. 14 Plot of ground state, singlet, and triplet redox potentials versus the conduction and valence bands of TiO 2. The difference in wateroxidation potential at pH 0 and 7 is shown. 1. [Ru(bpy) 3]2+ (bpy = 2,2 0-bipyridine); 186,196,211 2. [Ru(dcbpy)(bpy) 2]2+ (dcbpy = 4,4 0-dicarboxy-2,2 0-bipyridine); 219 3. [Ru(dpbpy) 3]

2+ (dpbpy = 4,4 0-diphosphonic-2,2 0-bipyridine); 196 4. Ru[{( m-dpp)Ru(bpy) 2}3]8+ (dpp = -2,3 0-bis(2 0-pyr-

idyl)pyrazine) 220 5. ZnBPFP (BPFP = 5-(4-carbomethoxyphenyl)-15-(4-carboxyphenyl)-10,20-bis-(pentauorophenyl)porphyrin). 221 wSinglet– triplet state splitting estimated via the method of Vlcek et al .211

O

View Online



17/28


data was not reported, but electrochemical water oxidationcurrent was observed.

Ultrafast ash photolysis studies of perylene derivativescoupled to a Cp* molecular Ir WOC revealed ultrafast electrontransfer ( o 10 ps) from the WOC to the excited perlyene. 230

Unfortunately backelectron transfer was also rapid and occurredwithin several nanoseconds. While photocatalytic activity wasnot observed due to the short charge separation lifetime, catalyticactivity was retained and demonstrated electrochemically.

3.3 Electron transfer mediators

The photochemical systems we have described thus far generallyinvolve a direct electron transfer between an oxidized sensitizermolecule and the WOC. As noted above, direct electron transferbetween P680 + and the OEC does not occur in photosystem II,instead the tyrosine–histidine mediator between them is anessential component of the electron transport chain. Themediator couples the fast electron transfer necessary to regenerateP680 + with the slower electron transfer from the OEC. Byanalogy, incorporating a mediator between an oxidized, unstablesensitizer and WOC is a promising strategy for use in a water-splitting DSSC.

Use of mediators in articial photosynthesis has beenpresaged for some time. Molecular triads used in electrontransfer studies represent the early examples of mediator-likesystems. The initial example featured a carotenoporphyrin– quinone triad which underwent rapid ( o 100 ps) initialcharge separation to yield a long-lived charge separated state

(t1/2 = 170 ns).231 In the context of water splitting, triads have

been designed to closely mimic some of the structural andfunctional features of photosystem II. Johansson et al. pre-pared triads containing a [Ru(bpy) 3]2+ core, a tyrosine ethylester, and a variety of electron acceptors. 232 They observed acharge separation yield of B 10% with a fraction of the chargeseparated states persisting for microseconds. Extending thetriad concept to a tyrosine-modied ruthenium trisbipyridylcomplex adsorbed to the nanocrystalline TiO 2 lm, Pan et al.noted ultrafast electron injection into the TiO 2 and B 90%internal quantum yield for transfer of an electron from thetyrosine. 233 A multiexponential charge recombination processwas observed with a component on the millisecond time scale.

Though it was not used to drive photochemical watersplitting, a molecular triad containing a high potential porphyrinbearing two pentauorophenyl groups demonstrated a chargeseparated state that was thermodynamically capable of wateroxidation. 234 This triad featured a benzimidazole–phenol (BiP)electron donor, which mimics the tyrosine–histidine mediatorfound in the OEC. Unlike most phenols, which undergoirreversible oxidation, hydrogen bonding between the histidinein BiP and the phenolic proton allows for the proton to beshuttled upon oxidation of the phenol. 235 An ionizable protoncontrols the electrochemical potential of the mediator, withthe potential 1.19 V vs. SCE in acidic conditions and 0.21 V vs.SCE under basic conditions (Fig. 17). 236

Recently Zhao et al. coupled the BiP mediator to an IrO xnH 2O colloid and co-absorbed a ruthenium polypyridinesensitizer onto a TiO 2 electrode (Fig. 18). 237 With the mediator-capped catalyst, substantially higher and sustained photocurrentwas measured compared to mediator free catalyst. It is importantto note that in this system there was no direct linkage between thesensitizer and the mediator, which were separately adsorbed ontothe high surface area TiO

2 electrode. The kinetics of bleaching

recovery of the sensitizer, studied by nanosecond ash photolysis,were multi-exponential as expected for a kinetically complexsystem with a range of distances between the electron donorand acceptor. Using a Pt collector electrode, they veried thatthe Faradaic efficiency (percentage of charge passed convertedto oxygen) was close to unity, i.e., that the observed photo-current corresponded to water oxidation.

As an alternative to using tyrosine derived mediators, theMeyer group has focused on the use of ruthenium polypyridylsas electron transfer mediators. The rate limiting step in Ce( IV)driven catalytic water oxidation by the blue dimer is oxidationof a Ru IVQ O to Ru V or oxidation of terminal peroxido group

Fig. 15 Zinc 5-(4-carbomethoxyphenyl)-15-(4-carboxyphenyl)-10,20-bis-(pentauorophenyl)porphyrin (ZnBPFP).

Fig. 16 Photochemical water oxidation by a ruthenium bipyridyl sensitizer (red) coupled to a single site ruthenium WOC (black) through a 2,6-di-(10,8 0-napthyrid-2 0-yl)pyrazine (blue). Reprinted with permission from ref. 226. Copyright 2012 American Chemical Society.

O

View Online



18/28


Ru IV –OOH to give oxygen. In either case, the these steps arerate-limiting due to slow Ce( IV) electron transfer. Rutheniumpolypyridyls, for example [Ru(bpy) 3]

2+ or Ru(bpy) 2(bpm)]2+ ,

can be added to function as redox mediators, undergoingrapid oxidation by Ce( IV) and increasing the rate of wateroxidation by 30-fold in the case of or Ru(bpy) 2(bpm)]

2+ .238

To extend this work to electrochemical systems, they adsorbed[Ru(bmpbpy) 2(bpy)] 2+ (bmbpy = 4,4 0-bis-(methyl)phospho-nato-2,2 0-bipyridine) onto an ITO electrode and left the bluedimer in solution. Without surface modication, a 3e /3H +

oxidation wave corresponding to oxidation of Ru IV –OH andRu III –OH 2 to Ru

VQ O is not present. When the surfaceis modied by [Ru(bmpbpy) 2(bpy)] 2+ , this wave appears.Sustained electrocatalysis at 1.46 V vs. NHE resulted in17 500 turnovers of [Ru(bmpbpy) 2(bpy)] 2+ and 1.25 turnoversof the blue dimer at a Faradaic efficiency of 95%. Higherturnovers of the blue dimer were limited by precipitation of thecatalyst onto the electrode surface. 239 The issue of catalyst pre-cipitation was circumvented by preparing a single site rutheniumcatalyst directly coupled to the phosphonate-modied electron

transfer mediator through a 2,2 0-bipyrimidine. When attachedto an ITO electrode held at 1.8 V vs. NHE, more than 28 000turnovers could be observed with a TOF of 0.6 s 1 and no lossof catalytic activity (Fig. 19). On a nanocrystalline TiO 2electrode, a Faradaic efficiency of 98% for oxygen productionwas obtained. 240

3.4 Photoanode materials

The anode in a dye-sensitized solar cell serves two majorfunctions, to facilitate light absorption and effectively collectelectrons. Efficient light absorption with ruthenium-basedsensitizers requires electrodes with surface areas hundreds of times greater than

Documents

Design and Development of Photoanodes for Water-splitting Dye-sensitized