Solar energy conversion using first row d-block metal coordination compound sensitizers and redox mediators

The use of renewable energy is essential for the future of the Earth, and solar photons are the ultimate source of energy to satisfy the ever-increasing global energy demands. Photoconversion using dye-sensitized solar cells (DSCs) is becoming an established technology to contribute to the sustainable energy market, and among state-of-the art DSCs are those which rely on ruthenium(ii) sensitizers and the triiodide/iodide (I3−/I−) redox mediator. Ruthenium is a critical raw material, and in this review, we focus on the use of coordination complexes of the more abundant first row d-block metals, in particular copper, iron and zinc, as dyes in DSCs. A major challenge in these DSCs is an enhancement of their photoconversion efficiencies (PCEs) which currently lag significantly behind those containing ruthenium-based dyes. The redox mediator in a DSC is responsible for regenerating the ground state of the dye. Although the I3−/I− couple has become an established redox shuttle, it has disadvantages: its redox potential limits the values of the open-circuit voltage (VOC) in the DSC and its use creates a corrosive chemical environment within the DSC which impacts upon the long-term stability of the cells. First row d-block metal coordination compounds, especially those containing cobalt, and copper, have come to the fore in the development of alternative redox mediators and we detail the progress in this field over the last decade, with particular attention to Cu2+/Cu+ redox mediators which, when coupled with appropriate dyes, have achieved VOC values in excess of 1000 mV. We also draw attention to aspects of the recyclability of DSCs.


Introduction
Why solar energy?
The United Nations Member States adopted the 2030 Agenda for Sustainable Development in 2015. This recognizes seventeen sustainable development goals (SDGs), of which SDG7 has the aim to "ensure access to affordable, reliable, sustainable and modern energy for all" by 2030. 1 Renewable energy incorporates biomass, wind, hydroelectric, solar and geothermal technologies. Because of their unlimited and cost-free supply, solar photons are an ideal source of energy to satisfy the everincreasing global demands. Moreover, in contrast to fossil fuels, solar energy poses no direct threat to the environment.
The solar spectrum ( Fig. 1) peaks in the visible region, and the latter accounts for ca. 40% of the total radiation; 55% falls in the infrared (IR) region, and the remaining 5% in the ultraviolet (UV). When light falls on an n-type semiconductor and the photons possess energies equal to or greater than the band gap, electrons are excited from the valence to the conduction band of the semiconductor. Photoenergy conversion then follows to transform light into electrical current. Naturally, semiconductors such as silicon are optimal for such applications, but have the disadvantage that the material is not optically transparent and the majority of the photoelectric effects occur at the surface. This prompted investigations of optically transparent materials which, of course, do not absorb visible light. For electron excitation to occur, wide-band gap semiconductors such as TiO 2 (band gap ¼ 3.2 eV for anatase) must absorb photons with energies in the UV region. From Fig. 1, it is clear that pristine wide-band gap semiconductors are not appropriate for efficient photoenergy conversion. Although a reactive titanium-terminated anatase surface phase with a band gap of <2 eV has been discovered, 2 the most convenient method of utilizing longer wavelength radiation is to functionalize the surface of the semiconductor with a material that absorbs in the visible region. Such materials are termed sensitizers or dyes and critically, the ground state (S) of the sensitizer must lie below the conduction band of the semiconductor, and the excited state (S*) above the conduction band (Fig. 2a).
The dye-sensitized solar cell: a general overview The Grätzel n-type dye-sensitized solar cell (DSC) was developed in the early 1990s, and the use of sintered nanoparticles of TiO 2 to produce an enormous surface area while maintaining a small device is crucial to the design. [3][4][5][6][7][8] The principle of the working device is shown schematically in Fig. 2a with detrimental recombination processes shown in Fig. 2b; Fig. 2c shows a typical laboratory device. The conducting glass must be transparent and is typically colourless glass coated with uorine-doped tin oxide (FTO). The processes at the photoanode (Fig. 2a) are the sequential photoexcitation of the dye, electron injection into the semiconductor, and electron transfer from the reduced form of the redox shuttle to the oxidized form of the dye. Aer dye-excitation and electron injection, the dye is formally in an oxidized state. The redox mediator (also referred to as a redox shuttle or couple) is responsible for transferring electrons from the counter electrode through the cell to regenerate the ground state of the dye. A note at this point about terminology: it is important to distinguish between the redox mediator and the electrolytethe electrolyte comprises the redox mediator and additives in a solvent. The three essential processes mentioned above (dye photoexcitation, electron injection and dye regeneration) compete with non-benecial electron transfers (Fig. 2b): decay of the excited state dye back to the ground state (i.e. no net electron injection), recombination of the injected electron with the oxidized dye (again, no net electron injection), and recombination of the injected electron with the oxidized form of the redox shuttle (once again, no electron injection). Minimizing recombination processes (back-reactions) at the interface between the semiconductor, dye and redox mediator is essential, and common ways to address this are through the use of co-adsorbents and additives. A popular additive is 4-tert-butylpyridine (TBP) which is added to the electrolyte in the DSC and leads to a raising of the conduction band (E cond , Fig. 2a) with a concomitant increase in the open-circuit voltage (V OC , Fig. 2a). The addition to the dye of co-adsorbents such as chenodeoxycholic acid (cheno, Scheme 1) decreases the aggregation of dye molecules, and enhances electron injection. 7,9 Computational studies play an important role in the development of structure-property relationships for molecular sensitizers, and interactions between dye molecules and between dye and coadsorbent species. [10][11][12] In an n-type DSC, the counter electrode functions as a catalyst (the Pt coating shown in Fig. 2a) for regeneration of the redox mediator and as a contact in the electrical circuit. In this type of DSC, light harvesting is governed by the dye adsorbed on the n-type semiconductor. In a double-junction or tandem DSC, both the photoanode and photocathode can be functionalized with dyes, and the solar energy conversion in the cell could, theoretically, reach ca. 40%. 13 However, progress in the development of p-type DSCs is hampered by the lack of effective combinations of wide-band gap p-type semiconductors and sensitizers. The last decade has seen an explosion of interest in the use of quantum dot sensitized solar cells and perovskite solar cells. These areas are out of the scope of the present review, and readers are directed to the following articles and references therein. [14][15][16][17][18][19][20][21][22][23][24][25] In this review, we focus on n-type DSCs and, critically, on the need to develop DSCs incorporating sustainable materials. 26 Ideally, all components in a DSC should utilize sustainable materials. Use of TiO 2 for the photoanode ts this criterion, with titanium having a natural abundance in the Earth's crust of ca. 5600 ppm. 27 The major use of TiO 2 is as a white pigment, and the United States Geological Survey (USGS) reported in 2021 that world resources of titanium minerals exceed two billion tons and that there is currently no recycling of TiO 2 . 28 To date, many of the best-performing DSCs have incorporated ruthenium(II) dyes and iodine-based redox mediators. Since ruthenium has an extremely low abundance in the Earth's crust (ca. 0.001 ppm), 27 dependence on this metal for large-scale DSC production is not sustainable. Perhaps less well recognized is the low crustal abundance of iodine (ca. 0.14 ppm), 27 and, as discussed later, electrolytes containing the I 3 À /I À redox couple possess intrinsic corrosive properties. The highest DSC photoconversion efficiency (PCE, h) values 29-35 are achieved by optimizing not only the molecular design and performance of the sensitizer. [36][37][38][39][40][41] Tuning the composition of the electrolyte is critical, [42][43][44] as are optimizing both the fabrication of the photoanode, and the materials and fabrication of the counter electrode. 45 State-of-the-art dyes for n-type DSCs are typically ruthenium(II) coordination compounds, 46-52 zinc(II) porphyrinato or phthalocyanato complexes, 46,48,[53][54][55][56][57] and metal-free organic dyes. 47,48,58 Natural pigments have also been thoroughly investigated, but their photoconversion efficiencies are limited. 59,60 We end this introduction with several general comments concerning the need for consistency in reporting data. Critically, DSCs should be fully masked to prevent the overestimation of their performance. 61,62 Wherever possible, we have only made direct comparisons between DSCs fabricated under the same or similar conditions. A second problem in overviewing the DSC literature is knowing the reproducibility of cell performances. Not all researchers report data for multiple devices. This appertains, not only to J SC and V OC values, but also to electrochemical impedance spectroscopic (EIS) data. We recently explored the reproducibility of EIS data for DSCs sensitized with N719 and SQ2 (Scheme 2). Whereas data for DSCs with N719 were reproducible, SQ2 proved to be an instructive example of a dye for which the EIS parameters can be rather variable within one set of DSCs with identical components and fabricated by the same person and in the same fashion. 63 A general note about light intensity is necessary. For most routine evaluations of the performances of DSCs, devices are illuminated under a light intensity of 1 sun ¼ 1000 W m À2 ¼ 100 mW cm À2 . † However, photoconversion efficiencies increase when lower light intensities are used. Such low or diffuse light sources are relevant to indoor applications of DSCs, and it is important to note that DSCs are reliable devices even under diffuse lighting conditions. 64,65 A nal comment to the introduction is that, although this review strives to cover the literature as broadly as possible, it is not fully comprehensive. We have chosen not to include studies in which the metal complexes used as sensitizers were not adequately characterized, or in which insufficient information was provided about cell fabrication.
An unfriendly chemical environment within the DSC: enter the first row metals The I 3 À /I À redox mediator works: why change it?
For efficient photoconversion efficiency, a critical factor for a redox mediator is that it can regenerate the ground state of the dye on a faster timescale than recombination events which negate electron injection. The I 3 À /I À couple fulls this requirement and has become an established component of most DSCs. 66 A disadvantage, however, is that using an I 3 À /I À redox mediator limits values of the open-circuit voltage, V OC (Fig. 2) to 700-800 mV. 29,67 Furthermore, its use creates a corrosive chemical environment within a dye-sensitized solar cell, limiting DSC stability. 66,68 Thus, the past decade has seen the development of alternative and less corrosive redox mediators having more positive reduction potentials than I 3 À in order to increase values of V OC . [69][70][71][72] The most notable are those based on Co 3+ /Co 2+ and Cu 2+ /Cu + couples. In this review, we focus mainly on the use of rst row d-block metal-ion redox mediators with dyes containing rst row d-block metals, and the discussion in this section on the use of Co 3+ /Co 2+ and Cu 2+ / Cu + redox mediators with other dyes is limited to introductory Scheme 2 Structures of the dyes N719 and SQ2.
comments and selected highlights, as well as reviews to lead the reader into the relevant literature.  2+ , long alkyl chains were introduced into the cobalt complex, but it proves benecial to incorporate them into the dye structure rather than the cobalt redox mediator. On the other hand, Mozer and coworkers have shown that the electron lifetime increases considerably when both the dye and the Co 3+ / Co 2+ shuttle contain alkyl chains. 86 A dramatic improvement in PCE to 11.9% was achieved by combining the [Co(bpy) 3 ] 3+ /[Co(bpy) 3 ] 2+ shuttle with the donorp-bridge-acceptor zinc(II) porphyrin dye YD2-o-C8 (Scheme 4), and co-sensitization with Y123 (Scheme 4) led to a further enhancement to 12.3%. With the related dye GY50 (Scheme 4), a record PCE of 12.75% was attained. Key to these successes are the high values of V OC (965 mV for a DSC with YD2-o-C8, and 885 mV with GY50), 29,67,87 and Fig. 3 illustrates the effect on E redox (dened in Fig. 2a  /I À . The DSC performances are also dependent upon the thickness of the TiO 2 layer. 88 The addition of electron-donating TPAA (TPAA ¼ tris(4-methoxyphenyl)amine) to the electrolyte is benecial. TPAA (like (2,2,6,6-tetramethylpiperidin-1-yl)oxyl, TEMPO 89 ) acts as an intermediate redox species, increasing the rate of dye regeneration. Electrons are transferred from TPAA to the oxidized dye in an extremely fast process (100-1000 ps), and are then transferred from [Co(bpy) 3 ] 2+ to the oxidized form of TPAA (TPAAc + ). DSCs with the organic dye LEG4 Scheme 3 Structures of bpy, phen, bpm, dbbip, bpy-pz, 4,4 0 -Me 2 bpy, 4,4 0 -t Bu 2 bpy and 4,4 0 -di-tert-butyl-2,2 0 -bipyrimidine (4,4 0 -t Bu 2 bpm).  3 Schematic illustration of the relative E redox levels (in red) for two representative Co 3+ /Co 2+ redox mediators with respect to the I 3 conventional Ru(II) dyes due to dominant recombination processes. Intermolecular interactions between the ruthenium(II) sensitizer and cobalt(III)/(II) species must be limited, and the use of a coadsorbent such as cheno, 93 and/or a shi in design of the ruthenium dye were required. [94][95][96] We provide selected examples here, and otherwise direct the reader to reviews that are focused on this topic. 71 3 ] 2+ redox shuttle gave V OC ¼ 735 mV and J SC ¼ 6.5 mA cm À2 and overall h ¼ 3.6%. Upon going to the more sterically demanding dye TT-230 (Scheme 5), a higher V OC was attained (774 mV) which could be boosted to 804 mV with the addition of the coadsorbent cheno. However, this was at the expense of J SC (3.3 and 3.0 mA cm À2 , without and with cheno). 98 Thiocyanatefree ruthenium(II) dyes are a promising route forward for reducing recombination and enhancing compatibility with cobalt-based redox mediators, 99,100 and cyclometallated ruthenium(II) sensitizers have also been investigated. 101 106 but little further progress was made until a report from Bai et al. in 2011. 107 This was followed by highly promising results from Li et al., 108 Magni et al., 109 Freitag et al. 110 and Saygili et al. 111 Homoleptic bis(diimine)copper(I) complexes such as [Cu(bpy) 2 ] + and [Cu(phen) 2 ] + are tetrahedral, while the corresponding copper(II) compounds are tetragonal. Flattening of the coordination sphere upon oxidation means that the Cu + to Cu 2+ potential is shied to more positive values when substituents are introduced into the 6,6 0 -positions of bpy or the 2, 9- /I À redox mediator achieved higher J SC (13.74 mA cm À2 ) but signicantly lower V OC (714 mV), leading to a lower overall h value of 6.5%. 107 These results were pivotal in delineating the use of copper(II)/(I) redox mediators, but at the same time, Bai et al. also commented upon the very low electron-transfer rates at the counter-electrode interface and the need for careful choice of counter-electrode materials compatible with copper(II)/(I) redox couples. 107  In an investigation of the effects of the Lewis bases TBP, 2,6bis(tert-butyl)pyridine, 4-methoxypyridine and 4-(5-nonyl)pyridine, Hagfeldt and coworkers concluded that the optimization of the pyridine base used in DSC electrolytes containing a copperbased redox shuttle depended upon a balance of basicity and coordination capacity. 113 2 ] + . Additionally, the performance of DSCs over a 46 day period suffered from a signicant decrease in the llfactor which has its origins in reduced charge transfer at the counter electrode and slow mass transport associated with the sterically demanding [Cu(Me 2 phen) 2 (TBP)(NCMe) x ] 2+ cations. 117 The pros and cons of using TBP as an additive continue to be debated. Recently, Fürer et al. presented a detailed investigation that conrms the critical benets of adding strong Lewis bases such as TBP or 1-methylbenzimidazole (NMBI) to the electrolyte, but importantly, these results distinguish between the formation of 5-coordinate complexes [Cu(Me 2 phen) 2 (LB)] 2+ (LB ¼ Lewis base) and ligand exchange to give [Cu(LB) 4 ] 2+ . The latter is exemplied with the redox mediator [Cu(Ph 2 phen) 2 ] 2+ /    4 ] 2+ has a detrimental effect on the regeneration of the reduced form of the redox mediator at the counter electrode (Scheme 7) and, therefore, limits the current output of the DSC. 119 With the aim of sterically protecting the copper centre from attack by a Lewis base and, at the same time, increasing the stabilities of the Cu(I) and Cu(II) species, Sun 123 An alternative approach to overcome the detrimental effects of TBP is to develop TBP-free electrolytes. One interesting direction has been to design double-stranded helical dicopper complexes with a redox process based upon equilibria (1)-(4); the relative importance of each equilibrium depends upon the ligand, L. With L ¼ dbpy (Scheme 8), the highest PCE achieved for DSCs sensitized with Y123 (Scheme 4) was ca. 8%, suggesting that the use of these dinuclear species is a promising way forward. 126 were also fabricated. In keeping with the more positive redox potentials of the copper-containing redox mediators, DSCs with the latter out-performed those with the cobalt-based couples. Table 3 2 ] + , and these DSCs represent state-of-the-art combinations of ruthenium(II) dyes and copper-based redox mediators. Factors contributing to the performances include relatively long electron lifetimes, slow recombination processes and rapid dye regeneration. However, the relatively low ff values in Table  3 are noteworthy. 138 Other first row M n+ /M m+ redox mediators While Co 3+ /Co 2+ and Cu 2+ /Cu + redox couples have been investigated in detail as alternatives to I 3 À /I À , the rst row of the dblock offers a number of other redox-active metals among which V, Mn, Fe and Ni have received some attention.
While vanadium-based redox mediators have gained minimal attention, the results that are available in the literature demonstrate promise. However, to the best of our knowledge, the long-term stability of these electrolyte components remains untested.

Mn 4+ /Mn 3+ and Mn 3+ /Mn 2+
The range of oxidation states offered by manganese makes it an attractive target for use in redox mediators. In addition, it is abundant in the Earth's crust (ca. 950 ppm) 27 3 ] redox mediator with both ruthenium(II) and organic dyes, the electron lifetimes of these DSCs were shorter than those for analogous cells containing I 3 À /I À and [Co(bpy) 3 Fig. 5 has also been screened for use in redox mediators. The structure of the [Mn(bdmpza) 2 ] + cation is shown in Fig. 5a. Although electrochemical properties of the complexes appeared promising, the solubilities of the Mn(II) compounds were low in polar solvents. 144 We return to the iron complexes in the next section.
Before closing this section of manganese-based redox mediators, we note that [Mn(HBpz 3

Fe 3+ /Fe 2+
With its high natural abundance (ca. 41 000 ppm of the Earth's crust) 27 and redox active properties, iron is the dream metal upon which to base DSC redox mediators in addition to DSC sensitizers (see later). An overview of Fe 3+ /Fe 2+ redox couples in DSCs was given by Pashaei et al. in 2015. 71 The ferrocenium/ ferrocene (Cp 2 Fe + /Cp 2 Fe, E ¼ +0.62 V vs. NHE) couple is a standard reference redox couple in non-aqueous solvents, 146 but in a DSC, it was found to suffer from rapid recombination of electrons from the semiconductor. 147 One approach to suppressing this pathway is to passivate the TiO 2 surface. 148 For example, surface treatment with MeSiCl 3 produces a blocking layer of poly(methylsiloxane) and leads to an improvement, but this is offset by slower regeneration of the oxidized dye. 149 Remarkable progress with the Cp 2 Fe + /Cp 2 Fe redox shuttle was made by Bach, Spiccia and coworkers in 2011. They demonstrated that DSCs sensitized with the organic dye Carbz-PAHTDTT (Scheme 12) and containing an electrolyte comprising Cp 2 Fe + /Cp 2 Fe and TBP in MeCN, attained PCEs of up to 7.5%. The dye was selected because of its light absorption over a wide visible-wavelength range; best performances are gained with thin TiO 2 electrodes. The data in Table 4 illustrate the effects of adding the co-adsorbent cheno, and compare the use of Cp 2 Fe + /Cp 2 Fe with the standard I 3 À /I À redox mediator. 150 Exclusion of O 2 from the DSCs is essential when the Cp 2 Fe + / Cp 2 Fe couple is employed, and this makes cell fabrication less convenient than with many other redox mediators. Just as the redox potentials of the Co 3+ /Co 2+ and Cu 2+ /Cu + couples can be tuned by choice of ligand (see earlier), an advantage of the Cp 2 Fe + /Cp 2 Fe couple is that the redox potential can readily be shied to higher or lower potentials through functionalization of the cyclopentadienyl rings. Tris(2,2 0 -bipyridine)iron(III)/(II) based couples have been investigated both as redox mediators and co-mediators, the [Fe(bpy) 3 3 ] 2+ was added as a comediator. The improvement has its origins in electron-transfer between the Co 3+ /Co 2+ and Fe 3+ /Fe 2+ couples which creates an electron cascade between oxidized dye, electron co-mediator and electron mediator. 151 As Fig. 2a and 3 illustrated, high values of V OC are achieved by careful tuning of E redox , and by judicious matching of dye and redox couple energy levels. Earlier, we described the efficient combination of [Co(bpy) 3 ] 3+ /[Co(bpy) 3 ] 2+ with the donor-p-bridge-acceptor triphenylamine dye D35 (Scheme 4). 85 Starting with D35, Delcamp and coworkers 152 designed dye RR9 (Scheme 12) to be energetically compatible with the [Fe(bpy) 3 ] 3+ /[Fe(bpy) 3 ] 2+ redox mediator. Note that the arylamine group in D35 was replaced by an aryl-centred unit in RR9 to achieve a lower energy ground-state oxidation potential. Upon going from a combination of D35 and [Co(bpy) 3 ] 3+ / [Co(bpy) 3 ] 2+ to RR9 and [Fe(bpy) 3 ] 3+ /[Fe(bpy) 3 ] 2+ , the maximum theoretical increase in V OC is 810 mV. In practice, DSCs with these dye-redox couple combinations achieved values (average for two cells) of V OC of 760 and 1420 mV, respectively. TiO 2 layer thickness (2.7 mm) proved critical. For masked DSCs with RR9 and [Fe(bpy) 3 ] 3+ /[Fe(bpy) 3 ] 2+ , average values of J SC ¼ 2.8 mA cm À2 , ff ¼ 47%, and h ¼ 1.9% were reported. This work is also of note for the fabrication of sequential series multijunction (SSM)-Scheme 12 Structures of the metal-free dyes Carbz-PAHTDTT and RR9, and the ruthenium(II) dyes [Ru(tpyCO 2 H)(ttpy)][PF 6 ] 2 (tpy ¼ 2,2 0 :6 0 ,2 00 -terpyridine) and the commercially available Ruthenizer-505.  2 ] 2+ (pytpy ¼ 4 0 -(pyridin-4-yl)-2,2 0 :6 0 ,2 00terpyridine) were tested under illumination of 1000 W m À2 , and compared with analogous DSCs using an I 3 À /I À redox shuttle.
The J-V characteristics of the Fe-based electrolyte were notably poorer than those with I 3 À /I À , and this was explained in terms of the slower reduction kinetics of the oxidized dye for the [Fe(pytpy) 2 ] 3+ /Fe(pytpy) 2 ] 2+ mediator. 153 Although there has been progress with the use of Cp 2 Fe + / Cp 2 Fe and [Fe(bpy) 3 ] 3+ /[Fe(bpy) 3 ] 2+ -based redox couples in DSCs, there is signicant scope for further exploration and improvements. Among the other classes of iron complexes considered for potential redox mediators are those with scorpionate ligands, i.e. tridentate (tripodal) ligands which lead to metal complexes with high stability constants. Such ligands are preorganized to bind to a metal ion in a fac-mode, as in [Fe(bdmpza) 2 ] + (Fig. 5b) First row d-block metals: from redox mediators to dyes So far, we have focused on the highly promising approaches to replacing the I 3 À /I À redox couple by mediators based upon rst row d-block metals, in particular cobalt and copper. As we have discussed, some outstanding photoconversion efficiencies have been achieved using Co 3+ /Co 2+ or Cu 2+ /Cu + couples and metalfree or zinc(II) porphyrin dyes. Nonetheless, the synthetic complexity associated with many state-of-the-art organic dyes is a disadvantage for upscaling for commercial applications. In contrast, rst row d-block metal coordination compounds containing synthetically-accessible ligands and which absorb in the visible region are readily prepared. Complexes of copper(I) and iron(II) are especially promising candidates for use as sensitizers in DSCs.
The excited states of both [Ru(diimine) 3 ] 2+ and [Cu(diimine) 2 ] + complexes are mainly metal-to-ligand chargetransfer (MLCT) in character, and arise from the excitation of an electron from metal d-orbitals to antibonding p*-orbitals localized on the diimine ligand. There is, however, a signicant difference between the ruthenium(II) and copper(I) compounds. Upon excitation of an octahedral [Ru(diimine) 3 ] 2+ (d 6 ) species, there is negligible change in the equilibrium geometry as the metal formally undergoes oxidation from Ru(II) to Ru(III). A d 10 [Cu(diimine) 2 ] + complex is tetrahedral (or distorted tetrahedral) and the excited MLCT state is formally a d 9 copper(II) species for which a tetragonal arrangement of donor atoms is preferred. Excitation is therefore accompanied by a attening of the copper coordination sphere and, unless this is mitigated through steric effects (see later), solvent interactions with the Cu(II) metal centre result in shortening of the excited-state lifetime.
Another important difference between [Ru(diimine) 3 ] 2+ and [Cu(diimine) 2 ] + complexes is the lability of the ligands. The d 6 conguration of ruthenium(II) leads to a kinetically inert metal centre. In contrast, the d 10 6 ]. 177 Ligand lability is a key issue that has been addressed by the 'surfaces-as-ligands, surfaces-as-complexes' (SALSAC) 178 and the heteroleptic 1,10-phenanthroline Cu(I) complexes (HET-PHEN) 179 approaches which are discussed in detail below.

Development of copper(I) dyes: homoleptic complexes
A number of reviews provide an entry into the area of bis(diimine)copper(I) sensitizers for DSCs. 174,175,[180][181][182][183] In 1994, Sauvage and coworkers were the rst to demonstrate the combination of a homoleptic copper(I) complex as a dye with a wide-band-gap semiconductor for photoconversion. 184 A number of features of their dye (Scheme 13) are relevant for an understanding of the design of ligands for bis(diimine)copper(I) sensitizers. Firstly, the choice of phen rather than bpy is advantageous because the phen metal-binding domain is preorganized for coordination whereas bpy requires a conformation change from s-trans to scis (Scheme 13). Consequently, [Cu(phen) 2 ] + -based complexes possess stability constants which are typically one or two log K units greater than analogous [Cu(bpy) 2 ] + complexes. On the other hand, this has to be offset against the fact that a greater range of functionalized bpy ligands is synthetically accessible than functionalized phen ligands. The second feature of the copper(I) complex shown in Scheme 13 is the presence in the phen ligands of sterically demanding 2,9-substituents to prevent attening of the copper coordination sphere upon excitation (see above). Thirdly, the carboxylate units are introduced to act as anchoring domains to attach the dye to the semiconductor surface. From this milestone report, little progress was made 185 (Fig. 7) illustrated that the 6-and 6 0 -methyl groups are sufficiently large to protect the Cu(I) centre. 186 Note that while dyes with alkyl ester functionalities may bind to TiO 2 as a result of hydrolysis to the corresponding carboxylic acid, 186 use of the ester-protected anchoring domains is usually detrimental to DSC performance. 187 In order to overcome this problem, Soo and coworkers pre-treated the TiO 2 -electrodes with a THF solution of KO t Bu for two days. Attachment of [Cu(4) 2 ] + was then possible through reaction of the ester groups in 4 (Scheme 14) with the activated surface. However, a nonoptimized DSC containing [Cu(4) 2 ] + and the I 3 À /I À redox mediator with TBP, GNCS, and DMII additives (GCNS ¼ guanidinium thiocyanate, DMII ¼ 1,3-dimethylimidazolium iodide) in a MeCN/valeronitrile-based electrolyte only achieved a value of J SC ¼ 0.0338 mA cm À2 and V OC ¼ 339 mV. Slight improvement was obtained by using [Cu(5) 2 ] 3À in which [5] 2À (Scheme 14) contains sulfonate anchors. 188 The effects of structural variation in the anchoring domain are known to have Scheme 13 The sensitizer designed by Sauvage and coworkers, and the preorganized nature of the phen metal-binding domain compared to the conformational change required by bpy. a signicant impact on dye performance in DSCs, 41 and Wills et al. 189 have shown that the introduction of a thienyl spacer between a 6,6 0 -Me 2 bpy metal-binding domain and a carboxylic acid anchoring group leads to enhanced PCE for DSCs containing the homoleptic dye [Cu (6) The extended conjugation with respect to related bpy complexes leads to the absorption maximum extending to longer wavelengths. In solution l max ¼ 564 nm, and for the dye adsorbed in TiO 2 , l max ¼ 552 nm. DSCs sensitized using [Et 3 NH] + or Na + salts of [Cu(4,4 0 -(HO 2 C) 2 biq)(4,4 0 -(O 2 C) 2 biq)] À or with [Cu(4,4 0 -(HO 2 C) 2 biq) 2 ] + coupled with an I 3 À /I À redox mediator were tested. Values of J SC and V OC were in the ranges of 0.197-0.235 mA cm À2 and 499-629 mV, respectively, leading to PCEs # 0.1%. Removing TBP from the electrolyte did not lead to enhancement of DSC performance. 190 A possible reason for these low performances is the short excited-state lifetime caused by exciplex formation. 191 From homoleptic to heteroleptic complexes: the SALSAC approach Enhancement of the photoconversion efficiency of DSCs based upon homoleptic copper(I) complexes is limited 175,180 because of the lack of the desired 'push-pull' effect which is the key to the success of donor-p-bridge-acceptor metal-free dyes (see earlier discussion). To improve the performance of homoleptic copper(I) dyes, one strategy is to optimize the electrolyte components. 192 A second strategy of using surface-bound heteroleptic dyes allows broad scope for capitalizing upon structure-property relationships. Fig. 8 illustrates the components of a heteroleptic copper(I) dye that should be present to facilitate electron transfer across the dye from the redox shuttle to semiconductor. In 2008, we established a protocol for ligand exchange reactions between [Cu(N^N) 2 ][PF 6 ] (N^N ¼ functionalized bpy ligand) and TiO 2 -anchored ligands, 177 and later screened a wider range of anchoring ligands to 7-10 (Scheme 15). 193 The dye assembly process involves ligand exchange between a surface-anchored diimine ligand, L anchor , and a homoleptic complex [Cu(L ancillary ) 2 ] + in the dye-bath (Fig. 9). This is the basis of the SALSAC approach which has also been developed to include stepwise assembly of the dye. 178 The strategy takes advantage of the lability of bis(diimine)copper(I) complexes in solution. Critically, once attached to the semiconductor surface, the heteroleptic complex does not suffer from rapid ligand dissociation. The formation of surface-bound heteroleptic complexes was conrmed by using MALDI-TOF mass spectrometry and solid-state absorption spectroscopy. While not achieving PCEs greater than 1.51% for unmasked DSCs, our investigation in 2011 was pivotal in revealing that, for copper(I) dyes, phosphonic acid anchoring groups were more benecial than carboxylic acids. 193 We later showed that the presence of a phenylene spacer in anchoring ligand 11 (Scheme 15) enhanced DSC performance with respect to analogous DSCs containing dyes incorporating anchoring ligand 10. 194 The observed bene-ts of using phosphonic rather than carboxylic acid anchors are consistent with the known adsorption strength of a phosphonic acid on TiO 2 being ca. 80 times greater than that of a carboxylic acid. 195 In 2013, Ashbrook and Elliott reported a stepwise assembly of heteroleptic copper(I) complexes of which [Cu(12)(tmpDMP)] + (Scheme 16a) is representative, coupled with the use of a [Co(4,4 0 -t Bu 2 bpy) 3 ] 3+ /[Co(4,4 0 -t Bu 2 bpy) 3 ] 2+ redox mediator. 196,197 Around the same time, we also showed that a [Co(bpy) 3 ] 3+ /[Co(bpy) 3 ] 2+ redox couple could replace I 3 À / Scheme 14 Structures of the bis(arylimino)acenaphthene compounds 4 and Na 2 [5], and ligand 6. I À with no loss in DSC performance. 198 Both studies used heteroleptic copper(I) dyes (Scheme 16) and were important in establishing the compatibility of copper(I) dyes and cobaltbased redox shuttles, paving the way for a shi away from the I 3 À /I À mediator. In the work from Ashbrook and Elliott, a DSC containing [Cu(12)(tmpDMP)] + performed the best in the series, but achieved only a modest value of J SC (0.54 mA cm À2 compared to 1.74 mA cm À2 for a reference DSC with the ruthenium dye N3). 196,197 The performances of masked DSCs sensitized by the dye shown in Scheme 16b combined with a [Co(bpy) 3 ] 3+ /Co(bpy) 3 ] 2+ redox mediator were inuenced by the thickness of the TiO 2 layer on the photoanode and by posttreatment with H 2 O-TiCl 4 . The highest value of J SC obtained was 5.11 mA cm À2 which contributed to h ¼ 2.08% (compared to h ¼ 6.90% for a DSC with N719 and I 3 À /I À ). Note that the design of ancillary ligand in the complex in Scheme 16b incorporates an electron-donating Ph 2 N unit and long alkyl chain to inhibit electron recombination. 198 We return to similar dyes later. Ashbrook and Elliott also demonstrated that Lewis bases in the [Co(4,4 0 -t Bu 2 bpy) 3 ] 3+ /[Co(4,4 0 -t Bu 2 bpy) 3 ] 2+ -based electrolyte (e.g. TBP or solvent) interact with the oxidized form of the surface-bound dye, inhibiting dye regeneration. They noted that this can be circumvented by careful choice of the ancillary ligand. When the dye was [Cu(12)(tmpDMP)] + , the 2,4,6,8-tetramethylphenothiazine groups in the ancillary ligand rapidly reduce the oxidized dye, thus precluding it from being kinetically trapped by coordination of Cu(II) with TBP. 196,197 Despite the promising performances achieved using combinations of bis(diimine)copper(I) dyes and Co 3+ /Co 2+ redox mediators, the vast majority of DSC investigations involving copper(I) dyes continue to employ the I 3 À /I À couple. We shall discuss the role of Cu 2+ /Cu + mediators in a later section. By using the SALSAC approach, it is possible to screen a wide range of [Cu(L anchor )(L ancillary )] + sensitizers containing different ancillary ligands with a common anchoring domain, or a range of anchoring ligands with a common L ancillary . As already detailed, we have found that for copper(I) dyes, ligand 11 is the anchoring ligand of choice. The protonation state of the ligand adsorbed on the TiO 2 surface remains undened, although the addition of one equivalent of base during surface functionalization with 11 can lead to an increase in DSC efficiency. On the other hand, addition of $3 equivalents of base results in poorer DSC performances. 199 Replacing the phenylene spacers in 11 by 2-thienyl spacers bearing the phosphonic acid in the 5-or 4positions (ligands 13 and 14, Scheme 17) leads to slight performance enhancement, and V OC is higher when the PO(OH) 2 group is in the 4-rather than 5-position. 200 However, the improvement in PCE is offset by the easier synthetic route to 11 compared to the thienyl derivatives. Similarly, there is no benet to replacing the PO(OH) 2 groups in 11 by cyanoacrylic acid or (1-cyanovinyl)phosphonic acid anchors (ligands 15 and 16, Scheme 17). 201 Table 5 presents DSC parameters for a wide range of heteroleptic copper(I) dyes in which the anchoring ligand is 11 and the ancillary ligands are dened in Schemes 18-20. So that comparisons are legitimate, we have only included data for DSCs which were fabricated in a similar manner, in which DSCs were fully masked, and in which the electrolyte comprised LiI (0.1 M), I 2 (0.05 M), 1-methylbenzimidazole (0.5 M) and 1-butyl-3-methylimidazolinium iodide (0.6 M) in 3-methoxypropionitrile (MPN). In most cases, the solvent in the dye bath for the ligand exchange process (Fig. 9) was CH 2 Cl 2 (column 2 in Table  5). So that the photoconversion efficiencies can be compared for different investigations, values of h are accompanied in Table 5 by values relative to h for a reference DSC sensitized by N719. This is especially important when DSC performances are measured using different sun simulators (see footnotes b and c in Table 5) where absolute values of J SC , V OC , and h may differ, but the relative values of h are comparable. 187 This is evident from the rst two entries in Table 5 for DSCs sensitized with Fig. 9 The SALSAC approach to in situ assembly of a heteroleptic copper(I) dye on an electrode surface using ligand exchange.
[Cu(11)(Me 2 bpy)] + , and for the DSCs containing [Cu(11)(22)] + . All parameters in Table 5 refer to the performances recorded on the day on which DSCs were fabricated. Where parameters for multiple cells were reported for a given dye in the original publications, the best performing DSC from each set is given in Table 5.
The simplest ancillary ligand in Scheme 18 and Table 5 is Me 2 bpy. The 6,6 0 -dimethyl substituents are primarily introduced to hinder attening of the copper coordination sphere upon photoexcitation (see earlier discussion). The sensitizer [Cu(11)(Me 2 bpy)] + can therefore be taken as a reference point to illustrate how DSC performance can be improved as a consequence of structural modication of the ancillary ligand. To achieve the donor-acceptor properties of the dye (Fig. 8) Table 5, for DSCs measured on the same instrument) and this is also reected in high external quantum efficiencies (EQEs) of 51% for [Cu(11)(17)] + and 46% for [Cu(11)(18)] + (EQE l max ¼ 480 nm). 202 On going from Me 2 bpy to the 4,4 0 -diphenyl derivative 19, 203 a small increase in PCE is observed as a consequence of gains in both J SC and V OC . The introduction of the peripheral halogen substituents in ancillary ligands 20-23 produces a signicant rise in J SC , the optimum (J SC ¼ 6.92 mA cm À2 ) being for dye [Cu(11)(23)] + with iodo substituents. The value of V OC is also enhanced, leading to an overall h of 2.89%, which is 38.9% of the performance of a DSC with the reference dye N719. Despite the relative simplicity of the ancillary ligand, this remains one of the best performing copper-based dyes combined with an I 3 À /I À redox shuttle  (Fig. 9) rather than CH 2 Cl 2 is detrimental to performance, probably due to the Lewis basicity and coordination tendency of MeCN. On going from ancillary ligand 22 to 25, the 6,6 0 -dimethyl substituents are replaced by 6,6 0 -diphenyl groups. By comparing data from DSC measurements made with the same instrumentation (Table 5), it is clear that the more sterically demanding phenyl substituents lead to slightly lower performances. More detrimental still is the introduction of phenyl substituents into the anchoring ligand, i.e. replacing the 6,6 0 -dimethyl groups in 11 by a 6,6 0 -diphenyl substituent pattern (11-Ph). Although the extended conjugation achieved with the phenyl groups leads to improved light absorption towards longer wavelengths, dyes with anchor 11-Ph rapidly bleach when exposed to the I 3 À /I À -containing electrolyte solution. 205 We now move to dyes carrying peripheral methoxy groups. Interestingly, DSCs sensitized with [Cu(11)(23)] + bearing the peripheral iodo groups 204 outperform those using the related dye [Cu(11)(24)] + carrying electron-releasing methoxy groups (Scheme 18 and Table 5). 203 On going from 24 to 26 and 27, changes in the substitution pattern have a signicant impact on DSC performance (Table 5), consistent with the electronreleasing nature of 4-MeO groups, and electron-withdrawing properties of 3-and 5-methoxy groups. 203 Introduction of the asymmetrically substituted 28 and 29 ancillaries (Scheme 18) also give DSCs with rather high performances, and this is despite the fact that these ligands are based on a 6-Mebpy rather than a 6,6 0 -Me 2 bpy core. 206 An investigation of a series of heteroleptic copper(I) dyes incorporating Schiff base ancillary ligands 30-34 concluded that the presence of the imine bond prevents efficient electron transfer across the dye. 207 This can also be seen by comparing the performances of DSCs sensitized by dyes [Cu(11)(34)] + (Schiff base) and [Cu(11)(35)] + (no C]N unit). The presence of the C]N domain is detrimental to both J SC and V OC (Table 5). In [Cu(11)(35)] + and [Cu(11)(36)] + , the electron-donating NR 2 groups were expected to be benecial in terms of stabilizing the hole remote from the TiO 2 surface. 208 Indeed, DSCs sensitized with these dyes performed well with respect to cells with the N719 reference dye (relative h ¼ 33.4 and 36.0%, Table 5). However, as with the introduction of the imine functionality, inserting an alkyne unit between the phenylene and pyridine rings on going from ancillary ligand 35 to 37, or 36 to 38, resulted in signicant decreases in J SC values and in overall PCEs (Table 5). 208 Ancillary ligands 39 and 40 (Scheme 18) represent rst and second generation hole-transport dendrons. Going from [Cu(11)(39)] + to [Cu(11)(40)] + leads to enhanced light absorption towards lower energies, and although DSCs with [Cu(11)(40)] + outperform those sensitized with [Cu(11)(39)] + , there was no gain in overall performance compared to the more structurally simple dyes listed in Table 5. 209 With its sterically demanding and arene-rich second generation dendron, the dye [Cu(11)(40)] + is likely to suffer from performance loss arising from aggregation. Hence, DSCs were also fabricated with cheno added to the dye. This had a signicant impact on values of Scheme 18 Structures of ancillary ligands which are derivatives of bpy used in complexes in Table 5. Note the use of electron-releasing methoxy groups in some of the ligands (see text).
both J SC and V OC leading to a PCE of 2.23% compared to 1.50% without cheno. Interestingly, when acetone was used in the dye bath in place of CH 2 Cl 2 (the step shown in Fig. 9), better DSC performances were observed for [Cu(11)(40)] + although the enhancement gained by adding cheno was less noticeable. 212 Like 39 and 40, ancillary ligands 41-44 (Scheme 19) were designed with electron-donating peripheral groups. They contain a phen metal-binding domain, and the design of 42-44 incorporates n-octyl chains to militate against electron recombination. For the SALSAC dye assembly (Fig. 9), the solvent used in the dye bath was MeCN. Compared to the DSCs with dyes containing bpy-based ancillary ligands (Me 2 bpy and 17-40), Table 5 (11)(45)] + , the device exhibited a high chemical capacitance and a low recombination resistance. However, the latter is offset by a low transport resistance, leading to a high J SC and V OC . DSCs with [Cu(11)(48)] + have the lowest transport resistance of this family of dyes. 211 One of the major drawbacks of bis(diimine)copper(I) dyes is the spectral limitation of light absorption. The broad MLCT absorption of a simple [Cu(bpy) 2 ] + derivative typically has its maximum at ca. 460-480 nm, and although the absorption can be extended towards longer wavelengths by judicious functionalization, much of the region beyond 600 nm remains unharvested by the dye. A major improvement can be made by co-sensitization with a complementary organic dye. Suitable commercial dyes include SQ2 (Scheme 2) and Fig. 10 displays the solid-state absorption spectra of separate FTO/TiO 2 electrodes functionalized with [Cu(11)(45)] + and SQ2. Aer optimization of dye-bath conditions, the best performing masked DSC achieved values of J SC ¼ 9.56 mA cm À2 , V OC ¼ 493 mV, h ¼ 3.36% and relative h (relative to N719) ¼ 44.5%. Upon ageing the DSC for a week, further improvement was observed with J SC ¼ 12.26 mA cm À2 , V OC ¼ 515 mV, h ¼ 4.51% and relative h of 65.6% with respect to N719. 215 Since this report in 2017, the co-sensitization strategy has unfortunately not been exploited further, and we advocate this as a fruitful method of boosting the PCE of copper-based DSCs without the need for elaborate organic structure design.
Further application of the SALSAC approach is exemplied in the dipyrrin complexes 216 which are discussed later with porphyinato derivatives.

From homoleptic to heteroleptic complexes: the HETPHEN approach
In contrast to our own approach of in situ assembly of a heteroleptic dye on TiO 2 , Odobel and coworkers have approached the problem of labile bis(diimine)copper(I) dyes by applying the HETPHEN strategy. 179,217,218 The underlying principle of this approach is the use of a phen metal-binding domain bearing very sterically demanding mesityl substituents in the 2,9-positions. 219 This motif can be in either the anchoring or ancillary ligand (ligands 51-54 in Scheme 21). In 2013, Sandroni et al. reported  Table 5.   179 The structural, photophysical and electrochemical properties of related complexes had previously demonstrated the success of this synthetic strategy and also the stability of the heteroleptic species in solution. Fig. 11 shows the structure of a representative complex and highlights the p-stacking interaction between one mesityl group and the phen domain of the second ligand. 220 The solid-state absorption spectra of [Cu (51) Table 6; the redox shuttle was I 3 À /I À . The values of h compared with 7.36% for a reference DSC containing N719. The benecial effects of cheno in preventing dye aggregation on the semiconductor surface are clear, with higher values of J SC and V OC in all cases. The PCEs of 4.42 and 4.66% for DSCs with [Cu(54)(56)] + /cheno and [Cu(54)(56)] + , respectively, remain the highest reported for heteroleptic copper(I) sensitizers. However, we note that no comments were made in the original work about the use of a DSC mask. 217 The HETPHEN approach is extremely attractive for the preparation of robust copper(I) sensitizers. However, with the exception of work from Dragonetti et al. described later, 221 there appears to have been little further progress in applications in DSCs since the initial work from the Odobel group. Additional investigations of the photophysical properties of sterically congested bis(diimine)copper(I) species prepared by the HET-PHEN strategy have been reported. 222,223 Other heteroleptic copper(I) and copper(II) dyes

Scheme 19 Structures of ancillary ligands which are derivatives of phen used in complexes in
In this section, we give an overview of copper photosensitizers other than bis(diimine)copper(I) complexes. In comparison to the large literature focused on porphyrinatozinc(II) or phthalocyanatozinc(II) dyes, 46,48,[53][54][55][56] those containing copper(II) are rather sparse. Porphyrins and phthalocyanines exhibit intense Scheme 21 Anchoring and ancillary ligands used in the HETPHEN approach to copper(I) dyes.  Photoanodes were made with either a 3.3 mm thick layer of TiO 2 or a double layer (7.5 + 5 mm) TiO 2 architecture. Table 7 summarizes the performances of the DSCs, illustrating that CPICu-based DSCs outperform their zinc(II) analogues, and that the thickness of the TiO 2 layer signicantly affects performance. 229 An interesting contribution from Leung and coworkers describes the fabrication of photoanodes comprising TiO 2 sensitized with N719 and then covered with a 30 nm layer of CuPc. The design is predicated upon a cascade charge-transfer process involving the absorption of a near-infrared photon and creation of an exciton (electron-hole pair) which diffuses to the CuPc/N719 interface; electrons are transferred to the LUMO of N719 and then injected into the TiO 2 conduction band. This innovative design leads to an increase in J SC from 14.97 mA cm À2 (no CuPc) to 21.12 mA cm À2 (with CuPc) with negligible change in V OC . Overall, the PCE increases from 6.39 to 9.48%. The thickness of the CuPc layer was optimized to minimize quenching effects from the oxidized form of the I 3 À /I À redox couple. 230 In 2014, Robertson and coworkers screened a series of heteroleptic copper(I) complexes containing the dipyrrin ligands [58] À , [59] À and [60] À (Scheme 22) and anchoring ligand 1 (Fig. 1). The heteroleptic dyes were assembled using the SALSAC approach, 178 (62)] + , the Cu(I) ion is 5-coordinate although one Cu-N bond is longer (3.020(2) A) than is typical. DSCs were made using N719 alone as the dye, and with TiO 2 photoanodes co-sensitized with N719 and [Cu(PPh 3 ) 2 (61)] + , or N719 and [Cu(PPh 3 ) 2 (62)] + . Given that the amount of N719 in the dye baths for the latter was half that in the pristine N719-based cells, it is noteworthy that the DSC containing N719 and [Cu(PPh 3 ) 2 (62)] + reached a relative PCE of 73.1% with respect to the pristine N719-based cell. However, low ll-factors contributed to low PCEs for all devices. 233 In contrast to this co-sensitization approach, Robertson and coworkers designed a dimetallic complex 63 (Scheme 23) in which the Ru(II) centre is in an environment similar to that in N3 or N719, and the Cu(II) centre is in a cyclam cavity. The absorption range and molar extinction coefficients of the dye were enhanced relative to those of N719. However, combined with an I 3 À /I À redox mediator, the best PCE that was obtained under light irradiation of 1000 W m À2 was 2.55% relative to 6.4% for an N719 reference DSC. This was attributed to an energy mismatch between the TiO 2 conduction band and the LUMO of the dye combined with the instability of the dye in the electrolyte solution. 234 The neutral dinuclear copper(I) complexes 64-67 (Scheme 23) were designed with thienyl domains although these are not associated with the anchoring units. Solid-state absorption spectra of the compounds adsorbed on TiO 2 exhibit broad bands which are most red-shied for 66 and 67 which contain the terthienyl substituents. However, the absorption maximum for each of the four compounds is <400 nm. Theoretical studies   (Table 8), and the ll factors with the copper-based redox shuttles were also low. 236 236 we also demonstrated the successful combination of copper(I) dyes and Cu 2+ /Cu + redox mediators. 238 The SALSAC approach was used to functionalize photoanodes with the dyes [Cu (11) 238 On the other hand, given the susceptibility of Cu(II) to coordination by TBP and other Lewis bases (see earlier discussion), the role TBP with Cu 2+ /Cu + couples remains to be further explored in all-copper DSCs.

The holy grail: iron in DSCs
The promise of iron(II) sensitizers In 2004, in reviewing metal complexes as sensitizers, Polo et al. stated: "As iron is a common and cheap metal, it can provide a very economical alternative to ruthenium in sensitizing complexes". 239 However, progress in this eld has been slow and the reasons for this are effectively summarized by Wenger who, in 2019, posed the question: "Is iron the new ruthenium?". 240 Of metal-based photosensitizers, some of the highest photoconversion efficiencies are realized using ruthenium(II) compounds. However, as we have already noted, the very low crustal abundance of ruthenium (ca. 0.001 ppm) 27 makes its use on a commercial scale non-sustainable. Like Ru(II), Fe(II) has a d 6 electronic conguration. However, the photophysical behaviour of iron(II) polypyridine complexes does not mirror that of their ruthenium(II) counterparts. While polypyridyl Ru(II) complexes possess long-lived MLCT excited states in a characteristic range of 100-1000 ns, the MLCT excited states in polypyridyl Fe(II) complexes suffer extremely fast deactivation. 240 Although this militates against their use as photosensitizers, early investigations by Ferrere et al. 241,242 conrmed that functioning DSCs were possible using simple dyes such as [Fe(4,4 0 -(HO 2 C) 2 bpy) 2 (CN) 2 ] (a close analogue of N3, Scheme 3) in the presence of the co-adsorbent cheno. We have previously reviewed these and related early results in detail. 175 More recently, Jakubikova and coworkers published the results of in which X ¼ carboxylic acid, phosphonic acid, hydroxamate, catechol and acetylacetonate. The results suggest that hydroxamate anchors could be particularly benecial, 244 but to the best of our knowledge, this has not been conrmed in practice. Using a computational approach, Tsaturyan et al. investigated the effects of having one, two, three or four CO 2 H anchors in the dye [Fe(qtpy)(NCS) 2 ], the anchors being introduced into the 4-, 4 0 -, 4 00 -and 4 000 -positions of 2,2 0 :6 0 ,2 00 :6 00 ,2 000quaterpyridine (qtpy). In terms of the absorption spectrum and energies of the frontier MOs of the complex, it was concluded that the optimal number of anchors is two. 245 An overview of computational work focused on polypyridyl iron(II) complexes as sensitizers published in 2015 gives an excellent insight into the ground rules for ligand design, in particular in addressing the relative rates of intersystem crossing (ISC) and IET processes, the aim being to ensure that IET becomes more competitive with ISC. 246 A further investigation from Jakubikova and coworkers is relevant to the move from polypyridyl to N-heterocyclic carbene (NHC) complexes of iron(II) as dyes in DSCs (see below). Starting with [Fe(tpy) 2 ] 2+ and [Fe(dcpp) 2 ] 2+ (dcpp ¼ 2,6-bis(2carboxypyridyl)pyridine), the eld strength of the terdentate ligands was systematically altered by replacing the central pyridine ring with 5-membered (NHC, pyrrole, furan) or 6membered (aryl, thiazine-1,1-dioxide, 4-pyrone) units. For applications of these complexes as dyes, several design principles are advocated: (i) the presence of Fe-C bonds (i.e. NHC ligands are favoured), (ii) as ideal an octahedral environment around Fe(II) as possible (i.e. 6-rather than 5-membered ring as Table 9 DSC (masked cells) parameters for the best combinations of bis(diimine)copper(I) dyes and Cu 2+ /Cu + redox couples compared with N719 from the work of Karpacheva et al. 238  the central unit in the terdentate ligand), and (iii) short Fe-X ligand bond lengths. 247 Ashley and Jakubikova have also developed a computational approach to predicting the redox behaviour of polypyridyliron(II) complexes which should assist in the design of Fe(II)-based dyes for DSCs. 248 Dyes containing ferrocenyl units also received attention in the period up to 2013, although performances in DSCs were typically poor. 175 Several later investigations of ferrocenyl derivatives have focused on ferrocenyl dithiocarbamate metal complexes. [249][250][251][252][253] A masked DSC combining an I 3 À /I À redox mediator with the Co/Fe dye shown in Fig. 12a achieved values of J SC ¼ 7.23 AE 0.03 mA cm À2 , V OC ¼ 620 AE 10 mV and h ¼ 3.25 AE 0.04% under light intensity of 1000 W m À2 . 252 Yadav et al. have also reported the enhancement of N719-based DSCs by cosensitization with the ferrocenyl dithiocarbamate zinc(II) complex shown in Fig. 12b. An average value for three devices of h ¼ 7.10 AE 0.02% was obtained which compared to h ¼ 5.76 AE 0.04% for N719 alone with an I 3 À /I À redox shuttle. 251 The group of van Zyl has reported a series of ferrocenyl-decorated dithiophosphonate complexes of Ni(II), Zn(II) and Cd(II) which have been tested as single dyes and as co-sensitizers with N719 in DSCs. The anchoring domains are P-OH units (Fig. 12c). Using a commercial I 3 À /I À -based electrolyte, the performances of unmasked DSCs reached PCEs of 2.43, 3.58 and 3.12% (averages of three cells) for the Ni(II), Zn(II) and Cd(II)-centred dyes compared to 7.19% for N719. Co-sensitization with N719 led to slightly enhanced performances with respect to N719 alone for the Zn(II) and Cd(II)-based dyes. 254 Several examples of derivatives of 1,1 0 -bis(diphenyl)phosphanoferrocene (dppf) have been trialled as sensitizers in DSCs. 255,256 The dimetallic complexes [Ni(dppf)(S 2 C]C(H)NO 2 )] and [Ni(dppf)(S 2 C]C(COMe) 2 )] absorb in the visible with maxima ca. 430 nm when adsorbed on TiO 2 . The dyes were combined with an I 3 À /I À redox mediator in unmasked DSCs which were illuminated under a light intensity of 1000 W m À2 .
Compared to values of J SC , V OC and h ¼ 13.22 mA cm À2 , 730 mV and 6.91% for a reference N719 DSC, the ferrocenyl dyes performed with values of J SC ¼ 5.87 and 7.96 mA cm À2 , V OC ¼ 644 and 654 mV and h ¼ 2.32 and 3.21%. However, the authors conclude that a change in the anchoring domain from the nitro and diacetyl present in [Ni(dppf)(S 2 C]C(H)NO 2 )] and [Ni(dppf)(S 2 C]C(CMeO) 2 )], respectively, could be benecial. 256 In a series of papers,Özacar and coworkers reported the use of complexes of iron(II) incorporating tannin or quercetin ligands as sensitizers with either TiO 2 or ZnO coated photoanodes. [257][258][259][260] The N-heterocyclic carbene era arrives The bottleneck inhibiting the use of polypyridyl iron(II) complexes as dyes in DSCs is their fast deactivation from an MLCT to lower lying MC states. 240 This leads to inefficient electron injection into the semiconductor and, therefore, low J SC values. A review from Visbal and Gimeno in 2014, looked at the progress made in the eld of luminescent transition metal complexes with HNC ligands, and included applications in DSCs. However, at this stage, the focus was still on ruthenium(II) complexes. 261 Theoretical investigations in the period 2014-2015 highlighted the advantages of using cyclometallating or Nheterocyclic carbene ligands (which are strongly s-donating) with iron(II) in place of polypyridyl metal-binding domains. 247,262 In practice, 2013 saw the preparation of the rst homoleptic NHC complex of iron(II) (70, Scheme 25) which exhibited an extended 3 MLCT lifetime (9 ps). This pivotal work from Wärnmark and coworkers 263 increased the excited-state lifetime by a factor of 100 in comparison to previously reported polypyridyl iron(II) complexes. In addition, transient absorption measurements showed no signicant population of 5 MC states, consistent with deactivation pathways in NHC iron(II) compounds such as 70 being fundamentally different from those in polypyridyl iron(II) complexes. The long excited-state lifetime is a consequence of a signicant destabilization of both 3 MC and 5 MC states with respect to those in polypyridyl iron(II) complexes. 264 Replacing the NHC ligands in 70 by the corresponding benzimidazolylidene-based ligands to give complex 71 (Scheme 25) increased the excited 3 MLCT state lifetime to 16.4 ps. 265 Wärnmark, Sundström and coworkers also demonstrated that use of strongly s-donating 1,2,3-triazol-5ylidene domains in the heteroleptic complex [Fe(bpy)(btz) 2 ] 2+ (Fig. 13) led to an excited 3 MLCT state lifetime of 13 ps. 266 Another highly relevant publication around this time came from Kühn and coworkers 267 who demonstrated that the electronic structure of NHC iron(II) complexes is strongly inuenced by the number of NHC donors. Cyclic voltammetric data conrmed a linear correlation between the oxidation potentials of the iron(II) complexes and the number of NHC donors.
All of the investigations of NHC iron(II) compounds described above came closely on the heel of one another, and heralded the arrival of NHC iron(II) dyes. 268 In 2015 both Gros and coworkers 269 and Wärnmark and coworkers 270 reported the synthesis of the carboxylic acid anchor-containing complex 72 (Scheme 25). The introduction of the CO 2 H groups on going from 70 to 72 leads to an increase in the lifetime of the 3 MLCT excited state from 9 to 18 ps (in MeCN solution). 270 A similar elongation of the 3 MLCT lifetime (from 16.4 to 26 ps) is seen on introducing CO 2 H functionalities into 71 to give 73 (Scheme 25). 265 When complex 72 was anchored on an Al 2 O 3 surface, the lifetime was further extended to 37 ps. The latter was (in 2015), the longest excited-state lifetime reported for any mononuclear iron complex. This detailed investigation also provided evidence for efficient photo-induced electron injection from the 3 MLCT state into the conduction band of TiO 2 with a 92% yield of conversion to photoelectrons in the conduction band. Wärnmark and coworkers noted that a proportion of the injected electrons underwent fast electron-dye recombination. 270 We return to this problem later. As this latter work was in progress, 270 Gros and coworkers reported the rst application of complex 72 as a sensitizer in DSCs, and also compared the photophysical properties of 72 with the tpy derivative 74 (Scheme 13). Notably, while the solution absorption spectrum of 74 exhibits one MLCT band in the visible region (l max ¼ 569 nm), that of compound 72 shows two bands (l max ¼ 394 and 520 nm), assigned to MLCT involving the carbene and pyridine rings, respectively. This is a general feature of this family of dyes and contributes to better light harvesting in the visible region. Under illumination of 1000 W m À2 , DSCs containing 72 and the co-adsorbent cheno (Scheme 1) with an I 3 À /I À redox mediator achieved values of J SC ¼ 0.41 mA cm À2 , V OC ¼ 457 mV and h ¼ 0.13% (entry 1, Table 10). These values compared to 0.016 mA cm À2 , 250 mV and 0.01% for DSCs with dye 74. For comparison, a reference DSC with N719 gave a PCE of 6.1%. Data were reproducible for three DSCs per dye. 269 Although the performances were low, these results were encouraging, and provided motivation for further optimization of DSC components.
Improving the performances of DSCs with NHC iron(II) dyes: electrolyte tuning Starting with dye 72, several strategies have been followed to improve the performances of NHC iron(II)-based DSCs. The rst is optimization of the electrolyte composition. Until recently, 271 NHC iron(II)-based DSCs have only incorporated the I 3 À /I À redox mediator, and a commercial electrolyte (AN50) was used in the initial report from Gros. 269 Table 10 shows one of the best-performing cells. 274  on the TiO 2 surface. They therefore used a sequential treatment of the surface with 72 followed by cheno, rather than a single dye bath with both adsorbents. This, as well as the use of a dense TiO 2 underlayer (deposited by spray pyrolysis of a [Ti(O i Pr) 2 (acac) 2 ] solution) leads to improved performance as is appreciated by comparing entries 2 and 3 in Table 10. The increase in V OC is particularly marked. Reproducibility of the DSC parameters was checked by measuring four cells, and ranges of values are given in Table 10. 271 Further tuning of the electrolyte composition 275 focused on optimizing the concentration of I 2 coupled with different ionic liquids (ILs). For dye 72, values of J SC , V OC and ff were all affected by changes in the I 2 concentration, irrespective of the IL. Aer screening a range of N-alkyl substituted imidazolium-based ILs with chains of lengths from methyl to n-hexyl, it was found that   (Tables 10 and 11) for masked DSCs with the dye 72 in the presence of the coadsorbent cheno; cheno was added to the dye in a single dye-  bath. The reproducibility of the cells in Table 11 is noteworthy, and provides credibility to the improvement in DSC performance as a consequence of electrolyte tuning. 275  with respect to a reference N719 cell). In all cases, data extracted from electrochemical impedance spectra support the trends in J SC and V OC . 276 A trade-off between values of J SC and V OC is oen problematical when optimizing DSC performances for a given dye, and this is apparent when comparing entries 3 and 4 in Table 10. Both we 274 and Marchini et al. 277 have also observed that the addition of TBP is detrimental to DSC performance when the dye is 72.
Marchini et al. continued the focus on the homoleptic dye 72 with an informative study that looked at the effects of a change from a TiO 2 to SnO 2 semiconductor, and of further electrolyte tuning. 277 TiO 2 remains the favoured photoanode material. When adsorbed on either SnO 2 or TiO 2 , the NHC-dye 72 gives rise to photoinduced injection leading to a long-lived (millisecond regime) charge-separated state, but for SnO 2 , recombination is faster than for TiO 2 leading to poorer DSC performance for SnO 2 . The PCE of DSCs with 72 (with cheno) can be pushed towards 1% by using (i) an electrolyte composition of Li (Table  10), although there is no comment about masking the DSCs in the original publication. 277 An noted earlier, NHC iron(II) dyes have invariably been combined with the I 3 À /I À redox mediator. /I À redox shuttle militate against good DSC performances, (ii) homoleptic NHC iron(II) complexes do not exhibit the necessary directional electron ow towards the dye-semiconductor interface, and (iii) although heteroleptic NHC iron(II) complexes exhibit benecial interfacial charge separation, they do not show efficient rates of electron injection. 278 However, a change in electrolyte composition as shown in Table 10, combined with the use of an under blocking layer on the photoanode and modications of the counter electrode (PEDOT or Pt, Table 10), have a remarkable effect on the values of J SC and, therefore, on photoconversion efficiency. For masked DSCs, the values of h ¼ 1.29 and 1.44% are some of the highest recorded for NHC iron(II) sensitizers. A critical factor was the addition of Mg 2+ ions to the electrolyte. Computational modelling of the dye/TiO 2 interface and calculations of the electron injection/recombination revealed that the presence of Mg 2+ ions adsorbed on the semiconductor surface leads to a higher rate of electron injection for dye 75. 279 Our own investigations of electrolyte tuning for homoleptic dyes (see above) concluded that an optimum composition was LiI (0.18 M), I 2 (0.05 M) and DMPII (0.60 M) in MPN, and we maintained this composition for initial screening of heteroleptic dyes. Two factors which contributed to enhanced DSC performance were the time in which the photoanode was immersed in the dye bath, and the introduction of longer alkyl substituents in the ancillary ligand. Dye 78 (Scheme 27) was designed with n-butyl and methyl substituents on the ancillary and anchoring ligands, respectively, and aer optimization of dye-bath conditions, the range of PCE values for three DSCs with dye 78 was 0.93-0.95% which represented 14.6-14.9% Scheme 26 Structures of some ionic liquids (ILs) used in DSC electrolytes. relative to DSCs sensitized with N719. The parameters in Table  10 are the average values for three cells. Interestingly, the performances of DSCs with 78 were lower when the coadsorbent cheno was added to the dye bath (Table 10), 280 and this observation appears to be consistent with the competition between dye and cheno reported by Lindh et al. 271 However, the electrochemical impedance spectroscopic response was more strongly inuenced by the immersion time in the dye bath than by the addition of cheno. 280 Earlier discussions in this review illustrated how the nature of the anchoring ligand affects dye performance. This aspect of NHC iron(II) dyes has only recently been explored with a comparison of the performances of sensitizers 75, 79 and 80 (Scheme 27). 279 We have already noted that the performance of 75 (in the presence of cheno) is markedly enhanced by the addition of Mg 2+ ions to the electrolyte (Table 10), the presence of an under blocking layer on the photoanode, and the use of Pt rather than PEDOT-coated counter electrodes. Employing similar fabrication conditions to DSCs with 79 and 80 leads to the DSC performances shown in Table 10. Although the introduction of the phenylene or thienyl spacer in 79 and 80 results in enhanced light absorption in the visible region, the dye performances were not signicantly affected. 279 A recent investigation from Gros and coworkers focused on heteroleptic dyes based upon 75 (Scheme 27) and bearing different substituents in the 4-position of the pyridine ring of the ancillary ligand. 281 Substituents with varying electronic properties were selected and the most efficient dye was 81 (Scheme 27). DSCs were fabricated with 20 mm thick TiO 2 photoanodes, and the optimized electrolyte composition described above, i.e. LiI (0.1 M), I 2 (0.1 M), PMII (0.60 M), MgI 2 (0.1 M), Bu 4 NI (0.1 M) and GNCS (0.1 M) in MeCN; the counter electrode was coated with PEDOT rather than Pt, 281 even though in a previous study, Pt was shown to be more benecial than PEDOT. 279 The best performance was found for dye 81 which incorporates electron-donating MeO groups (Scheme 27 and Table 10). From the results of EIS studies and computational modelling, it was concluded that 81 combines good lightharvesting and benecial recombination kinetics. 281 Wärnmark and coworkers have recently reported the photophysical and electrochemical properties and DSC performances of heteroleptic dyes 82 and 83 (Scheme 27) which incorporate electron-donating ancillary ligands. Time-resolved spectroscopy conrmed ultrafast (<100 fs) interfacial electron injection from dyes 82 and 83 into TiO 2 . However, charge recombination between the injected electrons and the oxidized dye occurs rapidly, and only 5-10% of the injected electrons contribute to DSC current. Given this difficulty, it is exceptionally promising that PCEs of 1.31% for 82 and 0.93% for 83 were achieved (Table  10). For both dyes, cheno was added as a co-adsorbent, but since competitive anchoring of cheno and the dyes to the photoanode was observed, Wärnmark and coworkers used a sequential treatment of the surface with 82 or 83 followed by cheno. The DSC parameters for 82 and 83 in Table 10 are the best cells from sets of four for which performances were reproducible. 271 In summary, the last few years have witnessed signicant progress in the application of NHC iron(II) sensitizers in DSCs. The range of NHC iron(II) complexes remains limited, but the transition from homoleptic to heteroleptic dyes has now been made, providing synthetic strategies that can be adapted to diversify the families of dyes. The steadily increasing PCEs have been a result of (i) systematic tuning of electrolyte compositions, (ii) modication of electrode materials, (iii) ligand functionalization, and (iv) use of the co-adsorbent cheno, although competition between cheno and CO 2 H-anchored dyes must be considered. The eld is ripe for development.
Other first row transition metals in DSCs: still capitalizing on MLCT transitions

Vanadium and chromium
To date, no vanadium-based complexes have been used as sensitizers in DSCs. Chromium compounds have also received little attention. Perhaps the most promising candidates for future investigations are octahedral chromium(0) complexes incorporating chelating diisocyanide ligands. These complexes have been developed by Wenger and coworkers and are isoelectronic with [Fe(bpy) 3 ] 2+ . They are strong reductants and can exhibit long-lived 3 MLCT excited-state lifetimes. [282][283][284][285] To date, these species have not been tested in DSCs.

Nickel
The most important use of nickel in DSCs is in the form of NiO as a p-type semiconductor in p-type or tandem DSCs. [286][287][288][289][290][291][292][293] In 2012, we overviewed Ni(II)-containing sensitizers for n-type DSCs that had been reported up to that date, 175 and here we focus on Ni(II) dyes reported since 2012. A number of dithiolate and dithiocarbamate complexes of nickel and bearing peripheral ferrocenyl units were described earlier in this review. [254][255][256] Zhang et al. reported the formation and structure of the 1Dcoordination polymer [Ni(en) 2 (azobc)] n (H 2 azobc ¼ azobenzene-4,4 0 -dicarboxylic acid) (Fig. 14). FTO/TiO 2 photoanodes were prepared with this dye and N719 as co-sensitizers. Although solution and solid-state UV-VIS spectra are reported, the integrity of the coordination polymer in solution was not established. Compared to N719 alone, additional light-harvesting towards the UV region is observed when a combination of dyes was used, and a signicantly enhanced J SC (13.46 mA cm À2 vs. 8.78 mA cm À2 ) was reported. An increase in V OC suggested that cosensitization produced an upward shi of the conduction band edge of the TiO 2 semiconductor. 294 While nickel(II) phthalocyanine dyes in DSCs have been investigated, their performances are typically lower than cells sensitized by their zinc(II) analogues, 57 as exemplied in the work of Gorduk et al. 226,227 Zinc(II) as 'glue': ligand-centred chromophores rather than MLCT transitions Since Zn(II) possesses a d 10 conguration, zinc-based dyes rely not on MLCT transitions but on ligand-centred chromophores. The most investigated zinc(II)-containing sensitizers in DSCs are zinc(II) porphyrinato and phthalocyanato complexes and their structures and applications in DSCs have been thoroughly reviewed. 46,48,[53][54][55][56][57] In this section, we focus upon complexes in which the Zn 2+ ion functions as the 'glue' that connects anchoring and ancillary ligand domains. We also look at a number of zinc(II) compounds that have been used as cosensitizers, e.g. with N719. As in previous sections, our main focus is on post-2012 publications; for earlier work, we refer the reader to our 2013 review. 175 Zinc(II) dyes assembled in situ using the SALSAC strategy As detailed earlier, the SALSAC approach to assembling copper(I) sensitizers on TiO 2 (Fig. 9) is extremely benecial for effective screening of copper-based dyes. Like Zn(II), Cu(I) has a d 10 electronic conguration and the ligands in tetrahedral [Cu(N^N) 2 ] + complexes (N^N ¼ diimine) are labile in solution. Ligand exchange between two different homoleptic [Cu(N^N) 2 ] + species occurs immediately in solution to give a statistical mixture of homoleptic and heteroleptic complexes. 177 In contrast, octahedral [Zn(tpy) 2 ] 2+ (tpy ¼ 2,2 0 :6 0 ,2 00 -terpyridine or a substituted derivative) complexes undergo slow exchange of ligands in solution. Thus, although the SALSAC approach as shown in Fig. 9 can be applied to assemble [Zn(tpy) 2 ] 2+ -based dyes on TiO 2 , the procedure has been adapted to a stepwise assembly (Fig. 15) to allow optimal formation of anchored heteroleptic [Zn(L anchor )(L ancillary )] 2+ complexes. 295 In an initial investigation, FTO/TiO 2 electrodes were functionalized with anchoring ligands 84, 85 or 86 (Scheme 28), followed by treatment with Zn(OAc) 2 or ZnCl 2 , and then ancillary ligands 87 or 88 (Scheme 28). Combinations of these anchoring and ancillary ligands produced TiO 2 -anchored zinc(II) dyes with solid-state absorption spectra having maxima in the range 425-480 nm arising from ILCT. DSCs using [Zn (84)(87)] 2+ in combination with an I 3 À /I À redox mediator performed the best of the series, but all efficiencies were very low (h < 1%). Nonetheless, this study conrmed the potential for using the SALSAC strategy to assemble zinc(II) sensitizers on TiO 2 photoanodes. 295  /I À redox shuttle exhibited poor PCEs, the primary reason being extremely low values of J SC (<0.5 mA cm À2 ). 296 An important point to note is that DSCs containing FTO/TiO 2 photoanodes without any adsorbed dye generate small values of J SC and V OC which contribute appreciably to

Fig. 15
The SALSAC approach to in situ assembly of a heteroleptic bis(terpyridine)zinc(II) dye on an electrode surface using a stepwise strategy. ZnX 2 is typically Zn(OAc) 2 or ZnCl 2 .
parameters for poorly performing dyes. 297 In order to enhance light-harvesting of the [Zn(L anchor )(L ancillary )] 2+ complexes, ancillary ligands 92 and 93 (Scheme 28) with extended chromophores were incorporated into the dyes [Zn (84) 2+ . Masked DSCs with these dyes and an I 3 À /I À redox shuttle still performed poorly under light intensity of 1000 W m À2 with J SC < 1 mA cm À2 and V OC z 400 mV, although good ff values were achieved. Moving the electron-withdrawing thiadiazole units from the ancillary to the anchoring domain could be benecial. 297 This series of studies indicates that dyes based upon the {Zn(tpy) 2 } 2+ core are not optimal for sensitizers in DSCs, and no further investigations in this area have been reported.

Other zinc(II) sensitizers and co-sensitizers
Jing and coworkers designed the D-p-A zinc(II) complexes shown in Scheme 29. DSCs were made by combining FTO/TiO 2 / dye photoanodes and an I 3 À /I À redox mediator and were illuminated under a light intensity of 1000 W m À2 . The introduction of the second anchoring group on going from 94 to 99 was benecial both for J SC and V OC . Taking into account that there is no comment about the masking of the DSCs in the original paper, 298 their performances appear to be comparable with those of the bis(terpyridine)zinc(II) dyes described above.
There has been a number of investigations dealing with the use of Schiff base zinc(II) dyes as co-sensitizers with the ruthenium(II) dye N719, the aim of enhancing light absorption in the blue-violet and UV regions. This could potentially offset visible light absorption by I 3 À that competes with light-  (103)Cl 2 ] as a co-sensitizer would 'll-in' the higher energy part of solar light harvested by the DSC. Indeed, values of J SC increased from 13.26 to 17.36 mA cm À2 leading to an increase in PCE from 5.14 to 6.62%; V OC was little affected. 300 Yang and coworkers also demonstrated that the introduction of a peripheral electron-donating MeO substituent to give ligand 104 (Scheme 30) is benecial in co-sensitized DSCs with dyes N719 and 104. EIS measurements show that the use of the zinc(II) co-sensitizer led to a decrease of the electron transfer impedance and an increase in the rate of charge transfer. 301 Yang, Fan and coworkers have investigated the use of [Zn (105) 2 ] (H105 ¼ quinoline-3-carboxylic acid, Scheme 30) as a cosensitizer with N719. In the solid state, the zinc(II) complex forms a coordination network [Zn (105) 302 Interestingly, use of the zinc(II) complex of ligand [106] 2À (Scheme 30) as a co-sensitizer with N791 leads to an increase in J SC but a decrease in V OC . Again, the characterized complex was a coordination polymer, [Zn 3 (106) 2 (OH) 2 (OH 2 ) 6 ] n , 303 and we note that dissolution in EtOH for the dye-bath and photoanode assembly will result in disassembly of the polymer.

Recyclability
This review has focused on harvesting energy from a sustainable source, the Sun, using DSCs with sustainable components. However, the process of photoconversion is only truly sustainable if materials in the DSC can be recycled and/or the device can be regenerated aer degradation processes result in its failure. This aspect of the end-of-life processing of DSCs is not widely investigated, with most effort being invested on the reuse of TiO 2 or reclamation of platinum group metals. Most recently, electrolyte recycling has become of interest, with suggestions that copper-based electrolytes might be more recyclable and sustainable than those based on iodine. A common degradation process in a DSC is the loss of volatile solvent from the electrolyte and extending the lifetime of the device can be addressed by replacing the organic solvent by a less volatile medium. The pros and cons of aqueous-based DSCs have been critically reviewed by Bella, Grätzel and coworkers. 304 An alternative approach is the use of solid-state, ionic liquid or gel electrolytes. 44,305,306 Typical DSCs used in the research laboratory comprise glass/ FTO/TiO 2 and glass/Pt electrodes which are only ca. 2 cm Â 2 cm in size (Fig. 2c). Upscaling to commercial needs demands signicant changes to design, for example the development of exible devices with polymer substrates for roll-to-roll manufacturing. The question of the recyclability of DSC components has been addressed in several recent reviews, [307][308][309] and the recovery of silver from silicon photovoltaics 310 is also relevant for the silver employed in electrical contacts in DSCs.
Bonomo and coworkers highlight several points which are fundamental for approaches to recycling. 309 The rst is the use of critical raw materials e.g. Ru, Co, Ag and Pt, and we have addressed this issue throughout this review. The second is performance degradation arising from electrolyte instability, mainly linked to loss of volatile solvents (see above). The third point concerns the high energy-demanding conducting glass substrates. 309 While recycling of FTO or other transparent conducting oxide (TCO) glass from DSCs may be technically achievable, it is not perceived to be commercially viable in practice. 307 Bonomo and coworkers also comment on the sustainability associated with waste management. They also note that, based upon the quantities of materials used within a DSC, an assessment of sustainability factors should focus more on counter electrode and electrolyte materials than on dyes. 309

Conclusions
DSCs are now an established technology and are destined to contribute to the sustainable energy market. Ideally, all device components should be based upon sustainable materials, and the use of TiO 2 as the semiconductor on the photoanode is in line with these criteria. The use of metal-free (organic dyes) including natural dyes is also benecial and indeed, state-ofthe-art DSCs including certied devices are all based on organic dyes. From the point of view of sustainability, organic dyes provide the most efficient and stable DSCs. The emphasis of this review has been on metal-containing dyes, of which the best performing DSCs include those containing ruthenium(II) sensitizers. However, ruthenium is poorly abundant in the Earth's crust. Although the dye is a critical component of a DSC, the photoconversion efficiency is also dictated by the complementarity of the dye and the redox mediator. Combined with organic, zinc(II) porphyrin, zinc(II) phthalocyanine and ruthenium(II) dyes, one of the most established redox mediators is the I 3 À /I À couple. However, its use limits the values of V OC that can be reached and it creates a corrosive chemical environment within the DSC which impacts upon the long-term stability of the cells. With these specics as a starting point, we have presented developments in the eld of DSCs containing dyes and redox mediators based upon coordination compounds of rst row d-block metal ions. The focus has been primarily on progress over the last decade. First row d-block metal coordination compounds, especially those containing cobalt(III)/cobalt(II), copper(II)/copper(I) and iron(III)/iron(II) couples, have come to the fore as alternative redox mediators to I 3 À /I À . In contrast to the I 3 À /I À couple, the redox potentials of couples based on metal complexes can readily be tuned by use of different ligands. Of particular importance is the fact that use of Cu 2+ /Cu + redox mediators (which are typically bis(diimine) complexes) has led to V OC values in excess of 1000 mV. This is an exciting development which contributes to higher PCEs. However, the attening of the copper(I) tetrahedral geometry upon oxidation leads to the copper(II) state being susceptible to attack by Lewis bases. This includes TBP which is oen added to electrolyte solutions to produce a shi in the conduction band of the semiconductor. The effects of TBP in Cu 2+ /Cu + -containing electrolytes remain to be fully understood.
Turning to the dye and the fact that ruthenium is a critical raw material, we focused in this review on the use of coordination complexes of the more abundant rst row d-block metals, in particular copper, iron and zinc, as dyes in DSCs. A major challenge in these DSCs is an enhancement of their photoconversion efficiencies which currently lag signicantly behind those containing ruthenium-based dyes. Bis(diimine) copper(I) sensitizers exhibit promising PCEs, especially when used as a co-sensitizer with organic dyes such as SQ2, their light-harvesting ranges being complementary (Fig. 10). Both the SALSAC and HETPHEN strategies militate against ligand redistribution that is an inherent problem of bis(diimine)copper(I) complexes in solution. The crustal abundance of iron makes this the dream element on which to base a sensitizer, and the discovery of the long excited-state lifetimes of Nheterocyclic carbene iron(II) complexes has, in the last few years, opened up an exciting eld of iron-based dyes. The steadily increasing PCEs of DSCs incorporating NHC iron(II) dyes have been achieved by systematic tuning of electrolyte compositions, modication of electrode materials, and ligand functionalization. Although the PCEs of these DSCs are only now passing 1%, the rapid improvements in the last few years conrm that this eld is ripe for development.
If the use of DSCs on a commercial scale is to become a truly sustainable endeavour, either the device components must be recycled and/or it must be possible to revive spent devices. These aspects of sustainable DSCs have been reviewed by others and we draw attention to these publications to highlight the needs for the future.

Conflicts of interest
There are no conicts to declare.