Excited state decay of cyclometalated polypyridine ruthenium complexes: insight from theory and experiment

Deactivation pathways of the triplet metal-to-ligand charge transfer ((3)MLCT) excited state of cyclometalated polypyridine ruthenium complexes with [RuN5C](+) coordination are discussed on the basis of the available experimental data and a series of density functional theory calculations. Three different complex classes are considered, namely with [Ru(N^N)2(N^C)](+), [Ru(N^N^N)(N^C^N)](+) and [Ru(N^N^N)(N^N^C)](+) coordination modes. Excited state deactivation in these complex types proceeds via five distinct decay channels. Vibronic coupling of the (3)MLCT state to high-energy oscillators of the singlet ground state ((1)GS) allows tunneling to the ground state followed by vibrational relaxation (path A). A ligand field excited state ((3)MC) is thermally accessible via a (3)MLCT →(3)MC transition state with the (3)MC state being strongly coupled to the (1)GS surface via a low-energy minimum energy crossing point (path B). Furthermore, a (3)MLCT →(1)GS surface crossing point directly couples the triplet and singlet potential energy surfaces (path C). Charge transfer states either with higher singlet character or with different orbital parentage and intrinsic symmetry restrictions are thermally populated which promote non-radiative decay via tunneling to the (1)GS state (path D). Finally, the excited state can decay via phosphorescence (path E). The dominant deactivation pathways differ for the three individual complex classes. The implications of these findings for isoelectronic iridium(iii) or iron(ii) complexes are discussed. Ultimately, strategies for optimizing the emission efficiencies of cyclometalated polypyridine complexes of d(6)-metal ions, especially Ru(II), are suggested.

Due to the fact that the isoelectronic cyclometalated complexes of ruthenium(II) perform very well as sensitizers in dye sensitized solar cells, they have also received increasing interest in the last years. 16,17,19,[107][108][109] Additionally, cyclometalated bridging ligands enhance the electronic coupling between the redox centers in mixed-valent Ru/Ru complexes [110][111][112] and Ru/organic hybrid structures. [113][114][115][116] Despite the large variety of cyclometalated polypyridine ruthenium complexes synthesized up-to-date, however, no phosphorescence comparable to [Ru(bpy) 3  Her key research interests comprise functional complex systems based on coordination and organometallic compounds with special emphasis on molecular wires, light-harvesting systems, bistable systems, emitters, switches and sensors as well as on (biomimetic) catalysts. She has authored more than 130 international refereed papers.
properties between the isoelectronic cyclometalated complexes of iridium(III) and ruthenium(II) is still missing. [117][118][119] Hence, this issue will be addressed in this perspective. We will discuss the excited state deactivation processes of cyclometalated polypyridine complexes beyond the 3 MLCT/ 3 MC and energy gap law schemes by picking illustrative examples from the literature and elaborate why these are weak emitters at room temperature. Several decay pathways of the emissive 3 MLCT state are known for polypyridine ruthenium complexes, and their individual contribution to the excited state decay of cyclometalated polypyridine ruthenium complexes will be discussed ( Fig. 1): Decay path A: tunneling into high-lying vibrational levels of the singlet state. [86][87][88] This channel is always available and its efficiency depends on the Franck-Condon overlap of the vibrational wavefunctions of the 3 MLCT and singlet ground states ( 1 GS).
Decay path B: via a thermally accessible 3 MC state. Population of this state is followed by rapid surface crossing to the singlet potential energy surface via a close-lying minimum energy crossing point (MECP) and vibrational cooling. 56 The energy of the 3 MC state mainly depends on the ligand field strength.
Decay path C: direct surface crossing from the 3 MLCT state to the singlet potential surface via a low-lying surface crossing point. The energy of this MECP depends on the degree of 3 MLCT state distortion and the energy of the 3 MLCT state itself. A shallow 3 MLCT state potential surface around the minimum favours the occurrence of a surface crossing.
Decay path D: via other non-emissive triplet states or states with higher singlet character as in [Ru(tpy) 2 ] 2+ and [Ru(bpy) 3 ] 2+ that allow efficient tunnelling into high-lying singlet states. 77,83,91,120 Decay path E: phosphoresence. In the following, we will divide the discussion into tris-(bidentate) complexes with [Ru(N^N) 2 (N^C)] + coordination sphere and bis(tridentate) complexes with either [Ru(N^N^N) (N^C^N)] + or [Ru(N^N^N)(N^N^C)] + motifs to highlight similarities and differences between the different classes. Finally, we will suggest strategies how to improve the room temperature emission of such complexes and how our conclusions might impact the photophysics of cyclometalated polypyridine complexes of still elusive iron(II) emitters 6,121-123 and wellknown iridium(III) emitters 49,104,124 as well.
Additionally, the shift of the absorption bands induced by the functional groups attached to either the ppy ligand or the bpy units further underlines Grätzel's assignment of all bands as Ru → bpy transitions. 3,107,127,135 For most tris(bidentate) complexes of this type, weak room temperature emission in the range between 720 and 820 nm is reported (Table 1). 107,[127][128][129]133,135 The emissive state is considered to be a Ru → bpy 3 MLCT state. 107,133 This is corroborated by the influence of functional groups on the emission energy. In a series of [Ru(bpy) 2 ( ppy-R)] + complexes ([1e] + -[1h] + , Scheme 1) with functional groups in meta-position to the cyclometalating carbon atom, the emission energy decreases with increasing electron donating strength of the respective substituent as this destabilizes the metal d orbitals (Table 1). 128 Similarly, functionalization of the bpy ligands  with electron accepting substituents such as COOR groups shifts the emission bathochromically as a consequence of the lowered LUMO energy (Table 1). 133 However, missing lowtemperature emission data (λ em , ϕ, τ) currently impede a quantification of the respective effects.
To the best of our knowledge, no quantum yields have been determined neither at room temperature nor at 77 K for any of the reported tris(bidentate) cyclometalated complexes due to their very weakly emissive character and the associated instrumental limitations. 127,128 Castellano and coworkers estimated the phosphorescence quantum yields to be ϕ < 0.005, 127 Table 1). The lifetimes are in the nanosecond range for all complexes and correlate nicely with the emission energy: the excited state lifetimes become smaller with decreasing emission energy. In ( polypyridine)ruthenium(II) complexes, the emissive 3 MLCT state typically is depopulated to some extent via a 3 MC state ( path B, vide supra). 84,[89][90][91] However, cyclometalation substantially increases the 3 MC energy, as pointed out by Dixon 136 and van Koten, 137 efficiently retarding emission quenching via this pathway. 118 In fact, we were able to localize the respective 3 MLCT and 3 MC states of [1b] + via DFT calculations (Fig. 3). The 3 MC state (Mulliken spin population at Ru: 1.87) is located 66 kJ mol −1 above the 3 MLCT level while for [Ru(bpy) 3 ] 2+ , the 3 MLCT-3 MC energy gap was calculated to be −7.7 kJ mol −1 in favour of the 3 MC state. [138][139][140] Remarkably, the nitrogen donor atom N1 of the cyclometalating ligand as well as the trans nitrogen atom N4 are essentially decoordinated in the 3 MC state of [1b] + with Ru-N distances of 2.48 and 2.39 Å, respectively (Fig. 4). This tetragonal distortion along the N1-Ru-N4 axis underlines the strongly dissociative character of the 3 MC state similar to that described for biscyclometalated tris(bidentate) iridium complexes [141][142][143] and for [Ru(bpy) 3 ] 2+ . [138][139][140] It resembles the Jahn-Teller mode of d 7 low-spin Co II complexes due to the (e * g ) 1 electron configuration. 144,145 The transition state between the 3 MC and 3 MLCT state was localized on the potential surface with an energy of 69 kJ mol −1 above the 3 MLCT state (Fig. 3). Hence, in contrast to the isoelectronic complex [Ru(bpy) 3 ] 2+ , the 3 MC state of [1b] + is thermally inaccessible. The comparably low quantum yields of [1b] + are attributed to an increased thermal relaxation into the ground state. Non-emissive excited state decay occurs via vibronic coupling to high-energy oscillators ( path A) or via a thermally activated surface crossing to the ground state potential surface ( path C). 86 The latter requires a low-energy surface crossing point. We attempted to localize such a minimum energy crossing point (MECP) between the 3 MLCT and 1 GS potential surfaces. However, the lowest 3 MLCT → 1 GS MECP we could find was localized at 120 kJ mol −1 above the 3 MLCT level. An analogous 3 MC → 1 GS MECP on the other hand is localized at 72 kJ mol −1 merely 6 kJ mol −1 above the 3 MC state (path B, Fig. 3). However, since all of these states are thermally inaccessible at room temperature or below, surface crossing to the singlet ground state is irrelevant for the excited state deactivation of [1b] + . As a consequence, emission quenching in [1b] + appears to occur exclusively via tunnelling into highenergy oscillators of the ground state ( path A). According to the energy gap law, the vibronic coupling of the 3 MLCT state and 1 GS becomes stronger, the smaller the 3 MLCT-1 GS energy gap is. [86][87][88] Secondly, a more pronounced distortion of the 3 MLCT excited state compared to the ground state geometry increases the non-radiative decay rate as it results in a higher Franck-Condon overlap of the vibronic wavefunctions of the ground and excited state. [86][87][88]127,146,147 Indeed, inspection of the DFT-optimized geometries of the 3 MLCT and 1 GS states reveals a sizable distortion of the former (Fig. 4) allowing for an efficient radiationless deactivation. This distortion is mainly localized at the bpy ligand trans to Ru-C. The Ru-N2 and Ru-N3 bonds are significantly elongated as a consequence of the formal oxidation of ruthenium to +III in the 3 MLCT state and the trans influence of the cyclometalating phenyl ring. However, while the Ru-N2 and Ru-N3 bonds are elongated, the formal negative charge on the second bpy ligand compensates the repulsion yielding essentially unaltered Ru-N4 and Ru-N5 bond lengths.
In summary, the high energy of the 3 MC state is very favourable for efficient emitters because it eliminates one pathway for exited state deactivation and concomitantly prevents photodecomposition reactions, that typically occur from the dissociative 3 MC state. 90,148 However, in order to increase phosphorescence quantum yields of cyclometalated ( polpyridine)ruthenium, radiationless deactivation via vibronic coupling has to be suppressed. This can be approached in two ways: the distortion of the 3 MLCT state compared to the 1 GS has to be reduced and the emission energy has to be blueshifted as far as possible. Chou and coworkers provided a beautiful example successfully implementing both approaches. 130 Clever molecular design yielded systems with Ru → ppy MLCT states as lowest triplet excited states. This is straight-forwardly accomplished by making the cyclometalating ppy − ligand the strongest π-acceptor in the complex. Chou and coworkers achieved this with carbon monoxide and phosphanes which are rather poor π-acceptors towards Ru II in biscyclometalated complexes of the type Ru(bq) 2 (CO)(PPh 2 Me) 3 (bqH = benzo-[h]quinoline, Scheme 1). 130 As the LUMO of bq − is much higher in energy than that of bpy (Fig. 2), the emission from the corresponding Ru → bq 3 MLCT state is blueshifted substantially to 575 nm with a quantum yield of ϕ = 0.24. Additionally, as the excited state involves a cyclometalating ligand, its distortion compared to the ground state geometry should be less pronounced as in complex [1b] + . Similar observations were made for the isoelectronic osmium complexes. 149 These findings suggest that with careful choice of suitable, very weakly π-accepting polypyridine ligands, cyclometalated ( polypyridine)ruthenium complexes with similar emission behaviour arising from high-energy Ru → ppy 3 MLCT states are accessible. Another way of improving the emission behaviour of cyclometalated ( polypyridine)ruthenium complexes could be by introducing tridentate chelate ligands as this potentially supresses excited state distortions while maintaining the high energies of parasitic 3 MC states. This should yield nested states with poorer Franck-Condon overlap (weakly coupling limit) which reduces tunnelling processes into high-energy singlet states ( path A). We will discuss the possibilities and consequences of bis(tridentate) coordination spheres on the phosphorescence properties of cyclometalated ruthenium complexes in the next section.

Bis(tridentate) ruthenium complexes
Considering cyclometalated bis(tridentate) complexes, a distinction between [Ru(N^N^N)(N^C^N)] + and [Ru(N^N^N) (N^N^C)] + coordination environments is reasonable. The next two sections will highlight similarities between the two classes of bis(tridentate) complexes as well as important differences and compare these findings to those concerning tris(bidentate) complexes (vide supra).

The [Ru(N^N^N)(N^C^N)] + coordination sphere
Similar to tris(bidentate) complexes, the visible range of the absorption spectrum of bis(tridentate) complexes with central cyclometalation is dominated by intense and broad absorption bands that have been assigned to MLCT transitions. Again, two bands are observed, one in the range of 350-450 nm and a second between 470 and 650 nm. Van Koten 137,150 and Berlinguette 107 assigned the high energy MLCT band to Ru → dpb transitions and the low-energy band to Ru → tpy excitations based on relative orbital energies of the lowest π*-orbitals of the respective ligands (Fig. 2). Schubert and coworkers 117,151 on the other hand assigned the low-energy band to mixed MLCT and ligand-to-ligand charge transfer (LL′CT) transitions arising from a HOMO-LUMO transition while the blue absorption band was attributed to mixed MLCT/LL′CT and MC transitions. We examined several [Ru(dpb-R 1 )(tpy-R 2 )] + complexes and demonstrated experimentally (resonance Raman spectroscopy) and theoretically that the low energy absorption arises from both Ru → tpy and Ru → dpb excitations (Fig. 2). 118,152 Excitation into the tpy-centered LUMO is only possible from the HOMO−1, while the dpb-centered LUMO+2 can be reached from the HOMO as evidenced from timedependent DFT calculations. Both excitations occur at very similar energy and contribute comparably to the absorption band at 500 nm. The higher energy absorption features result from MLCT transitions targeting the higher π* orbitals of the tpy ligand. In fact, a LL′CT transition as suggested by Schubert 117 is symmetry-forbidden as it involves two mutually perpendicular π orbitals. This symmetry argument will become important for the emission properties as well.
Weak emission is observed under ambient conditions for most bis(tridentate) complexes with central cyclometalation in the range between 700 and 800 nm while the parent [Ru (tpy) 2 ] 2+ complex is virtually non-emissive at room temperature. 107,109,118,119,137,151 Van Koten 137 and we 118,119 determined extremely low quantum yields in the range of 10 −5 for complexes of type [Ru(dpb-R 1 )(tpy-R 2 )] + . In contrast to the tris-(bidentate) series, however, no excited state lifetimes were reported so far ( Table 2). Our attempts to obtain lifetimes suggested that they are in the picosecond range. 119 Van Koten 137 studied complexes of the type [Ru(dpb-R 1 ) (tpy-R 2 )] + bearing carboxy-substituents on either or both ligands (Scheme 2, [4a] + -[4d] + ). The effect of the functional groups on the respective emission energy ( Table 2) points to a 3 MLCT emissive state. In fact, a COOR substituent at the tpy ligand ([4b] + ) leads to complexes that are non-emissive at room temperature due to its stabilizing influence on the tpy-centered LUMO. Since the phosphorescence quantum yields decrease with increasing emission wavelength following the energy gap law, van Koten suggested vibrational relaxation as main source of emission quenching. 137 We extended this study by also introducing electron-donating substituents such as NH 2 and NHCOMe to the dpb ligand (Scheme 2, [5a] + -[5c] + ) and the tpy ligand ([5d] + ). 118,119 Using theoretical methods we showed, that the emissive state is in fact a 3 MLCT state. The two singly occupied orbitals (SOMOs) in this state correspond to the d xy (HOMO−1) and the tpy LUMO, respectively (Fig. 2). No dpb ligand participation was observed. However, we localized a second charge transfer state with LL′CT character (dpb → tpy, HOMO → LUMO orbital parentage) on the triplet potential energy surface. The mutually perpendicular SOMOs make this state spectroscopically undetectable and non-emissive (dark state). As third type of triplet states, the 3 MC states were found (Fig. 5 and Table 3). While the 3 LL′CT state is essentially undistorted compared to the ground state with only a small displacement of ruthenium towards C1 and a concomitant elongation of the Ru-N3 bond, the 3 MLCT state exhibits substantial distortions within the tpy ligand. It is inclined with respect to the plane perpendicular to the dpb ligand by 12° (Fig. 5). At the same time the Ru-C and Ru-N5 bonds are slightly elongated while the Ru-N2 bond shortens (Fig. 4). In the 3 MC state on the other hand, the peripheral pyridine rings of the tpy ligand are tilted away from the metal center forming dihedral angles of about 11°with the central pyridine ring, respectively. Furthermore, the all Ru-N tpy bonds Ru-N1, Ru-N3 and Ru-N4 are substantially elongated (Fig. 4) underlining the dissociative character of the 3 MC state.
The calculated relative energies of the triplet states (Fig. 5) of [5a] + -[5d] + as well as the transition states connecting them are summarized in Table 3. The emissive 3 MLCT state is flanked by two thermally accessible quenching states, namely the 3 LL′CT and 3 MC states. The lowering of the relative 3 MC energy from 66 kJ mol −1 in [1b] + to 10-30 kJ mol −1 in these bis-(tridentate) complexes is attributed to the smaller N-Ru-N bite angles and the weaker overlap of the nitrogen lone pairs with the e g metal orbitals which results in a smaller ligand field splitting. The distortion of the 3 MC state allows for tunnelling to the 1 GS and for a surface crossing point (MECP) that is just 9 kJ mol −1 above the 3 MC level providing an accessible non-emissive deactivation channel for the 3 MLCT state (path B, Fig. 5). Interestingly, a direct 3 MLCT → 1 GS MECP was found for [4a] + as well at a moderate energy (26 kJ mol −1 above the 3 MLCT state). It is qualitatively similarly distorted as the 3 MLCT state but the degree of the distortion is larger. Thus, the geometry of this crossing point can be regarded as a highamplitude distortional vibration along the 1 GS → 3 MLCT vibrational mode ( 3 MLCT → 1 GS reaction coordinate, Fig. 5).
Hence, the 3 MLCT distortion of [4a] + opens up this low-energy deactivation channel ( path C), that is absent for [1b] + . However, experimental evidence for such a quenching channel is difficult to obtain as its activation barrier is similar to that of the 3 MC deactivation channel and hence a similar temperature-dependent emission behaviour is expected.
The 3 LL′CT state is connected to the 3 MLCT state via a transition state with a very low activation barrier ( Table 3). As the 3 LL′CT state is barely distorted compared to the 1 GS geometry (vide supra) it is considered a nested state. Indeed, attempts to localize a 3 LL′CT → 1 GS MECP, that would provide a non-emissive decay channel, failed. Because emission from the 3 LL′CT state is symmetry-forbidden, its only decay pathway proceeds via tunnelling into the vibrationally excited singlet state followed by thermal relaxation ( path D, Fig. 5).
Hence both, the 3 LL′CT and 3 MC states (and potentially also the 3 MLCT → 1 GS MECP) are responsible for the efficient phosphorescence quenching at room temperature. This DFTbased interpretation was evidenced experimentally by recording the temperature dependence of the quantum yield. The respective ln(ϕ) vs. T −1 data of [4c] + and [5a] + were reproduced using a fit function accounting for two thermally activated de-  activation pathways. The activation barriers obtained from the fit are in excellent agreement with the computed transition state energies (Table 3). Schubert and coworkers 117 published very similar results on structurally related complexes such as [Ru(dtp)(tpy)] + (dtbH = 1,3-di-(1,2,3-triazol-4-yl)benzene, Scheme 2, [6] + ). Two Arrhenius-like activation parameters were required to properly reproduce the temperature-dependent lifetime data yielding very similar activation energies compared to our findings for [4c] + and [5a] + . In contrast to our assignments, however, they attributed the two deactivation channels of [6] + to an irreversible 3 MLCT → 3 MC surface crossing and an internal conversion to a higher-lying MLCT state of increased singlet character. The latter is a common feature of non-cyclometalated ( polypyridine)ruthenium(II) complexes. 77,83,91,153 In the light of our results and the electronic similarity of the studied structures, it seems plausible that the second deactivation channel in [6] + actually is via a state of 3 LL′CT nature as well. In fact, their DFT-optimization of a triplet state of [6] + afforded a 3 CT state with orthogonal SOMOs. 117 Even if the cyclometalating ligand does not contribute significantly to the spin density of this state, its emissive relaxation is symmetry-forbidden.
To summarize, the combination of orthogonal ligands with strongly differing electronic properties, one being an excellent π-acceptor and the second a strong π-donor typically yields a low-lying 3 CT state. Even though this state is not directly populated after optical excitation into a 1 MLCT state and subsequent intersystem crossing onto the triplet manifold it serves as a further low-barrier channel for radiationless deactivation of the emissive 3 MLCT state ( path D) besides the 3 MC state ( path B). Recent reports underline that these results are transferable to structurally similar complexes of other transition metals with orthogonal tridentate ligands. In fact, Williams and coworkers suggested that a 3 LL′CT state plays a key role in the excited state deactivation of isostructural and isoelectronic [Ir(N^N^N)(N^C^N)] 2+ complexes. 154 A beautiful theoretical study on the excited state deactivation pathways in biscyclometalated gold(III) complexes of the general formula [Au(C^N^C)(acetylide)] revealed a 3 LL′CT state which efficiently contributed to the radiationless deactivation of the emissive state. 56,57,155 Its substantial distortion compared to the ground state increased the rate of non-radiative decay.
In order to optimize the emissive properties of bis(tridentate) cyclometalated ruthenium complexes, circumventing the low-energy non-emissive 3 LL′CT state is a key objective, for example by removing the axial symmetry. This can be achieved using a N^N^C coordination mode on the cyclometalating ligand which will be discussed in the next section (vide infra). However, the kinetic parameters extracted by Schubert and coworkers 117 from the temperature-dependent lifetime data suggest, that at room temperature the 3 LL′CT state ( path D) with an activation barrier ΔE 2 of just 4.2 kJ mol −1 is only responsible for the quenching of about 25% of the excited molecules of [6] + and structurally similar complexes. At the same time, the 3 MC state with a considerably higher barrier ΔE 1 of 21.9 kJ mol −1 is responsible for about 75% of the excited state deactivation ( path B and potentially path C), while direct radiationless decay into the ground state only contributes 0.1% ( path A). Hence, even avoiding the 3 LL′CT state by clever molecular design will not per se yield strong emitters. The strong electronic coupling of the 3 MLCT and 3 MC states [k 0 ( 3 MLCT → 3 MC) = 10 11 -10 13 s −1 ; k 0 ( 3 MLCT → 3 LL′CT) ≈ 10 8 s −1 ] 117 renders the former the dominant deactivating state despite the substantially higher activation barrier. The appreciably weaker coupling of the 3 MLCT and 3 LL′CT states can be traced back to the two states still being electronically nearly orthogonal (vide supra). Concluding, despite the occurrence of a 3 LL′CT state in this kind of bis(tridentate) complexes the important states to manipulate for improving emission efficiencies remain the 3 MC and 3 MLCT states. Some wellthought-out examples have been provided in the recent literature employing ligand bite angle manipulation 94,95,156,157 and push-pull concepts 93,157,158 to increase the 3 MLCT-3 MC gap in bis(tridentate) ruthenium complexes which in principle are applicable to cyclometalated complexes as well. However, these conceptual approaches are beyond the scope of this perspective. Recently, Dixon and coworkers 159-161 suggested on a computational basis, that two cyclometalating sites in cis-position could be beneficial to increase the ligand field splitting of iron(II) complexes and provide a tool for controlling the relative 3 MLCT and 3 MC energies. This concept should be transferable to cyclometalated ruthenium complexes as well (cf. 3) although it is likely accompanied with synthetic challenges. 105,130,162,163 An alternative approach could again involve attaching very weakly π-accepting ligands trans to a N^C^N ligand to yield potentially highly luminescent Ru → N^C^N 3 MLCT states.

The [Ru(N^N^N)(N^N^C)] + coordination sphere
The absorption spectrum of [Ru(tpy)( pbpy)] + type complexes resembles that of [Ru(bpy) 2 ( ppy)] + with two absorption bands, one around 400 nm, and a second around 500-600 nm. Again, the low-energy band is composed of MLCT transitions both involving the tpy and the cyclometalating ligand. However, the π* orbital of the coordinating phenyl ring (LUMO+3) is not involved in any of the low-energy transitions, as its energy is substantially higher than the frontier orbitals (Fig. 2). 107 Due to the near-degeneracy of the three lowest unoccupied orbitals (LUMO-LUMO+2), the absorption band at 500 nm is markedly sharper than that of [Ru(dpb)(tpy)] + and [Ru(bpy) 2 ( ppy)] + complexes. The feature around 400 nm is dominated by an intense Ru → phenyl (LUMO+3) transition (Fig. 2). 107 Unfortunately, accounts on emission properties of these [Ru(tpy)( pbpy)] + complexes are very limited. 107 3 MLCT state. This is accompanied by an altered energy ordering of the lowest unoccupied orbitals and reflects their energetic similarity (Fig. 2). This example highlights that in [Ru(N^N^N)(N^N^C)] + type complexes Ru → N^N^C 3 MLCT states are obtained by tuning the respective frontier orbital energies of the ligands. However, the cyclometalating phenyl ring does not contribute as π-accepting moiety and the excited electron is entirely localized on the bipyridine fragment of the pbpy ligand. Consequently, the phosphorescence efficiency is not affected significantly (Table 4).
Interestingly, Berlinguette and coworkers reported on a series of [Ru(tpy)( pbpy)] + based complexes with diarylamine groups appended via a thiophene linker such as [8b] + . These are highly emissive (quantum yields in the range of 0.1-0.3) in some cases, but emit at much higher energy than typically observed for these complexes. 164 In fact, the analogous thiophene substituted complex [8a] + without the diarylamine functionality lacks the strong emission. 134 However, Berlinguette showed that the emissive behaviour of [8b] + actually arises from a singlet intraligand charge transfer state ( 1 ILCT) involving the diarylamine unit as electron donor and the polypyridine moiety as electron acceptor. 165,166 An identical emission energy was observed for the free ligand with even higher fluorescence quantum yields (ϕ = 0.91, τ = 3.4 ns) explaining the untypically high emission energy and quantum yield of [8b] + . [164][165][166] To get a better understanding of the states involved in the excited state deactivation of [Ru(tpy)( pbpy)] + complexes, we studied the triplet potential energy surface of [7a] + using DFT calculations. Inspecting the 3 MLCT state geometry and spin density of [7a] + reveals a striking similarity to [4a] + (Fig. 4, 5  and 6). In fact, a similar distortion of the tpy ligand with an offset central pyridine ring is found in both cases (vide supra). Additionally, the bond length changes of the 3 MLCT states compared to the respective 1 GS geometries of [4a] + and [7a] + are very similar (Fig. 4). Given the similar quantum yields of the isoelectronic classes of complexes [Ru(dpb)(tpy)] + and [Ru (tpy)( pbpy)] + this suggests excited state deactivation channels with similar barriers are dominant in both cases. However, the 3 MC state ( path B) is found to be 60 kJ mol −1 ( 3 MLCT → 3 MC transition state at 62 kJ mol −1 ) higher in energy than the 3 MLCT state and thus it is thermally inaccessible at room temperature. As a consequence, its contribution to the excited state deactivation of [7a] + is negligible. The marked increase of the 3 MC-3 MLCT energy gap by about 30 kJ mol −1 by exchanging N^C^N by N^N^C chelate ligands in bis(tridentate) complexes (Fig. 6) was also found for the isoelectronic iron(II) complexes by Dixon and coworkers. 161 They argued that the cyclometalating ligand does not only act as a strong σ-donor but also as a π-donor. In the iron(II) complex [Fe(dpb)(tpy)] + the π-donor strength is the dominant influence yielding a net reduction of the effective ligand field strength and hence a stabilization of the 3 MC state compared to the non-cyclometalated complex [Fe(tpy) 2 ] 2+ . In [Fe(tpy)( pbpy)] + on the other hand, the π-overlap between the peripheral cyclometalating phenyl ring and the metal d orbitals is not as pronounced. As a consequence of the σ-overlap an increased ligand field splitting and a higher 3 MC energy compared to [Fe(dpb)(tpy)] + are calculated. At the same time, the 3 MLCT energies of [Fe(tpy) ( pbpy)] + and [Fe(dpb)(tpy)] + are essentially identical yielding an overall higher 3 MLCT-3 MC energy gap by about 30 kJ mol −1 for [Fe(tpy)( pbpy)] + . As for ruthenium, the d orbitals are more diffuse than for iron, the destabilization of the t 2g orbitals via π-donor interactions is much less pronounced yielding 3 MC states well above the 3 MLCT level in all cyclometalating complexes, but the same principles apply explaining the trends we observe for the isoelectronic ruthenium complexes. 161 In principle, a 3 LL′CT state similar to that described for [Ru (dpb)(tpy)] + complexes (vide supra) is also conceivable for complexes of the [Ru(N^N^N)(N^N^C)] + class. However, an analogous symmetry restriction as discussed above for the former does not apply in this case due to the lowered molecular symmetry. Although we tried to localize such a 3   Since the 3 MLCT state of [7a] + is electronically very similar to the 3 MLCT state of the tris(bidentate) complex [1b] + , an argumentation based on emission quenching via vibronic coupling to the ground state ( path A) is insufficient to account for the substantially lower emission quantum yields of the former (10 −6 -10 −5 as compared to 10 −4 -10 −3 ). Additionally, deactivation channels via low-lying 3 MC ( path B) or 3 LL′CT states (path D) as found for [Ru(dpb)(tpy)] + complexes do not contribute to the efficient non-emissive excited state decay of [Ru (tpy)( pbpy)] + complexes. A surface crossing point between the 3 MLCT and 1 GS potential energy surfaces ( path C), however, similar to that found for [4a] + (Fig. 5), would provide a concise explanation for the marked difference between the tris(bidentate) and bis(tridentate) complexes. Indeed, we localized a thermally accessible 3 MLCT → 1 GS surface crossing point that is only 29 kJ mol −1 higher in energy than the 3 MLCT state (Fig. 6). Remarkably, the geometry and energy of this crossing point is similar to that of the 3 MLCT → 1 GS MECP of [4a] + . Again, the distortion of the 3 MLCT state provides an excited state deactivation pathway for polypyridine ruthenium complexes. This finding has some predictive value as well. The 3 MC state is thermally inaccessible at temperatures below 298 K in [7a] + and does not contribute to the excited state decay. Hence, the temperature dependence of the emission of [7a] + can provide information on the contribution of a minimum energy surface crossing point in proximity to the relaxed 3 MLCT geometry to the emission quenching. An increasing emission intensity upon cooling would support this hypothesis. Additionally, as the dissociative anti-bonding 3 MC state is out of reach at room temperature, no photosubstitution reactions should occur for [7a] + in contrast to [Ru(bpy) 3 ] 2+ which is very prone to such reactivity. 90 We will devote future work into elucidating these predictions.
In conclusion, it does not suffice to reduce the molecular symmetry and circumvent the parasitic 3 LL′CT state to increase phosphorescence quantum yields in bis(tridentate) cyclometalated complexes. Bis(tridentate) [Ru(N^N^N)(N^N^C)] + complexes with peripheral cyclometalation suffer from the same distortion and low energy of the 3 MLCT state as the analogous tris(bidentate) complexes. As the relative 3 MC state energy of [Ru(tpy)( pbpy)] + complexes is substantially higher than for comparable [Ru(N^N^N)(N^C^N)] + complexes, its contribution to the excited state decay can be neglected ( path B). Hence, similar strategies are applicable for increasing the luminescence quantum yields as suggested before for tris(bidentate) complexes. These should focus on reducing the excited state distortion yielding a nested emissive state and shutting down the deactivation via direct 3 MLCT → 1 GS surface crossing (path C). This might be achieved by shifting the LUMO to higher energies and making the π* orbital of the cyclometalating moiety the acceptor site of the lowest 3 MLCT state. At the same time, a lower excited state distortion would shift the energy of the 3 MLCT → 1 GS MECP to higher energies as well. This could potentially be accomplished by a combination of a very weakly π-accepting spectator ligand with a cyclometalating ligand that also contains a weakly π-accepting site such as an N-heterocyclic carbene. 167,168 Experimental section Density functional theory calculations DFT calculations were carried out using the ORCA program package (version 3.0.2). 169 Tight convergence criteria were chosen for all calculations (Keywords TightSCF and TightOpt). All calculations were performed using the hybrid functional B3LYP 170 and employ the RIJCOSX approximation. 171,172 Relativistic effects were calculated at the zeroth order regular approximation (ZORA) level. The ZORA keyword automatically invokes relativistically adjusted basis sets. 173 To account for solvent effects, a conductor-like screening model (COSMO) modelling acetonitrile was used in all calculations. 174 Geometry, transition state and minimum energy crossing point optimizations were performed using Ahlrichs' split-valence double-ζ basis set def2-SV(P) which comprises polarization functions for all non-hydrogen atoms. 175,176 Optimized geometries were confirmed to be mimina or first-order saddle points by subsequent frequency analysis (n imag = 0 or 1, respectively). Surface crossing geometries were subjected to SurfCrossNum-Freq calculations to confirm that they are minima in the 3N-7 dimensional subspace excluding the surface crossing reaction coordinate. Computed free Gibbs enthalpies were used to compare the relative energies of all structures. Explicit counterions and/or solvent molecules were not taken into account in any case. The 3 MLCT and 3 LL′CT states were localized by triplet geometry optimizations from the optimized 1 GS geometry. 3 MC states were found by elongating two opposite Ru-N bonds to 2.40 Å and subsequent geometry optimizations. Transition state optimizations were started from geometries obtained by averaging all coordinates of the starting and final state geometries using the exact Hessian matrix of the initial geometry. MECP geometries were obtained by starting at the respective optimized triplet state geometry ( 3 MLCT geometry for 3 MLCT-1 GS MECP, 3 MC geometry for 3 MC-1 GS MECP).

Conclusions
Cyclometalated polypyridineruthenium(II) complexes with N 5 C coordination sphere typically exhibit very weak room temperature emission in the near infrared range (700-800 nm) of the electromagnetic spectrum. The reasons for the weak emission are various and depend on the chelate coordination sphere around the metal center. In tris(bidentate) [Ru(N^N) 2 (N^C)] + complexes, the 3 MC state ( path B) that is typically a major channel for excited state decay in polypyridineruthenium(II) complexes, 84,89,90 does not contribute as it is thermally inaccessible (ΔE( 3 MLCT → 3 MC) = 60-70 kJ mol −1 ) at room temperature. Furthermore, no low energy 3 MLCT → 1 GS surface crossing point was found ( path C) which suggests that tunnelling into high-energy vibrationally excited singlet states is the main channel of excited state deactivation (path A).
In contrast, the emission quenching of [Ru(N^N^N) (N^C^N)] + complexes is dominated by two thermally accessible triplet states, that flank the emissive 3 MLCT state, namely the 3 MC state ( path B, ΔE( 3 MLCT → 3 MC) = 10-30 kJ mol −1 ) and a 3 LL′CT state ( path D, ΔE( 3 MLCT → 3 LL′CT) < 10 kJ mol −1 ). [117][118][119] The 3 LL′CT state is a peculiarity of C 2 -symmetric cyclometalated complexes and provides a second, unprecedented non-emissive deactivation channel. Additionally, a 3 MLCT → 1 GS surface crossing point provides another decay channel ( path C) whose contribution yet needs to be quantified. These three channels B, C and D are responsible for almost 100% of the emission quenching. Although significantly faster than the emission process itself, direct non-emissive decay via 3 MLCT → 1 GS tunnelling ( path A) only plays a subordinate role simply because excited state decay via paths B, C and D is so efficient. This is reflected by the very low excited state lifetimes below the nanosecond range.
In [Ru(N^N^N)(N^N^C)] + complexes, the 3 MC states are thermally inaccessible (ΔE( 3 MLCT → 3 MC) ≈ 60 kJ mol −1 ) and no quenching 3 LL′CT states are relevant. However, the triplet and singlet potential surfaces intersect close to the relaxed 3 MLCT state providing an efficient deactivation channel ( path C) with an activation barrier of only about 30 kJ mol −1 . This explains the similarly low emission quantum yields of bis(tridentate) complexes with central and peripheral cyclometalating site despite the markedly different triplet states relevant to the two systems. However, the amount of emission quenching via tunneling ( path A) in [Ru(N^N^N)(N^N^C)] + complexes remains unclear until temperature-dependent lifetime data are acquired.
Strategies for increasing the phosphorescence quantum yields are proposed. In [Ru(N^N) 2 (N^C)] + complexes emission quenching is dominated the very low emission energies and the efficient tunnelling into high-energy singlet states following the energy gap law ( path A). Hence, improving the emission efficiency is very challenging and only achieved by structurally restraining the already small excited state distortions or increasing the emission energy drastically. The latter is possible by making the cyclometalating ligand the π-accepting site of the 3 MLCT state as shown by Chou and coworkers. 130 In bis(tridentate) cyclometalated ruthenium complexes, emission quenching predominantly arises from the distortion of the 3 MLCT state compared to the ground state. The triplet potential energy surface is relatively flat around the 3 MLCT geometry leading to a 3 MLCT/ 1 GS surface intersection less than 30 kJ mol −1 above the emissive 3 MLCT state giving rise to a deactivation channel via a direct 3 MLCT → 1 GS surface crossing. Minimizing the excited state distortion via structural constraints could circumvent this channel. Additionally, by making the cyclometalating ligand the π-accepting site within the 3 MLCT state via tuning the energy levels of the lowest unoccupied molecular orbitals, a substantial increase of the 3 MLCT state energy can be achieved which would be beneficial for suppressing the 3 MLCT → 1 GS surface crossing as well. This will, however, also shift the emission well into the visible range of the electronic spectrum. Furthermore, it is crucial to avoid orthogonal π-donor and π-acceptor sites in trans position across the metal center as in [Ru(N^N^N)(N^C^N)] + because this inherently invokes orthogonal, non-emissive 3 LL′CT states as quenching channels and also lowers the energy of the 3 MC states into a thermally accessible region due to a lower ligand field splitting.
In order to elucidate the excited state properties of cyclometalated complexes, temperature-dependent excited state lifetime or emission quantum yield measurements provide an invaluable tool. 71,[117][118][119] Additionally, quantum chemical approaches can deliver lots of information about energies and geometries of relevant excited states. Tong, Che and coworkers (vide supra) demonstrated this on luminescent cyclometalated gold(III) complexes. 155 They assessed the radiative and nonradiative decay rates from a computational standpoint and quantified key processes that yield or prevent efficient emission in these species. Similarly, Dixon and coworkers studied mono-and bis(cyclometalated) iron(II) complexes using DFT calculations. [159][160][161] The 3 MC state in [Fe( pbpy)(tpy)] + with peripheral cyclometalation is substantially higher in energy than in [Fe(dpb)(tpy)] + with central cyclometalation, very similar to the results presented here for the ruthenium homologues. Furthermore, they highlighted, that bis(cyclometalated) iron(II) complexes such as [Fe(dpb)( pbpy)] and [Fe( pbpy) 2 ] have very low-lying 3 MLCT states that are, in the case of [Fe(dpb)( pbpy)], only marginally distorted compared to the ground state geometry. We suggest that these findings apply to the analogous ruthenium complexes potentially opening a route to highly luminescent near-IR emitters. However, since only very few bis (cyclometalated) polypyridine ruthenium complexes are known so far 130,162,163 and none of them contain tridentate ligands, the synthesis of Ru(dpb)( pbpy) and Ru( pbpy) 2 complexes might be challenging. We will devote future work to the design and synthesis of ruthenium-based emitters with cyclometalating ligands to improve and exploit their excited state properties.