P∩N Bridged Cu(I) Dimers Featuring Both TADF and Phosphorescence. From Overview towards Detailed Case Study of the Excited Singlet and Triplet States †

We present an overview over eight brightly luminescent Cu(I) dimers of the type Cu2X2(P∩N)3 with X = Cl, Br, I and P∩N = 2-diphenylphosphino-pyridine (Ph2Ppy), 2-diphenylphosphino-pyrimidine (Ph2Ppym), 1-diphenylphosphino-isoquinoline (Ph2Piqn) including three new crystal structures (Cu2Br2(Ph2Ppy)3 1-Br, Cu2I2(Ph2Ppym)3 2-I and Cu2I2(Ph2Piqn)3 3-I). However, we mainly focus on their photo-luminescence properties. All compounds exhibit combined thermally activated delayed fluorescence (TADF) and phosphorescence at ambient temperature. Emission color, decay time and quantum yield vary over large ranges. For deeper characterization, we select Cu2I2(Ph2Ppy)3, 1-I, showing a quantum yield of 81%. DFT and SOC-TDDFT calculations provide insight into the electronic structures of the singlet S1 and triplet T1 states. Both stem from metal+iodide-to-ligand charge transfer transitions. Evaluation of the emission decay dynamics, measured from 1.2 ≤ T ≤ 300 K, gives ∆E(S1-T1) = 380 cm−1 (47 meV), a transition rate of k(S1→S0) = 2.25 × 106 s−1 (445 ns), T1 zero-field splittings, transition rates from the triplet substates and spin-lattice relaxation times. We also discuss the interplay of S1-TADF and T1-phosphorescence. The combined emission paths shorten the overall decay time. For OLED applications, utilization of both singlet and triplet harvesting can be highly favorable for improvement of the device performance.


Syntheses and Structural Characterization
Previously, we have reported on the preparation of complexes 1-X (X = Cl, Br, I) and 3-I (Scheme 1) [34]. In the present work, we additionally report on the new complexes 2-X (with X = Cl, Br, I) and 3-Br. They were prepared analogously by reactions of the respective ligands 2 and 3 [97] with copper(I) halides in dichloromethane. For ligand 3, the pure chloride complex could not be obtained. 1-X and 2-X give yellow, while 3-X red powders. Once precipitated, the complexes are only slightly soluble in standard, weakly or non-coordinating solvents, hence, no NMR spectra could be measured. Therefore, the complexes were characterized exclusively by elemental analysis. In addition, for complexes 1-Br, 2-I and 3-I the crystal structures could be determined. Single crystals suitable for X-ray diffraction could be obtained by slow gas-phase diffusion of diethyl ether into the filtered reaction solution of 1-Br, 2-I and 3-I. To our knowledge, no bromide of this class of di-nuclear copper compounds has been structurally characterized. Therefore, the structure of 1-Br provides the first structural data and completes the series of accessible halides. Together with the crystal structures of 1-Cl and 1-I reported in reference [34], it is now possible to compare all three halides of a homologous series of complexes of the type Cu 2 X 2 (P∩N) 3 . All so far structurally characterized complexes of this type have in common a butterfly-shaped Cu 2 X 2 core surrounded by three P∩N ligands (Figure 1). Two ligands coordinate exclusively via the phosphorus atom to Cu(I), while the third ligand is bound in a bridging manner with both the nitrogen and phosphorus atoms to two Cu(I) centers. Interestingly, not all halides 1-X are isostructural (Table 6 below and [34]): 1-Br/Cl crystallize in the monoclinic space group P2 1 /n, whereas 1-I is triclinic P1. Nevertheless, all structures are similar and show the expected trend for a series chloride → iodide, with only very small differences between 1-Cl and 1-Br (Table 1). For example, the Cu-Cu distances are almost identical for 1-Cl (2.878(1) Å) and 1-Br (2.883(1) Å), whereas 1-I features a shorter Cu-Cu distances of 2.7694(5) Å. On first sight, it seems a contradictory observation that the biggest anion causes the smallest intermetallic separation. However, this fact is due the most acute angles Cu1-X-Cu2 of around 62 • C for 1-I, whereas for 1-Cl, we find around 73 • C. The respective angle for the bromide 1-Br lies between these values (≈69 • C). Together with a significant longer Cu-X bond distance (Cu-Cl: 2.39-2.44; Cu-Br: 2.51-2.57; Cu-I: 2.65-2.81 Å) this leads to the shortest Cu-Cu distance for 1-I. All Cu-Cu distances are around or slightly above the sum of the van-der-Waals radii (r(Cu) = 1.40 Å [98]) indicative of only neglectable cuprophilic interactions.
Although the isoelectronic pyridyl and pyrimidyl moieties are expected to have similar steric characteristics, the structure of 2-I (monoclinic) is not isostructural to 1-I (triclinic) [34]. 3-I crystallizes as the dichloromethane solvate. It should be noted that another form of complex 3-I has been reported previously which does not differ considerably in its structural parameters (compare Table 1) [34]. For all halides and P∩N ligands summarized in Table 1, the Cu-P and Cu-N distances lie in a very narrow range of 2.24-2.29 Å and 2.08-2.14 Å, respectively.

Computational Investigations
An overview of the emission data, as presented below, shows that the compound Cu 2 I 2 (Ph 2 Ppy) 3 , 1-I ( Figure 2) exhibits particularly interesting properties with respect to a distinct combination of TADF and phosphorescence and shows high emission quantum yield at short emission decay time. Accordingly, we will discuss this compound's photophysical properties in deeper detail below. In this section, we first investigate this material by computational methods to shed light on the electronic properties. Initially, the compound was optimized in the singlet ground state. The calculated results reveal good agreement with the data from the X-ray structure ( Table 2). Although the Cu-Cu bond length is slightly overestimated, the general structural motifs like the butterfly shape of the Cu 2 I 2 core with an equilateral CuI 2 triangle are nicely reproduced. Further geometry optimization was performed in the lowest triplet state using unrestricted DFT. In a next step, time-dependent DFT (TDDFT) calculations with and without selfconsistent spin-orbit coupling using the ZORA Hamiltonian [99] (SOC-TDDFT [100]) were carried out for the S 0 and T 1 geometries with ADF2014 [101] solving for the lowest six spinmixed excitations. Due to computational constraints, the basis set for these calculations was chosen to be DZP, a double-zeta plus polarization basis set [102]. The TDDFT calculations at the S 0 geometry show that the S 1 and T 1 states correspond to a nearly pure excitation (98% and 97%) from the highest occupied molecular orbital (HOMO) to the lowest unoccupied molecular orbital (LUMO) as similarly reported in [34]. As depicted in Figure 3, the HOMO is localized on the Cu 2 I 2 core with significant contributions of the iodides, while the LUMO is localized on the bridging organic ligand. This leads to a classification of the HOMO-LUMO transition as being of (I+M)LCT character, abbreviated shortly as metal-to-ligand charge transfer (MLCT) transition.  3 1-I as obtained from DFT calculations (B3LYP/TZVP) for the S 0 and T 1 minimum compared to the values from the X-ray structure determination. The atom numbering scheme is analogous to Cu 2 Br 2 (Ph 2 Ppy) 3 1-Br shown in Figure 1.  Taking spin-orbit coupling (SOC) into account, also the energetic separations between the three triplet substates, the zero field splittings (ZFSs) and the radiative lifetimes of these states can be assessed (Table 3). At the S 0 geometry, a moderate agreement with the experimental data is observable. The calculated vertical excitation energies are somewhat underestimated and the small computed singlet-triplet splitting, although predicting 1-I as a TADF material, underestimate the experimental ∆E(S 1 -T 1 ) gap. Clearly, several important aspects are not taken into account at this level of theory. First, the inclusion of the difference in zero-point energies of ground and excited state may alter transition energies by 0.1-0.2 eV, usually leading to red shifted emission energies. Second, one generally expects excited state geometry relaxation to occur [103][104][105]. To investigate its impact, simulations were also carried out at the optimized T 1 state geometry. As Table 3 shows, this leads to a strongly red shifted emission, probably being significantly influenced by the strong shortening of the Cu-Cu bond length (Table 2). Experimentally, a strong dependence of the emission wavelength on the molecular environment was observed. For example, related Cu 2 I 2 (P∩N) 3 [(P∩N) = (2-diphenylphosphino)-4-alkyl-pyridine] complexes, as presented in [34], show strong red shifts going from the powder material to PMMA (poly(methyl methacrylate)) matrices and, finally, to solvents (also compare [37]). In the latter environments, large scale atomic rearrangements can easily occur. We find that the computed bond length reduction at the T 1 optimized geometry is accompanied by a 30 • C rotation of the P∩N ligand around the Cu1-P2 axis ( Table 2). While the former geometry change could also be realized in a polycrystalline environment, the latter one would likely be sterically hindered. Another influence is the effect of the dielectric environment, which probably is important due to the charge transfer character of the emission. Since a realistic modeling of the environment of the complex within the powder material is challenging, we performed SOC-TDDFT calculations with the COSMO continuum [106] solvent model (parameters: dielectric constant of dichloromethane (DCM) ε = 8.9, radius of solvent molecules r(sol) = 2.94 Å) for the S 0 geometry to obtain information about the general trends. Estimates of the radiative lifetimes are based on transition energies and transition dipole matrix elements as outlined in [107]. Corrections due to the refractive index of the molecular environment (n(DCM) = 1.42) were accounted for by the empty spherical cavity model [108]. Table 3 reveals a strong blue shift of ≈0.4 eV (≈3000 cm −1 ) compared to a gas phase environment. The observed blue shift can probably be explained by the considerable dipole moment change by the charge transfer from the Cu 2 I 2 core to the bridging ligand. Table 3. SOC-TDDFT (B3LYP/DZP) vertical excitation energies [eV] and radiative lifetime calculated for 100% quantum yield in µs (in brackets) for the S 1 state and the three substates of the T 1 state of Cu 2 I 2 (Ph 2 Ppy) 3 , 1-I. Singlet-triplet splittings ∆(S 1 −T 1 ) [cm −1 ] obtained from the average triplet energy and zero field splittings ZFS [cm −1 ] are also given. Results are presented for calculations at the S 0 optimized geometry with and without solvent effects and at the T 1 geometry in gas phase.  The calculated data, presented in Table 3, show considerably different energies, depending on the model and the state geometry chosen [109]. For comparison to emission data, we focus on the T 1 state geometry. Although the calculated transition energies are underestimated (in gas phase environment), it is usually accepted that energy differences display the experimental situation more realistically. The corresponding emission data are presented below in Section 2.4 and in Figure 8. It is seen that the energy gap ∆E(S 1 -T 1 ) is very well reproduced in the T 1 state geometry, while the ZFS value ∆E(III-II,I) is overestimated. This might be related to the model restricted to a too small number of spin-state-mixing of higher lying states. However, the very small splitting of ∆E(II-I) or the almost degenerate situation for the two low-lying triplet substates I and II is well reproduced by the SOC-TDDFT model (Table 3 and Figure 8, below).

Luminescence Properties of Cu 2 X 2 (P∩N) 3 Complexes with X = Cl, Br, I. An Overview
In this section, we will discuss the emission behavior of eight Cu 2 X 2 (P∩N) 3 complexes studied at 300 K and 77 K and give an overview over the large scale of varying emission properties.
The complexes studied are not soluble in common, non-coordinating organic solvents. Therefore, all measurements were performed with powders. Usually, investigations of solid-state samples are not convenient for studies of detailed emission properties, due to the influence of processes like triplet-triplet annihilation or energy transfer. However, for Cu(I) complexes, usually a self-trapping mechanism takes place that leads to quasi-isolated molecules embedded in the neat material without any significant excited state resonance interaction with the environment [37,110,111].
As summarized in Table 4, all studied di-nuclear Cu 2 X 2 (P∩N) 3 complexes show relatively intense photoluminescence under UV excitation with broad emission bands in the green to red spectral range assigned to (iodide+metal)-to-ligand charge transfer, (I+M)LCT transitions, as predicted by DFT calculations presented in the previous section. The HOMO resides on the Cu(I)-halide core, while the LUMO is largely localized on the bridging (P∩N) ligand ( Figure 3). Thus, modification of the ligand, in particular, of an extension of the ligand's aromatic system leads to a red shift of the MLCT transition [34]. There is a clear trend of the emission maxima depending on the halides. In each series, the emission maxima are blue shifted from Cl to Br to I complexes. For example, Figure 4 shows the emission spectra of the 1-X series. For ambient temperature, the peak maxima are blue shifted from 1-Cl to 1-I by about 1200 cm −1 . The flank of the excitation spectra in the region above about 400 nm shows a similar trend. This observation is rationalized by a reduction of the ligand field strength of the halides in the series Cl − > Br − > I − (being contrary to the trend of electronegativity) and thus, by a larger HOMO-LUMO energy gap [34,112]. The photoluminescence quantum yields at T = 300 K lie between Φ PL = 9% and 81%. At ambient temperature, the emission decay times are found in the lower µs time regime between 1.2 to 8.8 µs. However, with respect to photophysical interpretations, it is usually better to compare radiative decay times τ r = τ/Φ PL . In the series of compounds given in Table 4, they lie between 4.7 × 10 4 s −1 (21 µs) and 1.25 × 10 5 s −1 (8 µs). a -excitation wavelength λ exc = 372 nm; b -accuracy at 77 K: ± 10%, at 300 K: ± 5 % (relative error), (excitation wavelength λ exc = 400 nm); c -determined according to k r = Φ PL /τ; d -determined according to k nr = (1-Φ PL )/τ; e -the decay curve deviates from mono-exponential behavior. The lifetime given represents the main component; f -long component.
Upon cooling to 77 K, the emission quantum yields increase remarkably, for example, for compound 2-I by a factor of more than five. Such a behavior is not unusual, since non-radiative deactivation processes frequently become less important with temperature reduction [113]. Moreover, the emission bands of all complexes are red shifted. Most affected is the 1-X series with shifts of about 440 cm −1 (54 meV, 1-Cl and 1-I) and about 710 cm −1 (88 meV, 1-Br), while for the compounds 2-X and 3-X, values below around 300 cm −1 (37 meV) are observed. These energy separations frequently display approximately the energy gap ∆E(S 1 -T 1 ) that is responsible for thermal activation of the TADF emission. Interestingly, the value of 440 cm −1 (54 meV) found for Cu 2 I 2 (Ph 2 Ppy) 3 , 1-I corresponds well to the calculated one of 411 cm −1 (51 meV, Table 3). Concomitantly, the radiative rates of the emission decrease strongly (increasing decay times) with cooling, for example, by a factor of about 16 for compound 2-Cl. These effects, red shift and radiative rate decrease, occurring upon cooling, are consequences of freezing out the additional emission decay channel via the energetically higher lying S 1 state. In other words, TADF emission is largely frozen out. Thus, at sufficiently low temperature, mostly at T = 77 K but sometimes only below T = 50 K [29], all compounds exhibit only phosphorescence. The spectral changes of the emission bands with temperature variation are shown in Figure 5a for compound 1-I. At low temperature (1.3 K to almost 77 K), only T 1 →S 0 phosphorescence is occurring, while with further temperature increase to T = 300 K, the emission band center is blue shifted, mainly observable by a blue-side flank growing in above around T = 70 K (Figure 5b). The corresponding additional band is assigned to the thermally activated fluorescence from the higher lying S 1 state. Apparently, already the spectral changes displayed in Figure 5 indicate that at ambient temperature, the emission consists of overlapping TADF and phosphorescence. The appearance of a phosphorescence contribution even at ambient temperature is additionally based on the relatively fast radiative rate of k r (T 1 →S 0 , 77 K). Such a behavior of combined TADF and phosphorescence is not very frequently observed [48], but see refs. [50,51,83,86,114]. Mostly, however, the TADF channel dominates strongly [48]. For a more detailed discussion see Section 2.4.
Of the series of compounds studied (Table 4), especially, the iodide containing complexes 1-I, 2-I and 3-I show fast radiative phosphorescence rates lying above k r (T 1 →S 0 ) = 2 × 10 4 s −1 (50 µs). This may be rationalized by two factors. First, significant SOC is induced by admixtures of higher lying singlet states S n to the T 1 state. Secondly, the large contributions of 5p-orbitals of iodide to the higher lying occupied orbitals (Figure 3) with the high SOC constant of iodide of ξ(I) = 5069 cm −1 [115] may also play a role in speeding up the T 1 →S 0 transition rate [116]. This latter effect is supported by the distinctly lower T 1 →S 0 rate found for the chloride compounds with ξ(Cl) = 587 cm −1 [115] (Table 4).
A deeper discussion of the data summarized in Table 4 will not give more detailed information, in particular, not the spectral features available. This is a consequence of the very broad emission bands of MLCT character with halfwidths of ≈3500 cm −1 (440 meV) even at T = 1.3 K. On the other hand, investigation of the emission decay behavior with temperature variation will lead to a detailed characterization of the compounds' electronic structure, especially, of the lowest excited states, as will be shown in the next section.

Detailed Case Study of Cu 2 I 2 (Ph 2 Ppy) 3 , 1-I. the Lowest Excited Triplet and Singlet States
For a deeper case study, we selected Cu 2 I 2 (Ph 2 Ppy) 3 , 1-I (powder) due to four reasons. First, this material exhibits the highest emission quantum yield of Φ PL = 81% of all compounds summarized in Table 4. Second, 1-I emits with one of the fastest radiative phosphorescence rate of k r (77 K) = 2.88 × 10 4 s −1 (τ r (77 K) = 35 µs). Third, the emission decay time is mono-exponential over the whole temperature range (apart from very low temperature) and fourth, the emission quantum yield changes only slightly with temperature. These two latter properties are required for a detailed characterization based on the τ(T) fitting procedure as discussed below. With respect to these properties, 1-I represents a remarkable material and is expected to show clearly the interplay of phosphorescence and TADF. Moreover, according to the fast rate, efficient SOC experienced by the T 1 state should lead to a well observable zero-field splitting (ZFS) of T 1 into substates I, II and III, as already predicted by the SOC-TDDFT calculations (Section 2.2). This should lead to specific relaxation properties within the manifold of the substates, in particular, such as effects of spin-lattice relaxation (SLR) [117].
For the detailed characterization of the compound's electronic structure and the related decay rates, we study the emission decay behavior of 1-I over the large temperature range of 1.2 ≤ T ≤ 300 K. Figure 6 displays selected emission decay curves measured at different temperatures. At T = 1.2 K, the emission decays bi-exponentially with two clearly different components of 10 µs and 174 µs, respectively. Such a behavior can be well rationalized, if the lowest triplet state exhibits distinct ZFS. Thus, the short component refers to spin-lattice-relaxation (SLR) processes within the triplet substate manifold [117], while the longer component is ascribed to the thermalized emission that is established after around 60 µs at T = 1.2 K (estimated from the 1.2 K decay curve reproduced in Figure  6). It is known that thermalization according to SLR processes is strongly temperature dependent and becomes faster with temperature increase [117]. Indeed, the observed short decay component decreases from 10 µs at T = 1.2 K to 1.7 µs at T = 10 K and is faster than detectable (with our equipment) at T = 15 K ( Figure 6). This means that SLR processes are getting much faster than the thermalized emission decay time. Therefore, shortly after the excitation pulse, the population numbers of the excited states behave according to a Boltzmann distribution. We will come back to SLR properties below in this section. The decay time of the long (thermalized) component is also drastically shortened with temperature increase (Figure 7). Due to the photophysical mechanisms involved, the temperature dependence can be classified into three ranges: (i) Up to T ≈ 20 K, the decay behavior is determined by the zero-field split T 1 substates. At T = 1.2 K, we find a long component of τ = 174 µs. This emission dominantly stems from the thermally equilibrated two lower triplet substates I and II. By temperature increase the decay is strongly shortened (by a factor of more than five) when reaching a plateau near 20 K. The plateau decay time amounts to about 32 µs. This decrease is resulting from population of the higher lying triplet substate III. Because this transition III(T 1 )→S 0 exhibits a much faster rate than the rates from the two lower lying substates I and II, an additional decay path is opened. (ii) Within the temperature range of the plateau, between about 20 K and about 70 K, the triplet substates I, II and III are thermally equilibrated and emit as T 1 →S 0 phosphorescence with an average decay time of the three substates of 32 µs (see also Equation (4), below). (iii) With further temperature increase from T ≈ 70 K to 300 K, the decay time decreases to τ(300 K) = 6.5 µs. In this temperature range, the photoluminescence quantum yield changes only slightly (Table 4). Hence, the decay time decrease can (largely) be ascribed to a rate increase by a factor of more than four from k r (77 K) = 2.88 × 10 4 s −1 to k r (300 K) = 12.5 × 10 4 s −1 (Table 4). This increase is induced by opening the TADF decay path via the S 1 →S 0 transition in addition to the still remaining phosphorescence decay path (see also below). For the mono-exponential ranges of the decay curves ( Figure 6), it can be concluded that the emitting states are in fast thermal equilibration with respect to the individual decay times of the states involved. This is not only valid for the SLR processes but also for down-and up-inter-system crossing (ISC) processes between the T 1 and the S 1 states. Down-ISC will probably occur within around 10 ps or even shorter [103][104][105][118][119][120][121], while the up-ISC time (also named RISC time) is strongly temperature dependent [122] and may be estimated very roughly to around 30 ns at 80 K and to about 0.2 ns at ambient temperature. For completeness, it is mentioned that very fast ISC processes are probably based on direct SOC of higher lying singlet states to T 1 substates as is displayed in a high allowedness of the T 1 →S 0 transition. This is in contrast to the situation of molecules with weak SOC with respect to the lowest triplet state, as it seems to be valid for most organic molecules. For these, spin-vibronic processes will probably dominate the ISC rate [123].
Thus, with respect to the emission decay times of many µs, fast thermalization is well realized for the Cu(I) compounds discussed here. Accordingly, it is justified to describe the temperature dependence of the emission decay time τ(T), as shown in Figure 7, by a modified Boltzmann distribution of the four thermally equilibrated excited states, the three T 1 substates I, II and III and the S 1 state [19,50,87,124].
Herein, ∆E(II-I) and ∆E(III-I) are the ZFS values and ∆E(S 1 -I) is the energy gap between the S 1 state and the T 1 state (for ∆E(S 1 -T 1 ) >> ∆E(III-I)). k(I), k(II), k(III) and k(S 1 ) are the transition rates of the respective states to the electronic ground state S 0 . k B is the Boltzmann constant.
If we assume constant Φ PL over the whole temperature range, as approximately justified (Table 4), very good fit of Equation (1) to the measured decay times is realized ( Figure 7). As photophysical fit parameters we find ∆E(II-I) ≈ 0 cm −1 (meaning < 1 cm −1 ) and ∆E(III-I) = 3 cm −1 (0.37 meV). These values are smaller than the ones obtained from our SOC-TDDFT computations, but the splitting pattern is well reproduced (Table 3). For the manifold of the three triplet substates we expect further that for T < 1 K, a lowtemperature plateau of the emission decay time τ(T) will be adapted near 210 µs. Although, the individual decay times of the substates I and II cannot be determined directly, we will roughly set τ(I) ≈ τ(II) ≈ 210 µs. This is not unreasonable, since any larger deviation from this approach would not fit to the set of experimental data and the fitting procedure. Moreover, from the fit, we obtain k(III) = 8.3 × 10 4 s −1 , corresponding to formally 12 µs (compare Figure 8, below and Table 5). Since the III→S 0 decay path is much faster than the I,II→S 0 decays, state III involvement is already of importance at T = 1.2 K by reducing the decay time from ≈ 210 µs to 174 µs as found experimentally ( Figure 6). Figure 8. Energy level diagram, decay data and rates for Cu 2 I 2 (Ph 2 Ppy) 3 , 1-I powder. The T 1 substates I and II emit independently at T = 1.2 K with τ(I) ≈ τ(II) ≈ 210 µs. The transition rate k(III→S 0 ) amounts to 8.3 × 10 4 s −1 , corresponding to τ(III) = 12 µs. The spin-lattice relaxation time at T = 1.2 K is determined to τ(SLR) = 59 µs. The T 1 →S 0 (0-0) energy is estimated from the blue energy flank of the phosphorescence band as displayed in Figure 5. At T = 300 K, the total emission is composed of phosphorescence (20%) and TADF (80%), leading to a combined decay time of 6.5 µs at Φ PL = 81%.
The short decay component of 10 µs observed at T = 1.2 K (Figure 6) is assigned to the direct mechanism of SLR [117] from substate III to both substates I and II. Using the fit data presented above, we can determine the SLR rate k(SLR) for these processes by Herein, k(obs) corresponds to the observed decay component of 10 µs and k(III→S 0 ) is the rate of the III→S 0 transition with 8.3 × 10 4 s −1 . Accordingly, we find k(SLR) = 1.7 × 10 4 s −1 corresponding to τ(SLR) = 59 µs. This means, the SLR process is relatively slow. Values of similar size have, for example, been reported for Ir(III)complexes [125,126].
The fitting procedure gives also ∆E(S 1 -T 1 ) = 380 cm −1 (47 meV), the energy gap that is crucial for TADF properties. The computed value for the T 1 state geometry fits relatively well (411 cm −1 , 51 meV, Table 3). Moreover, the rate of the S 1 →S 0 fluorescence, also resulting from the fit, amounts to k(S 1 →S 0 ) = 2.25 × 10 6 s −1 (445 ns, Figures 7 and 8). However, the corresponding prompt fluorescence decay cannot be measured directly, since the S 1 -T 1 ISC is orders of magnitude faster. It amounts to around 10 ps [103][104][105][118][119][120][121] as already discussed above. The computations lead to a value of 1/k(S 1 -S 0 ) = τ(S 1 -S 0 ) = τ(S 1 ) = 1.5 µs for the prompt fluorescence (Table 3), thus, being overestimated by a factor of about three. Even the experimental S 1 →S 0 transition rate is relatively slow for an "allowed" fluorescence, but this is in line with the distinct CT character of the transition. At small HOMO-LUMO overlap, one obtains small exchange interaction and hence, a small ∆E(S 1 -T 1 ) gap (as required for TADF materials), however, as well as a slow S 1 →S 0 transition rate [37,48]. For completeness, it is remarked that design of materials with faster S 1 →S 0 transition rates and small gaps ∆E(S 1 -T 1 ) is highly attractive and indeed, became possible recently [40][41][42][43][44]52]. Table 5. Energies, emission quantum yields and decay data of Cu 2 I 2 (Ph 2 Ppy) 3 , 1-I (powder). Note the different temperatures. Interestingly, the emission intensity at ambient temperature represents a combined phosphorescence and TADF. According to the Supporting Information of ref. [87] and [82], we find for the ratio of phosphorescence intensity Int(T 1 ) to the total emission intensity Int(tot)

Property Value
For the phosphorescence decay time τ(T 1 ), we can insert the value given by the plateau displayed in Figure 7 with τ(plateau) = 32 µs or we can calculate it from the average decay time determined from the three triplet substates according to [19,117,127].
Inserting into Equation (3) the mean value of τ(T 1 ) = 32 µs, τ(S 1 ) = 0.445 µs and ∆E(S 1 -T 1 ) = 380 cm −1 , we find for ambient temperature the percentage of phosphorescence intensity relative to the total emission intensity of ≈20%. Thus, the fractional emission intensity of the TADF-only emission amounts to ≈80%. Accordingly, compound Cu 2 I 2 (Ph 2 Ppy) 3 1-I shows two emission decay paths, as already indicated by the relatively fast phosphorescence rate and by the temperature dependent development of the emission spectra ( Figure 5). The rate of the TADF-only process is expressed by k r (TADF-only) = k r (com) − k r (T 1 ) With the values of the radiative combined rate of k r (com) = 1.25 × 10 5 s −1 (k r (com) = Φ PL /τ = 0.81/(6.5 µs), Table 4) and the radiative phosphorescence rate of k r (T 1 ) = 2.88 × 10 4 s −1 (τ r = 35 µs, Table 4), we obtain for the radiative TADF-only process k r (TADF-only) = 9.62 × 10 4 s −1 corresponding to τ r = 10.4 µs decay time. Hence, the radiative decay time of the combined process, phosphorescence and TADF, is significantly shorter than the TADF-only process. This is valid for the radiative as well as for the measured decays. The combined emission decays by 23 % faster than the TADF-only emission. This is a favorable result and may help in future design strategies to shorten the overall emission decay time of emitter materials to reduce OLED device stability problems [68].
Essential properties of Cu 2 I 2 (Ph 2 Ppy) 3 , 1-I worked out above are summarized in Table 5 and visualized by an energy level diagram in Figure 8.

Summarizing Conclusions
In this report, we present an overview over a series of P∩N linked Cu(I) dimers of the type of Cu 2 X 2 (P∩N) 3 with respect to their X-ray structures and with particular focus on photo-luminescence data. Within the series of eight different compounds, the emission color varies from green to red, the ambient temperature radiative decay time τ r covers a range from 8 to 21 µs and the emission quantum yield Φ PL lies between 9 and 81%. All compounds studied show combined TADF and phosphorescence at ambient temperature. This is due to relatively large spin-orbit coupling (SOC) experienced by the lowest lying triplet state T 1 . Around T = 70 K, the TADF component is largely frozen out and only phosphorescence is observed. The set of data described in the first part, leads us to select a specific compound, namely Cu 2 I 2 (Ph 2 Ppy) 3 1-I, for deeper photophysical characterizations by DFT and SOC-TDDFT computations and for emission measurements over the large temperature range of 1.2 ≤ T ≤ 300 K. Thus, the interplay of phosphorescence and TADF is clarified and properties of the lowest excited singlet and triplet states can be revealed in detail. For example, we determine the S 1 →S 0 transition rate, the ∆E(S 1 −T 1 ) gap, detailed triplet state features, such as the T 1 →S 0 transition rate as well as the rates from the substates, zero-field splitting (ZFS) of the T 1 state and the spin-lattice relaxation (SLR) rate between triplet substates. Interestingly, the investigation of the decay time as function of temperature allows us to determine electronic splitting features of the order of less than 1 cm −1 (0.1 meV), although the spectral halfwidth of about 3500 cm −1 (0.43 eV) is by a factor of 3500 larger.
Thus, this presentation proceeds from an overview over properties of a series of related Cu(I) dimers to detailed photophysical characterization at the state of art. Moreover, it becomes evident that the occurrence of a combined emission process at ambient temperature, consisting of distinct phosphorescence and TADF, or if applied in an OLED, consisting of combined singlet and triplet harvesting, leads to significant shortening of the overall photoluminescent decay time. This is a favorable result and may help in future design to improve the performance of OLED devices with respect to decrease of roll-off and stability problems and even increase of the external quantum efficiency (EQE).

Photophysical Measurements
The complexes were investigated as powders. Emission spectra and decay curves were measured by use of a Fluorolog 3 spectrometer (Horiba Jobin Yvon, Munich, Germany) equipped with a cooled photomultiplier tube. The spectra were corrected with respect to the wavelength dependence of the instrument. The decay behavior of the phosphorescence was recorded using a multichannel scaler card (P7887, Fast ComTec, Munich, Germany) with a time resolution of 250 ps. For excitation, the third harmonic of a pulsed Nd:YAG laser (355 nm, pulse width < 8 ns) was used. A Konti IT (CryoVac, Troisdorf, Germany) cryostat was applied for the variation of temperature between 1.2 K and 300 K. Quantum yield measurements at ambient temperature and at 77 K were carried out with an integrating sphere applying a C9920-02 system (Hamamatsu, Herrsching am Ammersee, Germany).

Computational Investigations
Density functional theory (DFT) geometry optimizations were performed with the Turbomole [134] package using a basis set of triple-zeta plus polarization (def2-TZVP) quality [135] and the hybrid B3LYP exchange-correlation functional [136]. For iodine the corresponding effective core potential [137] was employed. Time-dependent DFT (TDDFT) calculations were carried out with ADF2014 [101] as mentioned in the main text.

Syntheses
General Procedure for the syntheses of complexes 2-X and 3-Br according to a simplified literature method: The copper(I) halide salt (2 equivalents) and the ligand (3 equivalents) were suspended in dichloromethane (15 mL) and stirred 12 h under ambient conditions. To the filtered reaction solution diethyl ether was added. The formed solid was filtered off, washed with diethyl ether and dried in vacuum. As soon as the complexes were precipitated form the reaction mixture, they were not sufficiently soluble in common weakly or non-coordinating solvents to perform NMR-spectroscopy. Slow gas phase diffusion of diethyl ether into a solution which was obtained by filtration of the crude reaction mixture gave single crystal suitable for X-ray diffraction.  Acknowledgments: The authors thank Manfred Zabel for the determination of the crystal structure of Cu 2 I 2 (Ph 2 Piqn) 3 , 3-I and Philipp Edelsbacher for help with formatting the manuscript.

Conflicts of Interest:
The authors declare that they do not have any competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Sample Availability: Samples of the compounds are available from the authors.