Dipole Moment and Charge Reorganization in Photoredox Catalysts

We report evidence of excited-state ion pair reorganisation in a cationic iridium (III) photoredox catalyst in 1,4-dioxane. Microwave-frequency dielectric-loss measurements combined with accurate calculations of dipolar relaxation time allow us to assign both ground and excited-state molecular dipole moments in solution and determine the polarizability volume in the excitedstate. These measurements show significant changes in ground-state dipole moment between [Ir[dF(CF3)ppy]2(dtbpy)]PF6 (10.74 Debye) and [Ir[dF(CF3)ppy]2(dtbpy)]BAr 4 (4.86 Debye). Photoexcitation of each complex results in population of highly mixed ligand centered and metal-to-ligand charge transfer states with enormous polarizability. Relaxation to the lowest lying excited-state leads to a negative change in the dipole moment for [Ir[dF(CF3)ppy]2(dtbpy)]PF6, and a positive change in dipole moment for [Ir[dF(CF3)ppy]2(dtbpy)]BAr 4 . These observations are consistent with a sub-nanosecond reorganization with the PF− 6 counter-ion, which cancels the dipole moment of the lowest lying excited-state, a process which is absent for the BAr F− 4 counter-ion. Taken together, these observations suggest contact-ion pair formation between the cationic metal complex and the PF− 6 anion and, at most, solvent-separated pairing with BAr F− 4 . The dynamic ion pair reorganisation we observe with the PF − 6 counter-ion may substantially modify both the thermodynamic potential available for electron transfer and kinetically inhibit oxidative catalysis, as the anion moves to cover the positively charged end of the molecule, providing a possible mechanistic explanation for recently observed trends in the catalytic activity of these complexes as a function of anion identity in low-polarity solvents. These tunable ion-pair dynamics could prove to be a valuable tool for tailoring the reactivity of both new and extant photocatalysts.


Introduction
Photoredox catalysts are able to drive reactions with high kinetic or thermodynamic barriers by harnessing the energy of light. This in-situ preparation of highly oxidizing or reducing reagents from stable precursors allows gentler more selective reaction conditions, opening the doors to many new reactions such as the processing of lignin into stock chemical species to aryl-dehalogenation reactions. [1][2][3][4][5][6][7][8] Cyclometalated (C^N) and polypyridyl (N^N) complexes of d 6 metals are some of the most promising transition metal photoredox catalysts to date. Heteroleptic compounds incorporating (C^N) and (N^N) ligands, as in [Ir(ppy) 2 (bpy)] + complexes, benefit from independent tunability of HOMO and LUMO levels via ligand architecture. [9][10][11][12][13][14][15][16][17][18][19][20] Yet, the relationship between structural/environmental perturbations and changes in the intricate excited-state manifold of [Ir(ppy) 2 (bpy)] + complexes are hard to capture with existing spectroscopies, making it commensurately difficult to understand all the factors that modulate their photoredox activity. Tuning of the redox potentials, absorption energy, excited-state lifetime, and quantum yield, are well known criteria, but other factors can play an important role. [21][22][23][24] For instance, recent work suggests that the size and polarity of the counter-ion effects the efficiency of the catalytic cycle, [25][26][27][28][29][30][31] . While many reports note spectral changes due to ion-pairing in common photoredox catalysts, 30,[32][33][34][35][36] there are currently no direct quantitative methods that can probe the ground and excited-state electron density distribution to understand how ion pairing influences the excited state structure and evolution. Here, we introduce a modern implementation of time-resolved dielectric-loss spectroscopy (TRDL) to investigate how counter-ion identity influences the excited-state charge distribution within an important photoredox catalyst, [Ir[dF(CF 3 )ppy] 2 (dtbpy)]X. [37][38][39] We find large differences in both the ground and excited-state dipole moments depending on whether X is a small associating anion (PF − 6 ) that forms a contact-ion pair vs. a large one that either dissociates or forms a solvent-separated pair (BAr F− 4 ). These results suggest that ion pairing and the control it confers over both ground-state and excited-state structure could be a heretofore underappreciated factor that modulates reactivity in ionic photoredox catalysts.
In what follows, we describe our modern implementation of TRDL using quantitative methods to measure complex permittivity, and calculations of the dipole moment of both ground and excited-state molecules in solution. We use electromagnetic simulations to relate measured microwave resonator properties to the complex permittivity of the sample. Combined with new quantitative calculations of the dipolar relaxation time, we are able to extract the molecular dipole moment from dielectric-loss data. We verify the validity of our methods by studying the ground and excited-state dipole moments of N,N-dimethylaminobenzonitrile (DMABN) in 1,4-dioxane; a model compound with a well understood intramolecular charge transfer excited-state. Finally, we apply this approach to understand the excited-state charge reorganization in [Ir[dF(CF 3 )ppy] 2 (dtbpy)]PF 6

Dielectric-loss Spectroscopy Background
Flash-photolysis time-resolved dielectric-loss (fp-TRDL) spectroscopy is a perturbative technique developed to measure the time dependent evolution of a molecule's electron density distribution; it is one of only a few methods to measure a solution-phase molecule's excited-state dipole moment in situ without relying on emission, which can bias observations to bright states and limit the broad applicability of the technique, for instance with dark molecular triplet states. [40][41][42][43] In this technique, microwave radiation is used to measure the complex permittivity (ε r = ε − iε ) of a sample by monitoring the phase and amplitude changes of a transmitted waveform. Performing this measurement inside a resonance cavity increases the effective path length through the sample and thus improves the sensitivity to small changes in ε r . At the X-band frequencies used in these experiments (8.2 -12.4 GHz), changes in ε r from a solution are a function of solute concentration, a solute's dipole moment (µ), and its polarizability volume (V p ). Selectively exciting a photoredox catalyst with a pulsed laser allows measurement of ∆ε r due to the solute molecule's change in dipole moment and polarizability volume -the change in charge density and localization from the ground-state to the excited-state. Ultimately, measuring both ε r and ∆ε r captures the dipole moment and polarizability volume of a solution-phase sample, both in the ground and excited-states.
Our implementation of dielectric-loss spectroscopy consists of placing a sample inside a resonant microwave cavity and measuring the change in cavity characteristics caused either as a function of ground or excited-state solute concentration. Electromagnetic modeling is then used to relate the complex permittivity of the sample to the resonator properties, allowing quantitative permittivity measurements. The modeling is combined with quantitative simulation of the dipolar relaxation time to extract the molecular dipole moment and polarizability volume of the solute. [SI section COMSOL Simulation of Microwave Cavities] For this study, a new microwave resonance cavity (figure 1) was designed and fabricated to measure solution-phase samples based on a design by Fessenden et. al.. 44 The experimental set-up consists of a 5 mm diameter EPR tube placed in the cavity at one electric field maximum of the T E 102 standing wave, and a single slit in the end wall allowing for photoexcitation. A circular iris structure is placed between the feeding waveguide and the cavity in order to couple microwave radiation into the cavity, and define one end of the resonator. For all samples in this work, a 9.20 mm diameter iris was used unless otherwise stated. Maxwell's equations were solved using the computer aided drafting (CAD) model and radio-frequency (RF) simulation package COMSOL Multiphysics (v4.3a). This allowed us to optimize the sample position and iris diameter before fabricating the microwave cavity; this design process has been described in our previous work. 45 After manufacturing the cavity, the reverse process was used to model the manufactured cavity in COMSOL to account for manufacturing defects, and extract quantitative values for the sensitivity of the cavity characteristics to the complex permittivity of the sample (ε and ε ). [described in section COMSOL Simulation of Microwave Cavities of the SI] By iteratively stepping the simulated sample's ε and ε inside the EPR tube, we create a look-up table of simulation results that we then use as a numerical fit function to quantify the dielectric properties of solution-phase samples; this enables us to convert measured microwave power reflectance as a function of frequency about the cavity resonance to ε and ε .
The complex permittivity of a solution-phase sample is a function of both the dipole moment and polarizability of all the molecules therein, and for dilute solutions, it may be expressed as the sum of contributions from the solvent and the solute. This relationship is simplified if the complex permittivity can be separated into its real and imaginary components (described by equation (1)) as the imaginary permittivity is only a function of the dipole moment.
The remainder of this discussion will focus only on the solute's contribution to the permittivity as the solvent's contribution is subtracted out in the experimental procedure. Firstly, the real part of the solute's permittivity can be separated into two components which describe the dipole contribution (ε D ) and the electronic polarizability contribution (ε E ).
Both components are derived from Debye's polarity equations of a polar molecule rotating in an electric field. 46 The dipole contribution to the real permittivity is described as: Where N is the solute concentration in molecules/m 3 , ε 0 is the permittivity of free space, ε(0) is the DC dielectric constant of the solvent, ε(∞) is the high-frequency dielectric constant of the solvent (which can be approximated as the square-root of the refractive index, √ n), ω is the microwave angular frequency, µ is the molecule's dipole moment, k B is the Boltzmann constant, θ is the dipolar relaxation time of the solute [described further in the Molecular Rotation section of the SI], and T is temperature. The function f (ωθ ) describes the variation in experimental sensitivity due to the correlation between the microwave frequency and the frequency of the dipole's relaxation.
The electronic component of the real permittivity is a function of the solute's electronic polarizability (α e ).
In contrast to the real part, the imaginary part of the complex permittivity is only a function of the dipole moment and looks similar to equation (3), however, the denominator term (ωθ ) is not squared.

Ground-state dielectric-loss
We calculate the ground-state dipole moment of a solute molecule by performing a series of measurements of ε as a function of solute concentration. Rewriting equation (6) to reflect this experiment yields the following equation: Where now the solute concentration changes (∆N) thus changing ∆ε and the solvent's contribution subtracts out. Dividing both sides by ∆N in equation (7), provides an expression for the slope of this concentration-dependent study.
Rearranging equation (8) gives the expression for the ground-state dipole moment.
Once the slope of the ∆ε vs. concentration has been measured, the only remaining unknown in this equation is the dipolar rotational relaxation of the solute. We have implemented a numerical simulation package in Python which provides accurate values for the dipolar rotational relaxation time using the molecular structure, solvent viscosity, solvent dielectric, and temperature as input parameters [described further in the Molecular Rotation section of the SI] enabling accurate dipole moment measurements without the need for carefully selected standards or outside experiments as was the case in the past. 41,44 Due to experimental sensitivity constants, we are unable to use this same technique in order to measure the ground-state polarizability with much accuracy without going to unacceptably high solute concentrations.

Excited-state transient dielectric-loss
Photoexcitation is used to excite the sample and study the dipole moment and polarizability of the excited-states. In this case, equation 3 describing the dipole component of the real permittivity becomes: Where N * (t) is the excited-state solute concentration in molecules/m 3 as a function of time due to the sample's excited-state kinetics, and µ 2 * is the excited-state dipole moment. Equation (5) which describes the polarizability component of the real permittivity becomes:

4/19
Change in electronic polarizability is related to the polarizability volume (∆V p ) which qualitatively describes the localization volume of a molecular excited-state. 47 Finally, the imaginary component of the complex permittivity becomes the following equation.
For transient measurements, we fit a global kinetic model to our data (detailed in the Fitting TRDL Transients section of the SI) in order to get an excited-state concentration (N * (t)) for each kinetic component. Then, the ∆ε and ∆ε values that relate to those respective components are used to calculate the various changes in dipole moment and molecular polarizability volume. The absolute value of the excited-state dipole moment can then be calculated knowing the ground-state dipole moment.
Thermal artifacts are easily measured in dielectric-loss spectroscopy due to its high sensitivity. These arise from transient changes in the dielectric properties of the solvent as a function of local temperature. Since the TRDL signal is proportional to the concentration of the excited species, small thermal changes in the solvent easily overwhelm larger changes in the solute properties. To minimize this, TRDL must be performed in low dielectric-constant solvents so that thermally-induced changes in its dielectric constant also remain small. For all of the work in this paper, 1,4-dioxane has been used as it has shown no evidence of thermal artifacts in TRDL transients.

Validating our methods using N,N-dimethylaminobenzonitrile
A donor-acceptor molecule, N,N-dimethylaminobenzonitrile, was used to test the validity of our COMSOL simulations, dipolar relaxation time calculations, and overall analysis. DMABN has been extensively studied because of it's intramolecular charge-transfer (ICT) excited-state and resulting dual-emission; in particular, John Warman has reported on its kinetics, ground-state dipole moment and excited-state dipole moments using dielectric-loss spectroscopy, making it an ideal reference compound for our work. [48][49][50][51] DMABN has two singlet states in equilibrium, a locally excited-state (LE) and the intramolecular charge transfer state (ICT). The LE state has a dipole moment that is only slightly larger than that in the ground-state at 9.7 D; whereas, the ICT state's dipole moment is significantly larger than the ground-state at 15.1 D. 52 The initial LE-ICT singlet state relaxes through a triplet state that has a dipole moment approximately equal to that of the LE state, with its inter-system crossing yield being correlated to the solvent's polarity. The CT state is stabilized by the solvent's polarity as well, thus the ratio of CT:LE is higher in polar solvents and the CT state's dipole moment dominates the TRDL signal at early times when the singlet state is evolving. Figure 2(a) shows a plot of imaginary permittivity as a function of DMABN concentration in 1,4-dioxane. Using the slope of the linear fit, equation (9), and a θ of 33.4 ps, we found the ground-state dipole moment of DMABN to be 6.2 D. This is in close agreement with the ground-state value for DMABN reported by Warman using TRDL: 6.6 D. This result demonstrates the accuracy and precision with which we can measure the ground-state dipole moment of a molecule in solution using our instrument and data analysis methods, and validates the calculations used to determine θ (described in the Molecular Rotation section of the SI).  Here, the real (blue) and imaginary (red) components of the of the transient change in permittivity are separated using the procedure described in the Fitting COMSOL Models to Data and the Complex K-Factor section of our SI, giving quantitative values for ∆ε r . Notably, the noise level here is ∼100 ppb in ε r after averaging fewer than 100 laser shots (limited by the photostability of DMABN). The transient kinetics observed are similar to that reported by Warman when taking into account the differences in instrument response. Indeed, the kinetics are fit using an identical model to that employed by Warman, with certain key parameters fixed from his prior work: sequential conversion of the LE+ICT singlet to the triplet state in competition with ground-state recovery. 52 The black traces in Figure 2(b) show the kinetic trajectories assigned to the singlet (solid) and triplet (dashed) states. The conversion between the singlet and triplet states takes place too fast (∼ 10 10 s −1 ) for a TRDL experiment to properly constrain the kinetic fit, and we use TRPL results from the literature to fix this timescale in our model. [52][53][54] Likewise, the TRDL experiment does not directly observe conversion of the singlet LE and ICT states to the ground-state in competition with intersystem crossing, and we fix the triplet yield based on prior results as well. 54 Thus, the only free fit parameters in our global kinetic model (over real and imaginary signals) is the triplet state lifetime and the change in ε r per molecule. The rate constants associated with each state are summarized in (see Supplementary Table S3), and the extracted values for the transient dipole moments for the ground, singlet, and triplet excited-states are summarized in table 1. In agreement with the previous literature: (1) the singlet state of DMABN is dominated by the large dipole moment associated with the ICT state in 1,4-dioxane, and (2) the obtained triplet state dipole is substantially less than that of the singlet state.

5/19
The relatively weaker agreement (relative to literature) in the absolute excited-state dipole moments we calculate compared to the ground-state most likely arises from the difficulty in ascertaining the excited-state concentration, both in our case and in past measurements. Since the calculation of the transient dipole moment is coupled to the excited-state solute concentration, the accuracy of estimating this concentration introduces error into the resultant dipole value in cases such as this where much of the dynamics are instrument-response limited. Warman et. al. relied on internal actinometry for estimating the excited-state concentration and parallel TRPL measurements to fix kinetic parameters that were not measurable via TRDL. We use careful calibration of light absorption in our EPR-tube sample configuration and excited-state density simulations that account for saturable absorption effects [discussed in section Excitation Density Calculations in the SI]. Additionally, we observed sample degradation from laser excitation, which was minimized but possibly not eliminated, by handling all samples in a nitrogen glove box, limiting our analysis to averaging the first 100 transients, and refreshing sample intermittently throughout the measurement. The good agreement we obtain for the singlet state shows that these methods are effective, but the rather large discrepancy for the triplet dipole moment measurement suggests that the triplet yield may be substantially lower in our hands than the value found previously, which is fixed in our kinetic model at the literature value. 54

Contact-ion Pairing and Transient Dipole Moments in [Ir[dF(CF 3 )ppy] 2 (dtbpy)]X
Having established the accuracy of our experimental methods via the well-studied ICT excited-state of DMABN, we turn to the more complex topic of the cationic photoredox catalysts [Ir[dF(CF 3 )ppy] 2 (dtbpy)]X (abbreviated: Ir[tBu]-X). Here, we expect a metal-to-ligand charge transfer (MLCT) state that is analogous to the ICT state of DMABN, but the presence of multiple ligands, large-spin-orbit coupling due to Ir(III) (ξ c,Ir = 3909cm −1 vs. ξ c,soc,Ru(II) = 990cm −1 ) 55 , and ion pairing in a low dielectric constant solvent introduce new and unexpected dynamics. Below, we show that the highly mixed excited-states of our energetically similar ligands manifests as charge transfer processes from dF(CF 3 )ppy to dtbpy on the ns timescale. Meanwhile, contact-ion pairing introduces an intermolecular dipole moment between the two ionic species that dramatically changes the dipole of the excited-state through additional nuclear reorganization. These latter two effects vanish for a more weakly bound anion, whilst a flip-flopping inter-ligand charge transfer behavior at early times is consistent in both. should shift the dissociation equilibrium so that contact-ion pairs no longer form, as Ir[tBu] + and BAr F− 4 are relatively the same size. As with DMABN, measuring the characteristics of the microwave cavity as a function of the Ir[tBu]-X concentrations allows us to obtain their ground-state dipole moments. However, an additional complexity in this context is that it is necessary to make an assumption about whether contact-ion pairing takes place in order to calculate the rotational relaxation time of the complex, which can change significantly when the ion pair rotates as a unit. The approach we take here is to test each possibility (contact-paired vs. not) and observe whether the dipole moment we calculate is consistent with basic chemical theory and literature calculations. [56][57][58] This provides an excellent method for assessing ion pairing in the ground-state of these complexes. The results of these elementary calculations are displayed in table 2.  When considering the ligand framework of Ir[tBu] + , a ground-state dipole moment is expected on grounds of both symmetry and formal charge distribution. However, cyclometalation of Ir(III) with dF(CF) 3 ppy ligands produces a highly mixed HOMO (Ir d -dF(CF) 3 ppy) that lies relatively close to the metal. 12,59 This bonding environment implies that the ground-state intramolecular dipole in Ir[tBu] + should be relatively small. In contrast, the apparent dipole moment resulting from a contact-ion pair interaction is expected to be large, due to the relatively large separation between the ionic charge centers and the fact that the ions rotate as a single unit in solution. Directly measured dipole moments of [Ir(C^N) 2 (N^N)] + complexes have never been reported to our knowledge, either in the ground or excited-states. 10, 19-21, 23, 31, 33-36, 56-58 Using the slope from ε" vs the concentration of Ir[tBu]-PF 6 in 1,4-dioxane (figure 4(a)), equation (9), and a θ paired of 70.5 ps (calculated assuming contact-ion pairing as described in the Molecular Rotation section of the SI) we found the ground-state dipole moment of Ir[tBu]-PF 6 to be 10.7 D. If we assume loose or no ion pairing, the rotation time is modified to 68.26 ps and the calculated dipole moment is reduced to 10.6 D, only a small change occasioned by the very compact size of PF − 6 . In both cases, the magnitude of the ground-state dipole is substantially larger than what we would expect for the intramolecular ground-state dipole moment of just Ir[tBu] + (∼ 5.0 D, based on that of Ir(ppy) 3 (see SI section Non-ionic Iridium(III) Complex)). Yet, the calculated value is still smaller than we expect for a contact-ion pair Ir[tBu]-PF 6 (27.4 D). We posit three primary explanations: first, that the inter-and intra-molecular dipoles partially cancel in a contact-ion pair that self-organizes to minimize electrostatic energy, as depicted in Figure 5a. Second, that the small PF − 6 anion may lie ∼ 2.5 Å closer to the positive charge center than expected (27.4-5.0 = 22.4 D, integer charge separation produces a dipole moment of 4.8 D/Å), interdigitating with the ligand shell. Third, that contact ion pair is likely to be in equilibrium with solvent-separated pairs, which would reduce the average dipole moment that we measure.
An excellent test of these ideas is comparison with results for the more weakly pairing anion BAr F− 4 . Using the slope from ε" vs. the concentration of Ir[tBu]-BAr F 4 ( figure 4(b)), equation (9), and a θ unpaired of 68.3 ps (calculated assuming loose ion pair association); we found the ground-state dipole moments of Ir[tBu]-BAr F 4 to be 4.82 D, consistent with our expectations for the range of possible dipole moment of Ir[tBu] + . If we assume contact-ion pairing on the other hand, we obtain a much larger dipolar relaxation time of θ paired = 367 ps due to the large size of the BAr F− 4 anion. Using this value we calculate a dipole moment of 10.8 D, which in this case is far too small to be consistent with the large intermolecular dipole moment expected for a contact-ion pair of Ir[tBu]-BAr F 4 , even when the anti-parallel vector sum of the intramolecular and intermolecular dipole moments and similar charge displacements are considered. This, combined with the consistency of the unpaired dipole we measure here with that of Ir(ppy) 3 (see Supplementary Fig. S11), leads us to conclude that contact-ion pairs are absent for , and (N^N)= bpy or 1,10-phenanthroline (bpy)). 56 In this study, the ground and excited-state dipole moments were determined from absorption and emission solvatochromism data in 17 solvents spanning the polarity spectrum, but neglecting to account for ion pairing behavior. Thus, in low dielectric solvents, the dipole moment they observed should be characteristic of the ion-paired species [Ir(C^N) 2 (N^N)]Cl, whilst in high dielectric solvents it most likely represents that of the unpaired species. We conjecture that their measured value (µ Lit =6.7 D), represents an average of these two extremes, and compares favorably with an average of the dipole moments we find for Ir[tBu]-PF 6 and Ir[tBu]-BAr F 4 (µ Avg =7.7 D). Figure 5 illustrates the ion-pairing behavior we propose for the ground-state without contact-ion pairing (5(a)) and with contact-ion pairing (5(b)). If the ion pair is not in contact, the overall dipole moment measured using dielectric-loss will be equal to just the intramolecular dipole moment (purple). However, in the presence of contact-ion pairing, the overall dipole moment measured using dielectric-loss will be equal to the vector sum (equation (14)) of the intramolecular dipole moment (purple) and the intermolecular dipole moment (yellow). Dielectric-loss can only measure ion pairs that are bound sufficiently tightly for the pair to tumble as a single molecular unit in solution. In any situation where each complex can tumble separately the measured dipole moment will revert to the intramolecular component depicted in Fig. 5(a), even if the ions remain paired in a looser solvent-separated fashion. 23   Fig. 6 (a) (X − = PF − 6 ) and (b) (X − = BAr F 4 ). The data obtained were collected and separated into their complex permittivity components as detailed in the Fitting COMSOL Models to Data and the Complex K-Factor section of the SI. Both complex's kinetics were fit with a global two-Species sequential kinetic model similar to that used for DMABN. Here, however, there is less of a clear a-priori basis for assigning the identity of these kinetically identified Species from TRDL measurements. We simply designate the fast, early-time component as Species A and the slower, later-time component as Species B for now. The rate constants from the kinetic fits and the absolute dipole moments are summarized in table 4.  While TRDL measurements can't directly report on the identity of Species A and B, the obtained time constants and dipole measurements are interpreted with respect to the literature on similar Ir(III) species to help elucidate the photoexcitation mechanism. Species A for both Ir[tBu]-PF 6 and Ir[tBu]-BAr F 4 displays nearly instrument-response limited kinetics and is associated with a large increase in ∆ε which is a function of the polarizability volume and the dipole moment. Since this feature of Species A is present in both complexes, it is most likely a property of Ir[tBu] + , and not influenced by differences in ion pairing behavior. Moreover, the change in ∆ε is much smaller than ∆ε , indicating that the major contribution to ∆ε is a change in the polarizability volume. Two mechanisms can contribute to a change in polarizability volume; the electron cloud increasing in volume, or a resonance structure where a dipole moment flips between resonances much faster than the frequency of the microwave electric field (<100 ps period). Since the dF(CF 3 )ppy-based LUMO+1 and dtbpy-based LUMO are energetically similar in Ir[tBu] + (See SI, [Ir(C^N) 2 (N^N)]X Electronic Structure), we conclude that the change in polarizability volume is due to ligand-ligand charge transfer (LLCT) dynamics described by the dipole moment shifting between the dF(CF 3 )ppy and dtbpy ligand in the excited-state. This interpretation of the large ∆ε signal is consistent with previous reports on similar complexes indicating that large changes in dipole upon photoexcitation result from substantial geometric distortion in the MLCT bpy excited-states due to LLCT. 19,20,56,58,60,61 Moreover, we observe at least two emissive bands in the emission profile of both Ir[tBu]-PF 6 and Ir[tBu]-BAr F 4 (See SI, Absorbance and emission spectroscopy). The origin of multiple emission bands in [Ir(ppy) 2 (bpy)]PF 6 is elucidated in previous work, and the photoinitiated excited-state dynamics are described as follows:

11/19
(1) the Franck-Condon excited-states form in <30 ps then decay to (2) a 3 LLCT state formed within 2.6 ns, (3) a 3 MLCT ppy formed in <0.4 ns, and (4) a 3 MLCT bpy formed within 4.8 ns (all time constants are from toluene measurements). 58 This previous work describes emission from the 3 LLCT , 3 MLCT ppy , and 3 MLCT bpy in τ em = 617, 643, and 582 ns, respectively in toluene, with the intensity of each band depending on the relative population of each state (see Fig. 7). Considering the instrument response function of our system and the differing dielectric constants of toluene vs 1,4-dioxane, the literature mechanism described above agrees well with the measured time constant associated Species A's decay (2.78 · 10 8 s −1 , i.e. τ A = 3.60 ns), which likely represents a rough average of the time constants associated with processes (2) 58,60,61 The negative ∆ε from Ir[tBu]-PF 6 indicates that the observed excited-state dipole moment associated with the relaxed MLCT state is less than that of its ground-state. This striking observation is consistent with a sub-nanosecond excited-state reorganization mechanism where the ion pair reorients to cancel the intramolecular dipole moment of the MLCT state, as inferred for the ground-state in the preceding section. Indeed, contact-ion pairing combined with this sort of excited-state reorganization will always result in an overall negative change in the dipole moment. Simply by virtue of the fact that the counter-ion must remain further from the molecular center than the ligand shell, it is impossible for the intramolecular dipole moment to exceed that associated with the intermolecular dipole moment. The fact that we do not see this negative change in the B Species of the Ir[tBu]-BAr F 4 complex is confirmation that the negative transient dipole moment is most likely due to ion-pair reorganization. As described previously on similar complexes, ion-pairing imposes an electrostatic restriction on the spatial geometry of the complex. 35,36 This description agrees well with the differences we observe in the ground-state dipole moments and absorption spectra of Ir[tBu]-BAr F 4 vs Ir[tBu]-PF 6 (See SI, Absorbance and emission spectroscopy). However, the ion-pair rearrangement in the excited-state alters the geometry of the coordination sphere to accommodate the charge redistribution. This rearrangement results in a slight stabilization of the emitting states. This stabilizing effect is evidenced by small bathochromic shifts in the emission bands, differences in the population of the 3 MLCT dF(CF 3 )ppy/dtbpy (See Supplementary Fig. S8 and Supplementary  Table S7), and a reduction of the excited-state lifetime. The observed shifts in emission profile and reduction in the lifetime for Ir[tBu]-PF 6 vs Ir[tBu]-BAr F 4 also agree well with previous literature results. 19,32,33,35,36 In both complexes, ∆ε shows a much larger charge delocalization for Species A over Species B, as inferred from the large polarizability volume. 62 This makes sense in light of the highly mixed nature of the Frank-Condon and high-lying excited-states that result in LLCT. This could be important to understanding and improving the efficiency of a photoredox catalyst since the charges in Species A are more dispersed than in Species B. Extending the lifetime of Species A or chemically pre-associating the photoredox catalyst with target substrate could enhance the likelihood of charge transfer due to Species A having a larger dispersion of charge and thus improve the overall reaction.
Data from the Ir[tBu]-BAr F 4 complex provides the magnitude of the intramolecular dipole moment, which we use to compute absolute dipoles for each excited-state Species in table 4. We find that the dipole moment we observe for Ir[tBu]-BAr F 4 does not meet with our simplistic center-of-mass approximation, which suggests an MLCT-state dipole moment of 16.8 D. Instead, we observe a transient dipole moment of only 5.2 ± 1.6 D in the fully relaxed MLCT (Species B). This discrepancy could partly be accounted for if the ground and excited-state dipole moments oppose each other. However, even exact opposition would not make up the whole difference. We can eliminate the possibility that the small observed transient dipole moment is a more subtle manifestation of ion-pair reorganization by BAr F 4 , as the neutral homoleptic complex Ir(ppy) 3 (see SI section Non-ionic Iridium(III) Complex) displays an even smaller change in excited-state dipole moment (see Supplementary Table S6). Rather, we posit that the excited-state dipole moment is intrinsically small in these complexes, consistent with the high degree of mixing between the orbitals involved. This indicates that the simplistic picture of the lowest MLCT state as an integer charge localized on the bpy ligand is incorrect for this complex, and is likely similar in other related Ir(III) photoredox catalysts.
Knowing the intramolecular dipole moment and the observed dipole moment of the Ir[tBu]-PF 6 complex, the approximate distance between the iridium cation and phosphorous anion can be calculated using equation (14) for the ground-state and each excited-state Species. Table 5 summarizes this change in distance for Ir[tBu]-PF 6 . Distance calculations show that the PF − 6 counter anion is not effected by Species A, this could be because the ground-state dipole of the Ir[tBu] + is pointing in the same direction as the Species A dipole, or because Species A decays faster than the reorganizational response time of the PF − 6 counter anion. In Species B, the distance between the ions decrease by 0.4 Å which can be explained by the intramolecular dipole moment of Ir[tBu] + reorganizing to point away from the dF(CF 3 )ppy ligands as the excited electron settles closer to the dtbpy ligand. This change in dipole direction causes the PF − 6 counter anion to move between the two dF(CF 3 )ppy ligands of the Ir[tBu] + complex which, due to the smaller size of the dF(CF 3 )ppy ligands compared to the dtbpy ligand, allows the PF − 6 to get closer to the positively-charged iridium center.

Conclusion
In this work we have applied modern computational methods to revitalize time-resolved dielectric-loss spectroscopy as a foundational tool for studying the ground and excited-state electronic configuration of molecules in solution. Electromagnetic simulations of the microwave cavity response function and calculation of molecular dipolar relaxation time improve on the accuracy and broad applicability of the technique. We verify our methods by showing agreement with past literature on the model compound N,N-dimethylaminobenzonitrile. Application of TRDL to ([Ir[dF(CF 3 )ppy] 2 (dtbpy)]X), a photocatalyst of great interest in the modern renaissance of light escalated chemistry, reveals dramatic changes in the dipole moment in both ground and excited-states depending on the identity of the counter ion X − . Contact-ion pairing is observed in 1,4-dioxane when X = PF − 6 , and there is a negative change in the dipole moment upon photoexcitation. In contrast, when X = BAr F 4 we observe no evidence of contact ion pairs, and the change in dipole moment upon photoexcitation is positive, as expected. These results are consistent with an excited-state ion-pair reorganization mechanism that cancels the dipole moment of the lowest-lying MLCT state when X = PF − 6 . We suggest that this phenomenon may explain the recently observed dependence of photocatalytic activity on counter-ion identity in similar iridium(III)-based photocatalysts. 25 Ion pair reorganisation can be expected to reduce the excited-state potential energy and the proposed migration of the counter anion to the most positively charged region of the complex may limit the ability of these catalysts to drive oxidative chemistry. Finally, unrelated to ion pair dynamics, the measured dipole moment of the relaxed MLCT dtbpy state in [Ir[dF(CF 3 )ppy] 2 (dtbpy)]X is much smaller than naive point-charge calculations suggest, which, combined with a small polarizability volume indicate a rather localized final state of mixed metal-ligand character, in contrast with the highly polarizable initial MLCT states observed on the timescale of our instrument response. Future work should explicitly test to what extent ion pair reorganisation, the excited-state dipole moment and polarizability volume of photoredox catalysts may influences their reactivity.