Bathophenanthroline Disulfonate Ligand-Induced Self-Assembly of Ir(III) Complexes in Water: An Intriguing Class of Photoluminescent Soft Materials.

Strong evidence of concentration-induced and dissolved electrolyte-induced chromophore aggregation has been universally observed in numerous water soluble bis-cyclometalated Ir(III) photosensitizers bearing the sulfonated diimine ligands bathophenanthroline disulfonate and bathocuproine disulfonate. This new class of aqueous-based soft materials was highly photoluminescent in their aggregated state where detailed spectroscopic investigations of this phenomenon revealed significant blue shifts of their respective photoluminescence emission spectra with concomitant increases in excited-state lifetimes and quantum yields initiating even at micromolar chromophore concentrations in water or upon the addition of a strong electrolyte. A combination of nanoscale particle characterization techniques, static and dynamic photoluminescence spectroscopic studies, along with atomistic molecular dynamics (MD) simulations of these soft materials suggests the formation of small, heterogeneous nanoaggregate structures, wherein the sulfonated diimine ancillary ligand serves as a pro-aggregating subunit in all instances. Importantly, the experimental and MD findings suggest the likelihood of discovering similar aqueous aggregation phenomena occurring in all transition-metal complexes bearing these water-solubilizing diimine ligands.

Introduction

Cyclometalated Ir(III) complexes have realized a surge of recent interest because of their impressive photophysical properties, including substantial photoluminescence emission tunability and excellent photostability.[1,2] Various ligand coordination geometries make possible an array of Ir(III) chromophores featuring tailored properties, ideal for applications including lighting,[3−5] biological probes,[6,7] and photocatalysis.[8,9]

In order to make these complexes compatible with aqueous-based applications, appropriate solubilizing strategies are often necessary.[6,10,11] Bathophenanthroline disulfonate (BPS) disodium salt and bathocuproine disulfonate (BCS) disodium salt, first reported in the 1950s as water-soluble reagents for the spectrophotometric determination of iron[12] and copper,[13] respectively, are commercially available molecules commonly used as diimine ligands in transition-metal complexes to promote aqueous solubility. Examples in the literature include those of ruthenium,[14,15] platinum,[16,17] and iridium.[11,18] Here, we have synthesized bis-cyclometalated Ir(III) complexes 1–5 as shown in Figure , incorporating the BPS and BCS ancillary ligands.

Figure 1

Molecular structures of chromophores 1–7.

While these molecules are readily soluble in water, they also display unanticipated evidence of self-assembly. Dissolution of 1–5 in water immediately produces soapy solutions, illustrated in Figure S1, that are capable of forming photoluminescent bubbles upon nitrogen sparging. Additionally, these aqueous solutions display concentration-dependent photophysical properties, including significant enhancements in excited-state lifetime and quantum yield, suggesting a self-assembly phenomenon. Such photophysical enhancements are similar to those reported in tailor-made Ir(III) aggregates, which often incorporate synthetically complex ligands in order to induce self-assembly, such as metallosurfactants[19,20] and metallomesogens.[21,22]

Previously, we have reported these photophysical changes upon self-assembly in two of the molecules studied here, 1 and 3.[23] Expanding markedly on the scope of that work, several new Ir(III) chromophores are now included along with a Rh(III) structural analog, 6, and a sulfonated Ru(II) complex, 7. A combination of static and dynamic photophysical studies, particle characterization techniques, as well as detailed molecular dynamics (MD) simulations reveal heterogeneous aggregate formation in chromophores 1–6 occurring in the micromolar concentration range in water. Aqueous solutions of the BPS ligand also show evidence of similar self-assembly, suggesting that this ligand acts as a pro-aggregating unit when incorporated into these transition-metal complexes. The high quantum yields and long excited-state lifetimes of the Ir(III) chromophores enabled facile detection of this phenomenon across all concentration ranges investigated. Furthermore, the Ru(BPS)3 chromophore, 7, was also investigated, as it is a well-studied and commonly used water-soluble chromophore incorporating the BPS ligand.[14,15,24−26] Photophysical evidence of self-assembly in 7 was observed at much higher aqueous concentrations (in the millimolar range), consistent with its more hydrophilic molecular structure. Despite the popularity of this Ru(II) chromophore, it is likely that this self-assembly phenomenon has not previously been reported because these photophysical investigations were performed under optically dilute conditions, thereby concealing the effect. Thus, this ligand-induced self-assembly phenomenon likely occurs in a variety of transition-metal complexes that incorporate these sulfonated diimine ligands to promote aqueous solubility. The current work details the relevant parameters useful for identifying and characterizing these interesting classes of water-soluble transition-metal complexes.

Results and Discussion

Emission Enhancement in Self-Assembled Ir(III) Chromophores

Self-assembly in 1–5 is marked by blue shifts in their photoluminescence emission spectral profiles with increasing chromophore concentration in the micromolar concentration range in pure water, as can be gleaned from Figures and S2. Molecule 1 features the greatest overall blue shift in energy. In 2, the emission maximum does not shift; however, the photoluminescence spectrum becomes narrower with the increasing concentration, losing intensity on the red side of its profile. Complexes 3–5 display shifts that are intermediate with respect to that observed in 1 and 2. In methanol, no shifts in emission energy were observed in 1–5 with the increasing chromophore concentration, Figure S3, which is consistent with purely molecular behavior and no aggregation-induced phenomena being observed whatsoever. These concentration-induced photoluminescence emission spectral shifts are therefore indicative of a self-assembly phenomenon occurring exclusively in water. Similar spectral changes have been reported by the De Cola group upon aggregate formation in Ir(III)-based metallosurfactants[19] and have also been reported in mesogenic iridium complexes.[21] These shifts occur as the self-assembly process creates a less polar environment for the chromophores, effectively shielding them from the polar aqueous medium. Ir(III) complexes can be very sensitive probes of their environment because of their charge-transfer excited-state character.[21,27] In bis-cyclometalated diimine Ir(III) complexes incorporating a BPS ancillary ligand, the highest occupied molecular orbital is delocalized over the metal and cyclometalating ligand, rendering the emission an admixture of triplet ligand-to-ligand charge transfer (3LLCT) and triplet metal-to-ligand charge transfer (3MLCT), termed 3LLCT/3MLCT.[11] Depending upon the identity of the ligands, the triplet ligand-centered (3LC) character can be mixed into this excited state as well, yielding a more complex photoluminescence emission profile.[28,29] In complexes with such mixed excited states, it is possible that the change in polarity caused by self-assembly can destabilize the charge transfer excited state resulting in greater mixing of the 3LC character in the emission profile. Typically in bis-cyclometalated diimine complexes, the diimine ligand is associated with the 3MLCT transition and the cyclometalated ligands are associated with the 3LC transition;[28] however, Truong et al. have found through DFT calculations that in bis-cyclometalated Ir(III) complexes incorporating the BPS ligand, the diimine ligand may also contribute the 3LC character to the emission profile.[11] The high energy shoulders on the emission spectrum of 2 for example indicate some mixing of an 3LC state, which is not as susceptible to medium polarity effects. Hence, this molecule produces the smallest solvatochromic shift upon aggregate formation in this series of chromophores. The broad unstructured photoluminescence spectrum of 1 is consistent with possessing greater charge-transfer character in its lowest energy excited state. These solvatochromic effects were confirmed by examining the changes in photoluminescence emission spectra of 1–5 in solvents of greater and lesser polarity, Figure S4, which are comparable to those observed in the concentration-dependent spectra acquired in water.

Figure 2

(a) Concentration-dependent aqueous photoluminescence emission spectra and (b) concentration-dependent excited-state decays in air-equilibrated water.

Concentration dependent excited-state lifetimes also revealed this self-assembly phenomenon in aqueous solution. With increasing compound concentrations, lengthened excited-state decay profiles were observed in 1–5, Figures and S5. Additionally, complex photoluminescence intensity decay fittings were required at increased aqueous concentrations because single exponential functions were inadequate to model the kinetics. In 1, self-assembly was apparent in its excited-state decay profile at concentrations exceeding 20 μM in water; below this concentration, the excited-state lifetime was single exponential and remained constant with the increasing chromophore concentration, suggesting purely molecular behavior and no aggregation. From 50 to 100 μM in water, a long lifetime component was observed because of the self-assembly process, necessitating a bi-exponential fit to adequately model the intensity decay data. This suggested the likelihood of having two distinct chromophore environments under these experimental conditions. Further increasing the concentration led to excited-state intensity decay kinetics that were best fit using a stretched exponential function, indicative of a heterogeneous chromophore environment.[30−32] Similar trends in photoluminescence intensity decay fitting were observed in 3–5, where the specific fitting regimes occurred at different aqueous concentrations depending upon the identity of the chromophore. Complex 2 exhibited the hallmarks of self-assembly at the lowest aqueous concentrations of the molecules studied here; even in 2 μM solution, the photoluminescence intensity decay contained a minor long lifetime component, suggesting that some of the molecules are already being shielded from the solvent bulk. Selected kinetic fits and the associated residuals for 1 and 2 are presented in Figure S6. In methanol, excited-state decays for 1–5 were exclusively single exponential and do not show any concentration dependence, Figure S7. These dynamic results echo those from the static photoluminescence data for the same molecules measured in methanol (Figure S3), suggesting that no aggregation occurs whatsoever in this solvent in molecules 1–5.

Because of the complexity of the fitting required in the concentration range of interest, we have calculated the average lifetimes when making direct comparisons between samples. The average excited-state lifetimes, ⟨τ⟩, for a bi-exponential fit and a stretched exponential fit are given by eqs and 2, respectively. In eq , A1 and A2 are the pre-exponential factors and τ1 and τ2 are the corresponding lifetimes of the bi-exponential fit. This is also referred to as the intensity average lifetime and represents the average time that the chromophore spends in the excited state, with each population weighted by its contribution to the emission intensity. In eq , β is a fitting parameter indicative of the deviation of the fit from a single exponential function, where 0 < β ≤ 1, τ is the lifetime of the stretched exponential fit, and Γ is the gamma function.[33]

Sillen and Engelborghs have shown that for bi-exponential excited-state decays, calculations of radiative and nonradiative decay constants necessitate the use of a different average lifetime. ⟨τa⟩ is the amplitude weighted average lifetime[34] and is given in eq .

Photoluminescence emission enhancement was also observed in the photoluminescence quantum yields measured for 1–5. The overall enhancement was greatest in 1, increasing in the quantum yield from 3% in a 10 μM sample to 15% in a 1 mM sample. Concentration-dependent quantum yields for 1–5 measured in water are presented in Figure S8. These concentration-dependent quantum yields and their corresponding amplitude average excited-state lifetimes, calculated as discussed above, were used to calculate the rates of radiative (kr) and nonradiative (knr) decay for air-equilibrated solutions. The associated rate constants reveal sharp decreases in knr for 1–5 upon the increasing concentration, while kr remained relatively constant, Figure S9.

In self-assembled molecular systems, related enhancements in excited-state lifetime are often associated with processes that decrease the rate of nonradiative decay (knr). The De Cola group has observed such photophysical property enhancements in Ir(III) and Re(I) metallosurfactants upon self-assembly.[19,35] Szerb et al. have also observed extensions of excited-state lifetime upon liquid-crystal formation in Ir(III) mesogens.[21] There are clearly numerous pathways by which aggregate formation can suppress nonradiative decay in transition-metal chromophores. Self-assembly can induce the formation of a more rigid aggregated environment, suppressing relaxation from numerous nonradiative deactivation pathways such as collisions with solvent molecules. Additionally, the energy gap law may also be a contributing factor here, as the energy of emission increases with the extent of self-assembly. Another significant factor for 1–5 appears to be that the aggregate protects the excited state from quenching by diffusing oxygen. A comparison of air-equilibrated and air-free aqueous solutions of 1–5 shows minimal quenching of the excited states intensity decays with the introduction of oxygen, Figure S10, despite lifetimes in the hundreds of nanoseconds time regime. For example, in a 1 mM sample of 1, the intensity average excited-state lifetime in an air-free sample is 463 ns. Upon the introduction of oxygen to the solution, the intensity average excited-state lifetime only decreases to 393 ns. By comparison, in methanol where the complex behaves molecularly, significant quenching was observed upon the introduction of oxygen to a solution of 1, Figure S11, and the excited-state lifetime decreases from 566 to 100 ns.

Aggregate Characterization

Dynamic light scattering (DLS) was used to estimate the size of the aggregates formed in 1–5 in water as a function of concentration, Figure S12. While we do see evidence of self-assembly in the DLS experiments, it is important to consider the assumptions underlying data analysis in the DLS method. The Stokes–Einstein equation assumes the free diffusion of neutral spherical particles. Ionic particles, which diffuse faster because of Coulombic repulsions, can be markedly underestimated by this method. All Ir(III) complexes in this study showed evidence of a smaller particle size than anticipated using DLS experiments, with hydrodynamic radii roughly ranging between 1 and 10 nm. In addition, each molecule also showed evidence of larger aggregate formation with radii between 40 and 80 nm. However, this larger peak is not a likely representative of the sample on average, as the presence of such a large particle in any significant population would necessarily screen the signal of the smaller particle in DLS.[36] Additionally, a smaller aggregate is also consistent with the lack of Mie scattering observed in UV–vis absorption experiments on these same samples.

To further confirm the presence of aggregates, cryo-scanning electron microscopy (cryo-SEM) was used to obtain images of 1 in frozen water, Figure . This will not necessary be indicative of the self-assembly in a fluid solution and was merely used to obtain a snapshot of the average aqueous environment where significant evidence of self-assembly was occurring. The cryo-SEM data revealed spherical shaped particles with a radius of approximately 12 nm for the sample on average, qualitatively consistent with the DLS data. A few larger particles were also observed in the image, which are likely larger clusters of smaller aggregates.

Figure 3

Cryo-SEM image of 1 mM solution of 1 in frozen aqueous solution.

Atomistic MD simulations of 3 at varying molecule abundance (proportional to chromophore concentration) further elucidate the aggregate size as a function of transition-metal complexes available for assembly, consistent with the concentration dependence observed in photophysical studies. The resulting aggregates possess hydrodynamic radii between 3 and 6 nm, in agreement with DLS and cryo-SEM measurements. Figure a reveals the self-assembled cluster characteristics of three different chromophore abundances in aqueous solution. As the molecule abundance increased, the aggregation number (gray bar) and radii of both hydrodynamic (black square) and gyration (black dashed circle) of the largest cluster also increased. In order to explain the shape of the resultant aggregates, a shape factor was calculated as a ratio of radius of gyration to hydrodynamic, Rg/Rh. As the cluster size increased, the shape factor decreases, indicating that cluster growth occurred radially approaching a spherical aggregate (red triangle). Interestingly, the shape factor revealed a propensity for growth to occur spherically, but snapshots of the clusters revealed some directionality, Figure b. This can most easily be observed in the largest abundance calculated (128 molecules) where the aggregate appears to have segments of transition-metal complexes that merge together in multiple places forming junctures. Heatmaps of all clusters existing in the last 10 ns of the highest abundance simulation, Figure c-left, indicated volatility of the very weakly assembled aggregate in the presence of water. Between 111 and 114 ns, complexes or groups of molecules depart one aggregate and adhere to another stable aggregate or collide without permanently assembling. From the heatmap, a weighted ensemble average aggregation distribution reflects the relative dispersion of the resulting system of aggregates, Figure c-right. The inset in the upper right of the histogram is a rendering of the final morphology of the entire system, excluding water for clarity. All aggregation heatmaps and aggregation distributions can be found in Figure S13.

Figure 4

Summary of abundance impact on aggregation: (a) comparison of radius of gyration (black dash circle), hydrodynamic radius (solid black square), and shape factor (red triangle); (b) snapshots of the resulting largest cluster from each of the three different abundance simulations consisting of 64 (top), 96 (middle), and 128 (bottom) complexes forming 16, 27, and 36 complex aggregates, respectively; (c) cluster aggregation heatmap of a 128 molecule simulation (left) with probability distribution of aggregation number (right) for the last 10 ns including an inset snapshot of the resulting simulated system.

Effect of Salt (Dissolved Electrolytes) and Temperature on the Self-Assembly Process

In addition to the effect of the chromophore concentration, the self-assembly of 1–5 was also influenced by the ionic strength of the aqueous solution. The addition of sodium chloride produced enhancements in quantum yield and excited-state lifetime that are similar to those observed upon increasing the chromophore concentration in water. This effect was most pronounced at low chromophore concentrations; for example, the addition of sodium chloride to a 1 mM solution of 1, which was already aggregated, produces negligible effects on the excited-state lifetime, Figure S14. Similar results were obtained by the addition of potassium chloride and potassium nitrate, Figure S15.

This effect is similar to the salt effect well known in ionic surfactants, where increasing the ionic strength screens the Coulombic repulsions between the charged headgroups of surfactant molecules and lowers the critical micelle concentration. The surface tension of 1 and 3 was measured in the micromolar concentration range and did not show any changes with concentration, indicating that these are not surfactant-like molecules, Figure S16. However, as these are ionic complexes, it is likely that salt produces a similar effect, promoting self-assembly at lower chromophore concentrations.

This was further explored in MD simulations yielding a similar effect of ionic strength on self-assembly. Evident in the largest cluster snapshots comparing no salt with 0.75 M NaCl (Figure a), increasing ionic strength of the solvent promotes an increase in chromophore aggregation. Clustering heatmaps and aggregation histograms for all salt concentrations and abundances can be found in Figure S13. A cluster size analysis of the impact due to increasing salt concentration is shown in Figure b. Adding 0.25 and 0.5 M salt steadily increases the aggregation (Naggregation) and general size (Rg and Rh) of the largest cluster assembled. Upon increasing the salt concentration to 0.75 M, the aggregation number no longer increased, and its size actually decreases. This illustrates that the addition of salt also increases the propensity to form a more spherical assembly (Rg/Rh). Quantifying the number of sodium ions within the proximity of the charged headgroup (specifically the oxygen anion) at increasing ionic strength illustrates the screening effect created by the addition of salt (Figure c). The inset in Figure c is a radial pair distribution function (RDF) between oxygens atoms of the headgroup and sodium ions. The RDF shows a clear orientation preference, and focusing on the first and second peaks, the ensemble average of sodium ions can be counted within the closest and second closest proximities. As the ionic strength of the solution is increased, more headgroups participate in screening, further supporting the necessity of screening to promote self-assembly.

Figure 5

Summary of the salt concentration impact on aggregation: (a) largest cluster comparison between a simulation with no salt vs 0.75 M salt solution; (b) comparison of the radius of gyration (black dash circle), hydrodynamic radius (solid black square), and shape factor (red triangle); (c) number of sodium ions within 3 and 5.5 Å of a headgroup oxygen, inset is an oxygen-sodium ion RDF; (d) electrostatic and van der Waals pairwise interactions within a cluster and headgroup oxygens with solvent components; and (e) SASA comparison of each complex belonging to the largest cluster formed without salt and 0.75 M salt solution.

Nonbonded energetic contributions of the aggregates provided further means of understanding the impact of salt on the assembly process, Figure d-left. Electrostatically, the energy between complexes within a cluster increases as the aggregate grows; specifically, the contributions from the headgroup oxygen drive the assembly process. In essence, Coulombic repulsion between similar oxygen atoms in immediately neighboring molecules tends to destabilize aggregation, whereas Coulombic attraction between an Ir(III) center on one complex and an anionic oxygen on a neighboring molecule promotes aggregation. Prior to the addition of the NaCl electrolyte, the interactions are relatively mild between the headgroup oxygen and water. As NaCl is added to the solution, Na+ ions along with their hydration shell tend to screen these electrostatic contributions. Because there are more repulsive centers than attractive ones, the addition of the electrolyte and the resultant electrostatic screening favors aggregation by minimizing repulsions. Interestingly, another nonbonded interaction trend was the lack of change in the number of π–π stacks per complex as aggregation and NaCl concentration increased, Figure S17. Further, van der Waals interactions play a relatively small role in the energetic landscape of the assembled cluster, Figure d-right, highlighting the importance of ionic screening in the self-assembly process.

Figure e provides distributions of the solvent accessible surface area (SASA) for each individual complex participating in the largest cluster of the system without salt (left) and with 0.75 M salt (right). Highlighted colors on the histogram translate to snapshots of the cluster immediately to the right of the distributions. The blue shading indicates complexes with the smallest SASA, or most buried in the cluster, green shading indicates a medium SASA, and red shading highlights the most exposed complexes found in the aggregate. Interestingly, the inner most complexes or lowest SASAs are not found in the general center of the cluster but instead are located at key junctures where the directional growth of the aggregate splits or diverges. Additionally, the histogram belonging to the 0.75 M salt aggregate reveals a more Gaussian distributed assembly, highlighting the increased stability of the cluster by the presence of salt.

The self-assembly can also be manipulated through temperature variation. Excited-state decay measurements at increasing temperatures produce a decrease in the average excited-state lifetimes, suggesting that the aggregation can be broken up by the addition of heat; that is, self-assembly is exothermic. Temperature-dependent excited-state lifetimes for three concentrations of 1 and 3 are shown in Figures and S18, respectively. As the temperature approaches the boiling point of the water, the average lifetimes of each concentration converge. At a temperature of 25 °C, a 50 μM solution of 1 has a monoexponential excited-state decay. Upon heating this sample, a monoexponential lifetime of approximately 60 ns was observed at temperatures up to 90 °C, suggesting a monomeric, nonaggregated species. However, if we decrease the temperature of this sample below room temperature, we see the development of a longer lifetime component, necessitating a biexponential decay fitting, and suggesting aggregate formation. Repeating this experiment with a 100 μM sample of 1, which already evidences self-assembly at room temperature, showed a sharp decrease in the average excited-state lifetime with increasing temperature, and the monoexponential lifetime of 60 ns was now observed at temperatures of 70 °C through 90 °C. In a 200 μM sample of 1, a temperature of 90 °C was required to achieve monoexponential kinetics. Qualitatively similar behavior was observed in the temperature-dependent excited-state lifetimes of 3. These data strongly suggest that in these self-assembled molecular systems, the addition of heat shifts the equilibrium, forming predominately monomeric, de-aggregated species.

Figure 6

Temperature-dependent excited-state lifetimes of 1 in air-equilibrated water at the concentrations specified in the legend.

This temperature-dependent aggregation phenomenon was further exemplified by modeling the chromophores with MD at temperatures starting at 26.85 °C and increasing up to 176.85 °C. As the simulation temperatures were increased, clustering heatmaps show that the clustering becomes more unstable and aggregation decreases (Figure a). Observing the images inset in each cluster heatmap, it becomes somewhat apparent that the clusters are extending conformation as the temperature increases. This is further examined by similar cluster properties as with salt and abundance (Figure b). For the first increase in temperature to 77 °C, the aggregation of the largest cluster increases slightly to Naggregation = 38, but the Rg/Rh also increases demonstrating the increased penchant for elongation. As temperature increases further, aggregation steps down but elongation increases steadily. This reveals that as heat is added and the systems become more dynamic, the complexes are unable to stabilize into a large cluster relative to that it achieves at lower temperatures.

Figure 7

Summary of temperature impact on aggregation in water: (a) clustering heatmaps for increasing temperature; (b) comparison of the radius of gyration (black dash circle), hydrodynamic radius (solid black square), and shape factor (red triangle); (c) most probable aggregation number prediction map for all simulated environments including abundance, salt, and temperature.

Investigating the Self-Assembly Properties of the BPS Ligand

Aqueous samples of the BPS ligand also show evidence of self-assembly through particle characterization techniques. Light scattering was observed in DLS experiments in aqueous solutions of this ligand, Figure S19. Similar to the Ir(III) complexes, a small peak around 1 nm and a larger peak around 80 nm were observed in DLS experiments, however, the larger peak was much more prominent in this sample. Cryo-SEM images are also fairly consistent with the DLS data, showing a small, roughly spherical aggregate with an average radius of 12 nm, similar to that observed in 1. We also observed several larger aggregates in the images of the ligand, Figure , which were not observed in the cryo-SEM images of 1.

Figure 8

CryoSEM image of the 1 mM BPS ligand in the frozen aqueous sample.

Rh(III)- and Ru(II) BPS-Containing Chromophores

To better understand the self-assembly process and whether this phenomenon is likely universal, two other transition-metal chromophores incorporating the BPS ligand were investigated in water for evidence of a similar self-assembly. The bis-cyclometalated Rh(III) complex, 6, is isostructural to 1. Upon dissolution in water, similar formation of a soapy solution was observed. This complex is weakly emissive in air-equilibrated aqueous solution and is completely nonemissive in air-equilibrated methanol. Although not as well studied as their Ir(III) counterparts, the excited states of bis-cyclometalated diimine Rh(III) complexes are commonly reported in the literature as being predominately 3LC on the cyclometalating ligand, with some mixing of the 3MLCT state, and typically show structured emission profiles.[37−39] However, few examples of unstructured, or less structured, emission from bis-cyclometalated Rh(III) complexes has also been reported.[40,41] Ohsawa et al. have assigned such broad emission from bis-cyclometalated Rh(III) complexes to 3LC transition on the cyclometalating ligand based on a small red shift between the 77 K emission and room-temperature solution spectra; however, it is generally acknowledged that the strongly mixed emission spectra in these complexes are not simply classified. 6 shows an unstructured emission profile at room temperature in the aqueous and organic solvents studied here, suggesting mixing of a 3MLCT excited state, but does not show any solvatochromic shift upon changing solvent polarity, Figure S20, consistent with an 3LC excited state, while no shift in aqueous emission energy is observed as concentration is increased and self-assembly occurs, which is consistent with the absence of a solvatochromic shift.

Concentration-dependent excited-state decays of 6 in water showed evidence of aggregate formation similar to Ir(III) complexes 1–5 in the micromolar concentration range, with a lengthening of the excited-state decay profiles as concentration is increased and more complex intensity decay modeling required at greater aqueous concentrations. Corresponding enhancements in quantum yield are also observed with the increasing concentration, Figure S21. DLS data for 6 were comparable to those obtained for 1–5, Figure S22, suggesting a similar aggregate formation.

Additionally, the well-established Ru(BPS)3 chromophore, 7, was also studied to seek evidence of its self-assembly in water. This compound was first reported in 1985[24] and has since been widely studied as a water-soluble photosensitizer and O2 sensor.[14,15,25,26] This complex was of particular interest for this study because it contains the BPS ligand suspected of inducing self-assembly, but is structurally distinct from the bis-cyclometalated diimine complexes of Ir(III) and Rh(III). In the micromolar concentration range, a consistent and single-exponential excited-state decay is observed, and there is no shift of the emission energy with the increasing concentration, consistent with previous literature reports. However, upon increasing the concentration to 2.5 and 5 mM, we observed the development of a long lifetime component, and a concomitant red shift in the emission energy, Figure S23. Red-shifted emission spectra have been previously observed in the self-assembly of other Ru(II) chromophores.[42−46] Further, sodium chloride was added to a 100 μM solution of 7 to test the effect of salt on the system and produced enhancements in excited-state lifetime consistent with aggregate formation, Figure S24. The DLS spectrum for a 2 mM sample of 7 was similar to that observed in 1–6, further confirming the presence of an aggregate, Figure S25. While this self-assembly has not been previously reported, it is because photophysical measurements have never been performed at millimolar concentrations of 7 in water. The greater hydrophilic nature of this particular Ru(II) complex necessitates higher chromophore concentrations or ionic strength to induce the requisite photophysical property enhancing self-assembly. The observation of self-assembly in this complex further supports the identification of the BPS ligand as the pro-aggregating unit in the Ir(III) complexes studied here and suggests that this effect is not limited to the bis-cyclometalated diimine molecular structure.

Conclusions

A series of five water-soluble bis-cyclometalated Ir(III) photosensitizers have been synthesized, incorporating the sulfonated diimine ligands BPS and BCS. Molecules 1–5 show indisputable evidence of self-assembly in water in all instances, easily detected in the micromolar concentration range using standard photophysical techniques. In these Ir(III) complexes, self-assembly resulted in significant enhancements to their photophysical properties, with excited-state lifetimes and quantum yields increasing by up to a factor of 5 over the concentration range studied. Studies of the free BPS molecule in water also showed evidence of self-assembly, implicating the ligand in driving the self-assembly when incorporated into these Ir(III) complexes to promote aqueous assembly. This was further supported by the observation of self-assembly in the isostructural Rh(III) complex, 6, as well as in the structurally distinct Ru(II) chromophore 7, which incorporates three BPS ligands and shows evidence of self-assembly at much higher concentrations, consistent with its greater hydrophilic nature.

Increasing the ionic strength of the aqueous solutions in these complexes also results in swifter formation of self-assembled aggregates featuring enhanced photophysical properties. Cryo-SEM characterization experiments and MD simulations of the Ir(III) complexes revealed a small, heterogeneous aggregate formation induced by the sulfonated diimine ancillary ligand. MD simulations provided a unique insight into the self-assembly process, showing the growth of the aggregate being promoted by both increasing molecular concentration and by the addition of salt, through increased aggregate stability by ionic screening of the charged headgroups. Given the ubiquity of these sulfonated ligands for aqueous solubilization in transition-metal complexes and the combined results presented in this contribution, self-assembly must be considered when investigating the photochemical properties of transition-metal chromophores bearing sulfonated diimine ligands.

Experimental Section

Synthesis and Molecular Characterization

All reagents and solvents were purchased from commercial sources and used as received. Proton (1H) NMR spectra were recorded on a Varian 400 MHz spectrometer at room temperature unless otherwise indicated. Elemental analyses were performed by Atlantic Microlab, Inc. For 1–6, in the solid powder, each sulfonate group retains a sodium ion, and the complex contains a chloride counterion. This is manifest in the elemental analyses of these complexes and has also been reported previously in the literature.[47] Mass spectrometry was performed by the Michigan State University Mass Spectrometry Core.

Cyclometalated chlorobridged Ir(III) and Rh(III) dimers were synthesized according to literature procedures.[48,49] The F-mppy ligand was synthesized via the Kröhnke method following literature procedures.[50] The BPS and BCS ligands were purchased from commercial sources as a mixture of isomers with respect to the position of the sulfonates. Sulfonation can occur in either the meta or the para position of each phenyl, giving three possible ligand isomers (m–m, m–p, and p–p). The ratio of these isomers will vary by supplier and batch, giving different proton NMR signals and integrations. The use of these isomeric mixtures has been well established in the literature.[14,16,51] As it has been determined that these isomers cannot be distinguished through photophysical or electrochemical measurements,[18,26] commercially available isomeric ligands were used as received. No difference in the photophysical behavior of complexes synthesized was observed using different batches of ligands.

[Ir(ppy)2BPS] (1)

Bis-(μ)-chlorotetrakis(2-phenyl-pyridinato-C2,N)diiridium(III) and bathophenanthrolinedisulfonic acid disodium salt hydrate (BPS) (2 molar equivalents) were added to a round-bottom flask, along with 9:1 ethanol/water (30 mL). This solution was degassed by N2 purging for 20 min. The solution was then heated to reflux for 12 h under N2. Upon cooling, the reaction mixture was filtered, and diethyl ether was added to the filtrate to precipitate the product as an orange/yellow solid. The precipitate was collected by filtration and washed with diethyl ether. The product was recrystallized by vapor diffusion of diethyl ether in methanol. Yield: 83%. 1H NMR (400 MHz, CD3OD, δ): 8.39–8.41 (m, 2H), 8.21 (s, 2H), 8.12–8.14 (d, 2H), 8.0–8.04 (m, 4H), 7.8–7.87 (m, 6H), 7.61–7.72 (m, 6H), 7.02–7.06 (m, 2H), 6.91–6.97 (m, 4H), 6.41–6.43 (d, 2H); HR-ESI-MS m/z: (M–) calcd, 991.1236; found, 991.1211. Anal. Calcd for C46H30ClIrN4Na2O6S2·4H2O: C, 48.27; H, 3.35; N, 4.89. Found: C, 48.38; H, 3.45; N, 5.03.

[Ir(bt)2BPS] (2)

Reaction conditions for 1 were employed, using the bt-cyclometalated Ir(III) dimer as the starting material. Yield: 81%. 1H NMR (400 MHz, CD3OD, δ): 8.60–8.63 (m, 2H), 8.20–8.22 (m, 2H), 7.98–8.06 (m, 12 H), 7.66–7.75 (m, 4H), 7.30–7.33 (m, 2H), 7.14–7.18 (m, 2H), 6.99–7.03 (m, 2H), 6.92–6.96 (m, 2H), 6.53 (d, 2H), 6.06 (d, 2H); HR-ESI-MS m/z: (M–) calcd, 1103.0677; found, 1103.0698. Anal. Calcd for C50H30ClIrN4Na2O6S4·5H2O: C, 47.11; H, 3.16; N, 4.40. Found: C, 47.01; H, 3.30; N, 4.43.

[Ir(F-mppy)2BPS] (3)

Reaction conditions for 1 were employed, using the F-mppy cyclometalated Ir(III) dimer as the starting material. Yield: 64%. 1H NMR (400 MHz, CD3OD, δ): 8.41–8.44 (m, 2H), 8.25 (s, 2H), 8.00–8.07 (m, 7H), 7.87–7.92 (m, 4H), 7.66–7.76 (m, 7H), 7.36 (s, 1H), 6.78–6.83 (m, 1H), 5.97 (d, 1H), 2.04 (s, 6H); HR-ESI-MS m/z: (M–) calcd, 1055.1361; found, 1055.1331. Anal. Calcd for C48H32ClF2IrN4Na2O6S2·8H2O: C, 45.02; H, 3.78; N, 4.37. Found: C, 45.05; H, 3.68; N, 4.48.

[Ir(ppz)2BPS] (4)

Reaction conditions for 1 were employed, using the ppz-cyclometalated Ir(III) dimer as the starting material. Yield: 90%. 1H NMR (400 MHz, CD3OD, δ): 8.55–8.58 (m, 4H), 8.22–8.23 (m, 2H), 8.03–8.08 (m, 5H), 7.87–7.89 (m, 2H), 7.68–7.76 (m, 5H), 7.56–7.58 (m, 2H), 7.07–7.12 (m, 4H), 6.89–6.93 (m, 2H), 6.56–6.57 (m, 2H), 6.45 (d, 2H); HR-ESI-MS m/z: (M–) calcd, 969.1140; found 969.1142. Anal. Calcd for C42H28ClIrN6Na2O6S2·4H2O: C, 44.94; H, 3.23; N, 7.49. Found: C, 44.79; H, 3.50; N, 7.42.

[Ir(ppy)2BCS] (5)

Reaction conditions for 1 were employed, using the bathocuproine disulfonic acid disodium salt for the diimine ligand. Yield: 83%. 1H NMR (400 MHz, CD3OD, δ): 8.14–8.16 (m, 2H), 8.07 (s, 2H), 8.02–8.05 (m, 4H), 7.86–7.90 (m, 2H), 7.78–7.79 (m, 4H), 7.72–7.74 (m, 2H), 7.65–7.69 (m, 4H), 7.01–7.04 (m, 2H), 6.94–6.97 (m, 2H), 6.75–6.79 (m, 2H), 6.20 (d, 2H), 2.24 (s, 6H); HR-ESI-MS m/z: (M–) calcd, 1019.1549; found, 1019.1531. Anal. Calcd for C48H34ClIrN4Na2O6S2·7H2O: C, 47.00; H, 3.94; N, 4.57. Found: C, 46.97; H, 3.78; N, 4.58.

[Rh(ppy)2BPS] (6)

Reaction conditions for 1 were employed, using the bis-(μ)-chlorotetrakis(2-phenyl-pyridinato-C2,N)dirhodium(III) complex as the starting material. Yield: 78%. 1H NMR (400 MHz, CD3OD, δ): 8.45–8.47 (m, 2H), 8.15–8.20 (m, 4H), 8.02–8.05 (m, 4H), 7.90–7.94 (m, 5H), 7.84–7.87 (m, 2H), 7.66–7.73 (m, 5H), 7.60 (m, 2H), 7.10–7.14 (m, 2H), 6.98–7.04 (m, 4H), 6.44–6.46 (d, 2H); HR-ESI-MS m/z: (M–) calcd, 901.0662; found, 901.0677. Anal. Calcd for C46H30ClN4Na2O6RhS2·3H2O: C, 53.26; H, 3.50; N, 5.40. Found: C, 53.52; H, 3.66; N, 5.45.

[Ru(BPS)3] (7)

Synthesized according to literature procedures.[14]

Spectroscopy

All samples were measured in 1 cm2 quartz cuvettes in air-equilibrated HPLC-grade water or spectroscopy grade solvent, unless otherwise indicated. Aqueous air-free samples of 1–5 were prepared by slow argon sparging for 40 min.

Absorption spectra were recorded on a Shimadzu UV-1800 double beam spectrophotometer or an Agilent 8453 diode array spectrophotometer. Emission spectra were acquired using Edinburgh Instruments FS920 or FS980 spectrofluorometer equipped with a 450 W xenon-arc lamp as the excitation, and samples were excited into the lowest energy absorption feature. The FS920 has a Peltier cooled, red-sensitive PMT detector (R2658P, Hammamatsu), and the FS980 has a R928P Hammamatsu PMT. All emission spectra were corrected for the detector response.

Photoluminescence quantum yields were calculated using the relative actinometry method.[33] Air-equilibrated aqueous solutions of Ru(bpy)32+, having a reported quantum yield of 0.04,[52] were used as the reference. For concentration-dependent quantum yields, because of the range of concentrations measured, the optical densities at the appropriate excitation wavelengths are outside of the optically dilute range for many samples measured. To accurately measure these samples and account for inner filter effects, the absorbances of the standard and the sample were matched at the excitation wavelength.[33,53] For each quantum yield measurement, this experiment was performed using at least two different excitation wavelengths, and the resulting quantum yields were averaged.

Concentration-dependent excited-state decays for 1–5 and 7 in air-equilibrated solutions were measured on an Edinburgh Instruments Mini-Tau spectrometer using a 405 nm picosecond laser diode as the excitation source. A band-pass filter centered at 600 nm with a 45 nm width was used to select the emission. The recorded TCSPC decays were fit using Edinburgh Instruments FAST Software, and the best fit was determined after inspecting the residuals and considering the reduced χ2 value of the fit.

The excited-state lifetimes for bubble-degassed samples of 1–5, variable temperature lifetimes of 1 and 3, as well as air-equilibrated samples of 6 were measured using a transient recording setup, where a pulsed N2 dye laser was used for excitation. The emission was spectrally filtered using a monochromator and detected with a Hammamatsu PMT. A Tektronix oscilloscope was used to record the data, and 128 transients were averaged per measurement. Fitting was performed using IGOR Pro software; the best fit was judged by inspection of residuals. For samples 1–5, the excitation wavelength used was 443 nm. For 6, the excitation wavelength was 365 nm at aqueous concentrations of 100 μM and below, and 383 nm for concentrations exceeding 100 μM.

DLS experiments were performed using a DynaPro Nanostar instrument from Wyatt Technology. All measurements were recorded at 25 °C using 663 nm incident laser light. Light scatter was collected at a 90° scattering angle. Prior to measurement, aqueous solutions were filtered using a 0.4 μM pore size filter. Each experiment contained an average of 50 measurements, each using a 5 s acquisition time. The hydrodynamic radii were calculated by Wyatt’s Dynamics 7 software using the Stokes–Einstein equation.

MD Simulation

The MD simulations were conducted with complex 3 using Amber software[54] with interactions from the general Amber force field (GAFF).[55] GAFF is parameterized for small organic molecules and was previously tested for metalated organic ligands.[56,57] The electrostatic potential for the iridium interactions was calculated using the AM1-BCC charge model that emulates the Hartree–Fock theory with the 6-31G* basis set (HF/6-31G*).

Each simulated system was minimized for 500 steps by steepest descent in solution. Followed by minimization, each simulation was run for a production of 120 ns with a timestep of 1 fs in NPT using the Langevin thermostat. A total of 16 simulations were conducted. Computational modeling at the very dilute concentrations is unfeasible because of limits on system sizes; therefore, systems of increasing number of complexes were simulated to emulate the effect of the increasing concentration. Three different abundances of complexes were simulated, each with increasing the number of complexes while reducing the number of water molecules: 64 complexes with 91 051 water molecules; 96 complexes with 89 031 water molecules; and 128 complexes with 87 395 water molecules. Each abundance was also simulated with three additional salt concentrations of 0.25, 0.5, and 0.75 M. All systems described thus far were conducted at 26.85 °C. Additionally, the 128 abundant system without salt was simulated at three increased temperatures of 76.85, 126.85, and 176.85 °C.

Size and energetic analysis of the simulated trajectories was conducted using AmberTools,[54] Discovery Studio,[58] and in-house scripts. Key properties were measured on the largest cluster that resulted from each individual simulation. Because of the dynamic nature of clustering, an aggregate was defined to be any complex where iridium atoms are within 1.5 nm of each other, while the largest cluster was declared to be the largest aggregate that remained stable for a minimum of 1 ns over the last 10 ns of the simulation. The distance of 1.5 nm was chosen based on the first peak of the radial pair distribution function (RDF) of the iridium atoms, Figure S26. Specific properties of the largest cluster were calculated including the radius of gyration (Rg) and hydrodynamic radius (Rh), with the following eqs and 5, respectively

Because the Rh is size based on the assumption of a hard sphere, while Rg is measured from the center of mass of a cluster, in order to compare the changes in cluster shape, a dimensionless shape factor was calculated as a ratio of Rg/Rh. Smaller Rg/Rh reflects a more spherical cluster, while an increased value indicates a more elongated conformation. In order to evaluate complexes located on the interior of the aggregate versus the exterior, the SASA was calculated for each individual in the largest cluster.

Lastly, from the clustering analysis, heatmaps were generated where color reflects the fraction of clusters participating in a specific aggregation number at a calculate time for the last 10 ns. From the heatmaps, a weighted ensemble average for the last 10 ns was used to create a histogram showing the distributions of clusters in the entire system. All snapshots were gathered using VMD.[59]