We report a study of chromophore-catalyst assemblies composed of light harvesting hexabenzocoronene (HBC) chromophores axially coordinated to two cobaloxime complexes. The chromophore-catalyst assemblies were prepared using bottom-up synthetic methodology and characterized using solid-state NMR, IR, and x-ray absorption spectroscopy. Detailed steady-state and time-resolved laser spectroscopy was utilized to identify the photophysical properties of the assemblies, coupled with time-dependent DFT calculations to characterize the relevant excited states. The HBC chromophores tend to assemble into aggregates that exhibit high exciton diffusion length (D = 18.5 molecule2/ps), indicating that over 50 chromophores can be sampled within their excited state lifetime. We find that the axial coordination of cobaloximes leads to a significant reduction in the excited state lifetime of the HBC moiety, and this finding was discussed in terms of possible electron and energy transfer pathways. By comparing the experimental quenching rate constant (1.0 × 109 s−1) with the rate constant estimates for Marcus electron transfer (5.7 × 108 s−1) and Förster/Dexter energy transfers (8.1 × 106 s−1 and 1.0 × 1010 s−1), we conclude that both Dexter energy and Marcus electron transfer process are possible deactivation pathways in CoQD-A. No charge transfer or energy transfer intermediate was detected in transient absorption spectroscopy, indicating fast, subpicosecond return to the ground state. These results provide important insights into the factors that control the photophysical properties of photocatalytic chromophore-catalyst assemblies.
I. INTRODUCTION
Due to their chemical tunability and modular design, molecular photocatalysts that combine light-absorbing chromophores with metal-based electrocatalysts are extensively utilized in solar energy conversion.1–5 Time-resolved studies of energy and electron migration in these chromophore-catalyst dyads provide fundamental mechanistic insights into the factors that control overall quantum efficiencies of molecular photocatalysts. For example, the mechanism of electrocatalysis by cobalt-based electrocatalysts has been extensively studied using time-resolved optical and x-ray spectroscopies.6–13 Seminal work by Gray and Dempsey has resulted in the identification of the key Co(I) and Co(III)-hydride intermediates formed upon reduction or protonation of cobaloxime-based complexes and in the evaluation of the kinetic barriers associated with the competing heterolytic and homolytic catalytic pathways.6,7 Subsequent experiments probing the Co K-edge transitions in the x-ray region have resulted in the enhanced structural information regarding the reduced Co(II) and Co(I) species in cobaloxime, cobalt polypyridyl, and Co macrocyclic complexes.10,12 These studies have shown that the Co(I) intermediate loses axial ligands and adopts a square planar geometry. The dissociated axial ligand, when held covalently to the complex, was proposed to serve as a proton relay for the hydrogen evolution reaction.12 Time-resolved studies are also exceptionally useful in identifying the kinetics of photoinduced charge separation and undesired charge recombination in chromophore-catalyst assemblies.14–18 In some of these reports, short lifetimes for charge-separated species were observed and assigned to the fast charge recombination. Mulfort found that the axial/equatorial chromophore/catalyst geometry does not significantly change the rate of charge recombination.14,19 Wasielewski showed that this charge recombination occurs in competition with the cobaloxime dissociation from the chromophore assembly due to the loss of the axial pyridine ligand.18 A detailed understanding of structural and electronic factors that control recombination kinetics is still lacking.
Graphene quantum dots (GQDs), defined as two-dimensional polyaromatic hydrocarbon flakes of varying size, have emerged as promising chromophores for light harvesting applications. The optical bandgap of GQDs can be readily tuned either by changing the size of the aromatic flake, from 6 eV in benzene to 0 eV in graphene,20 by introduction of functional groups to GQD edges,21 or by introduction of twist to the aromatic plane.22 Recent developments in the bottom-up synthesis of well-defined GQDs23 and their self-assembly into one-dimensional wires with large aspect ratios24–26 have enabled the development of materials with excellent photoconductivity. Furthermore, the successful implementation of GQDs into the walls of metal–organic frameworks27 and branched macromolecules28 has provided a pathway toward three-dimensional GQD materials for light harvesting applications. The coupling of GQDs with other light harvesting elements, porphyrins, has shown that the directionality of energy transfer can be changed by varying the size of the aromatic GQD core.29 Preliminary studies involving the coupling of GQDs with the electrocatalytic metal center, Re-based catalyst for reduction of CO2 to CO, have shown that the photoinduced charge transfer to the metal can take place, leading to photocatalytic performance.30,31
Previously, we investigated the electrochemical and photophysical properties of GQDs.32,33 Light harvesting characteristics of GQD assemblies, such as alkyl-substituted hexabenzocoronene (HBC), were evaluated by determining the exciton size and dynamics using ultrafast pump–probe spectroscopy.32 The exciton coherence length was found to be 1–2 monomeric units, indicating a high degree of static and dynamic disorder in these assemblies. However, the exciton diffusion length measurements, associated with an incoherent hopping process, showed that up to 30 GQD molecules in a one-dimensional π-stacked assembly can be sampled within the exciton lifetime. In this work, we build on these studies by investigating the kinetics of photoinduced charge separation and recombination in chromophore-catalyst assemblies shown in Scheme 1. Being well-studied catalytic motifs, cobaloxime-based hydrogen-evolving catalysts Co-A and Co-B were selected for this study. Light-harvesting HBC chromophores were functionalized with pyridine moieties (HBCPy) to enable axial coordination of cobaloximes CoQD-A and CoQD-B. Using time-resolved optical and mid-IR spectroscopy, we show that the photogenerated excitons, initially centered on the HBC moiety, undergo a fast decay. However, the absence of spectroscopic signatures for charge or energy intermediates in our data prevented us from identifying whether it is the electron or energy transfer that resulted in the excited-state quenching. The results are discussed using Förster/Dexter models for energy transfer and the Marcus model for electron transfer. These results point to the challenges associated with obtaining long-lived charge-separated states in chromophore-catalyst assemblies.
II. EXPERIMENT AND THEORY
Model compounds were synthesized as described in the supplementary material. Chemicals and solvents were purchased from Sigma-Aldrich and were used without further purifications. Solution state 1H and 13C NMR were recorded on a Bruker DPX400 spectrometer operating at 400 MHz for 1H and 100 MHz for 13C. Solid-state NMR spectra were recorded on a Bruker DRX700 spectrometer operating at 176.0 MHz for 13C and 224.0 MHz for 11B with a spinning speed of 30 kHz. MALDI was recorded on a Bruker ultrafleXtreme MALDI-Tof-Tof. UV–vis spectra were recorded on either a Cary 300 Bio spectrometer or an Ocean FX spectrophotometer. Fluorescence spectra, fluorescence lifetimes, and quantum yields were recorded on a Horiba PTI QuantaMaster 8000 spectrometer. IR measurements were performed on a Thermo Scientific Nicolet iS5 FTIR spectrometer. XPS measurements were done on a Kratos AXIS-165.
Co L-edge and N K-edge x-ray absorption spectroscopy (XAS) measurements were done at beamline 8-2 at the Stanford Synchrotron Radiation Lightsource (SSRL). The powder samples were mounted on conductive carbon tape, and the XAS spectra were measured by scanning the incident x-ray energy and detecting the total electron yield, at room temperature in vacuum. The spectra are normalized to the incident x-ray flux measured before the sample and represent an average of 4–9 measurements per sample. The spherical grating monochromator resolution was ∼0.2 eV–0.3 eV, and the energy was calibrated to the L3-edge peak of NiF2 at 852.7 eV and its second harmonic at 426.35 eV. A second order polynomial baseline was subtracted from all spectra, which were subsequently normalized to the average intensity in the region >810 eV (up to 878 eV) at the Co L-edge or to the peak intensity in the range of 405 eV–420 eV at the N K-edge.
The cobalt K-edge x-ray absorption near edge structure (XANES) and extended x-ray absorption fine structure (EXAFS) were collected at beamline 12-BM of the Advanced Photon Source of Argonne National Laboratory. Si(111) crystals were used as the monochromator. Powder samples were mixed with boron nitride to fulfill the transmission detection requirements with the x-ray attenuation of 0.1–1 at the Co K-edge. FK102 Co(III) TFSI salt {tris(2-(1H-pyrazol-1-yl)pyridine)cobalt(III) tri[bis(trifluoromethane)sulfonimide]} was used as a reference sample for the oxidation state of cobalt. Energy calibration was performed using a Co foil behind the sample to collect transmitted x-ray signals after the sample. XANES/EXAFS data analysis was performed with the Athena and Artemis packages based on IFEFFIT and FEFF programs. Theoretical models were constructed using Gaussian; the theoretical EXAFS spectra were constructed using FEFF and were fit to the experimental data using Artemis.34,35
All electrochemical measurements were carried out under an argon atmosphere at room temperature in a three-electrode cell consisting of glassy carbon as a working electrode, an auxiliary platinum wire as a counter electrode, and Ag/AgCl as a reference electrode. Cyclic voltammograms were recorded on a GAMRY Instrument Interface 1010E. Solute concentrations were 1.0 mM for the cobaloxime (Co-A and Co-B) and 0.1M for the supporting electrolyte, tetrabutylammonium perchlorate (TBAP). For HBCPy, CoQD-A, and CoQD-B, thin films of the compounds were prepared on the glassy carbon electrode using 1 µl of 1 mM solution in chloroform. The thin films were coated with 2 µl of 10% Nafion solution in ethanol and were dried under argon before starting the electrochemical measurements. The scan rate for all measurements was 100 mV/s.
The setup used to perform UV–vis transient absorption measurements was described previously.36 In brief, 800 nm laser pulses of 100 fs duration were produced at 1 kHz repetition rate by a mode-locked Ti:sapphire laser and regenerative amplifier (Astrella, Coherent Inc.). The output from the Astrella was split into pump and probe beams. The pump beam was sent into a BBO crystal to double the excitation pulse photon energy to 400 nm. The probe beam was focused into a 4 mm CaF2 crystal that was continuously translated with a linear stage (Newport MFA-CC) to generate the white light continuum between 350 nm and 750 nm, which was focused into the sample. Care was taken to ensure that the CaF2 crystal axis matched the polarization of the incident light to avoid any wavelength polarization dependence. Thin film samples were excited by 400 nm light at pump intensities between 200 nJ and 1200 nJ per pulse in a nitrogen-purged 1 cm quartz cuvette, and a high angle of incidence between pump and probe (15°) was used to reduce pump scattering. Thin film measurements were performed at magic angle polarization between pump and probe. After passing through the sample, the probe continuum was coupled into an optical fiber and input into a CCD spectrograph (Ocean Optics, Flame-S–UV–VIS–ES). The data acquisition was achieved using in-house LabVIEW (National Instruments) software routines. The group velocity dispersion of the probing pulse was determined using nonresonant optical Kerr effect measurements.
Time-resolved mid-infrared absorption spectroscopy (TRIR) was performed using 800 nm laser pulses of 100 fs duration, produced at 1 kHz repetition rate by a mode-locked Ti:sapphire laser and regenerative amplifier (Astrella, Coherent Inc.). The output from the Astrella was split into pump and probe beams. The pump beam was sent into a BBO crystal to double the excitation pulse photon energy to 400 nm, and every other pulse was blocked using a mechanical chopper wheel synchronized to the pulses. The probe beam was sent to an optical parametric amplifier (TOPAS-C, Light Conversion) to obtain the mid-IR probe pulse, which was split into reference and probe beams. The probe was focused into the sample. The films were excited at 600 nJ per pulse in a nitrogen purged custom sample holder with 4 mm CaF2 glass windows. Thin film measurements were performed at magic angle polarization between pump and probe. After passing through the sample, the probe along with the reference was refocused into a Czerny–Turner spectrograph (Chromex Inc.), which separated the mid-IR colors in space and refocused the light onto two mercury cadmium telluride detector arrays of 32 pixels each (MTC-14-2 × 32, Infrared Systems Development Corporation) for the probe and reference beams. The data acquisition was achieved using in-house LabVIEW (National Instruments) software routines. Compressed air was filtered (PCRSBX1A64-FM, Puregas) to be water and CO2 free, and flowed into the instrument area that was completely enclosed.
Calculations for nitrogen K-edge XAS were performed using GAUSSIAN 16.37 The 6-31g(d) basis set was employed for all light atoms, while the LanL2DZ effective core potential and associated basis set38 were used for Co atoms. Density functional theory (DFT) calculations were performed with the CAM-B3LYP functional.39 This functional was chosen because it is range-separated, to better describe charge-transfer excited states.40 The IEFPCM41 solvation model for dichloromethane was used for all calculations. The dodecyl side chains of the GQDs were truncated to propyl groups. Energetic minima found by ground-state geometry optimizations were confirmed by normal mode calculations at the same level of theory. Excited states were evaluated using time-dependent density functional theory (TDDFT).
Calculations related to theoretical IR and UV–vis absorption were performed using the Gaussian 09 software42 package with the resources of the Extreme cluster at the University of Illinois at Chicago. Structure optimization and frequency calculations were performed using the B3LYP43,44 hybrid function and 6-31g(d,p) basis set for carbon, hydrogen, nitrogen, oxygen, boron, fluorine, and chlorine atoms, and the LanL2DZ basis set for cobalt. The solvation method used was the IEFPCM41 model for dichloromethane (DCM). Methyl groups were used instead of dodecyl chains to save computational time. Vibrational frequency analysis was done to show the absence of imaginary frequencies. A scaling factor of 0.966 was used in frequency calculations for the 6-31g(d,p) basis set.45
III. RESULTS AND DISCUSSION
A. Synthesis and characterization
The synthesis of HBCPy was performed starting with dibromohexaphenylbenzene (compound 1, Scheme S1, supplementary material), which was prepared via a Diels–Alder addition/decarbonylation sequence developed by Ito and co-workers.46 Subsequent Suzuki coupling47 between compound 1 and a pyridine-based boronic ester yielded a soluble polyphenyl derivative, which was further subjected to oxidative dehydrogenation48 to yield HBCPy (Scheme S1, supplementary material). Two types of cobaloximes were coordinated axially to HBCPy using the approach developed previously for pyridine based complexes Co-A49 and Co-B.50 These reactions yielded disubstituted CoQD-A and CoQD-B derivatives.
Due to the low solubility of GQDs shown in Scheme 1, we were not able to characterize these compounds using standard solution-based techniques. The solid-state 13C NMR of HBCPy reveals two broad peaks associated with the aliphatic (34 ppm) and aromatic (120 ppm) carbons (Fig. S1, supplementary material). Coordination with cobaloximes is supported by the appearance of an additional peak at 155 ppm, associated with the C=N group of the oxime ligands. Similarly, solid-state 11B NMR spectra identified the presence of oxime-based boron atoms, consistent with the CoQD-B structure (Figs. S2 and S3, supplementary material). Mass spectrometry was used successfully to characterize HBCPy, but molecular masses of CoQD-A and CoQD-B were not observed, likely due to the labile bond between HBCPy and cobalt.51 Infrared (IR) spectroscopy was combined with density functional theory calculations to identify key functional groups in model compounds (Sec. S2.2, supplementary material). The C=C stretching region of HBCPy consists of modes arising from the aromatic HBC core (1377 cm−1) and the pyridine moiety (1595 cm−1) in HBCPy. Upon coordination with cobaloximes, the pyridine-based C=C stretch shifts from 1595 cm−1 to 1616 cm−1 and 1613 cm−1 for CoQD-A and CoQD-B, respectively (Figs. S4 and S5). In addition, CoQD-A and CoQD-B exhibit C=N stretching modes arising from the equatorial oxime ligands (at 1571 cm−1 and 1624 cm−1 for CoQD-A and CoQD-B, respectively).
Additional characterization of GQDs was performed using core spectroscopies, namely, x-ray photoelectron spectroscopy (XPS) and x-ray absorption spectroscopy (XAS). Due to the small concentration of N elements in the sample, the XPS N1s region of HBCPy contains a very weak pyridinic peak at 398.7 eV (Fig. S9). Once coordinated with Co, the N1s region exhibits more intense peaks with binding energies of 398.5 eV and 399.1 eV (in CoQD-A and CoQD-B, respectively). These peaks are associated with N-atoms of equatorial oxime ligands and confirm that the coordination of Co to GQDs took place. Similarly, the Co2p XPS region of CoQD-A and CoQD-B contains two peaks associated with p1/2 and p3/2 peaks of the cobaloxime moiety. The N1s and Co2p peaks of CoQD-A and CoQD-B are shifted to slightly lower binding energies relative to the model compounds Co-A and Co-B (Figs. S9 and S10), indicating lower electron density on N and Co atoms in GQD samples.
The attachment of the Co(DMG)2 complex to HBCPy was confirmed using N K-edge XAS, combined with electronic structure calculations, as shown in Fig. 1. The pyridyl N-atoms of HBCPy have an absorption peak at 397.7 eV. Upon Co binding, the pyridyl N transitions shift to higher energy and overlap with the stronger dimethylglyoximate N peak, consistent with the depletion of the electron density in pyridinic N upon coordination. Cobaloxime N does not change significantly, as can be observed by comparing Co(DMG)2Cl2 and CoQD-A. The lack of a peak at 397.7 eV in the CoQD-A spectrum, therefore, indicates that there is no significant amount of free HBCPy present in the sample.
The oxidation and spin states of CoQD-A and Co-A were characterized using Co K- and L-edge XAS, as shown in Fig. 2. The L-edge spectrum consists of L3 and L2 edges formed due to the spin–orbit coupling of the 2p hole and some fine structure associated with the multiplet nature of the generated 2p53d7 final state [Fig. 2(d)]. The L-edge XAS spectra are highly sensitive to the Co oxidation and spin state, and the spectra of CoQD-A and Co-A are similar to those of previously reported low-spin Co(III) compounds.52,53 The low-spin character of CoQD-A and Co-A is expected, as previously shown using magnetic susceptibility measurements for cobaloximes.54 The Co K-edge XANES of Co-A and CoQD-A, shown in Figs. 2(a)–2(c), is also consistent with Co(III) metal centers. Figure 2(b) shows the pre-edges of the K-edge spectra, which consist mostly of quadrupole-allowed 1s → 3d electronic transitions. For the reference Co(III) complex (FK102 Co(III) salt), one pre-edge peak is observed at ∼7710 eV; this is expected for a low-spin, d6 metal, which in Oh geometry would have a t2g6 eg0 electronic configuration, which will yield a single-peak due to the excitation of the 1s core electron into the empty eg manifold. The Co-A and CoQD-A samples also show a peak at a comparable energy, along with an additional peak of ∼7712.5 eV. This peak is likely not the standard 1s → 3d transition but instead transition into a ligand π* orbital with mixed metal 4p character, which has been observed previously for low-spin, d6 Fe complexes.55 Based on the edge energies and pre-edge peaks, the Co-A and CoQD-A samples can be confirmed to have low-spin, Co(III) metal centers.
The extended x-ray absorption fine structure (EXAFS) was measured to determine the local structure about the Co center in each of the Co and QD samples (Fig. 3). Fourier transform of this interference pattern yields a pseudo-radial distribution function that is centered about the absorbing Co atom with peaks that reflect Co-atom scattering distances. Structural models were constructed in Gaussian and used to generate theoretical EXAFS spectra using FEFF according to the EXAFS equation. The EXAFS fits (fixing the optimized structural parameters and fitting for the mean squared variation in pathlength, σ2) are shown as dashed lines in Fig. 3, and the corresponding fit parameters are shown in Table S2 (supplementary material). The EXAFS spectra of Co-A and CoQD-A are adequately reproduced using the DFT structures, demonstrating that the DFT methods produce excellent structural models for these systems.
B. Electrochemistry
Cyclic voltammograms (CVs) of model Co-complexes and the corresponding GQDs are shown in Fig. 4. CoQD-A exhibits an electrochemically irreversible process at −0.25 V vs Ag/AgCl, which is assigned to the Co(III/II) reduction coupled with the loss of an axial chloride ligand, based on the comparison with the model Co-A and the previous literature report.49 The second reduction peak appears at −1.01 V vs Ag/AgCl, and it is electrochemically reversible and is assigned to the Co(II/I) reduction. While the Co(II/I) reduction is chemically reversible in the case of the Co-A model, the same process is chemically quasi-reversible in CoQD-A, as evidenced by the ratio of anodic/cathodic peak currents (Ia/Ic = 0.28). The loss of anodic current may occur due to the partial detachment of the cobaloxime moiety from HBCPy upon reduction. This assignment is supported by a relatively small binding constant (K = 200M−1 in DMF)50 reported previously for axial pyridine coordination to cobaloximes. In the case of B-series, the electrochemical behavior of CoQD-B was not consistent with the model compound Co-B: the Co(II/I) reduction that appears at −0.84 V vs Ag/AgCl for Co-B is barely visible in the case of CoQD-B. Again, we assign this behavior to the detachment of the Co(II) cobaloxime from HBCPy prior to the electrochemical experiment. HBCPy exhibits a chemically irreversible oxidation at +1.35 V vs Ag/AgCl, which is consistent with the oxidation of the aromatic core.33 The relevant standard reduction potentials are tabulated in Table I and are used in the subsequent text to evaluate the kinetics of photoinduced electron transfer from GQD to cobaloxime in CoQD-A.
ET . | λRE (eV) . | VDA (cm−1) . | a (V) . | b (V) . | E00c (eV) . | w (eV) . | ΔGET (eV) . | kET (s−1) . | τCo (ns) . | Experimental . |
---|---|---|---|---|---|---|---|---|---|---|
. | 1.1 . | 1.7 . | 1.35 . | −0.25 . | 2.6 . | 0.05 . | −0.95 . | 5.7 × 108 . | 0.9 . | . |
EnT | R0 (nm) | JF (nm4 M−1 cm−1) | τ (ns) | (s−1) | JD (cm) | (s−1) | kq (s−1) | |||
0.8 | 1.9 × 1013 | 10.8 | 8.1 × 106 | 3.0 × 10−3 | 1.0 × 1010 | 1.0 × 109 |
ET . | λRE (eV) . | VDA (cm−1) . | a (V) . | b (V) . | E00c (eV) . | w (eV) . | ΔGET (eV) . | kET (s−1) . | τCo (ns) . | Experimental . |
---|---|---|---|---|---|---|---|---|---|---|
. | 1.1 . | 1.7 . | 1.35 . | −0.25 . | 2.6 . | 0.05 . | −0.95 . | 5.7 × 108 . | 0.9 . | . |
EnT | R0 (nm) | JF (nm4 M−1 cm−1) | τ (ns) | (s−1) | JD (cm) | (s−1) | kq (s−1) | |||
0.8 | 1.9 × 1013 | 10.8 | 8.1 × 106 | 3.0 × 10−3 | 1.0 × 1010 | 1.0 × 109 |
Anodic peak potential (due to chemical irreversibility) referenced to Ag/Ag+.
The standard reduction potential referenced to Ag/Ag+.
E00 was obtained from the fluorescence maximum of HBCPy (λ = 479 nm). While E00 is usually obtained as an intercept between absorption and emission maxima, this approach was not used because this would cause an overestimate of E00 (absorption maximum of HBCPy corresponds to transitions to S3 and S4 states, not the S1 state).
C. Steady-state electronic spectroscopy
The absorption profile of GQDs obtained in chloroform at low concentration (∼2 µM) is shown in Fig. 5, along with electronic transitions and relevant orbitals calculated using time dependent DFT calculations. The experimental spectrum of HBCPy exhibits a peak with a maximum at 373 nm and the absorption tail that extends up to 500 nm. This spectrum is consistent with calculations, which predict intense transitions at 380 nm and 384 nm, as well as the low oscillator strength S1 transition at 445 nm. The orbitals involved in the S0 → S1 transition (HOMO-1 → LUMO and HOMO → LUMO+1) indicate that the S1 state exhibits a charge transfer character, where the electronic density moves from the hexabenzocoronene (HBC) core to the pyridine ring. Upon coordination with cobaloximes, the experimental absorption spectra of CoQD-A and CoQD-B undergo a small red shift of the intense bands and an increase in the oscillator strength in the tail region. Again, the experimental findings correlate well with calculations, which predict the red shift of the intense bands and an increase in the transition dipole moment for the HBC → Py charge transfer bands at 456 nm and 440 nm for CoQD-A and CoQD-B, respectively. In addition to these bright states, CoQD-A and CoQD-B also exhibit a number of dark states (Fig. S14, supplementary material). For example, CoQD-A exhibits six dark states below the lowest bright HBC → Py charge transfer band at 456 nm, most of which are localized on the cobaloxime moiety. The lowest S1 state involves a number of different orbitals, as illustrated in Table S3 (supplementary material), which complicates the characterization of this dark state. Overall, this transition exhibits a ligand-to-metal charge transfer (LMCT) character, where the electronic density is shifted from oxime, pyridine, and chloride ligands to the cobalt d-orbitals. This lowest energy dark state may be involved in the fast fluorescence quenching of HBC fluorescence observed in CoQD-A and CoQD-B, as described below.
To further examine the influence of coordination of cobaloxime, we performed fluorescence measurements of HBCPy, CoQD-A, and CoQD-B [Figs. 6(c) and 6(d), all samples exhibited the same optical density at the excitation wavelength of 377 nm]. HBCPy exhibits an emission spectrum with a maximum at 479 nm that we assign to the emission from the HBC → Py charge transfer state. Almost complete quenching of this fluorescence was observed for CoQD-A and CoQD-B, respectively [Figs. 6(c) and 6(d)]. In the case of CoQD-A, the experimental rate constant for fluorescence quenching (kq = 1.0 × 109 s−1, Table I) was calculated from fluorescence lifetimes of the HBCPy (τ) and CoQD-A (τCo) (Fig. S12) using the following equation:
As will be discussed below, this quenching is consistent with the photoinduced electron transfer from HBC to the cobalt centers in CoQD-A and CoQD-B. However, the observed fluorescence quenching may also be associated with the energy transfer from HBC to the cobaloxime-based excited states. The calculations discussed in the previous paragraph indicate the presence of cobaloxime-based LMCT states that could serve as energy acceptors. This possibility is supported by the fact that a broad emission band centered around 600 nm was observed for both CoQD-A and CoQD-B, which resembles the emission from the model compounds Co-A and Co-B, respectively.
D. Energy vs electron transfer
The fluorescence quenching observed in Fig. 6 could be associated with either an energy or an electron transfer process. To evaluate the possible mechanism, we compared the experimental quenching rate for CoQD-A with the rate constants estimated for Marcus electron transfer (ET), as well as Dexter and Förster energy transfer (EnT) processes (additional calculation details are presented in Sec. S3, supplementary material). The ET is expected to take place from the photoexcited GQD moiety to the cobalt center as follows:
The rate constant for ET (kET) in CoQD-A was evaluated using the Marcus ET theory,56 expressed as follows:
Here, λRE is the solvent reorganization energy, VDA is the electronic coupling matrix between QD and Co moieties, and ΔGET is the thermodynamic driving force for ET. The solvent reorganization energy, λRE, is obtained from the following equation:
Here, e is the charge of the electron; ε0 is the permittivity of free space; εo and εs are the optical and static dielectric constants of the solvent, respectively; rd and ra are radii of QD and Co, respectively; and r is the center-to-center distance between QD and Co obtained from the optimized structure of CoQD-A. The value for inner-sphere reorganization energy λi was approximated to be 0.1 eV.57 The electronic coupling factor (VDA) between QD and Co in Eq. (3) is given by58
Here, μtr is the transition dipole moment associated with the QD-to-Co charge transfer excited state. The value for μtr was obtained from TDDFT calculations for CoQD-A (S3 excited state, more information available in the supplementary material). ΔE is the energy of the QD-to-Co charge transfer excited state S3, while Δμ is the difference between the dipole moments for S3 and S0 states of CoQD-A, evaluated as er (where e is the electron charge and r is the center-to-center distance between QD and Co). The thermodynamic driving force for ET (ΔGET) was determined using the Gibbs energy of photoinduced electron transfer,59
where are the standard reduction potentials for QD/QD.+ and Co(III/II) processes, estimated from the cyclic voltammograms in Fig. 4. E00 is the energy of the QD excited state derived from the peak fluorescence of HBCPy (Fig. 6), while w is the work term associated with the distance of charge separation. The calculated free energies (Table I) indicate thermodynamic feasibility of PET from QD to Co in CoQD-A. Upon calculating the terms in Eqs. (3)–(6) (additional details in Sec. S3, supplementary material), a kET value of 5.7 × 108 s−1 was obtained for CoQD-A using the Marcus theory.
The energy transfer (EnT) can take place from π,π* states localized on the QD moiety to either ligand-centered (LC) or ligand-to-metal charge transfer (LMCT) states of Co,
We evaluate here the rates for both Förster60 and Dexter61 energy transfer mechanisms in CoQD-A. The rate of Förster EnT was determined using the following equation:
where τ is the fluorescence lifetime of the QD, ϕD is the fluorescence quantum yield of the QD, and JF is the Förster overlap integral. The orientation factor is described by the following equation:
Here, θD and θA are the angles made by the transition dipoles of donor and acceptor, respectively, along the line joining the centers and θT is the angle between the transition dipoles. This value was obtained from transition dipole moments obtained using TDDFT calculations (Sec. S3, supplementary material). The Förster overlap integral JF is defined by the following equation:
Here, FD is the area normalized emission spectrum of the QD, εA is the molar extinction coefficient of Co-A, and λ is wavelength in nm. The JF value was calculated to be in the order of ∼1013 nm4 M−1 cm−1, a relatively low value due to a poor spectral overlap between emission of HBCPy and absorption of Co-A (Fig. S11A, supplementary material). Furthermore, the orthogonal arrangement of transition dipoles of QD and Co moiety in CoQD-A (Fig. S11B, supplementary material) yields a κ2 value of 0.046, which is also very small. Due to these poor conditions for Förster EnT, the rate constant was found to be 8.1 × 106 s−1, a value that is several order of magnitudes lower than the experimental rate constant for fluorescence quenching in CoQD-A (Table I).
The Dexter EnT requires close proximity between donor and acceptor, which is the case in CoQD-A as the QD is axially coordinated to Co via a pyridine linkage. The rate of Dexter EnT was calculated using the following equation:58
Here, JD is the Dexter overlap integral calculated from the fluorescence spectrum of QD [] and absorption of Co-A [],
From Eq. (11), a rate constant of 1.0 × 1010 s−1 was obtained for Dexter EnT.
The experimental EnT rate constant is significantly higher than the calculated values for , indicating that the Förster EnT can be ruled out as a possible quenching mechanism. On the other hand, the calculated rate constants for both Marcus ET and Dexter EnT are of the same order of magnitude as the experimental value for fluorescence quenching, making it difficult to distinguish which process takes place. The negative ΔGET (−0.95 eV) value and close proximity of HBCPy and cobaloxime (∼2 Å distance between pyridinic nitrogen and cobalt in CoQD-A) makes Marcus ET and Dexter EnT possible in CoQD-A.
E. Time-resolved spectroscopy
To investigate the photophysical properties of CoQD-A and CoQD-B, time-resolved transient absorption (TA) experiments were performed by probing in UV–vis and mid-IR spectral regions. Due to the low solubility of CoQD-A and CoQD-B in common organic solvents, TA measurements were performed on drop-casted thin films onto transparent substrates. Preliminary experiments were performed on HBCPy to evaluate the excited-state dynamics of the GQD core in the absence of the metal moiety (Fig. 7). TA spectra of HBCPy consist of the ground-state bleach feature centered at 388 nm and a broad excited-state absorption feature that spans the entire visible range [Fig. 7(a)]. The transient signal decays with the same kinetics at all wavelength, which is also evident by the isosbestic point with zero ΔA at 475 nm. Most of the transient signal disappears within the first nanosecond after the excitation pulse, which is quite different from the fluorescence dynamics observed for HBCPy in solution (10.8 ns lifetime, Fig. S12). We assign the differences in photophysics to the presence of inter-chromophore interactions in thin films via π-stacking interaction between aromatic cores. A similar increase in the excited-state relaxation was observed in our previous studies of other GQD assemblies.32
The excited-state dynamics of HBCPy assemblies showed a pump-fluence dependence [Fig. 7(b)], consistent with the presence of annihilation of excitons that migrate along the chromophore stack. This pump intensity dependent dynamics was used to determine the exciton mobility in HBCPy aggregates, using the exciton–exciton annihilation approach we recently applied to a similar system of molecular aggregates.32 The experimental data in Fig. 7(b) were fit to a kinetic model involving two types of excitons: initially formed mobile excitons (with density N1, lifetime τ1, and diffusion coefficient D) and trapped excitons (with density N2 and lifetime τ2) that are formed from mobile excitons with lifetime τtr. The exciton–exciton annihilation of the mobile excitons was modeled using one-dimensional exciton diffusion,
Differential equations (13)–(15) were solved numerically using the initial exciton density as described previously.32 The fit of the experimental data was performed simultaneously for all pump fluences by varying only four parameters: τ1, τ2, τtr, and D. Figure 7(b) shows a satisfactory agreement between the experiment and theory. The fitted parameters are shown in Table II along with previously measured values for a similar thin film without pyridine groups (HBC) using the same model.32 The addition of the pyridine groups to HBC does not affect significantly the values for τ1 and τtr. The exciton trapping (associated with τtr) in similar organic chromophore assemblies is often assigned to interchromophore motions that ultimately lead to the formation of “relaxed” excitonic species that cannot undergo further hoping due to the poor energy mismatch between donor and acceptor states.62,63 Based on the similarity in τ1 and τtr values for HBC and HBCPy, it appears that the pyridine groups in HBCPy do not alter these relaxation pathways significantly and that the aggregate structure and dynamics are similar in two chromophore aggregates. On the other hand, the introduction of pyridine moiety causes D to double in HBCPy compared to HBC. The larger D in HBCPy indicates faster exciton hopping along the aggregate and is likely due to the improved value for the Förster energy transfer overlap integral JF relative to HBC. Larger overlap integral is consistent with the higher value calculated for the transition dipole moments for the S1 state in HBCPy (0.46 D) compared with HBC (0.20 D). This improvement in the exciton diffusion coefficients results in a larger exciton diffusion length in HBCPy (where 50 molecules are sampled within the exciton lifetime) relative to HBC (39 molecules sampled).
. | τ1 (ps) . | τ2 (ps) . | τtr (ps) . | D (mol2/ps)a . | LD (mol)a . | τMET (ps) . | τ3 (ps) . |
---|---|---|---|---|---|---|---|
HBC | 75 | 750 | 70 | 10 | 39 | ||
HBCPy | 67 | 160 | 76 | 18.5 | 50 | ||
CoQD-A | 67b | 160b | 76b | 18.5b | 51 | <0.5 | |
CoQD-B | 67b | 160b | 76b | 18.5b | 101 | <0.5 |
. | τ1 (ps) . | τ2 (ps) . | τtr (ps) . | D (mol2/ps)a . | LD (mol)a . | τMET (ps) . | τ3 (ps) . |
---|---|---|---|---|---|---|---|
HBC | 75 | 750 | 70 | 10 | 39 | ||
HBCPy | 67 | 160 | 76 | 18.5 | 50 | ||
CoQD-A | 67b | 160b | 76b | 18.5b | 51 | <0.5 | |
CoQD-B | 67b | 160b | 76b | 18.5b | 101 | <0.5 |
Mol = molecule.
Values were set to those found for HBCPy and kept constant during the fit.
Qualitatively, the TA spectra of CoQD-A and CoQD-B thin films are similar to HBCPy, with the presence of bleach signal at ∼388 nm and a broad excited-state absorption throughout the visible range (Fig. 8). The two thin films differ from HBCPy in that the signal decays faster (Fig. S13), consistent with the shorter fluorescence lifetimes obtained for CoQD-A and CoQD-B samples in solution (Fig. S12). The faster decay of CoQDs indicates that excitons are finding additional decay pathways that are not present in HBCPy, such as PET and EnT discussed earlier. The kinetic profiles at different pump fluences [Figs. 8(c) and 8(d)] were modeled using a kinetic scheme analogous to the exciton–exciton annihilation model used for HBCPy, with the addition of another decay pathway (via electron or energy transfer from mobile excitons N1 with lifetime τMET) to generate a new state (with density N3 and lifetime τ3),
It was assumed that the lifetimes τ1, τ2 and the diffusion coefficient D in CoQD-A and CoQD-B are the same as obtained for HBCPy, and the experimental data in Fig. 8 were fit using two parameters: τMET and τ3. The results of the fit are shown in Table II and indicate that the N3 state is very short-lived, which agrees with the lack of spectroscopic signature of this transient in our TA data. In specific, transient species is formed at a slower rate (τMET = 51 ps for CoQD-A) than the rate of its loss via the recombination pathway (τ3 < 0.5 ps), making it difficult to build up a detectable concentration of the transient species needed to surpass the detection limit of our instrument. Thus, while spectroscopic signatures for ET64 and EnT16 transients have been described previously, we were unable to detect them due to the short-lived nature of the N3 state. The excited states of cobaloximes, if formed via the Dexter mechanism, are expected to be short-lived, on the order of several tens of picoseconds, as reported previously by Guldi.16 The recombination lifetimes obtained in our fits (<0.5 ps) are much shorter, suggesting that the ET pathway is more likely.
No matter what the deactivation pathway is, it is clear that the long-lived charge separated states desired for photocatalysis applications do not form in CoQDs. It is interesting to note that our result correlates very well with numerous previous reports involving cobaloxime complexes coordinated with organic and inorganic photosensitizers. For example, Mulfort and Tiede found that Co-A and Co-B covalently attached to ruthenium tris bipyridine both showed faster decay kinetics than the chromophore alone, and only saw short-lived spectroscopic signature for CoI in one Co-A type compound.14,19 It is plausible that, at least in some reports on fast charge-separation in chromophore-cobaloxime systems, the quenching mechanism involves a Dexter energy transfer from photoexcited chromophore to a manifold of dark cobaloxime-based states with LMCT and dπ-dπ* character. These states are expected to undergo efficient non-radiative decay to ground state and are likely to outcompete the productive charge-separation pathways.
Time-resolved infrared (TRIR) spectroscopy was also utilized in the attempt to identify the spectroscopic signature of ET or EnT intermediates formed in CoQDs (Fig. 9). The ground state IR spectra of HBCPy and corresponding CoQDs (panels B, D, and F) were assigned using DFT-based vibrational mode analysis (Sec. S2.2, supplementary material). All three spectra consist of aromatic ring C—C modes of HBC (at 1300 cm−1–1400 cm−1) and pyridine (at ∼1600 cm−1) moieties. An additional strong vibrational mode at ∼1450 cm−1 is associated with the C—H bending modes of the alkyl group and are not expected to be relevant for the TRIR spectra. Importantly, the CoQD spectra contain an additional absorption band of ∼1550 cm−1 (for CoQD-A), which is assigned to the C=N stretching modes of the cobaloxime moiety. These modes serve as an excellent marker for the involvement of cobaloxime moieties in the photophysics of the model compounds: if the transient bleach of these modes were observed in TRIR, it would indicate that the energy or the electron transfer to the cobaloxime took place.
TRIR spectra of HBCPy consist of a broad feature that covers the entire probe window and peaks at 1545 cm−1. Similar broad transients have been reported previously by Asbury in perylene-imide chromophore assemblies,65,66 and the observed vibrational broadening was explained by the fast dephasing caused by the fact that a large number of molecules is sampled throughout the material, either through large exciton size or fast exciton hopping. Our previous study of exciton size and dynamics in HBC-based chromophore assemblies32 showed strong evidence that the exciton delocalization length involves 1–4 chromophore units. Furthermore, experimental and computational studies indicate that the dynamic exciton decoherence in organic chromophore assemblies take place at timescales shorter than 100 fs.67 Based on these studies, we do not expect that the mode broadening arises due to large exciton delocalization. On the other hand, the fast exciton hopping may be associated with the observed broadening, particularly since our kinetic data modeling implies a large exciton diffusion length (LD) of 50 molecules (Table II).
In the presence of cobaloximes, TRIR spectral features do not change significantly [Figs. 9(c) and 9(e)], but the transient signal decays faster, consistent with the excited state quenching observed in the steady-state and time-resolved measurements in the UV–vis range. Unfortunately, the decay of the broad vibrational features associated with the HBCPy unit was not accompanied with the concomitant growth of the cobaloxime-based modes at 1571 cm−1. This result is again explained by the unfavorably low population of ET or EnT products controlled by the τMET and τ3 kinetic parameters.
IV. CONCLUSIONS
In summary, our study of photophysical properties of dyads composed of cobaloxime catalysts and GQD chromophores shows that the fast nonradiative decay pathway exists in these systems. By comparing the experimental quenching rate with the rates estimated using models for ET and Ent, we conclude that the Marcus ET and Dexter EnT are both plausible mechanisms for GQD excited-state quenching. The EnT mechanism is also supported by TDDFT calculations, which predict the presence of several low-lying cobaloxime-centered dark excited states that could serve as Dexter EnT acceptors and mediate subsequent nonradiative decay to the ground state. The current study points to additional challenges associated with the achievement of long-lived charge-separation in chromophore-catalyst systems: the plausible energy transfer mechanisms need to be avoided. However, avoiding energy transfer can be a challenge for many metal-based catalysts, considering that they often exhibit low-energy metal-centered ligand states that could serve as energy acceptors.68
SUPPLEMENTARY MATERIAL
See the supplementary material for additional details on synthesis, characterization, and computational calculations.
ACKNOWLEDGMENTS
This work was supported by the U.S. Department of Energy (DOE), Office of Basic Energy Sciences, Division of Chemical Sciences, Geosciences and Biosciences, through Argonne National Laboratory under Contract No. DE-AC02-06CH11357. G.K. and A.A.C. were supported by the U.S. Department of Energy (DOE), Office of Basic Energy Sciences, Division of Chemical Sciences, Geosciences and Biosciences, through SLAC National Accelerator Laboratory under Contract No. DE-AC02-76SF00515. This research used resources of the Advanced Photon Source, a U.S. Department of Energy (DOE) Office of Science User Facility operated for the DOE Office of Science by Argonne National Laboratory under Contract No. DE-AC02-06CH11357. Use of the Stanford Synchrotron Radiation Lightsource, SLAC National Accelerator Laboratory, is supported by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences under Contract No. DE-AC02-76SF00515. K.D.G. acknowledges the National Science Foundation (Grant No. 1806388). We thank the beamline scientist Dr. Sungsik Lee for his help with cobalt K-edge XAS measurements.
DATA AVAILABILITY
The data that support the findings of this study are available from the corresponding author upon request.