Controlling photocatalytic activity by self-assembly – tuning perylene bisimide photocatalysts for the hydrogen evolution reaction

structures in solution, the nature of which depends on the local environment. Using a high-throughput photocatalysis setup, five PBIs are studied for the hydrogen evolution reaction (HER) under a range of conditions (pH and hole scavenger concentration) across 350 experiments to explore the relationship between supramolecular structure and photocatalytic activity. Using small angle X-ray scattering (SAXS), NMR spectroscopy and ultraviolet-visible (UV-vis) absorption spectroscopy, it is shown that photocatalytic activity is determined by the nature of the self-assembled aggregate that is formed, demonstrating the potential of self-assembly to tune activity. There is a clear correlation between the presence of charged flexible cylindrical aggregates and the occurrence of photocatalytic H 2 production, with UV–vis spectroscopy indicating that the most active structure type has a distinctive form of π -aggregation which is proposed to enable efficient charge separation across multiple PBI units. solvent mix, PBI functional group) which can all affect the self-assembly process. We completed approximately 350 photocatalysis studies in 2 weeks, in contrast a sample-by-sample approach would have required ≈ 6 months and it is likely that the optimal photocatalysis conditions would not have been discovered without this approach. This is important as it is feasible that viable photocatalysts are being prepared in labs world-wide but discarded after an initial activity analysis with a specific set of conditions showed low levels of activity. Using a combination of SAXS and NMR spectroscopies we have been able to show in our systems that formation of charged flexible cylindrical aggregates is required, but in-itself insufficient, for efficient photocatalysis. This can be rationalized by the clear correlation between UV–vis spectral features and photocatalytic activity that indicates that a specific form of local packing within the cylindrical aggregates is also essential for photocatalysis. It appears likely that in this study on supramolecular structures in aqueous solvent that the most active forms are those that lead to strong mixing of the Frenkel and charge transfer excitations. Future modelling and experimental studies will focus on confirming the specific packing motif and its role on the photo-physical properties of the material. However, it is already clear that control of the self-assembly process offers an exciting route to tuning activity of photocatalysts for the production of solar fuels.


(2 of 10)
© 2020 The Authors. Advanced Energy Materials published by Wiley-VCH GmbH have been reported. [24] Often PBIs are designed for solubility in organic solvents but they can be dispersed in water using a range of suitable functional groups including oligoamines, oligo(ethylene oxide)s, charged species such as sulfonates and amino acids at the imide position (Figure 1). [10,25,26] Many amino acid functionalized PBIs have been reported and can form selfassembled structures in aqueous solution or form hydrogels depending on the exact conditions used. [6,25,[27][28][29][30] Free carboxylic acids on the amino acids mean that the PBIs can be dispersed in water at a pH above the pK a and then often form hydrogels when the pH is lowered below the pK a . [31,32] The application of hydrogels and aqueous solvated self-assembled materials to photocatalysis is particularly interesting as we have recently shown that the solvent environment enables charge separation, overcoming a known limitation of many other organic photocatalysts with inherently low dielectric constants. [33] Stupp's group has shown a range of self-assembled PMIs can be effective as part of a photocatalytic hydrogen evolution reaction (HER). [20,[34][35][36] Recently, they have also published work studying hybrid polyelectrolyte and supramolecular systems in which PMIs are entrapped in a crosslinked polyelectrolyte hydrogel. [37] Once deposited in the hydrogel network, self-assembly of the PMI is triggered through a solvent switch and the photocatalytic hydrogen production of these systems studied. Interestingly, they note a pH dependence, where the H 2 production increases from pH 1.8 to 4.0. [37] In other work, visible-light-driven H 2 evolution from water has been performed using 1D nanofiber composites with dodecyl and/or dimethylaniline functionalized PBIs. [21] The PBI with the dimethylaniline substituent showed greater H 2 evolution, which was attributed to the electron donating characteristics of the dimethylaniline improving charge separation. [21] Some of us have used a phenylalanine functionalized PBI (PBI-F) as a self-assembled photocatalyst for the HER, [22] where the activity was greatly dependent on solution pH. PBI-F forms hydrogels at low pH (<pH 5.0), and this coincided with optimal activity in the HER at pH 4.0. The pH was lowered while stirring, resulting in the same self-assembled structure present in a hydrogel but not a self-supporting network. The evolved H 2 required the formation of 1D structures. Although this study allowed us to propose that the nature of the supramolecular structures formed was important in determining activity, our suggestion was based off this single example, and why a specific supramolecular structure was needed was not examined. To address this here, we significantly expand the library of PBIs and conditions studied to allow for identification of wider supramolecular structure-activity trends.
A challenge, and potential advantage, of developing selfassembled photocatalysts is that the nature of the self-assembled structure formed in such systems is often determined by (and sensitive to) a significant number of variables. These include pH, concentration, co-solvent (methanol is present as a sacrificial electron donor, but will also affect the solubility of the PBI), temperature, salts, and additives. [38][39][40] Therefore a single sub-unit can assemble into a wide range of photocatalyst structures. However, identifying and controlling the formation of the most active structures remains challenging and optimizing even a single system is extremely time-demanding and it is unsurprising that there is a tendency in the literature to focus on a small number of experimental variables. This may lead to effective photocatalytic systems being missed simply due to time constraints.
To address this, here we set out to use a high-throughput approach to allow us to access a wider range of experimental parameters. [41] From five PBIs we found three that form structures that are active for the photocatalytic evolution of H 2 . The three active PBIs show a pH dependent activity, with maximum hydrogen evolution at pH 5.0. The two other PBIs showed very little H 2 evolution but the activity in the HER still shows a pH dependence. Once optimal conditions for HER had been identified it was possible to deploy a range of characterization techniques and we show a clear correlation between the presence of charged flexible cylindrical aggregates and high levels of H 2 being formed demonstrating a wider link between supramolecular structure and photocatalytic activity.

Results and Discussion
A small library five different PBIs were prepared varying the amino acid at the imide positions was used for this study (Figure 1), but under a wide range of conditions leading to ≈350 photocatalytic experiments being carried out. Dispersions of these PBIs can be formed at high pH by the addition of a base to an aqueous suspension of the PBIs. [6,22] Samples at different pH were prepared by the addition of dilute acid to the stock solutions at high pH. These PBIs can form gels at low pH. [6,22] At pH >5.0, UV-vis absorption spectroscopy shows that the PBIs exist as aggregated structures due to the presence of strong π-π interactions. Only at concentrations of <10 −6 m are free PBI molecules seen ( Figure S2, Supporting Information). Therefore, at the concentrations used in this work, aggregated structures are present in vast excess.
A high-throughput method was used to study photocatalytic activity (Figure 2) which enabled the activity study to be completed in ≈2 weeks. In contrast if each sample was looked at in turn, we estimate it would take ≈6 months (based on 3 h sample −1 and a 40 h working week). Our aim here is not to report a one off, best-in-class material, but instead to survey a wide parameter space and identify wider structure-activity trends. Briefly in the high-throughput method, aqueous solutions of the PBIs (5 mg mL −1 ) at different pH were prepared in the presence of methanol (20 v/v%), with the inclusion of Pt nanoparticles (1 mol%, average hydrodynamic diameter 16.6 ± 1.1 nm, Section S2.2, Supporting Information). We use methanol as a sacrificial electron donor as it is photo-stable and allows for studies across a wide pH range. While PBIs can form gels at low pH, the pH of the solutions was adjusted while stirring. This resulted in the formation of supramolecular structures, but not a self-supporting gel network. Forty-eight samples at a time were placed on a mechanical roller and illuminated for 3 h under a solar simulator. Headspace gas analysis using an autosampler was used to quantify the amount of H 2 produced. H 2 production was seen in significant amounts for PBI-A, PBI-F, and PBI-V (Figure 3). Hydrogen was only produced in negligible amounts for PBI-H and PBI-Y. For PBI-A, PBI-F, and PBI-V, the H 2 produced varied significantly with the solution pH, where the maximum H 2 was produced at pH 5.0 for all three PBIs. Similar amounts of H 2 were produced at pH 5 for both PBI-F and PBI-V, at 11.6 ± 1.1 µmol g −1 h −1 and 12.7 ± 2.9 µmol g −1 h −1 , respectively. At pH 5, PBI-A produced 5.1 ± 1.8 µmol g −1 h −1 H 2 . As the pH was decreased from pH 5.0 to pH 2.0, the absolute amount of H 2 produced decreased, reaching a negligible amount at pH 2.0 for all three of these PBIs. At pH 6.0 and above, there was minimal H 2 produced for all three PBIs. For PBI-F, these results are consistent with previous results [22] although the absolute amount of H 2 evolved differs due to the photocatalysis setup used. PBI-F was included in the initial high-throughput screen to benchmark against conventional photocatalyst testing methods. As PBI-F has been previously reported and its behavior is reproducible and has been previously discussed, [22] for the remainder of this study, we focus the discussion and characterization on the new samples, in particular PBI-A as a representative of a photocatalytically active material and PBI-Y as an inactive material.  PBI-A shows both UV and visible light driven H 2 production ( Figure S3, Supporting Information) and we see no evidence for PBI degradation during a typical experiment (Figures S4 and S5, Supporting Information). We propose that excitation of the PBI leads to exciton formation, followed by rapid quenching with sacrificial electron donation occurring from the methanol present. Recent spectroscopic studies on related PBI-A hydrogel films and pH 9 solutions indicate that hole transfer can occur via the alanine group and this may also be occurring here prior to methanol oxidation. [33] Electron transfer to the Pt co-catalyst from the PBI structures can then occur, enabling H 2 production. Electrochemical measurements show that H 2 evolution is thermodynamically feasible following electron transfer from the singly (PBI ○− ) and doubly reduced (PBI 2− ) species below pH 8 (PBI ○− ) and pH 7 (PBI 2− ) for PBI-A, and below pH 7 (PBI ○− ) and pH 6 (PBI 2− ) for PBI-Y, (Section S4.2, Supporting Information). However, no correlation between the measured reduction potentials of the PBI units and the level of H 2 produced, which shows a peak at pH 5, is seen. This indicates that the pH dependence alone is not due to the driving force for charge transfer from the singly (PBI ○− ) and doubly reduced (PBI 2− ) species. [42,43] In addition to acting as an electron scavenger, methanol can change the solubility of PBI and its differing properties as a solvent can affect both the structure of the aggregates and the photophysics of the material. To explore the possible wider role of methanol concentration on photocatalytic activity and structure, we have expanded our study to use six methanol concentrations Regardless of the methanol concentration used, we find that both PBI-A shows the maximum level of photocatalytic H 2 evolution from pH 4.0 to 5.0, in agreement with the initial screen of PBIs and pH ( Figure 4). The activity of PBI-Y remains low at all pHs and methanol concentrations studied. Between pH 4.0 and 5.0, the rate of H 2 evolution of PBI-A shows a considerable methanol dependence. Considerably less H 2 is formed when there is no methanol present; this confirms the importance of the sacrificial electron donor indicating that in the absence of methanol the fast carrier recombination occurs preventing photoelectron transfer to the Pt co-catalyst, in agreement with past transient absorption studies on related systems. [33] The low-levels of H 2 seen at 0 v/v may indicate that in the absence of an electron donor complete water splitting can occur or more likely that self-oxidation of the photocatalyst can take place under these conditions. The maximum H 2 evolved occurred at 30 v/v methanol concentration; however, we do not see a simple trend for increased H 2 evolution with methanol concentration and a local maximum in H 2 production is observed around 5-10 v/v as well ( Figure 4).
We propose that the supramolecular structures formed at different pH and methanol concentrations are key in understanding the photocatalytic activity trends. To identify the supramolecular structures formed, small angle X-ray scattering (SAXS) was used. This is an excellent technique for studying structures in situ as it does not require drying of the samples, which can alter the supramolecular structures present. [31] SAXS data were collected for all of the PBIs at ten different pH values as well as a range of methanol concentrations for PBI-A and PBI-Y. From the results (Figures 3 and 5; extended discussion and data shown in Section S4.4, Supporting Information), it can be seen that the underlying self-assembled structure changes significantly across the pH range for each PBI. The samples prepared between pH 6.0 and 10.0 had no visible large aggregates; for these samples, the SAXS generally shows lower intensity scattering. Below pH 6.0, aggregates were visible by eye and the scattering intensity increased for all samples at low scattering vectors (Q), indicative of larger supramolecular structures forming at lower pH. A comparison between the rate of photocatalytic hydrogen evolution and the scattering intensity at low Q (0.01 Å −1 ; scattering at this Q is indicative of large structures) shows a correlation for PBI-A and PBI-V ( Figure 3). Between pH 6.0 to pH 5.0, we find the "switching on" of the H 2 evolution is accompanied by a large increase in scattering intensity at low Q. PBI-Y shows a similar pH dependent change in the scattering intensity at low Q, but is photocatalytically inactive. For PBI-H, the H 2 evolution is also very low (within error zero) and the scattering at low Q being similar from pH 7 to 3, with a significant decrease at pH 2. We therefore demonstrate that H 2 is only produced in appreciable quantities under conditions where large supramolecular aggregates are present in high concentrations and at pH values where our electrochemical measurements indicate that H 2 evolution is thermodynamically feasible (pH < 8). This is similar to our previous observations with PBI-F where H 2 evolution required the formation of large aggregated structures indicating a wider trend with these PBIs. [22] Fitting of the SAXS data to structural models was carried out using the SasView software. [44] Full fitting parameters for the PBIs are found in Section S4.4, Supporting Information and only the key points are highlighted here. For PBI-A at high pH , the data can be fitted to a sphere with a radius of 1 nm which is consistent with a small PBI aggregate of 2-6 stacked units rapidly tumbling in solution, based on an interlayer spacing of 3-4 Å. [7] The scattering is essentially unaffected until pH 6.0, where the data could be best fitted to a combined model of a sphere and a power law, representing a sphere of 1 nm radius and a power law of 2.6. The need to include the power law to fit the data at low Q suggests that more aggregation is beginning at this pH value. At pH 5.0, the data were best fitted to a combined model of a flexible cylinder (with a polydispersity in the radius) and a sphere. The polydispersity allows for a size distribution of cylinder radii to be accounted for in the fit. A good fit to the data could not be achieved with solely a cylinder model due the presence of a small bump at high Q. The requirement for a sphere model is attributed to the small aggregates that have yet to assemble into the larger structures that dominate the scattering data as pH is lowered further. At pH 5.0, the flexible cylinders have a length of 130.0 ± 0.5 nm, Kuhn length of 22.0 ± 0.1 nm, and a radius of 3.5 ± < 0.1 nm (polydispersity 0.25). At pH 4.0 and 4.5, the data were fitted exclusively to a flexible cylinder model with a polydispersity of radius. At pH 3.5 and below, the data were fitted to a combined model of a flexible cylinder and a power law. Here, the power law allowed for the fit to capture an increase in scattering at low Q. This additional scattering may be due to scattering off the fiber network, which is anticipated to form as the hydrogel forms. These data show that there is a structural transition, which begins at around pH 6.0, with 1D structures being formed at pH 5.0 and below. These data are consistent with our previous studies on PBI-A [45] and can be summarized by the cartoon in Figure 6. Fitting also reveals that for the other photocatalytically-active sample (PBI-V), the increase in scattering at low Q ( Figure 3) below pH 6 is due to the formation of 1D cylindrical structures ( Figure S15, Supporting Information).
The SAXS data for the inactive PBI-Y sample showed similar trends to PBI-A. At pH 10.0, the data were best fitted to spheres of 1 nm and a power law of 1.9 indicating the presence of larger aggregates. Again, as the pH was decreased larger aggregates form with data fitted to either a combined flexible elliptical cylinder and sphere model (pH 6.0 to 5.0) or solely a flexible elliptical cylinder (pH 4.5) model, (Figure 5b,d). Below pH 4.5 the scattering intensity drops significantly for PBI-Y likely due to precipitation, leading to less sample being in the beam. PBI-H is discussed in detail in the supporting information ( Figure S11, Supporting Information), briefly we also find that the data can be well fitted to a combination of flexible cylinders and spheres. The recognition that both PBI-H and -Y, both very poor photocatalysts, form the type of large flexible cylindrical structures seen to be active with PBI-A and -V is significant.
The SAXS studies demonstrate photocatalytic H 2 evolution requires the formation of long, flexible, cylindrical supramolecular aggregates. However, SAXS is unable to identify why we observe significantly lower H 2 evolution yields with PBI-Y and -H, both of which also form similar cylindrical structures, or why the system is so sensitive to the methanol concentration. The SAXS data for PBI-A at a range of methanol concentrations for pH 2.0, 4.5 and 7.0 (see Section S4.4.5, Supporting Information) show no significant changes in the type of large structures present.
Therefore, a UV-vis absorption study was conducted to study the local electronic environment of the self-assembled PBI's at different pH (Figure 7 and Figures S18-S21, Supporting Information). SAXS is an excellent probe of large structures whereas UV-vis absorption spectroscopy probes the assembly of the PBIs at the molecular level and can provide information on the nature of the packing and coupling between aggregated chromophores. Initially we examine the UV-vis spectra of PBI-A. Absorption data is recorded with PBI samples are at 5 mg mL −1 in the presence of 20 v/v methanol and 1% Pt to match photocatalytic conditions. At this concentration, solutions are very strongly colored and spectra in transmission mode could only be recorded by creating thin films by pressing samples between two optical windows without the use of a spacer making it impossible to maintain a constant pathlength between samples. Therefore, the absorption data is normalized. The spectra for PBI-A at 3 pH values is initially discussed (10,4,2), Figure 7a. The spectra of PBI-A at pH 10 at a concentration of 5 mg mL −1 shows a decrease in the ratio of the first two vibronic peak intensities in the spectrum of PBI-A compared to that recorded at 10 −6 m ( Figure S2, Supporting Information), a feature often taken to indicate H-aggregate formation [46] ; however, no hypsochromic shift of the peaks occurs and there is minimal spectral broadening suggesting that the degree of electronic coupling between PBI sub-units at this pH is not large. [47] Previous studies have reported the presence of relatively sharp PBI absorption bands when small PBI aggregates are present, [48] and our findings are in agreement with the fitting of the SAXS data at pH 10.
At pH 4.0, we observe significant spectral broadening and shifts in peak positions, with both hypsochromic shifted bands (502 nm at pH 10 to 478 nm pH 4) and a broad bathochromically shifted band at 571 nm appearing. These spectral features have been reported to be indicative of the formation of a faceto-face π stacked aggregates. [47] It is well known that for 1D π-π aggregated molecules that bathochromic or hypsochromic shifts in the vibronic structure of the S 0 to S 1 transition are indicators of J or H aggregate formation, respectively. Rotational displacement of substituted PBI's can accompany face-to-face stacking, leading to a relaxation of the selection rules and both bathochromic and hypsochromic shifts can be observed in the aggregates.
At very low pH (pH 2.0), we see a further change in the UVvis spectrum indicating additional changes in the local PBI structure within the flexible cylindrical structures present. Further broadening of the UV-vis spectrum occurs at pH 2 when compared to pH 4, likely in part at least, due to the high levels of aggregation between the supramolecular structures. Such behavior was identified in the SAXS data where it was necessary to introduce a power law component to account for the formation of networks of fibers. Using an approach recently demonstrated for the fitting of peptide modified PBI's to identify the presence of multiple aggregate types [11] we find that the spectra of PBI-A at any pH (A pH (λ)) can be modelled as a linear combination of the spectra at these 3 pH values, Equation (1).
where C x is the coefficient of each component spectrum (A x ) corresponding to the type of aggregate present at pH 10 (which we label as type 1), pH 4 (type 2), and pH 2 (type 3), Figure 7b. A plot of C x versus pH gives an estimate of the relative concentration of each PBI aggregate type, Figure 7c. At the pH values where H 2 evolution occurs the dominant spectral contribution is from type 2 aggregates. The UV-vis absorption spectra of PBI-V also shows very similar behavior with the data being fitted to a combination of 3 contributing spectra ( Figure S22, Supporting Information). For PBI-V we also find a correlation between level of photocatalytic H 2 evolution and the concentration of an aggregate species that has a UV-vis spectrum similar to the type 2 form seen with PBI-A. In contrast, for photocatalytically inactive samples PBI-H and PBI-Y the pH dependent UV-vis absorption spectra show different pH dependent behavior and we do not see evidence of the type 2 aggregate that is shown to be photocatalytically active for PBI-A and -V within the UV-vis spectra of either PBI-Y (Figure 7) or PBI-H ( Figure S23, Supporting Information). Instead the spectra of PBI-Y can be well fitted to a linear combination of just two spectra (Figure 7e, residual plots and further details of the fitting accompanies Figures S24 and S25, Supporting Information). At high pH (10 to 6), the UV-vis absorption spectra of PBI-Y 5 mg mL −1 with 20 v/v methanol shows a small decrease in the ratio of the first two vibronic peak intensities when compared to a solution of PBI-Y at 10 −6 m ( Figure S26, Supporting Information) but minimal broadening and shift in peak positions, indicating weak electronic coupling between PBI sub-units at this pH (type 1', Figure 7d). [47] At the lowest pH studied (pH 2), the UV-vis spectra of PBI-Y is broadened and under these conditions SAXS experiments show that the PBI aggregates form fiber networks of flexible cylinders (type 3', Figure 7d). Spectra between pH 6 and 2 have an increasing contribution from the type 3' aggregates as the flexible cylinders begin to form at lower pH's. The differences between the UV-vis spectra of PBI-A and -Y at pH 5 (20% methanol) shows that although both systems forming flexible cylindrical structures the PBI's within these structures are aggregated in different forms. Figure 8 compares the UV-vis absorbance spectra of PBI-A at pH 5.0 as the methanol concentration is changed. Again, the spectra can fitted to a linear combination of 3 spectra. SAXS shows that flexible cylinders are present in high levels at all methanol concentrations. In agreement with the SAXS data there is only a small contribution to the UV-vis spectra from the type 1 (small/weakly coupled) aggregates at all methanol concentrations ( Figure S27, Supporting Information). Instead the spectra of PBI-A mainly consist of type 2 and 3 aggregates. A plot of the relative contribution of the type 2 aggregates to the UV-vis spectrum as the methanol concentration is changed shows excellent agreement to the measured H 2 yield (Figure 8b). The UV-vis data clearly demonstrates that there is a specific sub-set of PBI-A and -V cylindrical structures with a distinctive UV-vis spectrum (type 2) that are photocatalytically active which are only formed under certain solvent/pH conditions.
Self-assembly of photoactive units in high dielectric solvents (e.g., water) provides an environment that effectively screens Coulomb potential lowering the energy of long-range charge transfer states (where the exciton is separated not just over the nearest neighbor but over across multiple PBI units). It has been reported that under charge screening conditions mixing of light induced Frenkel excitations, which are typically strongly allowed, and charge transfer excitations can enable the efficient formation of weakly bound delocalized excitons. [49] This is particularly beneficial for photocatalysis as it enables charge transfer to either the Pt catalyst (electron) or the sacrificial electron donor (hole) and hence efficient photocatalytic H 2 evolution. [36,49] Calculations on face-to-face stacked PMI photocatalysts, similar to the PBI's examined here, have also reported that the UV-vis absorption spectrum can be used as an indicator of the degree of Frenkel and charge transfer excitation mixing and hence an indicator of ease of electron/hole separation. Systems showing the fastest and most efficient charge separation were characterized by the asymmetry of the absorption line-shape and peak splitting leading to spectral broadening. These reported signatures of strong Frenkel and charge transfer excitation mixing match those observed here in the absorption spectra of the type 2 aggregates of PBI-A and PBI-V that are photocatalytically active. This suggests that through control of pH and methanol content it is possible to tune the local packing of the PBI units within the flexible cylinder leading to the formation of a face-to-face π-aggregated structure that enable efficient charge separation. In contrast, in PBI-Y and -H we see none of the previously described indicators of strong mixing of the Frenkel and charge transfer excitations within the absorption spectrum and it is proposed that excitonic states formed within the flexible cylinders remain localized on PBI units. [49] Finally, we turn to the pH dependence of the PBI-A and -V photocatalysts between pH 5 and 3. SAXS spectroscopy shows no changes in the supramolecular structure and the UV-vis spectra indicates that the type 2 aggregation form is present at all these pH values. It is important to consider that the change in pH will also change the surface charge of the structures formed. [50,51] Using NMR spectroscopy we have studied the effect of pH on surface charge by the study of chemical probes (such as Na + and dioxane-d 8 ) that interact with 1D PBI structures that align in the magnetic field as reported previously. [48,50,51] The results (Section S4.6, Supporting Information) suggest that the greatest interaction between Na + and the 1D structures, and therefore greatest negative surface charge on the structures, is around the apparent pK a . The apparent pK a is around pH 5.0 for both PBI-A in-line with the peak of H 2 evolution and the drop in photocatalytic activity mirrors the decrease in surface charge of the photocatalytic structure measured by NMR. It may be that the surface charge plays an additional role facilitating either charge separation; however, we see no change in the UV-vis spectrum. Alternatively, the electron transfer to the Pt co-catalyst may be facilitated by interaction between the charged surface and polar PVP-capped Pt particles.

Conclusions
Self-assembled structures represent a greatly underexplored field in photocatalysis. Whilst design of molecular chromophores is commonplace, there has been less attention paid to how these building blocks assemble and the role of the assembly process on photocatalysis. Here we focus, not on achieving record levels of H 2 production, but instead on understanding how and why a single set of starting molecules can give rise to profoundly different levels of photocatalytic activity. The use of high-throughput photocatalysis allowed for the simultaneous studies of multiple variables (e.g., pH, Figure 8. UV-vis absorption spectra of a) PBI-A at pH 5 at a range of methanol concentrations (0-40%). Solid lines indicate experimental data, dashed lines are the fit curve generated from a linear combination of the UV-vis spectra of PBI-A 20 v/v methanol at pH 2 (type 3), 4 (type 2), and 10 (type 1) using Equation (1). Panel b) shows the correlation between the measured H 2 evolution rate (red) with the presence of the type 2 aggregate (black) as the methanol concentration is changed. solvent mix, PBI functional group) which can all affect the self-assembly process. We completed approximately 350 photocatalysis studies in 2 weeks, in contrast a sample-by-sample approach would have required ≈6 months and it is likely that the optimal photocatalysis conditions would not have been discovered without this approach. This is important as it is feasible that viable photocatalysts are being prepared in labs world-wide but discarded after an initial activity analysis with a specific set of conditions showed low levels of activity. Using a combination of SAXS and NMR spectroscopies we have been able to show in our systems that formation of charged flexible cylindrical aggregates is required, but in-itself insufficient, for efficient photocatalysis. This can be rationalized by the clear correlation between UV-vis spectral features and photocatalytic activity that indicates that a specific form of local packing within the cylindrical aggregates is also essential for photocatalysis. It appears likely that in this study on supramolecular structures in aqueous solvent that the most active forms are those that lead to strong mixing of the Frenkel and charge transfer excitations. Future modelling and experimental studies will focus on confirming the specific packing motif and its role on the photophysical properties of the material. However, it is already clear that control of the self-assembly process offers an exciting route to tuning activity of photocatalysts for the production of solar fuels.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.