UvA-DARE (Digital Academic Repository) Using supramolecular machinery to engineer directional charge propagation in photoelectrochemical devices

Molecular photoelectrochemical devices are hampered by electron–hole recombination after photoinduced electron transfer, causing losses in power conversion efficiency. Inspired by natural photosynthesis, we demonstrate the use of supramolecular machinery as a strategy to inhibit recombination through an organization of molecular components that enables unbinding of the final electron acceptor upon reduction. We show that preorganization of a macrocyclic electron acceptor to a dye yields a pseudorotaxane that undergoes a fast (completed within ~50 ps) ‘ring-launching’ event upon electron transfer from the dye to the macrocycle, releasing the anionic macrocycle and thus reducing charge recombination. Implementing this system into p-type dye-sensitized solar cells yielded a 16-fold and 5-fold increase in power conversion efficiency compared to devices based on the two control dyes that are unable to facilitate pseudorotaxane formation. The active repulsion of the anionic macrocycle with concomitant reformation of a neutral pseudorotaxane complex circumvents recombination at both the semiconductor– electrolyte and semiconductor–dye interfaces, enabling a threefold enhancement in hole lifetime. Artificial photosynthesis aims to create photoelectrochemical (PEC) devices for the conversion of solar energy into fuel, using the natural photosynthetic process as a blueprint 1 . One of the challenges in PEC devices is efficient charge separation with concomitant suppression of competing charge recombination 2 , required for the generation of both photo-current and redox potential to drive energetically uphill

Molecular photoelectrochemical devices are hampered by electron-hole recombination after photoinduced electron transfer, causing losses in power conversion efficiency.Inspired by natural photosynthesis, we demonstrate the use of supramolecular machinery as a strategy to inhibit recombination through an organization of molecular components that enables unbinding of the final electron acceptor upon reduction.We show that preorganization of a macrocyclic electron acceptor to a dye yields a pseudorotaxane that undergoes a fast (completed within ~50 ps) 'ring-launching' event upon electron transfer from the dye to the macrocycle, releasing the anionic macrocycle and thus reducing charge recombination.Implementing this system into p-type dye-sensitized solar cells yielded a 16-fold and 5-fold increase in power conversion efficiency compared to devices based on the two control dyes that are unable to facilitate pseudorotaxane formation.The active repulsion of the anionic macrocycle with concomitant reformation of a neutral pseudorotaxane complex circumvents recombination at both the semiconductorelectrolyte and semiconductor-dye interfaces, enabling a threefold enhancement in hole lifetime.
Artificial photosynthesis aims to create photoelectrochemical (PEC) devices for the conversion of solar energy into fuel, using the natural photosynthetic process as a blueprint 1 .One of the challenges in PEC devices is efficient charge separation with concomitant suppression of competing charge recombination 2 , required for the generation of both photocurrent and redox potential to drive energetically uphill, fuel-forming reactions.The natural photosynthetic apparatus promotes effective charge separation through the organization of pigments and electron acceptors in specific geometries via supramolecular interactions (Fig. 1a).Photosystem II (PSII) uses the plastoquinone/hydroquinone (Q B /QH 2 ; Fig. 1a) redox couple to spatially remove electrons after photoinduced charge separation at the reaction centre.The terminal electron-accepting Q B is fixed by hydrogen bonding within the PSII protein, close to the plastoquinone A (Q A ; Fig. 1a) 3,4 .After two consecutive proton-coupled electron transfer events, Q B is reduced to hydroquinone QH 2 , and the affinity for the binding pocket of PSII is lost 5 .The liberated QH 2 diffuses away to participate in subsequent redox chemistry (at cytochrome b 6 f; Fig. 1a), and the PSII binding pocket is occupied by another Q B for the subsequent photocycle.Mimicking this photoinduced supramolecular control of docking-and-release events of a redox mediator may represent a viable strategy for reducing charge recombination, improving the power conversion efficiencies (PCEs) in PEC devices.
Dye-sensitized solar cells (DSSCs) are PEC devices with operational principles that parallel natural photosynthesis; (1) light absorption is achieved by molecular components (Fig. 1c), (2) photocurrent generation is initiated by a photoinduced electron transfer and (3) − instead of the Q B /QH 2 employed by PSII 6 .While the TiO 2 -based n-type DSSCs (n-DSSCs) exhibit a PCE of up to 14.3% (ref. 7), those of the complementary, NiO-based p-type DSSCs (p-DSSCs) are typically 5-10 times lower (current record at 2.51%) 8 .This disparity precludes efficiency improvements in tandem DSSCs and artificial photosynthetic PEC devices.The origins of PCE differences between n-and p-DSSCs is attributed to the charge carrier characteristics of the semiconductor, with very slow (4 × 10 −8 cm 2 s −1 ) 9 charge (hole) diffusion in NiO compared to that in TiO 2 (electron diffusion, 10 −4 cm 2 s −1 ) 10 .As a result of this slow hole transport, charge recombination at the semiconductor-dye interface and the semiconductor-electrolyte interface is a much larger issue in NiO-based p-DSSC (Fig. 1c, Pathways 5 and 6) 11,12 .The natural photosystem circumvents undesirable recombination pathways by preorganizing the redox mediator Q B at the electron-accepting docking site, followed by unbinding of the reduction product QH 2 , effectively separating the charges spatially.By contrast, the DSSC relies on collisional electron transfer under diffusional control, making the restrictive process the mass transfer of the reduced redox-mediator species away from the semiconductor surface to the counter electrode for regeneration.Former studies imply that dye-mediator interactions between dye and redox mediator could have a favourable effect for the overall PCE of both n-and p-DSSC [13][14][15][16] .
The stimulated binding and unbinding events found in the natural photosystem are of central importance in the field of artificial molecular machines.These include extraordinary examples of functional architectures including molecular pumps 17,18 , propellors 19,20 , robotic arms 21 , molecular muscles 22,23 and a nanocar 24 .Essential for the function of these molecular machines is the reversible bond, whose dynamic nature allows for molecular motion upon a chemical, electrochemical or photochemical stimulus 25,26 .An example relevant to the work at hand is the photoelectrochemical trigger that leads to the reduced affinity of a macrocycle for a binding site in (pseudo)rotaxane structures, resulting in molecular ring launching or shuttling events 27,28 .
The question that we address in this paper is if pseudorotaxane motifs can be used as molecular machinery, engendering the preorganization and launching of reduced redox mediators in a p-DSSC, and if such a launching mechanism is sufficiently fast to reduce charge recombination.For this we utilize a macrocyclic redox mediator (3-NDI-ring) that threads onto the dye P STATION to form the P STATION ⊂3-NDI-ring pseudorotaxane (Fig. 1b).Directional electron transport in the PEC device is established by the inbuilt free-energy impetus that actively shuttles reduced redox mediator 3-NDI-ring •- away from the thread on a relevant timescale as determined by time-resolved spectroscopy, promoting movement away from the semiconductor-dye interface.Finally, the thread favourably binds the next neutral 3-NDI-ring molecule to reform the pseudorotaxane.Application of this concept in a p-DSSC results in a PCE increase by a factor of 16 compared to the reference dye P PEG4 (a control that cannot form pseudorotaxanes) and by a factor of 5 for the P1 (the benchmark system for p-type DSSC), both attributable to reduced interfacial charge-recombination phenomena (Fig. 1b,c).

Design and synthesis
The p-DSSCs in this study are based on the well-documented P1 dye (Fig. 1e), and as such, the design of the molecular machinery started with this molecular scaffold.The dye P STATION is an analogue of P1 where the terminal (dicyano)vinyl electron acceptors are replaced with cyanoacrylate esters to facilitate introduction of a glycol-tethered 1,5-dioxynaphthalene (DNP).The DNP unit acts as a binding station (recognition site) for electron-deficient molecular rings through the formation of pseudorotaxane suprastructures.The influence of implementing bulky cyanoacrylate esters instead of (dicyano)vinyl electron acceptors is investigated via P PEG4 , a dye that is sterically analogous to P STATION , but lacking the DNP recognition sites and therefore the ability to form pseudorotaxanes with 3-NDI-ring.
In a preliminary study, the dye P STATION was demonstrated to bind cyclobis(paraquat-p-phenylene) as a ring, leading to enhanced photocurrents when applied in DSSCs.However, redox properties of this ring are suboptimal, and the binding of this highly charged ring was too strong, leading to dye detachment from the electrode surface at higher concentrations of this ring, preventing its use as the sole redox mediator in the device 16 .The naphthalene-diimide-based macrocycle 3-NDI-ring (Fig. 1e) also binds to the DNP recognition sites of P STATION , and was designed in an analogous fashion to NDI-based macrocycles previously reported to form pseudorotaxanes with DNP recognition sites at the surface-liquid interface 29,30 .As the 3-NDI-ring functions as a redox mediator in the envisioned p-DSSC, its redox properties are of key importance, and these compare favourably to the typically used I − / I 3 − (vide infra).Thus, the proposed photosensitizer, P STATION ⊂3-NDI-ring pseudorotaxane (Fig. 1e), is anticipated to improve the PCE of the DSSC device in two ways 16 .First, the 3-NDI-ring redox mediator is preorganized close to the dye by the DNP recognition sites of P STATION at the surface-electrolyte interface, favouring charge propagation (Fig. 1b 1) after photoexcitation of the pigment P680 and pheophytin (Phe; 2) leading to electron transfer to quinone Q A and terminal quinone Q B (4).Conversion of Q B to hydroquinone leads to its replacement by a new quinone from the quinone pool (Q pool) (5).The hydroquinone is regenerated by the cytochrome b 6 f complex.b, Schematic representation of the pseudorotaxane-based DSSC.The pigment in this artificial system is the P STATION dye with docking stations for the redox mediator, the macrocycle (3-NDI-ring).Upon photoexcitation of the dye (1) and hole injection into NiO (2), the electron (e -) is transferred, reducing the 3-NDI-ring (3), which is subsequently replaced by a neutral ring (purple), effectively removing the charged ring from the NiO-dye interface, thereby preventing charge recombination (4).The regeneration of the electron acceptor occurs at the counter electrode of the device (5 and 6).FTO, fluorine-doped tin oxide; D, dye (P STATION ); RC, redox couple; CE, counter electrode.c, Schematic representation of the forward electron propagation (1-4) and the recombination pathways (5 and 6).d, Schematic energy diagram for the p-DSSC based on the P STATION pseudorotaxane dye.The excited-state reduction potential (D*/D − ) was estimated by E D * /D − = E D/D − + E 0−0 (Table 1) 48 .Energy levels are represented in volts (versus NHE).VB, valence band; CB, conduction band.e, Molecular structures of P STATION and the 3-NDI-ring, forming the pseudorotaxane-based molecular machine (top), and the control dyes P1 (benchmark dye for p-DSSC) and P PEG4 (pseudorotaxane formation inhibited), which cannot form pseudorotaxanes (bottom).

Table 1 | Summary of the optical and electrochemical properties of P1, P PEG4 and P STATION (0.5 mM) in MeCN 16
Molecule To all samples, ferrocene/ferrocenium (Fc/Fc + ) was added as an internal redox standard to determine the redox potentials versus NHE (E 1/2 Fc/Fc + = 630 mV in MeCN 46 and 700 mV in DCM 47 ).E 0-0 is determined from the intersection between the normalized absorption and the emission spectrum.ΔG° is the driving force for electron transfer to the 3-NDI-ring.ε, molar absorptivity; E 0-0 , zero-zero transition energy; E red , reduction potential; E ox , oxidation potential.The P1 dye 32 and the P STATION dye 16 were synthesized according to the literature.The dye P PEG4 was synthesized according to a protocol described in Supplementary Section 1.2.The absorption maximum (λ max ) of the P STATION (λ max = 455 nm) and P PEG4 (λ max = 453 nm) experiences a slight blueshift compared to P1 (λ max = 472 nm; Fig. 2a).This particular absorption, derived from an intramolecular charge transfer (CT) in the dyes, highlights the decrease in respective acceptor strength (that is, cyanoacrylate in P STATION and P PEG4 versus (dicyano)vinyl in P1) between molecules.The 3-NDI-ring was synthesized in two steps, employing Mitsunobu coupling to effect ring closure between pyromellitic diimide and the 3-NDI fragment in 31% isolated yield (Supplementary Section 1.2).
Binding studies of 3-NDI-ring to (the DNP recognition site within) P STATION were hampered by the limited solubility of the dye; therefore, the recognition site moiety DNP-thread (Fig. 2d) was used to analyse pseudorotaxane formation by 1 H NMR titration.A typical upfield shift (0.5 ppm) in the 1 H NMR spectra for the aromatic protons of the 3-NDI-ring in CD 2 Cl 2 was observed (Supplementary Fig. 9) upon titration against DNP-thread 29 .Fitting the DNP-thread⊂3-NDI-ring titration curve to a model for 1:1 binding revealed an association constant (k a ) of 210 M −1 .Pseudorotaxane formation between the electron-rich and electron-deficient components typically leads to a CT band at visible wavelengths.UV-visible (UV-Vis) spectrophotometry in a valeronitrile/MeCN (15:85) solution of DNP-thread⊂3-NDI-ring (10:1) revealed the characteristic CT band evolving at 460 nm in line with pseudorotaxane formation (Fig. 2b) 33 .The k a ascertained from 1 H NMR was complemented by spectrophotometry by probing the formation of the DNP-thread⊂3-NDI-ring complex by UV-Vis titration (Fig. 2d).The spectral overlap of the P STATION CT absorption (wavelength, λ = 455 nm) precluded observation of the CT evolving from pseudorotaxane formation (λ = 460 nm) over the course of the titration.Monitoring the absorption at 460 nm and fitting to a 1:1 binding model afforded a k a = 160 M −1 for the DNP-thread⊂3-NDI-ring pseudorotaxane (Fig. 2c and Table 1).The difference between k a values derived from NMR and UV-Vis experiments is due to the difference in solvent (CD 2 Cl 2 versus valeronitrile/MeCN (15:85), respectively).Immobilization of P STATION onto NiO electrodes (vide infra; Supplementary Fig. 14) and immersion into a 3-NDI-ring solution (20 µM in MeCN) led to a decrease in 3-NDI-ring absorption intensity at λ = 378 nm.Given that the control experiment with P1 in place of P STATION (Supplementary Fig. 15) experienced no absorption drop at 378 nm, we could ascribe the absorption decreases to the binding of 3-NDI-ring to the DNP recognition sites of the P STATION NiO.
Cyclic voltammetry (CV; Supplementary Fig. 10) of 3-NDI-ring revealed four reductions, attributed to two independent reduction events at the naphthalene diimide (NDI) and two at the pyromellitic moieties of the 3-NDI-ring.The redox events were chemically reversible, demonstrating sufficient stability, which is an important requirement for redox mediators in DSSCs.The first reduction potential of the 3-NDI-ring (−0.35 V versus normal hydrogen electrode (NHE)) is 0.55 V lower than that of P STATION − (Fig. 1d), facilitating exergonic electron transfer from P STATION − to 3-NDI-ring.CV of the DNP-thread⊂3-NDI-ring complex shows that binding of the DNP-thread to the 3-NDI-ring has a small effect on the reduction potential (40 mV).Importantly, scan-rate-dependent CV experiments demonstrate that reduction of 3-NDI-ring in the model DNP-thread⊂3-NDI-ring prompts a loss of affinity and unbinding (that is 'ring launching') from the DNP-thread (Supplementary Fig. 11) 34 .This 'ring-launching' effect, reflected by the 40 mV reduction potential decrease, is observed only for the first reduction event of the 3-NDI-ring when bound to the DNP-thread (−0.39 V versus NHE, compared to −0.35 V versus NHE for the free 3-NDI-ring (Supplementary Table 2).Importantly, the absence of this typical shift in the three subsequent reduction events shows that the mono-reduced ring 3-NDI-ring •− unbinds the DNP-thread after the first reduction.
https://doi.org/10.1038/s41557-022-01068-y the P1 benchmark and those based on the P PEG4 control system.All photovoltaic parameters are better compared to the DSSC based on control dye P PEG4 , with 165 mV enhancement in V OC (P PEG4 = 166 mV versus P STATION = 331 mV) and a six times higher J SC (P PEG4 = 0.063 mA cm 2 versus P STATION = 0.388 mA cm 2 ), leading to a 16-fold increase in PCE (P PEG4 = 0.003% versus P STATION = 0.048%).
The pseudorotaxane-based DSSC using P STATION also outperforms the benchmark system based on P1, despite the 50% difference in dye loading, with 123 mV enhancement in V OC (P1 = 208 mV versus P STATION = 331 mV in) and a tripling of J SC (P1 = 0.143 mA cm 2 versus P STA- TION = 0.388 mA cm 2 ), leading to a fivefold increase in PCE (P1 = 0.009% versus P STATION = 0.048%).
The photocurrent action spectrum of P STATION reveals a higher incident photon-to-current conversion efficiency (IPCE) across the spectrum (Fig. 3b) with a maximum of 5.3% at 472 nm compared to the P PEG4 -based system (maximum IPCE of 0.75% at 470 nm; P1 maximum IPCE = 2.6% at 500 nm).The improved fill factor in the P STATION ⊂3-NDI-ring-based cells compared to the P PEG4 control system, in combination with the higher V OC , suggests reduced charge recombination resulting from preorganization of the redox mediator to the dye via pseudorotaxane formation.
Whereas the better performance of P STATION compared to P1 could in principle be partly explained by the difference in size, which can have a positive effect on suppressing recombination at the semiconductorelectrolyte interface 15 , this is not the case for P STATION and P PEG4 , as these are similar in size (head-to-tail size, based on PM3 calculations, is 16, 29 and 35 Å for P1, P PEG4 and P STATION , respectively).The poor performance of the control p-DSSCs based on the P PEG4 implies that the performance enhancements observed for P STATION ⊂3-NDI-ring compared to both the P PEG4 -based and the P1-based devices cannot be attributed to the molecular size differences between P STATION and benchmark P1.Furthermore, as P STATION and P PEG4 differ only in the presence of the DNP recognition site, the substantial improvement in P STATION ⊂3-NDI-ring devices must stem from its ability to engage in supramolecular interactions.
Chopped light amperometry experiments were performed, where the light is switched on and off in periods of 10 seconds with an increasing illumination density starting from 5 mW cm −2 to 50 mW cm −2 at short current conditions.For both solar cells, the J SC increases with the light intensity, as expected for these p-DSSCs (Fig. 3c), and the J SC enhancements for P STATION ⊂3-NDI-ring-based cells is in line with the improved PCE.Interestingly, the shape of the photocurrent response for the P1 shows a tailing behaviour that increases with light intensity.This indicates mass transfer limitations of the redox mediator through the mesoporous electrode (Fig. 3c inset), expected for large molecules like 3-NDI-ring at low concentrations 35,36 .This tailing behaviour is not observed for P STATION -based DSSCs, entirely consistent with preorganization of the redox mediator and efficient replacement of reduced, ring-launched 3-NDI-ring •-for neutral 3-NDI-ring 0 , leading to high local concentrations of 3-NDI-ring at the dye-electrolyte interface even at low (25 mM) concentrations of redox mediator.
Differences in solar cell performance originating from pseudorotaxane formation were further probed by electrochemical impedance spectroscopy (EIS).Performing EIS under varying light intensities affords insight into electron-hole recombination at the semiconductor-dye interface through determination of the hole lifetime (τ h ) as a function of V OC (Fig. 3d) 37 .At any given V OC , the hole lifetime for P STATION (624 ms at 0.1 V) is at least three times longer than for P PEG4 (198 ms at 0.1 V; P1 is 324 ms at 0.1 V), implying that less recombination occurs in the pseudorotaxane system.This could arise either from a difference in recombination resistance (R REC ) or from a change in chemical capacitance (C µ ), originating from a valence band shift 38 .The C µ (Supplementary Fig. 22d) shows no dependency on the applied voltage and a minimal shift between the P1, P PEG4 and P STATION DSSCs, expected given the similarity of the systems.For this reason the valence bad shift cannot be the reason for V OC enhancements in P STATION ⊂3-NDI-ring p-DSSCs.The measured R REC for the P STATION system (3.20 × 10 5 Ω cm −2 at 0.1 V) is higher than for P PEG4 (1.31 × 10 5 Ω at 0.1 V; P1 is 2.93 × 10 5 Ω cm −2 at 0.1 V), meaning that the difference in hole lifetime originates from a lower recombination at the semiconductor-electrolyte interface.The suppression of recombination is known to lead to an improved V OC 39-41   .The improved V OC that we observe for P STATION indicates that in these systems, the recombination is inhibited compared to the system based on the control dye P PEG4 and the benchmark system P1.This effect further supports the active charge rectification bestowed by introducing molecular machinery in the P STATION ⊂3-NDI-ring p-DSSCs to influence the preorganization and replacement of the redox mediator in the solar cell.
Further mechanistic insight was gained by performing femtosecond transient absorption (TA) spectroscopy measurements on fluorine-doped tin oxide (FTO) plates with NiO|P STATION in a supporting electrolyte (1.5 ml, 1 M LiTFSI valeronitrile/MeCN, 15:85).TA spectra obtained at various time delays after 500 nm excitation are presented in Fig. 4, and further detailed in Supplementary Section 3. In the absence of the 3-NDI-ring redox mediator (Fig. 4a), a photoinduced absorbance around 575 nm (black dashed line) is detected along with a ground state bleach, which especially dominates the TA signal at <540 nm.The photoinduced absorbance around 575 nm can be assigned to a mix of P STATION * and P STATION

•-
, the latter formed by photoinduced hole injection into the NiO either within the instrumental response time (100-150 fs) or a few picoseconds [42][43][44] , which can explain the minor redshift observed over time.The TA signal has mostly decayed at 500 ps due to charge recombination.
Upon introduction of the 3-NDI-ring redox mediator, additional phenomena become apparent as a result of P STATION ⊂3-NDI-ring pseudorotaxane formation (Fig. 4b).The spectra feature a new absorbance band evolving around 615 nm (red dashed line).This band can be assigned to the generation of the reduced redox mediator (that is, 3-NDI-ring •-) and is consistent with the spectroelectrochemistry of the 3-NDI-ring •-(Fig.4c).Importantly, this new absorbance feature coincides with the development of a negative differential absorbance below 580 nm, evolving over time after photoexcitation (indicated by the arrow in Fig. 4b).
The origin of this negative differential absorbance at <580 nm was further investigated by comparing the kinetic traces at 556 nm in the absence and presence of the 3-NDI-ring (Fig. 4d).In both cases, the signal at early times is dominated by a photoinduced absorbance by P STATION * and P STATION •-, which decays in time due to slow (a few picoseconds) hole injection into the NiO (P STATION *) and charge recombination (P STATION •-).Interestingly, for the system with 3-NDI-ring present, a bleach is observed that develops in time and does not decay within the experimental time window, demonstrating that the system is not returning to the initial P STATION ⊂3-NDI-ring pseudorotaxane ground state in this time frame.The development of this bleach in time can The average performance (N = 5 for P PEG4 ; N = 9 for P1 and P STATION ) is provided with the best performing cell in brackets.
https://doi.org/10.1038/s41557-022-01068-ybe rationalized by comparison of the steady-state absorption spectra of P STATION and the P STATION ⊂3-NDI-ring pseudorotaxane (Fig. 4e).In both spectra, the P STATION component features a strong absorption at ~450 nm, originating from a donor-to-acceptor intramolecular CT within the dye.Introduction of the 3-NDI-ring to a solution of P STATION results in a broadening of this band (Fig. 4e, red), as evolution of an additional intermolecular CT band at 460 nm is prompted upon formation of the P STATION ⊂3-NDI-ring pseudorotaxane (that is, Fig. 2b; vide supra).For this reason, the absorbance spectrum of the semiconductor-supported analogue (that is, NiO|P STATION ⊂3-NDI-ring) can also be regarded as a superposition of both P STATION and the CT band from pseudorotaxane formation.With this knowledge in hand, we can ascribe this bleach at <580 nm and evolving in time after photoexcitation to the loss of the CT band upon launching of the 3-NDI-ring from the station.This CT band is not restored within the 500 ps experimental time window, as diffusional transport of a new 3-NDI-ring (given a diffusion constant of ~1.4 × 10 10 m 2 s −1 and the 5.9 mM concentration) from the bulk solution would likely occur in ~100-200 ns (ref. 45), explaining the long-lived bleach observed for the system with the 3-NDI-ring present.As the absorbance around 615 nm is already visible at early times after photoexcitation, reduction of the 3-NDI-ring as pseudorotaxane (NiO|P STATION ⊂3-NDI-ring) is likely an ultrafast process occurring within the instrumental response time (100-150 fs) simultaneous with ultrafast hole injection into the NiO (τ 1 ; Supplementary Table 16).Further slower (a few picoseconds) hole injection into the NiO [42][43][44] accompanied by 3-NDI-ring reduction is likely responsible for the further redshift in differential absorbance in this time window (τ 2 ; Supplementary Table 16).The ring-launching event commences after generation of the 3-NDI-ring •-radical anion and is completed within ~50 ps (Fig. 4d).Hence, we observe that both reduction of the 3-NDI-ring and dissociation of the 3-NDI-ring •-radical anion are apparent in the film.As observed in TA spectroscopy, the CT band characteristic to the pseudorotaxane has completely disappeared within 50 ps, which is faster than the charge-recombination time (>500 ps) or any competing diffusion-based process (that is, a nanosecond timescale 45 ).This in turn leads to the improvement in J SC , validating the proposed mechanism (Fig. 4f) of operation in this DSSC.

Conclusion
Charge recombination is one of the key issues to solve in p-DSSCs.Inspired by the binding/unbinding events of redox mediators in natural photosynthesis, we studied if molecular machinery can be implemented in p-DSSC to facilitate directional electron transport and reduce charge recombination.The P STATION dye based on the P1 benchmark system is equipped with a docking station for preorganization of the ring-shaped redox mediator (3-NDI-ring) that forms pseudorotaxanes.Reduction of the 3-NDI-ring by electron transfer in P STATION -⊂3-NDI-ring pseudorotaxane prompts disassembly of the supramolecular complex, resulting in ring launching of the reduced mediator, making space for a new neutral redox mediator to bind.The p-DSSCs based on P STATION ⊂3-NDI-ring pseudorotaxanes exhibit enhanced performance across all photovoltaic parameters in comparison to devices based on both P1 and the P PEG4 control dye, which do not facilitate preorganization of the 3-NDI-ring mediator.Chopped light amperometry and EIS under varying light intensities showed that both preorganization and ring launching contribute to lowering recombination and a threefold extension of hole lifetimes, leading to a higher V OC and a 16 times increase in PCE in p-DSSC.
Femtosecond TA studies presented herein provide direct evidence for the ring-launching mechanism through loss of the pseudorotaxane-derived CT band after generation of the 3-NDI-ring •- radical anion, commencing photoinduced hole injection into the NiO, occurring either ultrafast (<100-150 fs) or in a few picoseconds.Ring launching is fast, as the CT band completely disappears within ~50 ps, and is faster than charge recombination (>500 ps) or any other competing process.We envision that this bio-inspired approach to integrate artificial molecular machinery in p-DSSCs for supramolecular CT rectification is a strategy that could be expanded to other PEC devices for solar energy conversion technologies.

Online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41557-022-01068-y.

Fig. 1 |
Fig.1| Similarities between electron propagation in PSII and pseudorotaxane-based DSSC.a, Electrons are extracted from water (1) after photoexcitation of the pigment P680 and pheophytin (Phe; 2) leading to electron transfer to quinone Q A and terminal quinone Q B(4).Conversion of Q B to hydroquinone leads to its replacement by a new quinone from the quinone pool (Q pool)(5).The hydroquinone is regenerated by the cytochrome b 6 f complex.b, Schematic representation of the pseudorotaxane-based DSSC.The pigment in this artificial system is the P STATION dye with docking stations for the redox mediator, the macrocycle (3-NDI-ring).Upon photoexcitation of the dye (1) and hole injection into NiO (2), the electron (e -) is transferred, reducing the 3-NDI-ring(3), which is subsequently replaced by a neutral ring (purple), effectively removing the charged ring from the NiO-dye interface, thereby https://doi.org/10.1038/s41557-022-01068-yits affinity for the DNP recognition site of P STATION and is replaced by a neutral 3-NDI-ring from the bulk electrolyte.The reduced 3-NDI-ring •- is thus actively repelled from the NiO-dye interface (launching effect; Fig. 1b, Step 4), preventing charge recombination (Fig. 1c, Pathway 6).After ring release, a new 3-NDI-ring binds to the DNP recognition site of P STATION .As a result, the semiconductor surface is more shielded, thereby inhibiting recombination at the semiconductor-electrolyte interface (Fig. 1c, Pathway 5).Such a shielding effect through the introduction of insulating butoxyl chains on the dye was previously reported by Hagfeldt and coworkers 31 .The launched 3-NDI-ring •- is regenerated at the counter electrode, leading to photocurrent (Fig. 1b, Steps 5 and 6).It is therefore anticipated that the creation of unidirectional charge propagation at a molecular level should translate to macroscopic charge rectification in the device, which should inhibit both recombination pathways (Fig. 1c, Pathways 5 and 6) resulting in enhanced open circuit voltage (V OC ) and short circuit current density (J SC ) and therefore improved PCE.

Fig. 3 |
Fig.3| Photovoltaic performances of the DSSCs based on the P1 dye, P PEG4 dye and P STATION dye with the 3-NDI-ring.a, J-V curves (J, photocurrent density) of the devices based on the P1 dye (blue squares), P PEG4 dye (black triangles) and the P STATION dye with the 3-NDI-ring as redox mediator (red dots; 25 mM in 1 M LiTFSI), which can only form pseudorotaxanes with the P STATION dye.b, A photocurrent action spectrum of the P STATION -based system (red), P PEG4 -based system (black) and P1-based system (blue).c, Chopped light amperometry at different light flux values varying from 5-50 mW cm −2 with on/off cycles of 10 seconds.The inserted graph displays the decay effect (tailing behaviour) observed in the P1-based system for the photocurrent at 40, 45 and 50 mW cm −2 .d, Hole lifetime τ h as function of V OC obtained through EIS measurements under varying light intensities measured on the P1 system (blue), P PEG4 system (black) and P STATION system (red).For clarity, the figure presents data from a single experiment (N = 1), with replicate measurements presented in Supplementary Fig.23.

Fig. 4 |
Fig. 4 | TA (λ exc.= 500 nm) data of P STATION on NiO in supporting electrolyte (1.5 ml, 1 M LiTFSI valeronitrile/MeCN,15:85) in the absence and in the presence of the 3-NDI-ring (5.9 mM).a, TA spectra at given time delays in the absence of the 3-NDI-ring.The photoinduced difference in absorbance (ΔA) around 575 nm is indicated with a black dashed line.b, TA spectra at given time delays in the presence of the 3-NDI-ring, featuring a new absorbance band evolving around 615 nm (red dashed line).The negative signal that emerges is indicated with the blue arrow.c, Spectroelectrochemistry of the 3-NDI-ring and the DNP-thread⊂3-NDI-ring pseudorotaxane shows an absorption band around 615 nm when the 3-NDI-ring •-is formed, visible in NiO|P STATION in the presence of the 3-NDI-ring but not detected in the absence of the 3-NDI-ring (indicated with the red dashed line), supporting the idea that formation of the 3-NDI-ring •-takes place in the NiO|P STATION film.Cartoon represents reduced ring.d, Kinetic traces at 556 nm with (red) and without (black) 3-NDI-ring present.The red trace shows a clear negative signal developing in time after photoexcitation, indicative of ring launching.e, Broadening of the absorption of P STATION when 3-NDI-ring is added, attributable to pseudorotaxane formation, rationalized as a superposition of absorption from P STATION and the CT band from the P STATION ⊂3-NDI-ring complex.f, Implications of the TA results, where the loss of the CT band indicates ring launching.The bleach at 556 nm implies the loss of the CT band, corresponding to ring launching.Dotted lines are to guide the eye.