Photocatalytic Hydrogen Generation by Vesicle‐Embedded [FeFe]Hydrogenase Mimics: A Mechanistic Study

Abstract Artificial photosynthesis—the direct photochemical generation of hydrogen from water—is a promising but scientifically challenging future technology. Because nature employs membranes for photodriven reactions, the aim of this work is to elucidate the effect of membranes on artificial photocatalysis. To do so, a combination of electrochemistry, photocatalysis, and time‐resolved spectroscopy on vesicle‐embedded [FeFe]hydrogenase mimics, driven by a ruthenium tris‐2,2′‐bipyridine photosensitizer, is reported. The membrane effects encountered can be summarized as follows: the presence of vesicles steers the reactivity of the [FeFe]‐benzodithiolate catalyst towards disproportionation, instead of protonation, due to membrane characteristics, such as providing a constant local effective pH, and concentrating and organizing species inside the membrane. The maximum turnover number is limited by photodegradation of the resting state in the catalytic cycle. Understanding these fundamental productive and destructive pathways in complex photochemical systems allows progress towards the development of efficient artificial leaves.


Introduction
The transition to as ustainable, green energy economyn ecessitates the development of ac orrespondingt echnology through which abundant and renewable resources can be used to generate transportablef uels, such as hydrogen gas. The most straightforward process for the generation of hydrogen is the photolysis of water with sunlight by using bioinspired synthetic photocatalytic systemsk nown as artificial photosynthesis. Since the first working example reported by Fujishimaa nd Honda in 1972, [1] research into the design and developmento f these "artificial leaves" [2] has made spectacular progress, [3][4][5] up to the point where synthetic and biological systems have started to merge. [6][7][8][9][10][11][12] Many artificial photosynthetic systems are inspired by the naturallyoccurring[FeFe]hydrogenase enzyme, which catalyzes proton reduction with high efficiencya nd turnover numbers (TONs). [13] Research by the groups of Darensbourg and Lubtiz expanded mechanistic and structurali nsights into the activity of [FeFe]hydrogenase and showedt he importance of the essential protein environment surrounding the [FeFe]c ore. [14][15][16][17] Simultaneously,aw ide variety of synthetic mimics based on the [FeFe] core of hydrogenase have been developed with differentb ridgeheads that evolve hydrogen followingd ifferent mechanisms. To apply hydrogenasem imics in artificial leaves, these catalysts have been studied for their performance in light-driven hydrogen evolution, but mainly in organic solvents due to the apolar nature of these complexes.T ypically,t hese photocatalytics ystemsc onsist of three components:ap hotosensitizer (PS), as acrificial electron donor (SED), and the catalyst itself. [18] The first report on this system by Sun et al. showed that, after reductive quenching of the excited state of [Ru(bpy) 3 ] 2 + (bpy = 2,2'-bipyridine),e lectron transfero ccurred to an azadithiolate-bridged Fe 2 S 2 clusteri ns olution, [19] leading to ap hotocatalytic system that yieldedu pt o5turnovers, [20] which was improved upon by Ott et al. through the introduction of ad ichlorobenzenedithiolate-bridged Fe 2 S 2 cluster, which yielded 200 turnovers. [21] To avoid the use of organic solvents that might interfere with catalysis, the group of Wu studied this photocatalytic system by using water-soluble Fe 2 S 2 analogues,w hich gave similarT ONs and, similar to previous systems,d eactivation of the system after 1t o2h. [22,23] On the contrary,t he use of intact [FeFe]hydrogenase-containing cells from Thiocapsa roseopersicina in the same [Ru(bpy) 3 ] 2 + system leads to uninterruptedh ydrogen evolution for 12 h, with only slow decomposition observed for the isolatede nzymei np hosphatidylcholine vesicles. [24] Evidently,t he matrix in which the Fe 2 S 2 cluster is embedded has an important role in stabilization of the catalystd uring photocatalysis, and as ynthetic matrixs hould ideally mimict he functiono ft he originale nzyme-plus-cell structure. Systems in which hydrogenase mimics were embedded in matrices, such as micelles, [25,26] amphiphilic polymers, [27] vesicles, [28] proteins, [29] metal-organic frameworks, [30] and hydrogels, [31] displayed TONs below 1000, but generally ap ositive influence of the matrix on the overall efficiency of the catalytic system wasreported.
Moreover,p hotodriven reactions in nature are performedi n the presence of am embrane, which consists of amphiphilic molecules that self-assemble into bilayers in an aqueous environment. The membrane functions can be described as organizing, localizing, and concentrating reactive complexes to enhance reactions and suppress side reactions. In light of this, Kçnig andc o-workers assembled both membrane-embedded water oxidationc atalysts and protonr eduction catalysts, and showedw ater oxidation and hydrogen evolution in presence of a[ Ru(bpy) 3 ] 2 + as the PS. [28,32,33] Inspired by these results, we were interested in how preorganization of the catalyst and PS components in al ipid bilayer affected the mechanism of (photo)catalytic hydrogen evolution. Herein, we investigated [Ru(bpy) 3 ] 2 + (Ru 2 + + )a sadyea nd [(m-bdt)Fe 2 (CO) 6 ]( 1;b dt = benzene-1,2-dithiolate) as ah ydrogen-evolving catalyst embedded in l-a-phosphatidylcholine (PC)-based vesicles ( Figure 1), and used ac ombinationo fe lectrochemistry and time-resolved spectroscopy,v isible spectroscopy,a nd IR spectroscopy to investigate light-driven proton reductioncatalysis. [34] Results and Discussion Preparation and characterization of the vesicles We investigated as upramolecular system in which [FeFe]-benzodithiolate catalyst 1 served as ap rotonr eduction catalyst that wasembedded in lecithin-based vesicles,which were similar in terms of compositiona nd concentrations to the vesicular system reported by Kçnig et al. [28] This system self-assembles in aqueous buffer solution in either the presence or absence of 1 (detailsi nt he Experimental Section). Throughout this research, vesicles are studied under av ariety of conditions and the presence of vesiclesw as confirmed by meanso fd ynamic light scattering (DLS). The vesicles formed were similar over ap H range from 4t o7 ,i nt he presence of 1 (concentrations 0/0.1/ 0.5 mm), by using ab uffer medium (0.1 m ascorbate/phosphate). Small deviationsi nv esicle diameter are attributed to variations during preparation (Table S1 in the Supporting Information).
Inclusion of 1 in the lipid bilayer was confirmed by using IR spectroscopy ( Figure 2). Complex 1 does not dissolve in water and PC is used to solubilize this apolar complex in aqueous medium. The iron-carbonyl bands are clearly visible; the three absorption bands of the stretching modes are located at ñ = 2078, 2043, and2 004 cm À1 .B ecause the width of these bands correlates linearly with the polarity and polarizability solvent parameter p*, [35] we could use this information to probe the chemicale nvironment of 1 inside the vesicle. The IR spectrum is similart ot he spectrum of 1 in ethyl acetate, and quite distinct from that in pentane; this suggestst hat the carbonyl fragments are in the bilayer in closep roximity to the polar head groups of the lipids.

Electrochemistry
Electrochemical experiments on a0 .1 mm solution of 1 in as olution of vesicles containing 0.1 m sodium phosphate buffer show electrochemical responses similart ot hat in organic solvents.C omplex 1 evolvesh ydrogen in organic solvents in the presence of acid through double electront ransfer followed by protonation (EEC) to form 1H À . [36] In the presence of strong acid, 1H À is protonated to evolve H 2 ;h owever,aweak acid is not able to protonate 1H À ,a nd therefore, as econd reduction step must take place before protonation can occur.C omplex 1 inside vesicles shows ar eduction wave with ap eak potential at À0.8 Vv ersus an ormal hydrogen electrode (NHE;F igure 3A). Anodic reoxidation at thesep otentials was not observed; this is indicative of ap rotonation step after reduction. Ar eoxidationw ave was observed at potentials around À0.1 V  versus NHE, which is roughly 0.7 Vm ore positive than that of the reduction event. This behavior is in line with the two-electron reduction of 1 in organic solvents in the presence of weak acids. [36,37] Because one step involves protonation, the effect of pH wastested by performing cyclic voltammetry (CV) measurements at pH 4t o7( Figures S1-S5 and Tables S2 and S3 in the Supporting Information). The half-wavep otentialo ft he redox processes was determined by meanso fD PV ( Figure 3D)a t À0.73 Vv ersusNHE (whichwet entatively assign to at wo-electron reduction process, based on the similarity of the CV results to that of 1 in organic solventi nt he presence of aw eak acid) and À0.14 Vv ersus NHE (the reoxidation process), irrespectiveo ft he pH of the bulk solution ( Figure 3E). The peak potentials and peak currentso btained from CV were also independento fp H. In ap reviously reported micellar solution of 1, the reduction process did show ap Hd ependenceo nt he peak potentials and currents. [38] This is in contrastw ith our finding, which indicates that the membrane environmento ft he vesicles under study provides ac onstant local effective pH to the catalyst, as observed previously in similar lipid bilayers. [39] During electrochemical measurements, vesicles appeared to be adsorbed onto the glassy carbon working electrode. This was demonstrated by removing the electrode from the solution after af ew scans, rinsing with deionized water,a nd submerging in fresh buffer solution, after which similar voltammograms were recorded. In the cyclic voltammograms,t he cathodic peak currents, i p ,s cale linearly with scan rate to the power 0.75 (n 0.75 ;F igure 3B), which is behaviorb etween that of freely diffusing( i p proportionalt on 0.5 )a nd surface-adsorbed (i p proportional to n)r edox-actives pecies. [40] This partial diffusion behavior can be attributed to electron hopping throught he surface-adsorbed lipid bilayer mediated by 1 because there is no significant uncompensated for solution resistance during the measurements. [41][42][43] Overall, the electrochemistry of 1 in vesicles is similar to that of 1 in organic solvent, withr espect to redox potentials and protonation behavior. [36] Photocatalytic hydrogen production Photocatalytic hydrogen evolution from 1 in vesicles was studied by using Ru 2 + + (in our case [Ru(bpy) 3 ]Cl 2 )a st he PS and ascorbic acid as the SED. In these experiments, as olution (5 mL) containing 0.1 mm 1 in 0.9 mm PC vesiclesa nd 0.2 mm Ru 2 + + in 0.1 m ascorbate buffer wasi rradiated with l = 450 nm lightemitting diodes (LEDs)a nd hydrogen formation wasq uantified with an in-line GC setup. Hydrogen evolution startedi mmediately after switching on the LED and ceased within 30 min (Figure 4A). Controle xperiments, in which one of the components (Ru 2 + + , 1, PC or ascorbate) was omitted, did not yield any detectable amountso fhydrogen and demonstrated the necessity of every component in the mixture.
As discussed in the Electrochemistry section, the vesicles provideaconstant pH to the catalyst;hence photocatalytic hydrogen evolution in bulk solution can be optimized by finetuning the conditions for the (photoinduced) electron-transfer steps independently of the hydrogen-producing part of the system.T he pH of the ascorbate buffer wasv aried between 4 and 7a nd an optimum TON of 67 was obtaineda tp H4.5. We hypothesize that this pH optimum arises from an increased amount of ascorbate (versus ascorbic acid) at high pH and increasedp roton reduction activity by 1 at low pH. [44] We need a high ascorbate concentration because this acts as an electron donor,w hereas the protonated species, ascorbic acid, does not. The pH optimum is therefore found to be close to the pK a of ascorbic acid, 4.2. [45] The maximum TON was in all cases achieved within 30 min and was limited by catalystd ecomposition, as indicated by IR spectra recorded after the reaction, which no longer showedc haracteristic iron-carbonyl bands. Also, the addition of 0.5 mmol 1 to the solution after 60 min resulted in extra hydrogen formation (Figure4B).

Time-resolved UV/Vis and luminescence spectroscopy
To confirmo ur findings from electrochemical studies on weak acid catalysis and to gain an insighti nt he elementary steps of catalysis, we studied the system by using time-resolved UV/Vis and luminescence spectroscopy.D uring these experiments, we aimed to observe the key intermediates shown in Scheme 1: the [ 3 Ru(bpy) 3 ] 2 + speciesa fter excitation, the [Ru(bpy) 3 ] + species after quenching by ascorbate, and reduced 1 after the reaction with [Ru(bpy) 3 ] + .Excitation of 50 mm Ru 2 + + in D 2 Ow ith a l = 450 nm, 2mJ, nanosecondp ump pulse gave rise to the characteristic triplets pecies [ 3 Ru(bpy) 3 ] 2 + ,w hich decayedw ith at ime constant of 677 ns. The same experiment in the presence of the SED ascorbate (0.1 m,p D4.5) showedt he characteristic [Ru(bpy) 3 ] + species (l = 510 nm) and yieldedaquenching rate of k Q = 1.03 10 7 m À1 s À1 ,w hichw as identicalt op reviously reported rates. [46] Conducting the same measurement in the presence of 1 in vesicles showedn oa ppreciable difference in the spectra or in the associated decay curves, whereas a faster [Ru(bpy) 3 ] + decay was expected if electron transfer from [Ru(bpy) 3 ] + to 1 were to occur in appreciable quantities.
To detectr educed 1 after excitation of the Ru 2 + + PS, time-resolved IR spectroscopy experiments were performed at the same excitation wavelength (450 nm). The characteristic spectra andk inetics of the PS could againb eo bserved, this time in the CN stretching region between ñ = 1400 and 1500cm À1 , and matched with data reported in the literature. [47] No change in the CO stretching region at ñ % 2000 cm À1 was detected for the sample containing 1 in vesicles. This indicates that no sig-nificante lectron transfer occurs in the time window of these experiments (4 ms), most probably owing to slow diffusion of the photochemically generated reductant, [Ru(bpy) 3 ] + ,t ot he vesicles.
To circumvent this diffusion limitation, and with the aim of probingt he reducedf orms of 1 after excitation of the sample, we synthesized an amphiphilic analogue of the PS Ru amph 2 + + (as [Ru(bpy) 2 {5,5'-(CONHC 12 H 25 ) 2 bpy}]Cl 2 ;F igure 1) that could be incorporated into vesicles by supramolecular self-assembly,a s shownb yK çnig et al. [28] Because the UV/Vis absorption spectrum of this Ru 2 + + complexs hifts upon changing the polarity of its chemical environment, [48] we could track the uptake of Ru amph 2 + + by the vesicles (up to 0.9mm PC to 0.1 mm Ru amph 2 + + ). The D H at maximum intensity of the PC vesiclesw as 103 nm, which increased to 112nmu pon addition of the PS. This increasec annotb ep urely attributed to av esicle size increase, but must be accompaniedb yac hange in the hydrodynamicity of the vesicles upon incorporation of Ru amph 2 + + .T he combined UV/Vis and DLS resultsc onfirm that the amphiphilic ruthenium PS, Ru amph 2 + + ,s elf-assembles with the PC vesicles upon simple mixing( Figure 5).
The photophysical properties of the ascorbate-Ru amph 2 + + -vesicle-1 system were studied by means of time-resolved luminescence and UV/Vis experiments,b oth in H 2 Oa nd D 2 Oa t pH/pD 4.35, containing 0.1 mm Ru amph 2 + + .U pon excitation of Ru amph 2 + + with a1 .0 mJ, l = 485 nm, nanosecond pulse, the decay of 3 Ru amph 2 + + and the formation of Ru amph + + are followed over time ( Table 1).
The triplet lifetime, t T ,of 3 Ru amph 2 + + was determined by fitting the luminescence decay with am onoexponential function (Figure 6a nd Figures S6-S9 and S12-S15 in the Supporting Information). The phosphorescenced ecay rates, k d = 1/t T ,o f 3 Ru amph 2 + + were measured in neat (heavy)w ater and are summarized in Ta ble 1. It can be seen that the decay is faster in H 2 Ot han that in D 2 O, as is the case for the parent complex Ru 2 + + . [49] The presenceo fv esicles slightly increases the triplet  lifetime, as observed previously for similarr uthenium complexes. [50] The reductive quenching rate, k Q = (1/t T Àk d )/[ascorbate],w as calculated from the triplet lifetime in the presence of ascorbate by using the phosphorescenced ecay rates determined in the absence of quencher.T he datas ummarized in Ta ble 1s how the effect of the hydrogen isotope and the presence of PC vesicles on the quenching rate.
The presence of 1 does not seem to affect the quenching rate of 3 Ru amph 2 + + ,w hich indicates that oxidative quenching of 3 Ru amph 2 + + by 1 does not occur to as ignificant extent.T his is in line with the drivingf orce for various pathways and prior studies on the ruthenium-ascorbate couple. [51] After reductive quenching, the one-electron reduced form of Ru amph 2 + + , Ru amph + ,i sp resent in solution, which can result in charge recombination with the oxidized form, Asc + ,o fa scorbate or reduce precatalyst 1. [46] This first process is back-electron transfer with bimolecular rate constant k b .B ecauset hese primary electron-transfer eventsh appen after the reductive quenching of the Ru amph 2 + + triplet state;t his chemistry can be followed by probing the decay of Ru amph + through time-resolved UV/Vis spectroscopy.B ecause [Ru amph + ] % [Asc + ], the value for k b of all samples in which Ru amph + + is generated can be estimated with an umerical fit (see Figures S10, S11, S16, and S17 in the Supporting Information).
Ta ble 2s hows the estimation of k b .F or samples in which 1 is absent,t he decay directly translates to the back-electron transfer k b rate constant.F or samples in which 1 is present,t he decay is ar esult of k b and the electron-transfer rate to 1,a nd indeed the decay is faster in the presence of 1.T his observation, in combination with the fact that k Q does not change in the presence of 1 ( Table 1), indicates that the increased decay is due to an additional decay pathway for Ru amph + through electron transfer to 1.A lthough phosphorescence decay and the quenching rate constants differ quite substantially between H 2 Oa nd D 2 O, the measured k b values do not. Moreover, the differenceb etween k b in the presence and absence of 1 is almosti dentical for H 2 Oa nd D 2 O; therefore, the decay pathways of Ru amph + most probably do not involve the transfer of protons. To determine the rate constant, k ET ,f or electron transfer from Ru amph + to 1,w eassume that the differencei nt he measuredd ecay values in the absence and presence of 1 is directly related to the electron-transfer rate via [Asc + ]( 2.13 10 10 Àk b ) = [1]k ET .W ith [Asc + ]b eing approximately 1 mm during the monoanion decay,t he electron-transfer rate [1]k ET is in the order of 10 4 s À1 .T he concentration of [1]i nt he PC vesicles is approximately 70 mm (a concentration increaseo fo ver 1000 times;s ee the Supporting Information for more details), and a lower limit to the bimolecular rate constant, k ET ,i nt he order of 10 5 m s À1 was determined.

Time-resolved IR spectroscopy and the overall photocatalytic cycle
To probe the formation of 1 À À and/ors pecies formed from 1 À through follow-up reactivity,t ime-resolved IR spectroscopy was performed on asample containing 0.1 mm 1 in PC vesicles and 0.1 mm Ru amph 2 + + in ascorbate buffer in D 2 O ( Figure 7). To avoid catalyst decompositionf rom the excitation light, the sample was pumpedt hrough the cell with as yringe pump. In contrastt oe xperiments in which Ru 2 + + was used, in this case, we did see ad epletion of the lowest energy CO vibration band of 1 at ñ = 2078 cm À1 and growth of ab road band between ñ = 1930 and 2000 cm À1 ( Figure 7A). Because depletion of the bands at ñ = 2043 and 2004 cm À1 belonging to 1 was not observed, the new species formed most likely also contained these bands in its spectrum. Comparing the as-formed species( ñ = 2043, 2000-1930 cm À1 )w ith reported reduced and protonated species of 1 gave au nique match with the direduced, monoprotonated species 1H À ,w hich has ar eported spectrum of ñ = 2045, 1996,1979,1963, and 1935 cm À1 (Figure 7B). [52] This 1H À species, [(m,k 2 -bdt)(m-H)(m-CO)Fe 2 (CO) 5 ] À ,i s known for being as table intermediate during weak acid  [49] k d = 1.05 10 6 s [49] k Q = 0.10 10 8 m s À1 [46] k Q = 0.10 10 8 m s À1   proton reduction catalysis with 1. [36] At high initial concentrations of 1,t he formation of this bridging hydride speciesi s thought to occur through as equence of monoreduction,d isproportionation of two monoanionst og ive 1 and 1 2À ,a nd subsequentp rotonation of 1 2À to form 1H À .B ecause the timeresolved IR data at delay times shorter than 250 ns are masked by shock waves, we cannote lucidate the chemistry beforet he formation of 1H À .H owever,t he time-resolved IR results are in line with the observed electrochemical formation of 1H À in PC vesicles, and thus, we propose the photocatalytic cycle depicted in Scheme 2.
The catalytic steps observed herein are different from those observed by Lomotha nd co-workersi nadifferent Fe 2 S 2 system in organic solvent (acetonitrile), with which 1 is oneelectron reduced by photogenerated [Ru(dmb) 3 ] + (dmb = 4,4'dimethyl-2,2'-bipyridine) to the monoanion 1 À .I nc ontrast to our findings, this species [Fe 2 (bdt)(CO) 6 ] À does not disproportionate and persists on the timescale of seconds to reactw ith as trong acid to form 1Hi nstead. [34] Conclusion As elf-assembled system in which ar uthenium PS and ad iironbased proton reduction catalysta re preorganized in vesicles was studied in detail to elucidate preorganization effectsint he photocatalytic formation of hydrogen by [FeFe]hydrogenase mimics. Electrochemical experiments indicate that the behavior of 1 inside vesicles is similar to that of 1 in organic solvents in the presence of weak acids. Upon irradiation, the PS (Ru 2 + + )i s excited, reductivelyq uenched by ascorbate, and an electron is transferred to 1 to initiate the hydrogen-evolution cycle. The effects of the vesicle matrix around 1 during catalysis are twofold:1 )weh ypothesize that the increased local concentration of 1 leads to faster disproportionation of 1 À to 1 2À ,a nd 2) the constant neutral pH provided by the vesicles prevents protonation of 1 À .T his meanst hat preorganization of the molecular components in vesicles controls the reaction pathway by which the catalystsystem photogenerates hydrogen.
Photochemical hydrogen formation at pH 4.5 gives 67 turnovers,a nd is limited by photodecomposition of the catalyst. The photostability of the hexacarbonyls (and thereby,T ON) can possibly be improved upon by substitution of ac arbonyl for ap hosphorous ligand, [53] but at the expense of amore negative reduction potential. Photodecomposition can more easily be circumvented by choosing PSs that operate at wavelengths at which the catalytic (resting) species is transparent. As such, the use of [Ru(bpy) 3 ] 2 + with hydrogenase mimics seems to be inappropriate if high TONs are required.

Experimental Section General
All syntheses were carried out under an itrogen atmosphere by using standard Schlenk techniques. All purifications involving column chromatography were performed in air with non-degassed solvents. All commercially available chemicals were used as received. l-a-Phosphatidylcholine (PC;f rom egg yolk, type XVI-E, ! 99 %, lyophilized powder) was obtained from Sigma-Aldrich and stored at À20 8C. Compounds 1 and Ru amph 2 + + were prepared through procedures reported in the literature. [28,54,55] Solutions of PC vesicles were prepared freshly each day.A scorbate buffer solution was prepared freshly each week and stored at 5 8C; the correct pH/pD was set by mixing of a0 .1 m solution of sodium ascorbate with a0 .1 m solution of ascorbic acid, in which pD was measured with ac onventional pH meter by using pD = pH* + 0.4, in which pH* is the observed pH value. For the preparation of vesicles, aV ibracell VCX 500 probe-tip sonicator was used.

Preparation of PC vesicles
For a1mL solution:T oafinger flask was added PC (5 mg), ethanol (1 mL), or as tock solution of 1 in ethanol (depending on the sample). All solvent was removed on ar otary evaporator until a film was observed. The film was further dried under vacuum and subsequently hydrated by adding buffer solution (1 mL;p hosphate buffer or ascorbate buffer) of the desired pH by using av ortex mixer at room temperature. The suspension was transferred to an Eppendorf tube and sonicated by using ap robe-tip sonicator for 1 to 2min, at 10 so n/5 so ff intervals, until the suspension was clear to the eye.

Steady-state spectroscopy
The 1 HNMR spectra were measured on aB ruker AV400 spectrometer.F TIR measurements were conducted on aB ruker ALPHA FTIR spectrometer.U V/Vis measurements were conducted on aH PAgilent 8453 UV/Vis spectrometer.

Electrochemistry
Cyclic voltammograms and differential pulse voltammograms were performed by using aM etrohm/Autolab PGSTAT128N instrument. The working electrode was a2mm diameter glassy carbon disk and ap latinum wire counter electrode. The reference electrode (Ag/AgCl) was calibrated against the ferrocyanide couple to obtain potentials versus NHE (see the Supporting Information). Hardware iR compensation was employed for all CV measurements. DPV was performed by using as tep potential of 5mV, am odulation potential of 25 mV,am odulation time of 50 ms, and an interval time of 500 ms. Half-wave potentials were determined from the peak potentials by addition of 12.5 mV (half the modulation potential).

Photocatalysis
Photocatalysis was performed on as olution (5 mL) in ac ustombuilt setup, in which the cell headspace (ca. 200 mL volume) was continuously pumped through the sampling valve (25 mLs ampling volume) of aG lobal Analyzer Solutions CompactGC 3.0 gas chromatograph and sampled every 5min. Irradiation was performed with eight LEDs (l = 450 nm;4 .54 Wt otal power) mounted on aircooled heat sinks.

Dynamiclight scattering (DLS)
The DLS setup was based on an ALVD LS 5000 goniometer with a digital correlator and a l = 633 nm HeNe laser (35 mW) to minimize fluorescence. At ypical DLS run was 120 sl ong and measurements took place at 20 8C. Scattered photons reaching the two photodetectors were cross-correlated to give one intensity correlation function per measurement. The single-angle DLS measurements were made at 908.T he multiangle DLS measurements were conducted at 60/70/80/90/95/100/105/110/115/1208 and fitted to as et of weighed exponentials by using an onlinear least-squares algorithm.

Time-resolved luminescence and UV/Vis spectroscopy
In this transient spectroscopy setup, an Ekspla NT342B Nd:YAG laser was used for the generation of the pump light pulse. The probe light was generated by an Excelitas Te chnologies FX-1160 high-stability short-arc xenon flash lamp, the pulses of which were timed by using am odified PS302 controller from EG&G. The spectrograph used was aP rinceton Instruments SpectraPro-150 instrument. The reference and signal beam were recorded by using a gated, intensified Princeton Instruments PI-MAX3 charge-coupled device (CCD) camera. The timing of the excitation pulse, the flash lamp, and the gate of the camera was achieved with aStanford Research Systems DSG535 delay generator.T he samples were measured in as eptum-capped 1cmq uartz cuvette under continuous agitation by am agnetic stirrer bar.E xcitation of 50 mm Ru 2 + + in D 2 O with a l = 450 nm, 2mJ, nanosecond pump pulse gave rise to the characteristic triplet species, [ 3 Ru(bpy) 3 ] 2 + (l = 370/565/820 nm), which decayed with at ime constant of 677 ns. Excitation of the samples with a1 .0 mJ, l = 485 nm, nanosecond pulse gave a 3 Ru amph 2 + + concentration between 1a nd 3 mm (depending on the sample) at t = 0. For all samples, both luminescence and absorption spectra were recorded at time delays up to 4( luminescence) or 40 ms( absorption). Kinetic analysis was performed by global fitting of the spectral data with as et of equations derived from the system of chemical reactions or by numerically solving of the system of ordinary differential equations inside the target function for the nonlinear least-squares routine (see the Supporting Information for af ull explanation and derivation of formulae). As such, all (e.g.,p re-exponential) constants had physical meaning, and the data were fitted with ap hysical description of the system (see below). Time-resolved absorption spectra can be found in Figures S6-S19 in the Supporting Information. The contents of the samples are summarized in Ta ble 3.

Time-resolved IR spectroscopy
Ac ommercial Spectra-Physics OPA-800C BBO-based optical parametric amplifier (OPA) was pumped by aS pectra-Physics Hurricane Ti :sapphire laser (l = 800 nm;4 80 mJ) with ar epetition rate of 1kHz. IR probe pulses were generated by ad ifference-frequency mixing signal and idler from the OPAi naAgGaS 2 crystal. The nanosecond visible pump pulses (l = 475 nm;2 5mJ) were generated in aG WU versaScan-L BBO-based optical parametric oscillator (OPO) pumped by aS pectra-Physics Quanta-Ray INDI Nd:YAG laser with ar epetition rate of 20 Hz. The sample cell with CaF 2 windows spaced by 250 mmw as placed in the IR focus and the sample was pumped through the cell at af low rate of 10 mLmin À1 by using a syringe pump. Ac ustom-built 30 pixel HgCdTe( MCT) detector coupled to an Oriel MS260i spectrograph was employed to record the transient spectra by subtracting nonpumped absorption spectra from the pumped absorption spectra. Background correction was performed by subtracting the time-averaged spectra obtained at negative time delays.

UV/Vis study on the incorporation of Ru amph 2 + + intov esicles
Because the UV/Vis absorption spectrum of this complex was sensitive to the polarity of its chemical environment, [48] we tracked the UV/Vis spectral changes upon the addition of vesicles (up to 0.9 mm PC)t oan aqueous solution of 0.1 mm Ru amph 2 + + .( Ap roper supramolecular titration was not possible because the absolute absorption values were irreproducible due to light scattering from the vesicles.) The peak position of the MLCT transition at l % 480 nm was used as an indicator for the binding of Ru amph 2 + + to the vesicles (see the Supporting Information, ESI 2). [56] Concentration of compounds in/on liposomes The concentration of PC in vesicles was [PC] = 1000 mL L À1 / (768 gmol À1 0.99 mL g À1 ) = 1.3 m. [57] The concentration of 1 in PC

Chemical models and kinetic equations
Samples containing only Ru 2 + + :F or the samples only containing the PS, there were only two chemical species present, namely,t he ground state (G) and the excited state (E). From t = 0o nward, the reactivity was E!Gw ith decay constant k d .
The rate equations and boundary conditions are given in Ta ble 4, and the resulting concentration time dependence is given in Equations (1) and (2).
E ! G, with decay constant k d ð3Þ Because [Asc] @ [Ru 2 + + ], we could assume that [Asc] = Qw as constant over time. Moreover,b ecause Asc + was generated and consumed stoichiometrically with M, we could equate [Asc + ](t) = [M](t). The rate equations and boundary conditions are given in Ta ble 5.
The resulting concentration time-dependence could only be expressed analytically for E(t)[Eq. (6).
However,g iven values for the constants k Q Q, k d and k b ,a nd setting M 0 = 0, the function M(t)c ould be evaluated numerically,f or which we used the Matlab function ode45. After evaluation of M(t), the ground-state function, G(t), could be evaluated through G(t) = ÀE(t)ÀM(t).
Fittingo ftime-resolved luminescence and UV/Vis data All acquired time-resolved data was fitted in Matlab by using the nonlinear least-squares function lsqcurvefit in the physical model outlined above. The experimental time-resolved data matrix A exp (l;t)w as fitted to A fit (l;t), which was al inear combination of time-dependent species spectra as given by Equation (8).
A fit ðl;tÞ¼e G ðlÞGðtÞþe E ðlÞEðtÞþe M ðlÞMðtÞð 8Þ The spectra e species (l)a nd constants k Q Q, k d ,a nd k b were determined by the fitting procedure, if they had not already been determined in previous experiments. This yielded an approach in which the set of experiments were designed in such aw ay that every experiment generated as et of spectra and rate constants that could be used as fixed values in the next experiment, to provide minimum freedom during the fitting procedure, and thereby,maximum accuracy in the determination of kinetic rate constants. To minimize the amount of parameters in the fitting procedure even further,w eo nly determined difference spectra with respect to Gb ecause G(t) = ÀE(t)ÀM(t)[ Eq. (9)].
A fit ðl;tÞ¼½e E ðlÞÀe G ðlÞEðtÞþ½e M ðlÞÀe G ðlÞMðtÞð 9Þ This redundancy of e G (l)w as general and held for all analyzed chemical systems. The values of the obtained rate constants can be found in Ta bles 1 and 2. The Supporting Information contains cyclic voltammograms and experimental, fitted, and error plots for all time-resolved spectroscopy experiments. Table 4. As ummary of the rate equationsa nd boundary conditions of samples containing only Ru 2 + + .

SpeciesRate equation
Boundary condition