Molecular Dye-Sensitized Photocatalysis with Metal-Organic Framework and Metal Oxide Colloids for Fuel Production

Colloidal dye-sensitized photocatalysis is a promising route toward efficient solar fuel production by merging properties of catalysis, support, light absorption, and electron mediation in one. Metal-organic frameworks (MOFs) are host materials with modular building principles allowing scaffold property tailoring. Herein, we combine these two fields and compare porous Zr-based MOFs UiO-66-NH2(Zr) and UiO-66(Zr) to monoclinic ZrO2 as model colloid hosts with co-immobilized molecular carbon dioxide reduction photocatalyst fac-ReBr(CO)3(4,4′-dcbpy) (dcbpy = dicarboxy-2,2′-bipyridine) and photosensitizer Ru(bpy)2(5,5′-dcbpy)Cl2 (bpy = 2,2′-bipyridine). These host-guest systems demonstrate selective CO2-to-CO reduction in acetonitrile in presence of an electron donor under visible light irradiation, with turnover numbers (TONs) increasing from ZrO2, to UiO-66, and to UiO-66-NH2 in turn. This is attributed to MOF hosts facilitating electron hopping and enhanced CO2 uptake due to their innate porosity. Both of these phenomena are pronounced for UiO-66-NH2(Zr), yielding TONs of 450 which are 2.5 times higher than under MOF-free homogeneous conditions, highlighting synergistic effects between supramolecular photosystem components in dye-sensitized MOFs.


Introduction
Solar fuel production has emerged as a pertinent possibility on route to address growing energy challenges and shift fossil fuel dependency toward sustainable sources [1]. It aptly merges solar energy harvesting with subsequent energy conversion to form valueadding products [2]. At this cross-section, molecular coordination complexes can play a key role as discrete nature-mimicking photosystems combining light sensitizing and chemical reactivity [3]. Previous advances include molecular catalyst and dye engineering to improve catalytic parameters such as activity and selectivity on the one hand, as well as light harvesting and antenna effects on the other [4,5]. Additionally, immobilizing such species to host materials can yield beneficial sustained performance, with dye-sensitized photocatalysis (DSP) emerging as a selected approach with distinct advantages [6,7]. Specifically, this methodology bridges the fields of molecular (photo)catalysis and material chemistry by coupling a catalyst and a dye via co-anchoring onto a semiconductor particle-effectively employing the latter as a solid-state electron mediator and a scaffold [6]. While this working principle has been studied for several host particle compounds in colloidal DSP [8,9], applying metal-organic frameworks (MOFs) as the matrix component can provide similar benefits and is an area with burgeoning interest [10][11][12][13]. MOFs combine metal-based nodes with organic multitopic linkers to form porous coordination polymers, thus unlocking adjustable chemical properties, topologies, porosities, and molecular complex hosting capabilities [14]. Particularly regarding dye-sensitization possibilities, MOFs can substantially influence guest photophysics through approaches such as confinement effects as well as scaffold-directed photochromicity [15]. However, their organic/inorganic building principle can result in limited stability or conductivity [16]. This presents the question how UiO-66 is well-studied due to its versatility and chemical stability [18]. Additionally, a few reports showed UiO-66(Zr)'s applicability in DSM using Pt nanoparticles for H2 evolution [19,20]. Functional group implementation through linker variation offers further advantages [18]. For example, UiO-66-NH2(Zr), constructed from 2-aminoterephthalic acid linkers, allows improved gas and molecular guest adsorption capabilities, as well as shifts the optical absorption through amine group incorporation [10,21]. Due to its small pore apertures and diameters, molecular species are typically loaded onto the MOFs' surface [10,22]. ZrO2 has also been previously applied as a support for DSP primarily enabling so-called "on particle" electronic communication between PS and catalyst due to the high energy level of the conduction band [3,23].
UiO-66 is well-studied due to its versatility and chemical stability [18]. Additionally, a few reports showed UiO-66(Zr)'s applicability in DSM using Pt nanoparticles for H 2 evolution [19,20]. Functional group implementation through linker variation offers further advantages [18]. For example, UiO-66-NH 2 (Zr), constructed from 2-aminoterephthalic acid linkers, allows improved gas and molecular guest adsorption capabilities, as well as shifts the optical absorption through amine group incorporation [10,21]. Due to its small pore apertures and diameters, molecular species are typically loaded onto the MOFs' surface [10,22]. ZrO 2 has also been previously applied as a support for DSP primarily enabling so-called "on particle" electronic communication between PS and catalyst due to the high energy level of the conduction band [3,23].
Ligands with carboxylic acid groups were chosen to enable outer surface anchoring to MOF nodes, free amines for UiO-66-NH 2 (Zr), as well as to ZrO 2 surface hydroxyl groups ( Figure 1b) [11,12,30]. This approach has previously yielded stable host-guest colloidal assemblies, allowing for photoinduced electron transfers from the SED to the sensitizing units and subsequently to vicinal catalysts for CO 2 reduction [10][11][12][13]30]. Herein, we report the synthesis, thorough characterization, and photocatalytic CO 2 reduction performance of DSM UiO-66-NH 2 (Zr) and UiO-66(Zr), in direct comparison to DSP with their metal oxide counterpart ZrO 2 . This revealed key working principles and considerations for selecting suitable hosts in solar DSP and DSM.

Materials and Methods
A detailed overview on analysis methods, step-by-step synthesis routes, and characterization techniques is available to the reader in the Supplementary Material (SM). The most relevant material syntheses are described here.

Molecular Complex Immobilization
In a typical experiment, a 0.09 mM solution of 1 in acetonitrile (MeCN) and a 0.05 mM solution of 2 in MeCN (quantities for both varied according to the desired ratio, see SI, Table S1) were added simultaneously to the respective pristine host powder (10.0 mg). The supernatant was removed after 24 h, and the resulting powder was washed with pure MeCN (3 × 20 mL). Then the powder was dried at 80 • C overnight.

Host-Guest Photosystem Assembly
Following the pristine host characterization, we sensitized the materials with photocatalyst 1 and dye 2. This was achieved by soaking from a corresponding MeCN solution. The respective pristine host (10.0 mg in all cases) was immersed in a solution of 1 (and 2) in MeCN (quantities in Table S1) for 24 h, then washed with pure MeCN and dried. Molecular species are expected to load specifically at the surface for all materials, as the MOF pore sizes, channels, and windows (8.0 Å) are substantially smaller than both molecular diameters (1: 12.0 Å; 2: 14.5 Å) [10]. Surface anchoring was monitored via supernatant UV-Vis  Figure S7). While 66-NH 2 and 66 provided comparable spectra, ZrO 2 samples illustrated a smaller absorption decrease, suggesting a lower loading per mass of host. Further, measuring the washing solution after 8 h revealed no absorption in all cases, indicating no leaching and stable complex anchoring to the host surfaces ( Figure S7).
Average maximum molecular surface loadings of~58,~58, and~32 nmol mg −1 were obtained for 66-NH 2 , 66 and ZrO 2 -based assemblies, respectively, from inductively coupled plasma mass spectrometry (ICP-MS) measurements for Ru and Re content (see SM). The lower loading for ZrO 2 -based assembly is in-line with the reduced absorption decay in the sensitizing solution compared to MOFs ( Figure S7) and is rationalized by the higher crystal density of monoclinic zirconium oxide compared to porous UiO samples. Calculating surface areas and maximum coverages from SEM particle sizes (Figure 2b,c) with DFToptimized 1 and 2 gave values close to experimental ICP-MS results (SM, pages S4-S5), supporting that maximum surface coverage is attained by immobilization from solution. Exact metal contents further allowed molecular loading and photosensitizer to catalyst ratio calculations (R, Equation (1), Tables S1-S4). For each host in turn, assemblies with an R value of 0.5 and 2.1 were synthesized (Tables S1-S4), i.e., a sample with double the molar catalyst quantity compared to the dye, and vice versa. R = n(2)/n(1) = n(Ru per mg host)/n(Re per mg host) (1) 1-loaded, as well as 1-and 2-loaded assemblies, denoted as Re-Host and ReRu-host, respectively, were analyzed by PXRD, dynamic light scattering (DLS), attenuated total reflectance infrared (ATR-IR) spectroscopy, and solid-state UV-Vis spectroscopy to further assess the impact of molecular anchoring.
PXRD data of the functionalized hosts (R = 0.5) showed matching reflex positions as pristine hosts, suggesting retained sample crystallinity after the anchoring process ( Figure 2a). Importantly, DLS distributions revealed comparable hydrodynamic diameters of pristine and loaded hosts (~158,~201, and~284 nm for 66-NH 2 , 66 and ZrO 2 -based samples), highlighting that MOFs and nanoparticles do not agglomerate in MeCN solution and during molecular complex immobilization ( Figure S8). The received hydrodynamic diameters are larger than actual particle diameters determined by SEM (Figure 2b,c) due to the hydration shell additionally detected in DLS measurements.
Solid-state UV-Vis spectroscopy of complex-containing samples displayed additional bands between 250 and 600 nm not present in pristine host spectra, matching 1 and 2 ( Figure 3a). Further, ATR-IR spectra for all 1-loaded assemblies showed additional bands at 1917 and 2025 cm −1 , which are characteristic of the Re(CO) 3 moiety (Figure 3b). These support retained molecular catalyst integrity upon anchoring [11,44].
Finally, N 2 gas adsorption experiments were conducted to evaluate the porosity of the dye-sensitized MOF samples ( Figure S2). ReRu-66-NH 2 (R 2.1) and ReRu-66(R 2.1) demonstrated a significant reduction in uptake in respect to their pristine counterparts, however, with a non-zero BET area of 338 ± 1 and 360 ± 1 m 2 g −1 , respectively. This is in-line with previous studies, which showed that surface-anchored molecules of this size partially block outer MOF pores [10]. As for ZrO 2 , ReRu-ZrO 2 (R 2.1) illustrated negligible nitrogen adsorption.
BIH has been shown to give increased catalytic activity and stability compared to TEOA in homogeneous and heterogeneous photosystems [11,38,45]. The resulting sus- pension was first saturated with CO 2 ([CO 2 ] ≈ 0.28 mol L −1 for MeCN) [46], and then vigorously stirred under irradiation at 450 nm (Blue LED LXZ1 PR01 at 5.1 W). The potential evolution of H 2 and CO was analyzed via micro gas chromatography (GC) via reaction headspace sampling.
As a benchmark, homogeneous samples with pure catalyst, as well as dye and catalyst (ratio 2.0), were examined with both SEDs. For TEOA-based experiments, a solution of 0.5 µmol pure 1 showed selective CO formation under irradiation that plateaued after 15 min reaching 11.2 ± 0.3 TONs per Re catalyst (Table S5, TON calculation in SM). Additionally, 0.5 µmol 1 with 1.0 µmol of 2 resulted in a longer activity of 1.5 h before deactivation, while affording similar TONs as the PS-free experiments (Table S5). A comparable trend with higher absolute performances was observed with BIH as the SED, with 0.5 µmol pure 1 reaching TONs of 173 ± 7 over 3 h, and 0.5 µmol 1 with 1.0 µmol 2 yielding 182 ± 15 over 10 h. Next, 1-loaded colloids were investigated. All three host materials showed limited activity with TONs below 5 and 100 for TEAO and BIH, respectively (Table S5). Subsequently, ReRu-host hybrid systems were examined (Figure 4a,b, Table S5). In all cases, the detected activity increased from ReRu-ZrO 2 , to ReRu-66, and finally to ReRu-66-NH 2 samples. When using BIH all host-guest systems reached significantly higher TONs than under homogeneous conditions whereas TEOA afforded similar or lower activities. Additionally, higher TONs compared to R = 0.5 samples were obtained with R = 2.1 with the highest activity (TON = 450 ± 15; BIH) reached for ReRu-66-NH 2 (R = 2.1).
BIH has been shown to give increased catalytic activity and stability compared to TEOA in homogeneous and heterogeneous photosystems [11,38,45]. The resulting suspension was first saturated with CO2 ([CO2] ≈ 0.28 mol L −1 for MeCN) [46], and then vigorously stirred under irradiation at 450 nm (Blue LED LXZ1 PR01 at 5.1 W). The potential evolution of H2 and CO was analyzed via micro gas chromatography (GC) via reaction headspace sampling.
As a benchmark, homogeneous samples with pure catalyst, as well as dye and catalyst (ratio 2.0), were examined with both SEDs. For TEOA-based experiments, a solution of 0.5 µmol pure 1 showed selective CO formation under irradiation that plateaued after 15 min reaching 11.2 ± 0.3 TONs per Re catalyst (Table S5, TON calculation in SM). Additionally, 0.5 µmol 1 with 1.0 µmol of 2 resulted in a longer activity of 1.5 h before deactivation, while affording similar TONs as the PS-free experiments (Table S5). A comparable trend with higher absolute performances was observed with BIH as the SED, with 0.5 µmol pure 1 reaching TONs of 173 ± 7 over 3 h, and 0.5 µmol 1 with 1.0 µmol 2 yielding 182 ± 15 over 10 h. Next, 1-loaded colloids were investigated. All three host materials showed limited activity with TONs below 5 and 100 for TEAO and BIH, respectively (Table S5). Subsequently, ReRu-host hybrid systems were examined ( Figure  4a,b, Table S5). In all cases, the detected activity increased from ReRu-ZrO2, to ReRu-66, and finally to ReRu-66-NH2 samples. When using BIH all host-guest systems reached significantly higher TONs than under homogeneous conditions whereas TEOA afforded similar or lower activities. Additionally, higher TONs compared to R = 0.5 samples were obtained with R = 2.1 with the highest activity (TON = 450 ± 15; BIH) reached for ReRu-66-NH2(R = 2.1). Control experiments performed without irradiation or a SED yielded no detectable product formation under homogeneous conditions as well as colloidal photocatalysis (Table S5). Furthermore, CO2 was confirmed as the sole source of CO as 13 C-labelled CO2 Control experiments performed without irradiation or a SED yielded no detectable product formation under homogeneous conditions as well as colloidal photocatalysis (Table S5). Furthermore, CO 2 was confirmed as the sole source of CO as 13 C-labelled CO 2 produced only 13 CO ( Figure S9). Both experiments highlight the central role of each photosystem constituent in selective CO 2 photoreduction.

Discussion
The obtained homogeneous photocatalysis data matches previous studies with pure 1 showing modest activity and fast deactivation with TEOA, illustrating its instability under photocatalytic conditions [47]. Upon adding 2, the apparent CO evolution rate decreases, while not yielding effective electron transfers to the catalyst, apparent from unchanged TONs. This is ascribed to a too large mean distance between the photosystem components, as shown previously where higher dye concentrations or multinuclear complexes were required [4]. BIH was the more effective electron source as it quenches the excited state 3 2* more efficiently and detrimental TEOA radicals limit molecular dye/catalyst stability [11,45,47].
Comparing homogeneous results with 1 to heterogeneous 1-loaded colloids reveals a drop in performance for the latter independent of the SED. Similar performance decreases have been previously observed upon immobilizing fac-[ReCl(CO) 3 (bpy)]-derivatives on SiO 2 or TiO 2 for photocatalysis and were assigned to reaction environment change and mass transport limitations [29,48]. This motivates dye-sensitization to drive efficient electron transfers to the catalyst units.
Upon providing surface proximity for 1 and 2, exergonic electron transfers from the photosensitizer to the catalyst can occur (relying on statistical molecular distribution), allowing rapid CO 2 reduction and improved performance [10,34]. This working principle highlights the advantages of using dye-sensitized materials, as pure host-photocatalyst assemblies result in activity decreases. For all ReRu-host systems, an increased dye-to-catalyst ratio R yielded higher TONs as excess dye increases the probability of each CO 2 reduction catalyst being in proximity to a photosensitizer [11]. Distinct differences were recorded between the three anchoring materials, with the performance increasing from ReRu-ZrO 2 (TON~331, R = 2.1, BIH), to ReRu-66 (~395), to ReRu-66-NH 2 (~450) (Figure 4a,b). As all pristine host materials are photosilent at the irradiation wavelength of 450 nm, exhibit similar particle size, and bear similar dye-to-catalyst ratios, this activity trend does not stem from a discrepancy between direct host light absorption (Figures 2c, 3a, 4a,b and S6).
To gain more insights into the difference in the catalytic performance, we examined the reduction and oxidation potentials of the ground and excited molecule states, as well as the Zr-based materials' conduction band energy levels. Reductive quenching, i.e., photoreduction of the PS from the SED followed by electron transfer to the CRC, is the main electron pathway in DSP systems with non-conducting hosts. As the scaffolds' conduction band (CB) edges were previously evaluated at −3.3 eV (66-NH 2 , −0.9 V SCE ) [49], −3.1 eV (66, −1.1 V SCE ) [50], and −2.1 eV (ZrO 2 , −2.1 V SCE ) [51], the triplet state 3 2* (−3.1 eV, −1.1 V SCE ) is potentially able to inject electrons into the CB of 66-NH 2 through oxidative quenching, followed by regeneration from the SED. This could enable the MOF to act as a solid-state electron mediator, allowing electron hopping to distant catalyst units as E(1/1 − ) = −3.3 eV (−0.9 V SCE , Figure 5a,c). Electron hopping in MOFs is a topic of ongoing interest featuring reports of tailor-made MOFs with intrinsic conduction mechanisms and photoinduced electron injection by a Ru dye into a UiO material [52][53][54]. Although less thermodynamically favored, a similar electron cascade should also be possible with 66based DSM especially as Ru dyes have been reported to perform electron injection from unrelaxed states [55]. The difference in band positions between 66-NH 2 and 66 has been assigned to the amine-functionalized linkers, effecting a change of the sp 2 bonding in the aromatic carbon ring [56]. In contrast, oxidative quenching is prevented for ZrO 2 -based DSP due to the high CB energy value (Figure 5a). This implies that the reductive quenching cascade is the sole viable electron cascade for catalysis, potentially limiting the final activity of the system in comparison to MOF-based systems [23]. S2). To determine implications for photocatalysis, CO2 gas adsorption experiments at room temperature were conducted with the three hosts (Figure 5b). As a non-porous crystalline solid, ReRu-ZrO2 is not able to adsorb carbon dioxide within its structure and is therefore reliant on the diffusion of available CO2 in solution. This differs for MOFs, as their intrinsic porosity could allow gas uptake when the reaction solution is saturated with CO2 prior to irradiation. Noteworthy is that ReRu-66-NH2(R 2.1) shows markedly higher uptake than ReRu-66(R 2.1), despite comparable N2 adsorption isotherms (Figures 5b and  S2). This has been previously observed and attributed to interactions/reactions between the polar linkers' amine groups and carbon dioxide (Figure 5c) [57,58]. Thus, 66-NH2 provides an increased local CO2 concentration available during the course of photocatalysis, thereby reducing diffusion limitations, as well as potentially (c) Schematic representation of dye-and catalyst-sensitized 66-NH 2 under photocatalytic conditions, yielding increased performance due to electron hopping, intrinsic porosity-based CO 2 uptake, and interactions between polar linker NH 2 groups and CO 2 .
As photocatalysis proceeds in colloidal solution, the local CO 2 concentration available for catalyst centers could impact the final activity. Despite maximum molecular loading on the scaffolds' surfaces, MOFs still provide a certain permanent porosity ( Figure S2). To determine implications for photocatalysis, CO 2 gas adsorption experiments at room temperature were conducted with the three hosts ( Figure 5b). As a non-porous crystalline solid, ReRu-ZrO 2 is not able to adsorb carbon dioxide within its structure and is therefore reliant on the diffusion of available CO 2 in solution. This differs for MOFs, as their intrinsic porosity could allow gas uptake when the reaction solution is saturated with CO 2 prior to irradiation. Noteworthy is that ReRu-66-NH 2 (R 2.1) shows markedly higher uptake than ReRu-66(R 2.1), despite comparable N 2 adsorption isotherms (Figures 5b and S2). This has been previously observed and attributed to interactions/reactions between the polar linkers' amine groups and carbon dioxide (Figure 5c) [57,58].
Thus, 66-NH 2 provides an increased local CO 2 concentration available during the course of photocatalysis, thereby reducing diffusion limitations, as well as potentially activating CO 2 [59]. This is potentially shown by increased turnover frequencies (TOFs) from ZrO 2 -based, to 66-based, to 66-NH 2 -based assemblies ( Figure S10). Such reactant-hostguest synergies are a unique benefit of DSM over conventional DSP (Figure 5c), yielding higher performances. This increase in efficiency is also apparent from obtained AQEs which improve from 2.2 ± 0.2% for ReRu-ZrO 2 (R 2.1) to 8.0 ± 0.2% for ReRu-66-NH 2 (R 2.1) under ideal conditions (Tables S6 and S7), allowing incident photons to be converted more effectively to CO. The obtained performance and TONs for ReRu-66-NH 2 are comparable to state-of-the-art dye-sensitized semiconductors with Re-based molecular catalysts. For instance, dye-sensitized TiO 2 nanoparticles with ReCl(CO) 3 (bpy)-derivatives reached maximum TONs between 165 and 570 in DMF with BIH, depending on irradiation conditions [26,27]. This system showed an AQE of 2.1% at 436 nm [27], and a recent report with organosilica nanotubes and anchored 1 and 2 yielded AQEs between 0.4% and 15.1% at 450 nm, depending on reaction conditions [60]. Accordingly, our results are well within this benchmark range for DSP with colloidal host-guest systems.

Conclusions
Herein we designed a model study to compare dye-sensitized metal oxides to metalorganic frameworks for colloidal photocatalytic CO 2 reduction. Thus, monoclinic ZrO 2 was evaluated against Zr-based MOFs 66 and 66-NH 2 , all with surface co-immobilized molecular carbon dioxide reduction photocatalyst fac-ReBr(CO) 3 (4,4 -dcbpy) and photosensitizer Ru(bpy) 2 (5,5 -dcbpy)Cl 2 . Photocatalytic CO 2 -to-CO performance increased from ZrO 2 , to UiO-66, and UiO-66-NH 2 , respectively. While the former provides fixed proximity for the molecular photosystem constituents, the studied MOF scaffolds offer additional synergistic host properties. Particularly 66-NH 2 could enable host-mediated electron hopping, paired with uniquely increased CO 2 adsorption capacity from its permanent porosity and NH 2 -CO 2 interaction. These combined phenomena enable an AQE of up to 8% for the studied DSM.
Overall, these findings highlight how DSM can combine multiple functionalities to drive photocatalysis and provide tunable material composites. Our results competitively position dye-sensitized MOFs as a hybrid material for efficient solar fuel production, which could be further exploited with alternative optimized components toward higher performance in the future.