The Emergence of 2D Building Units in Metal‐Organic Frameworks for Photocatalytic Hydrogen Evolution: A Case Study with COK‐47

Metal‐organic frameworks (MOFs) are promising materials for photocatalytic water splitting reactions, but examples of visible light‐responsive, catalytically active, and stable MOFs are still rare. A detailed investigation is conducted for COK‐47 – a recently described MOF comprising 2D Ti‐O6 secondary building units (SBUs) – toward a photocatalytic hydrogen evolution reaction (HER), showing how overall particle morphology, surface area, and missing ligand defects are central parameters governing the material's ultimate performance. The newly synthesized COK‐47ISO is among the most active MOFs to date, yielding HER‐rates of 8.6 µmol h−1, and an apparent quantum yield (AQY) of 0.5% under visible light illumination. Optoelectronic and photoluminescence investigations, supported by theoretical calculations, enable the unraveling of its electronic structure along with charge transfer and recombination kinetics. A wavelength‐dependent reaction mechanism is proposed involving ligand to metal charge transfer (LMCT) and the main challenges for visible or UV photoexcitation are identified, demonstrating that the unique 2D layered structure aids charge separation and is key to the high performance. This work introduces COK‐47 as a promising alternative to the well‐known MIL‐125 family and offers directions for future studies


Introduction
The world's growing need for renewable and clean energy makes it imperative to develop technologies for the generation of green fuels such as hydrogen (H 2 ) when obtained from sustainable energy sources. In contrast to the established two-step process for green H 2 production, which involves a combination of a photovoltaic cell and water electrolysis, direct photocatalytic water splitting allows to generate H 2 in a single process without such sophisticated infrastructure, which in turn lowers the end-price of H 2 . [1,2] Still, to this date, photocatalytic water splitting for H 2 production has proven challenging for ordinary metal-oxide-based catalysts. [3] In order to increase the overall efficiency and stability of the water splitting photocatalysts, more tuneable and structurally well-defined photosystems need to be developed.
In this light, metal-organic frameworks (MOFs) have been widely investigated by the scientific community as (photo)catalysts due to their excellent and designable properties: high porosity, crystallinity, surface area, and accessibility to the catalytically active sites. [4][5][6] For the application in water splitting reactions, however, their hydrolytic stability often becomes a limitation and only some stable MOF familiesincluding UiO, ZIF, and MIL, which framework structure is governed by the hard-and-soft-acid-and-bases (HSAB) principle [7][8][9] -remain of interest. Among these, MOFs of the MIL family stand out due to their ability to promote light-driven H 2 evolution reaction (HER), often attributed to the presence of TiO 2 -like secondary building units (SBUs). [10][11][12][13] SBUs are paramount to the structural, optoelectronic, and catalytic properties of a MOF and are typically comprised of 0D metal-oxo clusters. [14] Recent works have reported that infinite 1D Ti-O SBUs can be more advantageous in terms of conductivity and catalytic performance compared to cluster SBU MOFs. [15][16][17][18] Higher dimension SBUs (rods in 1D or sheets in 2D) are, however, still in their infancy understanding stage, but are already speculated to enable more unique physical and chemical properties. [19] Inspired by this, we assumed that higher dimensional SBUs could be beneficial for charge separation and extraction relevant to photocatalysis, which led us to study COK-47 (COK = Centrum voor Oppervlaktechemie en Katalyse), which -together with IEF-1 [20] -represents one of the few examples of a 2D SBU Ti-based MOF reported so far.
The structure of COK-47 comprises sheets of TiO 6 octahedra as 2D SBUs connected by a biphenyl-dicarboxylate (bpdc 2− ) organic linker. [21] Dimensions and number of stacks of the sheets as well as defect structure can be tuned through variation of synthetic conditions. It is expected that the morphology of the COK-47 particle (dimensions, number of stacks, particle size) influences its surface area and ultimately affects charge diffusion length. Moreover, missing ligand defects may be beneficial as they can further increase intraparticle porosity facilitating reactant access, [22] contributing to higher surface areas [23] and improve host-guest interactions at the catalytic site. [24,25] Herein, we investigate the photocatalytic properties of a series of COK-47 MOFs toward HER under diverse illumination conditions and compare them to benchmark Ti-based MOFs. Aided by a range of characterization techniques, we discuss the effects of synthesis conditions on morphology, surface area, and defects (i.e., missing ligands) of the final MOF and show how these characteristics ultimately govern the photocatalytic performance of COK-47. Our results demonstrate that COK-47 ranks among the best performing visible light active photocatalysts for HER, which renders it a promising alternative to the commonly used MIL-125 (Ti) family.

Morphology and Structure
We reproduced the reported synthesis procedures [21] to obtain COK-47 L (L = large particle) and COK-47 S (S = small particle).
The morphological, structural, and stochiometric characteristics of these samples are comparable to the ones reported by Smolders et al.: [21] COK-47 L has (close to) stochiometric metal-to-ligand ratio and is made of comparatively large crystallites (corresponding to its relatively low surface area of 94 m 2 g −1 ), due to the slower heating and cooling rates of the classic solvothermal process. COK-47 S , in contrast, exhibits smaller dimensions of primary particles (corresponding to its higher surface area of 322 m 2 g −1 ) and additionally contains missing ligand defects that result from the microwave-assisted synthesis at higher temperatures and shorter reaction times. In order to maintain small dimensions but also avoid missing ligand defects, in this work we developed a new microwave-assisted synthesis protocol (details in Section 1.2, Supporting Information) with comparatively lower temperatures and longer reaction times. This optimization yielded a high-surface-area, defect-free, and size-isotropic material denoted here as COK-47 ISO (ISO = isotropic particle).
Scanning transmission electron microscopy (STEM) revealed that the new sample COK-47 ISO consists of small nanocrystallites (Figure 1a red dashed circles) with dimensions of ≈9 × 9 × 9 nm (a = axis perpendicular to 2D Ti-O SBU; b x c = plane in which 2D Ti-O SBU grows). These isotropic particles are different from the anisotropic COK-47 L and COK-47 S samples with primary crystallite dimensions of ≈30 × 100 × 100 and 30 × 5 × 5 nm, respectively (see Section 2 and Figures S1 and S2, Supporting Information). COK-47 ISO also consists of a layered structure revealed by STEM in Figure 1b: bright columns of 5-7 TiO 6 -sheet-layers separated by a distance of 1.5 nm can be visualized, which matches the bpdc 2− ligand length (more in Section 3, Supporting Information) and confirms the formation of the 2D-SBUs.
The electron diffraction pattern in Figure 1c shows diffuse rings originating from the small size of randomly oriented crystallites present in the sample. The d-spacings match well with the expected diffractions of COK-47. [21] It is important to note that the pattern disappeared almost instantly upon electron irradiation, which is expected from the presence of highly sensitive, small crystallites.
Further analysis by powder X-ray diffraction (PXRD, Figure 1d) confirmed that COK-47 ISO has the same crystal structure as the COK-47 L and COK-47 S reference materials, both matching well with the patterns reported previously. The broader diffraction peaks in the COK-47 ISO sample indicate a higher degree of disorder, which is attributed to the presence of smaller crystallites. According to Smolders et al., [21] the missing diffraction at 2 = 26°( black star) indicates a preferred orientation of the material in the 0k0 plane, as is also observed for COK-47 S . In contrast, the presence of this diffraction in COK-47 ISO confirms its isotropic morphology. We can also ensure the absence of the TiO 2 polymorphs in the synthesized products based on temperature programmed in situ PXRD, X-ray photoelectron spectroscopy (XPS), and Raman analyses (see Sections 4 and 5, Supporting Information).

Connectivity, Surface Area, and Missing Ligand Defects
Attenuated-total-reflection Fourier-transformed infrared (ATR-FTIR) and Raman spectroscopy provided further insights about the connectivity and potential defects of the as-prepared COK-47 ISO . The FTIR-spectrum in Figure 2a shows exclusively the COK-47 specific peaks, i.e., the symmetric and asymmetric vibrations of the carboxylate coordination bonds at v (COO-) = 1723, 1572, 1506, and 1416 cm −1 (black stars) and the fingerprint region, which matches well with reference spectra. [21,[26][27][28] The prominent v (C = O) vibration at 1674 cm −1 related to the carboxylic acid function of the free H 2 bpdc ligand (black line) is completely absent in COK-47 ISO , indicating the absence of leftover or undercoordinated ligands (details in Section 6, Supporting Information).
The Raman spectrum in Figure 2a (blue) shows that all 5 ligand-specific vibrations are in good agreement with a metalbpdc 2− -carboxylate bonding [29] (purple triangles). Three of the peaks also match the IR bands (1611, 1156, and 862 cm −1 ), while the other 2 peaks (1457 and 1285 cm −1 ) are complementary. All of these peaks are related to stretching or deformation of the carboxylate aromatic ring C ar -C ar , C ar -H, or C ar -C [30,31] (assignments in Section 7, Supporting Information). This confirms a complete metal-carboxylate connectivity without free or singly-coordinated ligands. Furthermore, the detailed Ti-O specific region (Figure 2a, Inset and Figure S15, Supporting Information) shows a unique pattern connected with the structure of the 2-D SBUs of COK-47, which further corroborates the absence of rutile or anatase phase [32] (green or orange triangles).
XPS revealed the expected oxidation state of +IV for Ti. [33][34][35] Moreover, the extracted Ti:O ratio of 1:4 matched well the theoretical value for the defect-free COK-47 (Section 5, Supporting Information).
Ar physisorption resulted in Type II isotherms (Figure 2b), which are typical for nonporous materials. The hysteresis hints at cavitation-induced evaporation, typical for nanoparticle aggregates, [36] or strong interaction between adsorbate and adsorbent, previously reported in flexible MOFs. [37] Note the stronger hysteresis for COK-47 S , which speaks for pore blocking in a narrower range of pore necks, likely caused by this sample's anisotropy and thus stronger preferential orientation. A negligible amount of 16 Å micropores could be observed in the NLDFT pore size distribution and can be explained by the missing ligand defects present in this sample (Section 8, Supporting Information). The defect-induced porosity and internal surface area, however, play a minor role in the overall porosity of this sample, as the t-plot method showed a negligible internal surface area value, suggesting pore blockage or inaccessibility even for Ar molecules. Table 1 shows the resulting external surface areas for all materials. The ratio of surface areas of COK-47 ISO to COK-47 S , i.e., 0.89, goes well in line with the theoretical ratio obtained from TEM and geometric calculations, i.e., 0.77 (details in Section 8, Supporting Information).
The content of missing ligand defects was quantitatively assessed by TGA upon heating to the temperature at which the MOF fully decomposes and TiO 2 is formed (more in Section 9, Supporting Information). Figure 2c and Table 1 show the organic mass losses at 600°C for COK-47 L , COK-47 S , and COK-47 ISO corresponding to 61, 56, and 63%, respectively. The theoretical organic mass loss for the stochiometric material Ti 2 O 3 (bpdc) is 63%, whereas the defect-rich formula Ti 2 O 3 (bpdc) 0.61 (MeO) 0.78 corresponds to a mass loss of 55%. [21] The % of defects can thus be calculated and is summarized in Table 1. The data confirms  that the longer reaction time and moderate temperatures of the new microwave-assisted synthesis procedure avoid missing ligand defects in COK-47 ISO ; at the same time, missing ligand defect contents in COK-47 L (9%) and COK-47 S (32%) are comparable to the ones previously reported [21] (more in Section 9, Supporting Information). Elemental analysis (Section 10, Supporting Information) confirms similar C/Ti atomic ratios as should be expected by these theoretical formulas.

Optoelectronic Properties
Absorption spectra obtained with solid-state diffuse reflectance spectroscopy (DRS) reveal a notable red shift of the absorption edge for COK-47 ISO compared to the COK-47 S and COK-47 L samples (Section 11, Supporting Information), thus opening up the possibility for visible light photoexcitation. DFT calculation (Figure 3a) showing the band structure along high symmetry points, accompanied by the Tauc functions (Figure 3b; Section 11, Supporting Information) reveal that all samples feature an indirect band gap (E g ) [38][39][40] estimated as 3.17, 2.98, and 2.85 eV for COK-47 L , COK-47 s , and COK-47 ISO , respectively. A small indirect band gap is generally preferred for photocatalytic applications due to the enhanced lifetime of photoexcited charge carriers, [41] thus rendering COK-47 ISO with a band gap of 2.85 eV (absorption onset of 435 nm) the most promising candidate for visible-lightdriven photocatalysis among this range of COK-47 materials. We used steady-state and time-resolved PL spectroscopy to gain more insight into the charge recombination kinetics. Steadystate PL ( Figure S24, Supporting Information) shows weak emission peaks for all COK-47 samples when excited at 405 nm. It is noteworthy that these PL peak intensities are much lower than those measured for other visible-light active MOFs, such as NH 2 -MIL-125 (Ti) (Section 12, Supporting Information). Generally, emission is indicative of charge recombination, which in the case of COK-47 originates from either insufficient mobility of the photoexcited electrons or the presence of missing ligand defects acting as recombination centers. [42][43][44] The almost complete quenching of emission in nanocrystalline defect-free COK-47 ISO -compared to COK-47 S and COK-47 L -suggests a more efficient charge separation, which potentially provides longer-lived charge carriers and avoids quick radiative recombination processes, as we have demonstrated in our previous work. [13] Time-resolved emission spectroscopy (TRES) measurements (Figure 3c) support this hypothesis showing that COK-47 ISO has the longest lifetime with = 2.17 ns, compared to COK-47 L and COK-47 S , with = 1.43 and 0.30 ns, respectively, which correlates well with the increasing defect content and larger particle dimensions found in the samples.

Light-Driven HER of COK-47
In this work, we tested COK-47 for the first time toward photocatalytic HER. We used a range of illumination sources: narrowband LED emitters, broad-band Hg lamp, and solar simulated light, with special focus on the 365 and 405 nm LEDs. The experiments were conducted in tailor-made glass reactors using an optimized amount of Pt (2 wt.%) as co-catalyst and various amounts of MeOH as hole scavenger (details in Sections 1.5 and 13.1, Supporting Information). Figure 4a shows the HER rate in the batch reactor after 1 h of irradiation with different light sources and reveals that COK-47 is highly active toward photocatalytic HER. UV irradiation (under narrow-band 365 nm light) yielded 29.5 μmol h −1 of H 2 when using 2 mg of photocatalyst, which corresponds to an apparent quantum efficiency (AQY) of 2.0% (calculation in Section 13.2, Supporting Information). Remarkably, under otherwise identical experimental conditions, visible-light HER activity (under 405 nm) reaches 8.6 μmol h −1 (AQY = 0.5%). Although comparing activities measured by different research groups is a complicated matter, [45][46][47]  In the interest of real-life catalytic applications of the optimal COK-47 ISO , we conducted HER experiments in a flow reactor using an online detection system (details in Section 1.5, Supporting Information) aiming to gain insights on the long-term performance of the catalysts. Figure 4c shows that the HER rates in the flow reactor are comparable to the ones obtained in the batch reactor. An appreciable deactivation of the Pt@COK-47 photosystem can be observed especially under 365 nm illumination in both short term (over 2 h, Figure 4c) and long term HER runs (over 14 h, Figure S31, Supporting Information). Based on the well-maintained PXRD pattern of COK-47 ISO after reaction (Figure 4e), we assign this effect to the early stage deactivation phenomenon common in Pt@TiO 2 , [48] and suggest that only a minor UV-induced oxidation of organic ligands, i.e., MOF corrosion, takes place under these experimental conditions (see Section 13.5, Supporting Information). In contrast to this, when using the 405 nm illumination source, COK-47 ISO displays a less pronounced deactivation proving to be more advantageous. Importantly, under both 365 nm and 405 nm light, the structure of the material is preserved even after such long time periods

DFT and Mechanistic Studies
Computational studies and density functional theory (DFT) calculations (details in Section 1.4, Supporting Information) revealed the electronic structure of COK-47, as the atomic projection of the total density of states (DOS) in Figure 5a shows. The resulting valence band is mainly originating from the p-orbital in the C (and O atoms), while the conduction band is primarily originating from states in the d-orbitals of the Ti atom, as Figure 5b shows.
The band structure analyses and the high symmetry points in the first Brillouin zone suggest that the nature of the band gap is indirect (theoretical value E g , electronic = 2.19 eV and experimental value E g, optical = 2.85 eV), well in line with the Tauc plot analyses. This indirect band gap likely contributes to suppressing electron-hole recombination processes, as revealed by PL measurements. Interestingly, COK-47 ISO possesses the smallest exciton binding energy (E b ), i.e., the difference between the ex-perimentally measured optical gap and DFT-estimated electronic gap (see Sections 18 and 18.1, Supporting Information), which can be another beneficial factor toward better photocatalytic performance.
Partial DOS assigned to the individual atoms (Section 18.2, Supporting Information) further reveal that two types of C atoms correspond to the ligand (those on benzene rings and those of the -COO) and that C atoms from the rings -which contributes to the valence band maximum -are likely to be responsible for the excitation of electrons near to the valence band edge. At the same time, partial DOS of the O atoms (Section 18.3, Supporting Information) indicate that many O states, which arise from the 2D SBU (TiO 6 octahedra), locate at deeper energies within the valence band. Such a complex VB profile can be responsible for a dual-excitation mechanism as a function of the incoming photon energy. Upon visible light illumination (less energy), electrons at the biphenyl ligand are excited, which are then transferred to the Ti-SBU, resulting in a band-to-band excitation via a ligand to metal charge transfer (LMCT) mechanism. Upon UVlight illumination (higher energy), electrons from deeper in the valence band, i.e., from the O atoms in the SBU's, can also be photoexcited. In this case, the resulting electrons and holes reside at the SBU and are not spatially separated, making recombination events more likely. Mechanistic studies were conducted to reveal the prevalent mechanism responsible for hole consumption during the photocatalytic process. Generally, the photo-generated holes can be consumed either via direct extraction of an adsorbed sacrificial reagent (SR), usual for MeOH on oxide surfaces, [49] or via the OH-radical ( • OH) pathway, usual for TEOA in solutions. [50] • OH-trapping experiments (details in Section 1.7, Supporting Information) help unraveling these consumption mechanisms: upon illumination of a COK-47 suspension, OH-radicals are generated and then trapped by terephthalic acid, forming hydroxyl terephthalic acid, which fluoresces at ≈420 nm. A quenching of this signal is an indication for the efficiency of charge transfer to the SR, which hinders the formation of • OH. Under visible illumination (Figure 5d), the PL quenching is similar for both SRs, speaking for a similar charge extraction mechanism, which is also reflected in similar HER-rates obtained when aqueous MeOH or TEAO were used (Figure 5c).
The DFT results suggested that the photogenerated holes in COK-47 reside at the ligand, hence, both MeOH and TEOA can undergo adsorption at the ligand and consume the holes. Under UV illumination (Figure 5e), the quenching is very strong for MeOH compared to that for TEOA, speaking for a facilitated charge extraction. In this case, however, the holes tend to reside at the TiO 6 -SBU on which MeOH adsorbs directly as methoxy [13] and consumes the charges more efficiently (as a result, we get high HER activity and less • OH generated). In contrast to this, TEOA is not able to directly adsorb at the SBU and is only con-sumed through the OH-radical pathway, which results in higher • OH generation and lower HER activity.
These findings speak for a wavelength-dependent HER mechanism for the novel COK-47 MOF (Figure 6), which might be caused by electrons being excited from different energy levels (and different locations of the framework, i.e., ligand or Ti-SBU) depending on the wavelength used, leading to different catalytic pathways, and most likely affecting adsorption equilibria of reactants. Hence, we suggest that the key challenge for the visible light triggered photoexcitation is the hole transfer to the sacrificial reagent (given the LMCT), whereas the rate limiting contribution upon UV illumination is recombination of nearly located electron-hole pairs.
Systematic comparison of all COK-47 samples, which feature controllable variations in structural and optoelectronic properties, further allows us to make some qualitative deductions with regard to which of COK-47's properties are beneficial to its HER performance. As Figure 4b shows: high surface area and narrow band gap result in a positive HER performance, while missing ligand defects decrease the HER rate. This seemingly contradicts recent works that demonstrated the beneficial aspect of similar defects in increasing surface area and facilitating access to active sites with the framework by introducing additional porosity. [22,23] In our work, however, the defects do not introduce micropores in the frameworks, thus restricting methanol to adsorb only on the external surface. Consequently, the missing ligand defects act predominantly as recombination centers, in line with our TRES results. Moreover, the lower activities of COK-47 L and COK-47 s likely arise from their lower ligand contents, resulting in states near the valence band edge composed of oxygen instead of carbon, as the DOS calculations show (Figures S44 and S45, Supporting Information). Under 405 nm excitation, a remarkably detrimental effect on visible light activity is observed for COK-47 L , which we attribute to this sample's broader band gap and larger dimensions, hindering its visible light absorption and increasing the charge transport distance for electrons (100 nm SBU vs 10 and 5 nm for the other 2 samples), respectively. Hence, the final trend COK-47 L < COK-47 S < COK-47 ISO results from an interplay of the sample's parameters, in which large surface area, short charge transport distances (small dimension of the 2D SBU), narrow band gap, and low defect contents result in higher activities, which suggests that COK-47 ISO possesses the optimal requisites for photocatalytic HER.

Conclusion
Here we report a first detailed investigation on the optoelectronic and photocatalytic properties of a novel Ti-based MOF featuring 2D SBUs toward hydrogen evolution reaction (HER). COK-47 is shown to be highly active under both UV and visible light illumination with H 2 generation rates of 29.5 μmol h −1 (AQY of 2.0%) and 8.6 μmol h −1 (AQY of 0.5%), setting it among the most active MOFs for -especially visible light driven -HER. We were able to identify the positive impact of larger surface areas and narrower band gaps, while missing ligand defects showed a detrimental effect on the HER activity, playing an important role in facilitating recombination by trapping electrons. We also unraveled important insights into the reaction mechanisms in terms of charge generation, transfer, and light dependent consumption of different sacrificial reagents, identifying the key challenges for visible or UV photoexcitation: under visible light, hole transfer to the sacrificial reagent is determining, whereas recombination is the rate limiting factor for UV excitation. Our experimental results are supported and go well in line with the theoretical DFT calculations. The 2D structure promotes an efficient charge separation due to fast electron transfer, hence, future studies focusing on COK-47 and other MOFs with high-dimensional SBU's can hold the key into advancing MOF-based photocatalysis.

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