Attraction in Action: Reduction of Water to Dihydrogen Using Surface-Functionalized TiO2 Nanoparticles

The reactivity of a heterogeneous rhodium(III) and ruthenium(II) complex-functionalized TiO2 nanoparticle (NP) system is reported. The ruthenium and rhodium metal complexes work in tandem on the TiO2 NPs surface to generate H2 through water reduction under simulated and normal sunlight irradiation. The functionalized TiO2 NPs outperformed previously reported homogeneous systems in turnover number (TON) and frequency (TOF). The influence of individual components within the system, such as pH, additive, and catalyst, were tested. The NP material was characterized using TGA-MS, 1H NMR spectroscopy, FTIR spectroscopy, solid absorption spectroscopy, and ICP-MS. Gas chromatography was used to determine the reaction kinetics and recyclability of the NP-supported photocatalyst.


Introduction
The increasing energy demand of the world population has led to the unsustainable consumption of fossil fuels and emission of greenhouse gases [1][2][3]. Within the next century, fossil fuels will be substantially depleted if consumption rates or energy sources do not change [4]. The development of cleaner, renewable, available, and less-expensive energy solutions has become a central societal and research imperative [5].
Dihydrogen is an outstanding candidate as fuel, possessing key advantages, including long term storage and carbon-free combustion. Liquid or high-pressure gaseous dihydrogen has high gravimetric energy density while lacking volumetric energy density compared to liquid fossil fuels [6]. Hydrogen is the most abundant element in the universe and the tenth most abundant element in the earth's crust by weight percentage. Terrestrially, only small amounts of elemental H 2 occur, with most hydrogen found within molecules, most commonly water [7]. This means dihydrogen for use as fuel must be generated through chemical transformations.
The most common processes for large-scale dihydrogen preparation are water electrolysis and steam methane reforming [8]. Both technologies require high amounts of energy and work best with non-sustainable metal catalysts [8][9][10]. An imperative is improving energy efficiency and using sustainable, recyclable, or easily recoverable catalysts [11]. One option is to combine energy harvesting and dihydrogen evolution using photocatalysts that work under sunlight irradiation [12].
Homogeneous photocatalytic systems can be very efficient, although catalyst recovery can be extremely challenging and cost or energy-intensive [13,14]. In particular, multicomponent systems are problematic. Heterogeneous catalysts are often easier to recover but have the disadvantage of inactive interior volumes with only surface sites being catalytically active [13][14][15]. An alternative is to surface-functionalize nanoparticle (NP) scaffolds composed of cheap and abundant elements with photocatalysts. Such immobilized photocatalysts offer greater catalyst-to-volume ratios than bulk heterogeneous catalysts, which We now describe the immobilization of components from a previously reported aqueous photochemical system for water reduction under simulated sunlight irradiation [27]. The system utilizes TiO 2 NPs supporting ruthenium(II) and rhodium(III) 2,2 -bipyridine (bpy) complexes as photosensitizer and relay species, respectively Scheme 1) [27]. The complexes at the desired surface ratio are assembled on [2,2 -bipyridine]-4,4 -diylbis(phosphonic acid) (1) functionalized NPs (1@TiO 2 ). Comparative studies with ZrO 2 NPs are also reported. ZrO 2 is an insulator and the NPs are commercially available. The change from TiO 2 (band gap = 3.2 eV) [28] to ZrO 2 (band gap = 5.1 eV) [29] was expected to yield insights into metal complex to NP surface bonding and electronic interactions.

TiO 2 NPs Functionalization
NP activation and functionalization with ligand (1) were carried out according to our previously published procedure [36,37]. The procedures were adjusted according to the molecular weight for anchoring ligand (1) (see following section).

Activation of Commercial P25 TiO 2 NPs
The commercially available NPs were activated as previously reported [36]. The procedure was scaled up as follows. Commercial P25 TiO 2 NPs (5.00 g) were dispersed by sonication for 15 min in dilute aqueous HNO 3 (70 mL, 3 M). The mixture was then stirred for 30 min. The suspension was centrifuged (10 min, 7000 rpm) and the NPs were washed once with milliQ water (70 mL). The NPs were added to milliQ water (50 mL) and dispersed by sonication for 10 min. The suspension was then stirred overnight. The suspension was centrifuged (10 min, 7000 rpm) and the NPs were washed with milliQ water (2 × 50 mL). The activated NPs

Rh@TiO 2
The metal complex was formed directly on the NP surface. 1@TiO 2 (200 mg), RhCl 3 ·3H 2 O (4.51 mg, 17.1 µmol) and 2,2 -bipyridine (6.81 mg, 4.36 µmol) were added to a vial. H 2 O (5 mL) and EtOH (3 mL) were added, and the mixture was thoroughly dispersed using sonication and stirring. The suspension was transferred to an autoclave PTFE liner with additional EtOH (2 mL). The autoclave was sealed and then heated in an oven with 320 • C/h to 160 • C. The autoclave was left at 160 • C for 1 h. After cooling down the autoclave was opened and the suspension was centrifuged (20 min, 7000 rpm). The resulting NPs were washed with H 2 O (3 × 10 mL) and EtOH (1 × 10 mL). Rh@TiO 2 was isolated as white powder. 1

rR-TiO 2
The metal complex was formed directly on the NP surface. Hence, 1@TiO 2 (241 mg), RuCl 3 ·3H 2 O (0.24 mg, 0.9 µmol), RhCl 3 .3H 2 O (5.16 mg, 19.6 µmol), and 2,2 -bipyridine (8.08 mg, 51.8 µmol) were added to a vial. H 2 O (5 mL) and EtOH (3 mL) were added, and the mixture was thoroughly dispersed using sonication and stirring. The suspension was transferred to an autoclave PTFE liner with additional EtOH (2 mL). The autoclave was sealed and then heated in an oven at a rate of 320 • C/h to 160 • C. The autoclave was left at 160 • C for 1 h. After cooling down the autoclave was opened and the suspension was centrifuged (20 min, 7000 rpm). The resulting NPs were washed with H 2 O (3 × 10 mL) and EtOH (1 × 10 mL). rR@TiO 2 was isolated as an orange powder. 1

RR-TiO 2
The metal complex was formed directly on the NP surface. 1@TiO 2 (340 mg), RuCl 3 .3H 2 O (2.39 mg, 9.1 µmol), RhCl 3 ·3H 2 O (5.16 mg, 19.6 µmol) and 2,2 -bipyridine (11.3 mg, 72.3 µmol) were added to a vial. H 2 O (5 mL) and EtOH (3 mL) were added, and the mixture was thoroughly dispersed using sonication and stirring. The suspension was transferred to an autoclave PTFE liner with additional EtOH (2 mL). The autoclave was sealed and then heated in an oven with 320 • C/h to 160 • C. The autoclave was left at 160 • C for 1 h. After cooling down, the autoclave was opened and the suspension was centrifuged (20 min, 7000 rpm). The resulting NPs were washed with H 2 O (3 × 10 mL) and EtOH (1 × 10 mL). RR-TiO 2 was isolated as a dark orange powder. 1

General Procedure
The system for dihydrogen generation used metal complex functionalized NPs [38] as photo-and electrocatalysts, triethanolamine as a sacrificial electron donor, K 2 [PtCl 4 ] as catalyst to facilitate dihydrogen formation (possibly by Pt NP formation), bpy as additive, aqueous H 2 SO 4 for pH adjustment, and milliQ water as solvent. As [Ru(bpy) 3 ] 2+ and [Rh(bpy) 3 ] 3+ are somewhat photolabile under the operating conditions, the additional bpy was added to regenerate ruthenium and rhodium surface-bound complexes.
In a 5 mL microwave vial TEOA (2.52 mmol, 376 mg), K 2 [PtCl 4 ] (1.7 µmol, 0.70 mg) and 2,2 -bipyridine (18.6 µmol, 2.91 mg) were added together with milliQ water and aqueous H 2 SO 4 (1M) to modify the pH. Experiments performed at pH 10 used no aqueous H 2 SO 4 (1M) and 6 mL milliQ water while experiments performed at pH 7.5 used 1 mL aqueous H 2 SO 4 (1M) and 5 mL milliQ water. Metal complex-functionalized NPs were added (114.1 mg). The vial was flushed with nitrogen and then sealed. The suspension was sonicated (10 min) and thoroughly shaken. Nitrogen was bubbled through the suspension for 10 min. The vial was irradiated for 4-8 h at a slight angle (5 • ) with a sun simulator generating 1200 W m -2 . The suspension was shaken hourly. Headspace samples for gas chromatography were collected using a syringe and transferred to a 10 mL GC vial for analysis. The measured GC integral was converted to mL of H 2 with a calibration made by injecting several known volumes of dihydrogen.

Kinetic Measurements
Using the general procedure, a kinetic run was performed at pH 7.5 using rR@TiO 2 as photocatalyst. Two vials were prepared and the H 2 evolution was measured hourly by collecting the headspace by syringe and transferring it to a 10 mL GC vial for GC analysis. After each collection, the suspension was bubbled with N 2 for 5 min and then irradiation was continued. The vials were irradiated in total for 8 h and 9 h respectively.

Recycle Measurements
Using the general procedure, a recycling experiment was performed preparing two vials at pH 7.5 using rR@TiO 2 as photocatalyst with 4 h irradiation for each cycle. H 2 evolution was recorded after each cycle by collecting the headspace by syringe and transferring it to a 10 mL GC vial for GC analysis. After each collection, the suspension was centrifuged, the supernatant was removed, the NPs were washed with water (4 × 10 mL). Subsequently, the NPs were dried under high vacuum. Following the general procedure, the recycled NPs were used instead for the next cycle.

Anchoring Ligand Functionalization
We commenced by establishing a reliable method for the preparation of 1@TiO 2 and its subsequent metallation. Anchoring ligand 1 was prepared according to the literature [31][32][33][34][35] with the phosphonic acid being chosen for stable binding to TiO 2 surfaces in neutral, slightly basic and slightly acidic conditions [36,37]. The bpy metal-binding domain in 1 is commensurate with the assembly of surface-bound {M(bpy) 3 } n+ (M = Ru, n = 2; M = Rh, n = 3) motifs. The previously developed method [36,37] for NP anchor functionalization with tpy metal-binding domains was adapted for anchoring ligand 1.
In this case, 32.2 eq. of activated NPs [38,39] were dispersed with anchoring ligand 1 in water and heated to 130 • C for 3 h in a microwave reactor (see Experimental section for full details).
The purchased ZrO 2 NPs were functionalized using similar techniques but without prior activation. The ZrO 2 NPs had a diameter of 100 nm changing the surface area to volume ratio significantly (from 28% to 6%). The functionalization method was optimized using our previously established formula [38] to give an adjusted ratio of 6.9 ZrO 2 eq. to 1 eq. anchoring ligand. The resulting NPs were carefully washed to avoid non-bound free ligand.

Ligand Functionalized NP Characterisation Methods
The 1@TiO 2 NPs were analysed using TGA-MS (see Figure S1) and showed a small weight loss (<1%) in two steps (<120 • C, <380 • C) attributed to the loss of physisorbed and chemisorbed water. This was further seen in the coupled MS, which showed peaks with m/z of 17 and 18 for HO and H 2 O respectively. A further weight loss of 2.9% over the range of 380-900 • C was measured and attributed to the decomposition of the anchoring ligand on the surface. The MS in this range showed the decomposition product of the ligand to be CO 2 . In contrast, activated NPs (see Figure S2) showed almost no weight loss over the range of 380-900 • C. To verify that the organic weight loss within the 380-900 • C range was not caused by non-covalently bound free ligand, the 1 H NMR spectrum of 1@TiO 2 was measured in D 2 O (see Figure S3) and showed no significant amounts of free ligand in solution (ligand bound to the NPs is not observed) [36]. Further characterisation with FTIR spectroscopy (see Figure S4) and solid-state absorption spectroscopy (see Figure S5) revealed peaks at 1630, 1590, 1540, 1500, 1480, 1430, and 1160 cm -1 and 400-670 nm respectively which are not observed for pristine activated NPs. The additional peaks found for 1@TiO 2 in FTIR spectroscopy did match to the pristine anchoring ligand. MALDI mass spectrometry showed peaks arising from species consistent with surface binding of the phosphonate ligand (See Figures S6-S8).
1@ZrO 2 was characterized similarly to 1@TiO 2 : TGA-MS, 1 H NMR spectroscopy, and FTIR spectroscopy (see Figures S9-S12) were measured. 1@ZrO 2 showed a weight loss of 1.7% over the range of 380-900 • C, which was attributed to the decomposition of the anchoring ligand on the surface. The MS in this range showed the decomposition product of the ligand to be CO 2 . The commercial NPs did not show any weight loss other than H 2 O. 1 H NMR spectroscopy in D 2 O was used to verify absence of non-covalently bound free ligand and no significant amounts of free ligand in solution were observed. FTIR spectroscopy showed several weak additional peaks compared to the commercial NPs within the 1600 and 800 cm -1 region (see ESI † Figure S12) originating from the anchoring ligand.

Nanoparticle Surface Complexation
Metal complexes were assembled directly on the NP surface using methods derived from literature procedures for heteroleptic ruthenium(II) complexes, replacing free ligand by 1@TiO 2 ( Table 2) [35]. Orange Ru@TiO 2 was isolated from the reaction of 1@TiO 2 NPs [38], RuCl 3 .3H 2 O, and bpy in 1:1 water-ethanol in an autoclave at 160 • C for 1 h (see Experimental section for full details). Ru@TiO 2 can also be formed by replacing RuCl 3 .3H 2 O and bpy with [Ru(bpy) 2 Cl 2 ], giving a similar product (TGA-MS, 1 H NMR spectroscopy, FTIR and solid-state absorption spectroscopies, MALDI mass spectrometry see Figures S13-S27) differing only slightly in colour and small weight loss differences during the TGA measurement. All subsequent metal complex-functionalized NPs ( Table 2) were prepared in the autoclave with H 2 O/EtOH as solvent using the one-pot (MCl 3 + bpy) procedure described in the experimental. The abbreviation rR@TiO 2 describes metal complex functionalized NPs with a small amount of ruthenium vs. rhodium (1:20) on the surface, while for RR@TiO 2 ruthenium and rhodium surface concentrations are comparatively similar (1:2). The first letter in rR and RR always refers to ruthenium and the second to rhodium.
The metal complexes Ru@ZrO 2 and rR@ZrO 2 were prepared using 1@ZrO 2 as starting material and followed the procedure described above with TiO 2 .
Emission spectra of the metal complex functionalized NPs dispersed in water were recorded (Figure 1). The suspension was excited at 450 nm. Ru@TiO 2 showed a broad emission at 634 nm while RR@TiO 2 showed a broad emission at 630 nm. Both Rh@TiO 2 and rR@TiO 2 were non-emissive. ruthenium and rhodium, the observed isotope distribution in the MALDI mass spectrum confirmed that the rhodium complex was dominant (Figures S36-S44). For full characterisation details, see Section 2 or ESI † Figures S36-S49.
Emission spectra of the metal complex functionalized NPs dispersed in water were recorded (Figure 1). The suspension was excited at 450 nm. Ru@TiO2 showed a broad emission at 634 nm while RR@TiO2 showed a broad emission at 630 nm. Both Rh@TiO2 and rR@TiO2 were non-emissive. The NPs were further investigated using ICP-MS (see Table 3). Ruthenium or rhodium functionalized NPs showed their respective elements. It was not possible to perform an exact surface concentration measurement of the elements with ICP-MS as even concentrated nitric acid did not remove all of the surface-bound species from the NP.  The NPs were further investigated using ICP-MS (see Table 3). Ruthenium or rhodium functionalized NPs showed their respective elements. It was not possible to perform an exact surface concentration measurement of the elements with ICP-MS as even concentrated nitric acid did not remove all of the surface-bound species from the NP. For Ru@ZrO 2 and rR@ZrO 2 , characterization was performed using TGA-MS, 1 H NMR spectroscopy, and FTIR spectroscopy (see Figures S12, S50-S52). TGA-MS of Ru@ZrO 2 and rR@ZrO 2 revealed a smaller weight loss of 2.1% (380-900 • C) compared to the TiO 2 equivalents Ru@TiO 2 (4.6%) and rR@TiO 2 (3.7%). This was within expectations considering the weight loss seen for 1@ZrO 2 . Importantly, an increase from 1@ZrO 2 to Ru@ZrO 2 or rR@ZrO 2 was still observed. 1 H NMR spectroscopy was used to verify the absence of non-bound anchoring ligand while the FTIR spectra showed only minor differences to 1@ZrO 2 within 1600-800 cm -1 .

Performance and Influence of Individual Components during Dihydrogen Generation
Experimental details are given in Section 2.3.1. The study revealed that 1@TiO 2 , Ru@TiO 2 , and Rh@TiO 2 all produce H 2 under irradiation (Table 4) but the gas generation is significantly higher when both ruthenium and rhodium are present on the surface. It is especially interesting that when a single batch of NPs was functionalized with both metal complexes (rR@TiO 2 or RR@TiO 2 ), the H 2 production was more than double compared to that using an equivalent mixture of Ru@TiO 2 and Rh@TiO 2 . We propose that the additional efficiency for rR@TiO 2 and RR@TiO 2 can be explained by an energy transfer from Ru to Rh which promotes the dihydrogen generation [27,40]. The recorded emission spectra of each species support this proposal as rR@TiO 2 is not emissive. It has to be noted that the energy transfer must be relatively inefficient since with higher ruthenium concentrations on the surface (RR@TiO 2 vs. rR@TiO 2 ) the metal complex functionalized NPs are emissive again. Further experiments were performed to test the influence of each component in the system. In the absence of K 2 [PtCl 4 ], there was a strong decrease in efficiency while K 2 [PtCl 4 ] on its own gave no H 2 generation. Removing bpy had a less significant influence on the H 2 generation but slightly lowered the efficiency. In contrast, increasing the bpy concentration had no effect on the efficiency. TEOA was essential for dihydrogen formation. Another important parameter was the pH; at pH 10 the system generated dihydrogen at half the rate observed at pH 7.5. A pH dependence is expected as the formal driving force of the reaction (∆G) will depend upon pH according to the Nernst equation (∆G = −nFE) since the key step in the formation of H 2 involves a proton. A significant improvement in H 2 formation (17%, TOF Rh = 3.4 × 10 -3 s -1 , TOF Ru = 7.4 × 10 -2 s -1 ) could be observed by simple stirring of the suspension instead of periodic shaking. This shows when the metal complex functionalized NPs settle and block the light, they partially hinder a successful water reduction on covered NPs. For this observation to be possible, a successful surface functionalization must have happened. A further experiment was performed by using normal sunlight, resulting in respectable dihydrogen generation (see Table 4, Entry 15 g ), especially since the weather conditions were not bright sunlight.
The dihydrogen generation experiment was expanded using ZrO 2 NPs as the metal complex carrier material. Since ZrO 2 is an insulator, the change from TiO 2 (band gap = 3.2 eV) [28] to ZrO 2 (band gap = 5.1 eV) [29] was expected to hinder the H 2 generation during the experiment if the metal complex is surface-bound and interacting electronically with the NP. Hence, the experiment (see Table 4, Entry 16 f ) yielding almost no H 2 generation during the irradiation was interesting and strongly implies mediation by the semiconducting TiO 2 nanoparticles.

Kinetics of rR@TiO 2 during Light Irradiation
The kinetics of the H 2 formation when using rR@TiO 2 in the general setup was investigated and the experimental results are plotted in Figure 2a) and shown in Table 5. Experimental details are given in Section 2.3.2. Analysis of the data in Figure 2 when TEOA was considered as a reactant (see Figure 2b) indicated first order kinetics (R 2 = 0.986) or second order kinetics (R 2 = 0.995). This observation is consistent with the mechanism proposed by Kirch et al. [27]. Using these data, a minimal TON of 86 (TOF = 2.7 × 10 -3 s -1 ) and TON 1844 (TOF = 5.7 × 10 -2 s -1 ) can be calculated for the rhodium and ruthenium surface-bound complexes respectively. Furthermore, the amount of H 2 produced after 9 h irradiation corresponds to a depletion of 32% of TEOA.

Recyclability of rR@TiO2
Our motivation was to develop a system that could be recycled multiple times and still generate dihydrogen upon irradiation (see Video 1 in the ESI † for visual representation). After an initial run, the NPs were separated from the solution and washed several times with water to ensure the complete removal of the solution components. The NPs were then dried under high vacuum and redispersed in milliQ water at pH 7.5 with fresh TEOA, bpy and K2[PtCl4]. After each run, the headspace was collected and measured using GC analysis (see Table 6). Overall, rR@TiO2 NPs were most suitable for multiple cycles with only a slight decline in efficiency (see Figure 3). We believe that the decline in efficiency is due to the loss of NPs in the recycling processing (10 wt.% after 8 cycles) and defunctionalization of the NP surface. Using the data, a minimal TON of 300 (TOF = 1.9 × 10 -3 s -1 ) and TON 6424 (TOF = 4.1 × 10 -2 s -1 ) can be calculated for the rhodium(III) and ruthenium(II) surface bound complexes, respectively.

Recyclability of rR@TiO 2
Our motivation was to develop a system that could be recycled multiple times and still generate dihydrogen upon irradiation (see Video 1 in the ESI † for visual representation). After an initial run, the NPs were separated from the solution and washed several times with water to ensure the complete removal of the solution components. The NPs were then dried under high vacuum and redispersed in milliQ water at pH 7.5 with fresh TEOA, bpy and K 2 [PtCl 4 ]. After each run, the headspace was collected and measured using GC analysis (see Table 6). Overall, rR@TiO 2 NPs were most suitable for multiple cycles with only a slight decline in efficiency (see Figure 3). We believe that the decline in efficiency is due to the loss of NPs in the recycling processing (10 wt.% after 8 cycles) and defunctionalization of the NP surface. Using the data, a minimal TON of 300 (TOF = 1.9 × 10 -3 s -1 ) and TON 6424 (TOF = 4.1 × 10 -2 s -1 ) can be calculated for the rhodium(III) and ruthenium(II) surface bound complexes, respectively.

Conclusions
We have demonstrated the immobilization of an established photochemical system for the solar generation of dihydrogen using sunlight by binding photo-and redox-active Rh and Ru complexes to TiO2 NP surfaces. We further show that binding both metal complexes to the same NP improves the photocatalytic efficiency. The kinetic rate order and

Conclusions
We have demonstrated the immobilization of an established photochemical system for the solar generation of dihydrogen using sunlight by binding photo-and redox-active Rh and Ru complexes to TiO 2 NP surfaces. We further show that binding both metal complexes to the same NP improves the photocatalytic efficiency. The kinetic rate order and recyclability were determined. The NPs could be recycled multiple times and retain dihydrogen generation capacity. TON and TOF of the system were determined and exceeded the previous reported similar homogenous system. The dihydrogen genera ration was monitored using GC while the NPs were characterized using various methods, including MALDI, FTIR spectroscopy, TGA-MS, solid-absorption spectroscopy, fluorescence emission spectroscopy, and ICP-MS.