*2.3. Photocatalytic Experiments*

Photocatalytic experiments were performed in a 30 mL double -mouthed heart-shaped reactor under UV light irradiation (UV Spotlight source LightningcureTM L8022, Hamamatsu, maximum emission at 365 nm) or solar irradiation (Newport, Xe lamp). Light was focalized on the sample compartment through an optic fiber. In a typical process, 5 mg of catalyst was dispersed into 5 mL of glycerol/water (10% *v*/*v*) solution. Reactions were performed under an inert atmosphere, achieved by bubbling a nitrogen flow (20 mL·min−1) for 30 min. The catalyst suspension was continuously stirred (800 rpm) and the reactor was thermostated at 20 ◦C. A picture of the photocatalytic reactor is shown in Figure 2.

**Figure 2.** Picture of the photocatalytic reactor used in the hydrogen production from glycerol photo-reforming.

Hydrogen was analyzed by sampling with a pressure-lock precision analytical syringe (Valco VICI Precision Syringes, 1 mL, leak-tight to 250 psi) from the head space after 3 and 6 h of irradiation. Analyses were performed on an Agilent Technologies 7890A gas chromatograph equipped with a Supelco CarboxenTM 1010 Plot column with TCD detector. The separation was performed at 70 ◦C for 2 min, followed by heating to 120 ◦C (ramp of 10 ◦C·min−1), and was left for 8 min (total analysis time, 15 min). All reactions were performed in duplicate, with the standard deviation being below 3%. The calibration plot (Figure S1) and a typical chromatogram for hydrogen quantification (Figure S2) are given in the Supplementary Materials.

## **3. Results and Discussion**

The synthesized catalysts were characterized from the structural and chemical point of view with a wide variety of techniques. The chemical composition was determined by ICP-MS and the results, presented in Table 1, evidenced a good incorporation of nickel, with values being quite close to the nominal content (0.5% by weight).

**Table 1.** Metallic content of the solids as determined by inductively-coupled plasma mass spectrometry (ICP-MS). DP: deposition–precipitation; IMP: impregnation.


X-ray di ffraction patterns were obtained and used to obtain structural information of the catalysts, and the results are presented in Figure S3. The Evonik P25 support clearly shows the di ffraction lines corresponding to anatase (80%) and rutile (20%) phases, which are not a ffected by nickel incorporation, independent of the synthetic method or reduction treatment. In addition, consistent with the small metal loading and the homogeneous nickel dispersion evidenced by TEM micrographs, no signals associated to nickel species are observed.

Temperature-programmed reduction (TPR) measurements were carried out for both Ni-0.5-DP and Ni-0.5-IMP catalysts, and the results are presented in Figure 3 (right). These results showed that the reduction peaks associated with nickel species begin at temperatures below 200 ◦C but extend to 400 ◦C.

**Figure 3.** UV-vis spectra (**left**) and temperature-programmed reduction profiles (**right**) obtained for the catalysts prepared in this work.

It is well known that NiO is usually the main surface species when nickel is deposited in high loads on metal oxide-type supports. However, certain species from the Ni–support interaction can be observed depending on the physicochemical properties of the support. It has been reported that the reduction of certain nickel species is di fficult in supported nickel catalysts, with this di fficulty being proportional to the strength of the Ni–support interaction [24]. In general, the Ni–support interaction falls into three categories: i) an absence of interaction, which occurs when the support acts as a mere dispersing agent, ii) a weak interaction, usually associated with the presence of small nickel nanoparticles deposited on the support; and iii) a very strong interaction, involving the formation of a

new surface species (creation of new chemical bonds). The degree of interaction depends on the nickel charge (particle size) and the calcination temperature of the catalyst [24].

It has been reported that the reduction of unsupported NiO takes place at temperatures of around 220 ◦C [24], while the presence of metal–support interactions extend the nickel reduction process to higher temperatures. However, Petrik et al. associated the observed reduction peak at 200 ◦C to the reduction of Ni2O3 (ions with formal oxidation state higher than +2) to NiO [25]. In this sense, Carley et al. demonstrated, through XPS studies, the massive formation of surface Ni3<sup>+</sup> species after the calcination of the solid at temperatures above 300 ◦C [26]. Finally, the reduction peaks observed at higher temperatures (300–600 ◦C) were associated with the reduction of the previously formed NiO species or to the reduction of small nanoparticles interacting with the TiO2 support [25].

Based on the above considerations, the observed reduction peak at about 200 ◦C could be associated with either the reduction of bulk NiO or with the reduction of Ni<sup>+</sup><sup>3</sup> species (nickel ions with a formal oxidation state higher than +2) [25,26]. Given the low nickel loading (0.5%) as well as the small particle sizes reported by TEM (2 and 4 nm for Ni-0.5-DP and Ni-0.5-IMP, respectively), it is more feasible to associate the reduction peak at 200 ◦C with the reduction of Ni<sup>+</sup><sup>3</sup> species present in the catalyst. Reduction peaks observed at higher temperatures would be associated with small NiO particles interacting with the titania support. According to these results, the temperature chosen for catalyst reduction was set to 400 ◦C.

Band-gap energy values of the semiconductors were determined from UV-Vis spectra. The method for the determination of band gap values is shown in Figure S4 using the example of Ni-0.5-IMP-Red. As can be seen, the modification of the reference titania material (Evonik P25) by nickel incorporation resulted in a slight decrease in the band gap (Table 2), with the absorption being shifted to the visible spectrum (Figure 3, left)


**Table 2.** Band-gap energy values of the solids as determined by UV-Vis spectroscopy.

TEM micrographs of the different solids are shown in Figures 4–6, and the particle size distribution is shown in Figure S5. Ni particle sizes were determined using ImageJ software. The deposition–precipitation method resulted in particles with an average size of 2 nm, whereas the impregnation method led to more heterogeneously-distributed sizes (Figure S5), with the average particle size being 4–5 nm. Particle sizes did not vary significantly after the pre-reduction treatment.

Furthermore, the Ni particle size did not vary significantly after the first use. In the case of the utilization of UV light, there were no changes either after the second use. On the contrary, when solar irradiation was applied, the Ni particle size in the Ni-0.5-DP sample increased up to 5 nm (Figure 6).

The surface chemical composition of the solids was studied by XPS, and the main results are summarized in Table 3. As far as the Ti (2p3/2) signal is concerned, there were no significant changes after the incorporation of nickel, with the signal appearing at ca. 458.5 eV, which is a typical value for Ti4<sup>+</sup> in TiO2. Regarding the Ni 2p3/2 signal, it has been reported that binding energies at 852.6, 854.6, and 856.1 eV correspond to Ni0, Ni<sup>+</sup>2, and Ni<sup>+</sup>3, respectively [25,27]. As we have commented previously, based on XPS data, Carley et al. unequivocally demonstrated the existence of Ni<sup>+</sup><sup>3</sup> species in nanostructured solids calcined at temperatures above 300 ◦C [26].

**Figure 4.** TEM micrographs of the fresh (unused) Ni catalysts.

**Figure 5.** TEM micrographs of Ni-0.5-DP solid after the reaction under UV (**A**) or solar (**B**) irradiation.

**Figure 6.** TEM micrographs of Ni-0.5-DP solid after the second reutilization using UV (**A**) or solar (**B**) irradiation.


**Table 3.** Ni (2p3/2), Ti (2p3/2) and Cl (2p) binding energies (eV) as determined by XPS.

Moreover, Petrik et al., working with nanosized nickel oxides, found that binding energies at 855.3 and 856.7 eV were typical for Ni/TiO2 systems, with the signal at around 856 eV being preferably associated with Ni2O3 rather than with NiO, whose signal appeared at around 855 eV. The authors thus speculated the existence of Ti–Ni–O interactions that were reflected as Ni<sup>+</sup><sup>3</sup> species in the XPS spectra [25].

In this work, the XPS data associated with Ni (2p3/2) signals are presented in Table 3 and Figure S6. The spectra showed a signal at ca. 856 eV for fresh unreduced solids which was assigned to Ni<sup>+</sup><sup>3</sup> species, whereas, after the reduction treatment, the signal shifts to 855 eV as a result of the reduction of Ni<sup>+</sup><sup>3</sup> to Ni<sup>+</sup><sup>2</sup> species. This is in agreemen<sup>t</sup> with the reduction peak observed in the TPR profile at around 200 ◦C. Moreover, in the XPS analysis of the non-reduced catalysts used in a photo-reforming process (both UV and solar), the Ni (2p3/2) signal appears at ca. 855 eV, indicating that during the photocatalytic process, the in-situ reduction of Ni<sup>+</sup><sup>3</sup> to Ni<sup>+</sup><sup>2</sup> species takes place (Figure S6). No signal associated to Ni metal was detected in XPS profiles, even for the reduced solids, and so the catalyst reduction at 400 ◦C would not be strong enough to carry out the NiO reduction to Ni0, or else the hypothetically formed Ni<sup>0</sup> would re-oxidize in contact with air. In this sense, Ju et al. have already reported the absence of the Ni (0) peak at 852 eV after the reductive treatment of nickel-containing absorbents, were was associated with the difficulty of reducing NiO to metallic nickel [28].

It is also interesting to note that XPS revealed the presence of chlorine from the precursor in fresh unreduced solids (0.56 and 0.80 atomic % for Ni-0.5-IMP and Ni-0.5-DP, respectively). Such chlorine atoms were eliminated either during pre-reduction treatment as HCl or during the photocatalytic reaction.

H2 production from glycerol photo-reforming on fresh unreduced catalysts after 3 and 6 h of UV (A) or solar (B) irradiation are given in Figure 7. For the sake of comparison, results obtained for the reference material (Evonik P25) have also been included.

A first conclusion from Figure 7 is that hydrogen production using UV light is always higher than that achieved with solar light. This is hardly surprising, considering that the former irradiation source is more energetic. Moreover, Ni incorporation to TiO2 (irrespective of the method) led to an increase in hydrogen production. For instance, when UV light was used, hydrogen production increased from 166 micromole·g<sup>−</sup><sup>1</sup> (Evonik P25) up to 534 (Ni-0.5-DP) or 551 (Ni-0.5-IMP) after *t* = 6 h. Such an increase is even more significant when visible light was used. The observed shift of UV-Vis absorption to the visible region on the introduction of Ni could account for this effect.

A comparison of hydrogen production on pre-reduced (Figure 8) and untreated systems (Figure 7) allows us to conclude that catalyst pre-reduction treatment significantly increases catalytic activity (3–5 fold or 8–9 fold for experiments under UV or solar irradiation, respectively). As with the untreated systems, there are no large differences in their catalytic behavior depending on the synthesis procedure (DP or IMP). In the pre-reduced systems, those synthesized by DP have 39% greater activity, which could be due to the more homogeneous particle size distribution of Ni.

**Figure 7.** H2 production from glycerol photo-reforming on fresh, unreduced solids using UV (**A**) and solar (**B**) irradiation.

**Figure 8.** H2 production via glycerol photo-reforming on untreated and pre-reduced solids using UV (**A**) and solar (**B**) irradiation.

The highest hydrogen production values (2606 H2 micromole·g<sup>−</sup>1) corresponded to Ni-0.5-DP-Red for *t* = 6 h. This value is similar to that achieved in previous studies on 0.2% Pt [29], which is quite promising considering that Pt is ca. 2000 times more expensive than Ni.

Some authors, such as Bahruji et al. [30], have described the influence of the metal oxidation state on the photocatalytic process. As can be seen in Figure 9B, the electron transfer from titania to NiO is thermodinamically impeded. On the contrary, the pre-reduction of the solid (Figure 9A) results in electron transfer from titania to Ni(0) being favored, with the metal thus acting as an electron trap and preventing electron–hole recombination.

**Figure 9.** Energy levels of (**A**) TiO2/metal and (**B**) TiO2/NiO, adapted from Bahruji et al [30].

Furthermore, Caravaca et al. [31], studying hydrogen photo-production from sugars on Ni-based catalysts, observed an induction period (in their case, 60 min) required for in-situ reduction of NiO to Ni. After that period of time, the hydrogen production rate of both the untreated and pre-reduced solid was the same. According to these reports, we asume the in-situ reduction of our catalysts during reactions, and so the electron transfer from titania to Ni(0) is favored [30].

Another possible explanation for the observed better catalytic performance of pre-reduced solids as compared to untreated systems is the presence in the latter solids of surface chlorides (a well-known poison for metals arising from the precursor and evidenced by XPS analyses). As mentioned above, those chloride species were not observed in pre-reduced systems as they were eliminated as HCl during hydrogen pretreatment.

In order to cast further light on the effect of Ni oxidation states and the presence of chloride species on catalytic performance, some reutilisation studies were carried out on the Ni-0.5-DP catalyst both under UV and solar irradiation. Therefore, after 6 h irradiation, the solid was recovered by filtration, washed with methanol and acetone and dried at 110 ◦C. The catalyst was labelled as Ni-0.5-DP 1 using UV and Ni-0.5-DP 1 using solar, depending on the irradiation source. The solids were tested in another reaction, and the catalytic results are shown in Figure 10. As can be seen, hydrogen production dropped from 503 to 212 micromole per gram of catalyst after 6 h of UV irradiation, whereas no significant deactivation was observed under visible light. In any case, catalytic results were far below those achieved with fresh, pre-reduced catalysts.

**Figure 10.** Comparison of hydrogen photocatalytic production on untreated Ni-0.5-DP after the first and second use using UV light (**A**) or solar light (**B**). For the sake of comparison, results obtained for the pre-reduced solid (Ni-0.5-DP-Red) have also been included.

TEM micrographs (Figure 5) did not evidence any significant increase in Ni particle size with the first use. After the second use, the metal particle size only slightly increased for solar irradiation studies.

XPS experiments (Table 3) showed that surface chlorides had already been eliminated after the first use, and it is assumed that there had been an in-situ reduction of nickel species. Thus, neither the presence of chlorides nor the in-situ reduction of NiO can account for the possitive effect of pre-reduction treatment at 400 ◦C on catalytic performance. It is possible that such a pre-treatment induced a strong metal–support interaction (SMSI) [32,33] which somehow favored the subsequent catalytic performance. This SMSI would be favored on homogeneously-distributed particles achieved by the deposition–precipitation method, which could explain the above-mentioned better catalytic performance of Ni-0.5-DP-Red as compared to Ni-0.5-IMP-Red. Nevertheless, these hypothetical Ni–support interactions were not detected by XPS measurements and therefore require further studies.
