Making H2 from light and biomass-derived alcohols: the outstanding activity of newly designed hierarchical MWCNT/Pd@TiO2 hybrid catalysts

A. Beltram a, M. Melchionna *a, T. Montini ab, L. Nasi c, P. Fornasiero *ab and M. Prato *ade
aDepartment of Chemical and Pharmaceutical Sciences, INSTM, University of Trieste, Via L. Giorgieri 1, 34127 Trieste, Italy. E-mail:
bICCOM-CNR Trieste, Via L. Giorgieri 1, 34127 Trieste, Italy
cCNR-IMEM Institute, Parco area delle Scienze 37/A, 43124 Parma, Italy
dNanobiotechnology Laboratory, CIC biomaGUNE, Paseo de Miramón 182, 20009 Donostia - San Sebastián, Spain
eIkerbasque, Basque Foundation for Science, 48013 Bilbao, Spain

Received 20th July 2016 , Accepted 15th August 2016

First published on 15th August 2016

Hydrogen evolution is among the most investigated catalytic processes given the importance of H2 from an industrial and an energy perspective. Achieving H2 production through green routes, such as water splitting or more realistically photoreforming of alcohols, is particularly desirable. In this work, we achieve a remarkable H2 productivity through photoreforming of either ethanol or glycerol as a sacrificial electron donor by employing a hybrid nanocatalyst where the properties of multi-walled carbon nanotubes (MWCNTs), Pd nanoparticles and crystalline TiO2 are optimally merged through appropriate engineering of the three components and an optimised synthetic protocol. Catalysts were very active both under UV (highest activity 25 mmol g−1 h−1) and simulated solar light (1.5 mmol h−1 g−1), as well as very stable. Critical to such high performance is the intimate contact of the three phases, each fulfilling a specific task synergistically with the other components.


H2 is a crucial molecule for many industrial processes including oil refining, hydrodesulfurization (HDS), hydrodenitrogenation (HDN), synthesis of NH3 and related fertilizers and many organic hydrogenation reactions in fine chemical production.1 In addition, H2 is indicated as one of the promising energy vectors of the future, even if in combination with a pool of different renewable fuels.2 From a sustainability perspective, utilization of H2-based energy, generated in a sustainable manner, implies a carbon-neutral footprint and therefore represents an extremely attractive option for a future switch to green methodologies. Surely, there are still many technical problems which are slowing down the implementation strategies, including the requirement for large infrastructure investments and safety issues related to H2 storage and distribution. Another highly interesting prospect is the in situ H2 photocatalytic production to harness the green synthesis of industrially relevant molecules, a recent example being the photo-driven synthesis of benzimidazole.3 In this complex scenario, development of efficient photocatalysts, in particular highly engineered nanocatalysts, for the generation of H2 has become a prominent field of research.4 Indeed, although there are several methods currently available for hydrogen production,5 particularly desirable are those strategies based on sustainable and environmentally friendly resources such as water and light.6–10 Pure photocatalytic water splitting, however, still suffers from a number of shortcomings that are not of easy resolution. A more realistic “clean” strategy relies on the utilization of biomass-derived species as sacrificial electron donors in the photocatalytic H2 production process.11–14

Ethanol and glycerol are two biomass-derived products, the former also known as bioethanol when sustainably obtained from next-generation biomasses, such as agriculture residues or lignocellulose, and the latter representing around 10 wt% byproduct of the biodiesel industry.15 Hence, the photoreforming of ethanol and glycerol catalyzed by heterogeneous catalysts opens a very attractive route to H2 production.16–18

As far as general heterogeneous photocatalysts are concerned, TiO2 still represents a benchmark material due to its high activity, low cost, large availability, low toxicity and stability to corrosion.19 However, its utilization still suffers from inherent limitations, above all the poor absorbance in the visible spectrum and therefore poor visible light photoactivity, as well as fast recombination of the photogenerated charge carriers (electrons and holes). The most popular method to improve hydrogen photo-catalytic production relies on the deposition onto the TiO2 surface of noble metal (Au, Pt, Pd, Ag) and base metal (Ni, Cu) nanoparticles, that are able to capture photoexcited electrons and retard the electron/hole recombination.20–22 There have been numerous reports on successful photoreforming of oxygenated compounds by TiO2 loaded with metal nanoparticles.11,23,24

One recent advance in the photocatalysis by TiO2-based catalysts is the combination of TiO2 with carbon nanotubes (CNTs) to form nanocarbon-inorganic hybrids with improved performances.25 CNTs have emerged as intriguing active supports in a range of catalytic applications due to their fascinating mechanical, electronic, optical and thermal properties.26,27 In particular, as far as photocatalytic processes are concerned, reports show that CNTs as active supports are able to scavenge away photoexcited electrons and retard charge recombination, therefore promoting reaction rates.28,29 Other possible mechanisms are however postulated, based for instance on a photosensitizing effect by the CNTs, which undergo the light-induced charge separation and inject the photoexcited electrons into the TiO2 conduction band30,31 or formation of intermediate states arising from Ti–O–C bondings such as those observed in carbon-doped TiO2.32,33 The effective result of the CNT–TiO2 hybridization, when appropriately achieved, is an increased photoactivity or a significant shift of the absorbance into the visible light, both important requisites for the further development of a TiO2-based material for H2 photo-production. Ahmmad et al. reported on alcohol photoreforming active catalysts composed of single-walled carbon nanotubes (SWCNTs) combined with TiO2, ascribing the improved activity to the electron sink effect generated by the SWCNT component.34 In another report, the main factor for the successful H2 production from triethanolamine solutions by multi-walled carbon nanotube (MWCNT)/TiO2 composites loaded with Pt nanoparticles was discussed in terms of the photosensitizing by the MWCNTs.35,36 A more recent advance in H2 production by MWCNT/TiO2 composites explored the loading with other metals such as Ir, Au, Pd as well as Pt.37 We have previously shown how a hierarchical system based on CNTs hybridized with a TiO2 shell incorporating Pd nanoparticles shows superior photocatalytic H2 production with respect to the conventionally separated components in a model reaction test of photoreforming of methanol. The synergy between the properties of the MWCNTs and the Pd@TiO2 structure is the key determinant for the good photocatalytic ability.38

Based on the above concept, in this work we have elaborated a new design for a hybrid nanomaterial incorporating functionalised MWCNTs, Pd nanoparticles and TiO2, which fully exploits the individual component's properties. As a result, we achieved high activity in photoreforming processes exploiting the use of industrially relevant biomass-derived alcohols, such as ethanol and glycerol. By careful tuning of the MWCNTs characteristics and by controlling the thermal formation of crystalline TiO2 around the MWCNT scaffold, we were able to prepare hybrid materials that can compete as novel benchmark catalysts in H2 photo-production. The catalyst is among the most active of those based on carbonaceous support. Remarkably, while the vast majority of photocatalytic systems for H2 evolution relies on the use of the generally most active Pt active phase, we were able to surpass H2 evolution rates by adopting the less rare and costly Pd, establishing a definite stride forward towards market feasibility. It is noteworthy that these novel catalysts can also generate H2 from ethanol and glycerol under simulated sunlight irradiation with excellent activity, thus further expanding the significance of these catalysts for possible solar driven H2 production.

Results and discussion

Synthetic procedures

The assembly of the nanocarbon-inorganic hybrid catalysts was carried out utilizing three building blocks: (a) benzoic acid-functionalised multi-walled carbon nanotubes (t-MWCNTs), (b) palladium nanoparticles functionalised with mercaptoundecanoic acid, and (c) titanium tetra-(n-butoxide). The synthetic scheme to access the final hybrid can be summarised in three phases (A, B and C in Fig. 1). The nanotubes building block was accessed via the radical addition of in situ formed diazonium salt of the benzoic acid (Fig. 1, Step A).39 As opposed to the oxidation treatment of CNTs, we chose this strategy for the modification of MWCNTs in order to have a more controlled functionalisation, endowing the nanotubes with COOH groups while minimising any extra damage of the polyaromatic framework by inserting additional oxygenated groups of no utility, such as epoxides or ketones. Freshly prepared Pd@TiO2 precursors were obtained via slow addition of Pd-MUA to a THF solution of Ti(n-OBu)4, causing the self-assembling of the Ti(n-OBu)4 around the Pd-MUA in a core–shell configuration (Fig. 1, Step B).40 The as-functionalised MWCNTs were then reacted with the Pd-MUA@Ti(n-OBu)4: ligand exchange between the alkoxide and the benzoic acid groups guarantees initial attachment of the inorganic nanoparticles to the t-MWCNTs sidewalls, with the t-MWCNTs then templating the growth of the oxide layer along the cylindrical scaffold. In contrast, a poorer interaction between MWCNTs and Pd-MUA@Ti(n-OBu)4 units was observed when unfunctionalised pristine nanotubes were employed, confirming that the attached benzoic acid groups serve as essential anchor points for immobilising the inorganic precursors. The relative amount of MWCNTs and Ti(n-OBu)4 was varied in order to have two nominal compositions with 10 and 20 wt% of nanotubes, while the % of Pd remains fixed at 1.5wt%. Final hydrolysis with H2O/THF affords the MWCNTs/Pd@TiO2: 10-CNTs/Pd@TiO2-fresh (10% MWCNTs), 20-CNTs/Pd@TiO2-fresh (20% MWCNTs) (Fig. 1, step C). As part of step C, the hybrid materials were subjected to calcination at 350 °C for 5 hours to give the corresponding calcined materials (10-CNTs/Pd@TiO2-calc, 20-CNTs/Pd@TiO2-calc). For reference, the Pd@TiO2 without MWCNTs was also prepared (Pd@TiO2-calc).
image file: c6gc01979j-f1.tif
Fig. 1 General Synthetic scheme of hierarchical CNTs/Pd@TiO2-fresh and CNTs/Pd@TiO2-calc. (A) Covalent functionalisation of the MWCNTs with benzoic acid molecules through radical addition of in situ made diazonium salt; (B) self-assembly of the Pd@TiO2 precursor; (C) hydrolysis to obtain the final hybrid, fresh and calcined.


TGA analyses under air confirm the successful organic functionalisation of the MWCNTs and the incorporation of the three building blocks in well-defined ratios in the final hybrids. The weight loss at ∼350 °C in the (Ph-COOH)-MWCNTs (about 3 wt%) is due to the removal of benzoic acid moieties on the nanotube sidewalls. TGA under air (Fig. 2, top) exhibits the weight losses due to the oxidation of residual organic ligands (mostly butanol) around 200 °C and of the MWCNTs around 500 °C, with the remaining weight given by the inorganic components. Consistently, the calcined samples are free from the first weight losses, as the organics have been removed during calcination prior to the analysis. The content of MWCNT was estimated as ∼9 and ∼14 wt% for 10-CNTs/Pd@TiO2-calc and 20-CNTs/Pd@TiO2-calc, respectively (Fig. 2, bottom).
image file: c6gc01979j-f2.tif
Fig. 2 TGA analysis under air of the p-MWCNTs, (Ph-COOH)-MWCNTs and fresh samples after drying (top) and of the samples after calcination at 350 °C in air (bottom).

Raman spectra of PhCOOH-MWCNTs show the characteristic bands of the CNTs: the disorder-induced D band (∼1300 cm−1) with its first (∼1600 cm−1, D′ band) and second-order related harmonics (∼2300 cm−1, 2D-band), and the G-band (∼1590 cm−1), due to the in-plane vibrational mode of the sp2 graphitic framework. After the hybridization with Pd@TiO2, the Raman spectra of the fresh samples (both 10-CNTs/Pd@TiO2-fresh and 20-CNTs/Pd@TiO2-fresh) suggest that TiO2 is predominantly present as an amorphous material, as no extra peak due to crystalline TiO2 is observed. Calcination induces the crystallization of the TiO2 shell, as the anatase fingerprint becomes evident in the Raman spectra, with an intense band at 146 cm−1 and minor bands at 198 cm−1, 395 cm−1, 513 cm−1 and 639 cm−1.41,42 In agreement, the sample Pd@TiO2-calc prepared without MWCNTs showed only the bands related to the anatase phase (Fig. 3, top).

image file: c6gc01979j-f3.tif
Fig. 3 Comparison of the Raman spectra of the PhCOOH-MWCNTs, CNT/Pd@TiO2-fresh, CNT/Pd@TiO2-calc (top) and powder XRD patterns of 10-, 20-CNT/Pd@TiO2-calc and Pd@TiO2-calc (bottom). CS = mean crystallite size.

In agreement with the Raman analysis, XRD patterns of the samples calcined at 350 °C show the typical reflections related to the anatase phase (Fig. 3, bottom). Notably, no reflections can clearly be related to Pd nanoparticles, reasonably because of the low metal loading and the small average size of Pd nanoparticles, as observed by HR-TEM. The mean crystallite size of TiO2 is not significantly affected by the formation of the hybrid materials with MWCNTs, indicating that the growth of anatase crystallites is mainly directed by the self-assembling around Pd nanoparticles.

TEM micrographs confirm the successful assembly of the hybrid materials, with the calcination treatment not affecting the intimate contact between the three phases, but only the crystallinity of the TiO2 layer. Thus, a clearly visible morphological difference between the fresh samples and the calcined samples is the diverse texture, with the latter appearing rougher, an indication of the crystallinity of the oxide layer (Fig. 4). A few small regions where the nanotubes have not been covered by oxide are also observed. A combined STEM and EDX analysis provides the compositional structure, confirming the incorporation of the three components (MWCNTs, Pd and TiO2) and their co-location (Fig. 5). The non-calcined samples bear almost only amorphous TiO2 while, as confirmed by Fast Fourier Transform (FFT) of a selected area, the calcined samples present a polycrystalline anatase phase for the TiO2 shell, whose particle size averages 10 nm (Fig. S1). The morphologies of 10-CNTs/Pd@TiO2 and 20-CNTs/Pd@TiO2, both fresh and calcined samples, appear similar. As expected, an average thicker layer of TiO2 in the former (150 nm for 10-CNTs/Pd@TiO2vs. 120 nm for 20-CNTs/Pd@TiO2) was measured via HAADF-STM analysis, resulting from the higher loading of oxide (Fig. S2). Despite being covered by the layer of TiO2, HR-TEM analysis of 20-CNTs/Pd@TiO2-calc evidences the crystalline Pd nanoparticles, whose size ranges between 3 and 5 nm (Fig. S3). The structure of the hybrids after catalysis remained completely unaltered as qualitatively observed by TEM.

image file: c6gc01979j-f4.tif
Fig. 4 Representative HAADF-STEM of 20-CNTs/Pd@TiO2-fresh (A) and 20-CNTs/Pd@TiO2-calc (B). Representative TEM of 20-CNTs/Pd@TiO2-fresh (C) and 20-CNTs/Pd@TiO2-calc (D). The morphology of catalysts 10-CNTs/Pd@TiO2-fresh and 10-CNTs/Pd@TiO2-calc are similar.

image file: c6gc01979j-f5.tif
Fig. 5 (A) Representative HAADF-STEM of a CNTs/Pd@TiO2-fresh; (B–E) EDX mapping showing the tight contact of the three phases: respectively palladium, oxygen, titanium, carbon; (F) superimposition of EDX of the C and Ti confirming their colocation.

The textural properties of the investigated materials have been analysed by N2 physisorption (Table 1). According to the IUPAC classification,43 all the samples present type IV physisorption isotherms, typical of mesoporous materials (Fig. 6, top). The contribution of micropores is negligible in all the investigated samples. Pristine and Ph-COOH functionalised MWCNTs present high surface areas and extended mesoporous textures, with large mesopores (10–100 nm) and high pore volumes. The textural properties of the hybrid materials calcined at 350 °C are different: pore sizes are much smaller (in the range of 2–4 nm) and, as a consequence, pore volumes are reduced while specific surface areas significantly increased (Fig. 6, bottom). Pd@TiO2-calc material presents textural properties comparable with that of the hybrid materials, confirming that the surface area and pore volume of the CNTs/Pd@TiO2 samples are mainly related with the inorganic component.

image file: c6gc01979j-f6.tif
Fig. 6 (top) N2 physisorption isotherms and (bottom) BJH pore size distributions for the various materials.
Table 1 Summary of the textural properties of the investigated materials
Sample Specific surface area (m2 g−1) CPV (mL g−1) D max (nm)
Pristine-MWCNT 142 0.634 3.4/22
(Ph-COOH)-MWCNT 117 0.741 37
Pd@TiO2 323 0.290 3.6
10% MWCNT 222 0.236 3.6
20% MWCNT 228 0.274 3.6

Accessibility of the Pd phase has been checked by H2 chemisorption. Samples were reduced at moderate temperature (see the Experimental section for details) to avoid significant deactivation of H2 adsorption on the Pd surface by strong metal–support interaction (SMSI) induced by the reduced TiO2,44,45 and the analyses were conducted at low temperature (around −90 °C using a solid/liquid acetone bath) to minimize H spillover to the reducible support.46 Under these conditions, H2 interaction with Pd nanoparticles results in the adsorption of H atoms on the surface and migration of H into the bulk of Pd forming Pd hydride (PdHX).47 The contribution of H adsorbed on the surface has been calculated by extrapolation of the linear part in the 10–20 mmHg range while the overall hydrogen consumption has been calculated after subtraction of the physisorbed H2 by linear extrapolation of the isotherms in the range of 200–400 mmHg (Fig. S4). The amount of hydrogen involved in PdHX formation has been calculated by the difference of the two quantities. Table 2 summarizes the main results of H2 chemisorption measurements. The accessibility and the dispersion of the Pd phase are significantly higher in the nanocarbon-inorganic hybrid materials with respect to the Pd@TiO2-calc material. Consistently, the formation of the hybrid structure allows a more efficient encapsulation and stabilization of Pd nanoparticles, avoiding their sintering. Notably, the amount of H absorbed in the formation of PdHX species is higher for the fully inorganic material. This result is in agreement with the dependence of the solubility of hydrogen in small Pd nanoparticles reported by Boudart and Hwang:48 the larger the Pd nanoparticles, the higher the solubility of H forming PdHX species.

Table 2 Summary of the metal textural properties of the investigated materials
Sample H surf/Pd Metallic surface area (m2 g−1) Apparent particle size (nm) X in PdHX
Pd@TiO2 0.20 1.4 5.5 0.77
10-CNT/Pd@TiO2 0.50 3.4 2.2 0.45
20-CNT/Pd@TiO2 0.52 3.5 2.1 0.33

Catalytic experiments

UV-Vis irradiation. Upon light irradiation, the PdO likely formed after the calcination treatment is reduced in situ leading to activation of the catalyst. As discussed in the introduction, the exact role of the carbon nanostructure in enhancing photocatalysis is still under debate. A plausible effect could be that of electron scavenging (Fig. 7). However, given the presence of bare parts of CNTs (see TEM), the photosensitizing could also be partially contributing, and therefore the CNTs could have multiple effects. The reaction generating H2 implies the reduction of H+ by the photogenerated electrons, that are injected from the TiO2 conduction band into the Pd nanoparticles, where they are accumulated and/or consumed for reactions. (Fig. 7):49
2H(ads)+ + 2e → H2

image file: c6gc01979j-f7.tif
Fig. 7 Scheme of a possible mechanism for H2 generation with concomitant scavenging of the photogenerated holes by the alcohol.

With this consideration in mind, in order for the reduction to be efficient, another important factor is therefore the rate of diffusion of the protons, which are produced on the TiO2 through the oxidation of the alcohol, to the Pd sites. All this implies a catalyst with very specific morphological characteristics. In this scenario, the MWCNTs would act as electron scavengers, retarding the charge recombination rates. Other possible roles for the MWCNTs, as explained in the introduction, are however plausible.

The benefits of the designed synthetic protocol have been first demonstrated by studying the photocatalytic H2 production from aqueous solutions containing methanol (Fig. S5). Despite not being the most sustainable choice in terms of sacrificial agent, the use of methanol allows the direct comparison of the activity of the present materials with the literature data (Table S1). Together with H2, the production of CO and CO2 has been observed (data not shown). Fresh catalysts show a moderate activity, due to the presence of amorphous/partially crystalline TiO2. After calcination, the H2 production strongly increases. A similar behaviour was already reported50 and can be related to the crystallization of the oxide shell, forming the nanocrystalline anatase with a high surface area and good accessibility of the Pd active phase. Calcination also results in a tighter contact between TiO2 and MWCNT and an increased number of heterojunctions which are key to the enhanced activity. Because of the positive effect of the thermal treatment on the performance of the present hybrid materials, only the calcined samples have been employed in the following photocatalytic studies for H2 production using more sustainable biomass-derived sacrificial agents, i.e. ethanol and glycerol.

Under UV-vis irradiation, the amount of produced H2 increases over time with all the investigated catalysts using both ethanol and glycerol aqueous solutions (Fig. 8), as already observed for catalytic systems based on metal modified TiO2.51–53

image file: c6gc01979j-f8.tif
Fig. 8 H2 production over time under UV-vis illumination using aqueous solution of ethanol (top) and glycerol (bottom) as the sacrificial donor. Activities are normalised by the catalyst surface areas reported in Table 1.

The H2 production reaches an outstanding value of 2.4 mmol mcat−2 for 20-CNTs/Pd@TiO2-calc after 24 h under UV-vis illumination, while the photocatalyst with a lower CNT content exhibits lower activity (2.0 mmol mcat−2). This is a clear indication of the beneficial effect of the carbon nanotube scaffold. By comparison, the activity of the reference catalyst Pd@TiO2-calc was only 0.5 mmol mcat−2. If reported by gram of catalysts (Fig. S5), activities are among the highest ever reported for catalysts based on titania and carbonaceous supports (Table S1), with values above 25 mmol g−1 h−1 with 20-CNTs/Pd@TiO2-calc and 20 mmol gcat−1 h−1 for 10-CNTs/Pd@TiO2-calc. The quantum efficiency (QE) at 365 nm of 10-CNTs/Pd@TiO2-calc and 20-CNTs/Pd@TiO2-calc, respectively 17% and 21%, confirm the promising performance of these hybrid materials.

Investigation of the reaction by-products reveals the complexity of the network involved in the photocatalytic H2 production. No O2 evolution was observed, implying that the water splitting was not significantly contributing to the photocatalytic H2 production under the present conditions. In fact, only a very small evolution of H2 (∼210 μmol g−1 h−1) was measured using pure water, while the addition of methanol lead to a considerable increase in H2 evolution rates (Fig. S8).

When ethanol is used as a sacrificial agent, the analysis of the gas phase evidences the presence of vapours of acetaldehyde and 1,1-diethoxyethane (its acetal with ethanol), CH4 and CO2 in equimolar amounts, together with ethane and traces of ethylene (Fig. 9 and Fig. S7). Importantly, we observed only very minor amounts of CO, this being an aspect of relevance for the application of photoreforming in fuel cells (Fig. 9). Moreover, the analysis of liquid solutions recovered after photocatalytic experiments showed that the major part of acetaldehyde and 1,1-diethoxyethane was accumulated in the solutions, together with minor amounts of acetic acid, 2,4,5-trimethyl-1,3-dioxolane, 3-hydroxy-2-butanone and 2,3-butanediol (Table S2).

image file: c6gc01979j-f9.tif
Fig. 9 Amounts of by-products in the gas phase accumulated in 24 hours with 10- and 20-CNT/Pd@TiO2 for the photoreforming of ethanol and glycerol.

The detection of large quantities of by-products confirms that, under the present experimental conditions, the reforming reaction is not complete. The photodehydrogenation of ethanol to acetaldehyde is the major process involved in H2 evolution. The stepwise oxidation of ethanol can proceed either through direct interaction with the photogenerated h+ or indirectly by interaction with ˙OH formed from the reaction H2Oads + h+ → ˙OHads + H+ads.54,55 Most likely, the two oxidation pathways are proceeding simultaneously. Oxidation of acetaldehyde to acetic acid is considered to be the rate determining step with TiO2 catalysts in alcohol photoreforming.56 Given the low adsorption capability of the acetaldehyde onto the TiO2, it can easily react with ethanol forming the acetal 1,1-diethoxyethane while the formation of traces of acetic acid is expected to occur through oxidation in solution by some radical species formed in situ (mainly ˙OH). The formation of CH4 and CO2 in equimolar amounts indicates that these gases are formed by direct decomposition of acetic acid or by decomposition of acetaldehyde to CH4 and CO, followed by oxidation of CO to CO2 through a “photocatalytic Water Gas Shift Reaction”. Together with methane and CO2, the presence of ethane could also arise through a radical mechanism, namely from the radical coupling of ˙CH3, or by hydrogenation of ethylene formed from ethanol on the acid sites of the oxide. The contribution of acetaldehyde photodecomposition was confirmed by performing a photocatalytic experiment using acetaldehyde as a sacrificial agent, added after 3 h of irradiation of the photocatalysts in pure water (Fig. S9). After acetaldehyde addition, H2 is produced in a major amount, together with CH4, CO2 and traces of ethane.

Among the other by-products accumulated in solution during ethanol photoreforming, 2,3-butanediol is probably formed through radical C–C coupling of two ethanol molecules. From this compound, 3-hydroxy-2-butanone is formed by dehydrogenation from one OH group while 2,4,5-trimethyl-1,3-dioxolane is formed by reaction of 2,3-butanediol with acetaldehyde.

In the case of glycerol, lower H2 production rates were observed for all the investigated catalysts (Fig. 8 and Fig. S6). Both 10- and 20-CNT/Pd@TiO2 displayed an activity normalised by the surface area of 0.6 mmol mcat−2, while that of Pd@TiO2-calc was only 0.3 mmol mcat−2. Analysis of by-products in the gas phase revealed that both CO2 and CO are present together with trace amounts of ethane, with their formation rate increasing over time (Fig. 9 and Fig. S10). Semi-quantitative GC/MS analysis of the solutions recovered after photocatalytic experiments evidenced the accumulation of non-volatile by-products. Hydroxyl acetaldehyde, 1-hydroxy-2-propanone and 1,3-dihydroxy-2-propanone were the most abundant, followed by formic acid and 2,3-dihydroxy-propanal and trace amounts of acetic acid.

Differently from the case of ethanol, the carbonylic compounds formed by the first dehydrogenation of glycerol – 2,3-dihydroxy-propanal and 1,3-dihydroxy-2-propanone – still contain OH groups, that allow their adsorption on the TiO2 surface and the competition with glycerol to be oxidised by the holes. The results clearly indicate that oxidation/degradation of 2,3-dihydroxy-propanal is much easier than that of 1,3-dihydroxy-2-propanone. CO and CO2 formation are usually related to the photocatalytic degradation of carbonylic and carboxylic intermediate compounds, formed by progressive dehydrogenation of the glycerol molecule, mainly through 2,3-dihydroxy-propanal and leading to smaller intermediate compounds (hydroxyl acetaldehyde, formic acid and acetic acid).52 On the other hand, 1-hydroxy-2-propanone is presumably formed on the surface of the titania via dehydration/hydrogenation from 1,3-dihydroxy-2-propanone, hardly oxidised/decomposed on the TiO2 surface by holes, or directly by dehydration of glycerol. The formation of added value organic by-products in both ethanol and glycerol photoreforming can lead to extra valorization of biomasses.18 However, this is only viable if the products can be easily separated or if the oxidation of the alcohol proceeds selectively. This aspect opens the doors to studies on contact times between the reactant and the catalyst, sparking exploration of the correct engineering of flow reactors.

Simulated solar irradiation. When irradiated with a simulated solar light, both 10-CNTs/Pd@TiO2-calc and 20-CNTs/Pd@TiO2-calc display an appreciable activity for the photoreforming of ethanol, although lower than the corresponding UV-irradiated samples. The drop in activity is somehow expected as the TiO2 can only absorb a small UV percentage of incident radiation. However, the detected activity provides a platform for these hybrid systems to be used with solar light (Fig. 10, top).
image file: c6gc01979j-f10.tif
Fig. 10 H2 production over time under simulated solar illumination using ethanol (top) and glycerol (bottom) as the sacrificial donor. Activities are reported normalised by the catalyst surface area as calculated in Table 1.

When EtOH is employed as a sacrificial electron donor, the catalyst 20-CNTs/Pd@TiO2-calc shows a H2 productivity of 120 μmol mcat−2 in 24 h, while for the 10-CNTs/Pd@TiO2-calc the productivity was slightly higher (140 μmol mcat−2). Interestingly, as compared to the UV experiments, the activity of the two catalysts with different CNT contents is inversed under simulated solar irradiation. Both catalysts are still more active than the reference catalyst Pd@TiO2-calc (87 μmol mcat−2). It is also worth noting that the two CNT-based catalysts undergo a progressive activation over the initial 4 hours before reaching a constant H2 evolution rate, while the CNT-free material shows stable activity from the beginning of the experiment. Remarkably, and in contrast to the UV experiments, the activity of all the catalysts with glycerol used as a sacrificial donor, does not decrease significantly (Fig. 10, bottom). In contrast, for 10-CNTs/Pd@TiO2-calc and Pd@TiO2-calc the H2 productivity is even slightly higher (149 μmol mcat−2 and 107 μmol mcat−2), while for 20-CNTs/Pd@TiO2-calc the productivity slightly diminishes to 95 μmol mcat−2. This suggests that the rate-determining step does not depend on the sacrificial electron donor or on the diffusion of the adsorbed protons from TiO2 to the Pd sites. Rather, it must be associated with the rate of formation of the charge carriers generated upon irradiation, which is considerably lower than that under UV. It is therefore acceptable to assume that once the fraction of electrons and holes are formed, they are immediately quenched by the (respectively) electron acceptors (protons) and electron donors (alcohol). As expected, when ethanol is used as a sacrificial donor, analysis of the liquid phase shows two main products, acetaldehyde and 1,1-diethoxyethane in amounts proportional to the produced H2 if compared to the UV experiments (Table S4). With glycerol we found, as the main liquid by-products, some of the expected first oxidation products, namely 1,3-dihydroxy 2-propanone, 1-hydroxy-2-propanone and hydroxyl-acetaldehyde, in analogy with the UV experiment (Table S5).


Functionalisation of the multi-walled carbon nanotubes (PhCOOH-MWCNTs). 100 mg of pristine MWCNT purchased from Nanoamor (20–30 nm diameter and 0.5–2 μm length) were dispersed in 100 mL of H2O by sonication for 20 minutes. Then p-amino-benzoic acid (3.3 equivalents with respect to carbon moles in the MWCNTs) was added and mixed by sonication for a further 10 minutes. The mixture was put under vigorous stirring and 2.25 mL of isopentyl nitrite were added just before starting the heating treatment at 80 °C for 6 h under reflux. Caution: diazonium salts should be handled with care since they can be explosive. The mixture was naturally cooled down at room temperature and collected by filtration on a 0.1 μm polytetrafluoroethylene Millipore membrane. Finally the material was washed with DMF, MeOH, H2O, EtOH and Et2O sonicating the solid for 5 minutes in each solvent and dried overnight at 80 °C.
Synthesis of MWCNT-Pd@TiO2 hybrids. The composite materials were prepared by employing 0, 10 and 20 mg of f-MWCNT and respectively labelled 10-CNTs/Pd@TiO2, 20-CNTs/Pd@TiO2and Pd@TiO2. The f-MWCNTs were dispersed in absolute ethanol (EtOH mL/f-MWCNT mg ratio: 2.5) by sonication for 30 minutes meanwhile a Pd-MUA THF solution (containing 1.5 mg of Pd) was slowly added to a THF solution of Ti(O-n-Bu)4 (containing 98.5, 88.5 and 78.5 mg of TiO2 respectively). Then the Pd@TiO2 precursor solution was slowly added under sonication to the f-MWCNT dispersion and the mixture was further sonicated for 30 minutes. Finally a 10% solution of H2O in EtOH (Ti(n-OBu)4/H2O molar ratio: 1/120) was dropped and the mixture was sonicated for 30 minutes. The materials were collected by filtration on a 0.45 μm polytetrafluoroethylene Millipore membrane, washed with ethanol and dried overnight at 80 °C. In order to eliminate the organic ligands and improve the crystallization of the TiO2 (see discussion below) the catalysts were subjected to calcination at 350 °C for 5 h (+3 °C min−1; −4.5 °C min−1).
Characterisation. X-ray diffraction (XRD) patterns were collected on a Philips X'Pert diffractometer using a monochromatized Cu Kα (λ = 0.154 nm) X-ray source in the range 20° < 2θ < 100°. The mean crystallite sizes of the TiO2 were calculated by applying the Scherrer equation to the (101) reflection of the anatase phase.

Thermogravimetric analysis (TGA) was performed on a TGA Q500 (TA Instruments) under air or nitrogen, equilibrating the temperature at 100 °C and heating at 10 °C min−1 up to 800 °C.

Raman spectra were recorded on a inVia Renishaw microspectrometer equipped with a Nd:YAG laser using an excitation wavelength of 532 nm. The dispersed samples were drop cast onto silicon wafers and analysed in 5 different points subsequently averaged.

TEM measurements were performed on a TEM Philips EM208, using an acceleration voltage of 100 kV. Samples were prepared by drop casting the dispersed particles onto a TEM grid (200 mesh, copper, carbon only). High resolution TEM (HRTEM) images were acquired on a JEOL 2200FS microscope operating at 200 kV, equipped with an Energy Dispersive Spectrometer (EDS), an in-column energy (Omega) filter, and a High-Angle Annular Dark-Field (HAADF) detector.

N2 physisorption isotherms were recorded at liquid nitrogen temperature using a Micrometrics ASAP 2020 automatic analyser. The samples were degassed under vacuum at 120 °C for 12 h before analysis.

H2 chemisorption was performed using a Micrometrics ASAP 2020C apparatus. A chemisorption stoichiometry H[thin space (1/6-em)]:[thin space (1/6-em)]M = 1[thin space (1/6-em)]:[thin space (1/6-em)]1 was assumed for the calculation of the exposed metal surface area. The materials have been pre-reduced in a flow of H2 (5%)/Ar (40 mL min−1) at 100 °C for 30 minutes and degassed at 250 °C for 5 hours. H2 chemisorption isotherms have been recorded in a 1–400 Torr pressure range at around −90 °C (solid/liquid acetone bath).

Catalytic tests and quantum efficiency measurement. It is known that the mechanism and therefore the H2 evolution depend on the sacrificial agent concentration.56,57 In the present experiments, ethanol and glycerol concentrations have been chosen in order to better represent aqueous solutions deriving from plants for the treatment of biomasses. Ethanol/water 50[thin space (1/6-em)]:[thin space (1/6-em)]50 solutions are representative of a solution coming from fermentation plants without being subjected to deep purification by distillation (a highly energy consuming industrial step). 1 M glycerol solution has a composition close to that deriving from biodiesel production plants. Moreover, the duration of each experiment has been optimized to minimize evaporation of the solution, in order to avoid significant variations of the alcohol/water ratios.

The materials were tested as photocatalysts for hydrogen production by photoreforming of ethanol (50% v/v) and glycerol (1 M) under two different light source apparatus: a Teflon-lined photoreactor illuminated with a Lot-Oriel Solar Simulator equipped with a 150 W Xe lamp and an Atmospheric Edge Filter with a cut-off at 300 nm and a Pyrex photoreactor illuminated with a 125 W medium pressure Hg lamp (model UV13F, Helios Italquartz, Italy). In a typical catalytic test 10 mg of the calcined material was first suspended in the photoreactor by sonication for 10 minutes in 60 mL of alcohol solution and subsequently purged from air with Ar flow of 15 mL min−1 for 40 minutes and thermostated at 20 °C. During the purge and the catalytic test the materials were magnetically stirred.

The on-line detection of volatile products was carried out using a gas chromatograph equipped with two analytical lines and a 10 way-two loop injection valve was employed for injection during on-line analysis of the gaseous products. In the former apparatus an Agilent 7890A Gas Chromatograph equipped with a Carboxen 1010 PLOT (Supelco, 30 m × 0.53 mm ID, 30 μm film) column followed by a Thermal Conductivity Detector (TCD) was used for gaseous products quantification using Ar as carrier and a DB-225 ms column (J&W, 60 m × 0.32 mm ID, 20 μm film) using He as carrier followed by a mass spectrometer (MS) HP 5975C was employed for the detection of the volatile organic compounds. In the latter apparatus an Agilent 6890N gas chromatograph equipped with a MolSieve 5 Å (Restek, 30 m × 0.53 mm ID) column followed by a Thermal Conductivity Detector (TCD) was used for gaseous products quantification using Ar as a carrier and a PoraPlot Q (Agilent, 30 m × 0.53 mm ID, 40 μm film) column using Ar as a carrier followed by a methanator and a Flame Ionization Detector (FID) was employed for the detection of the volatile organic compounds.

After the photocatalytic tests, the liquid phases were separated from the catalyst by filtration on a 0.45 μm PVDF Millipore membrane and subsequently analysed by GC/MS to detect the by-products accumulated. For a semi-quantitative analysis, 1-butanol and 1-hexanol were used as the internal standard in the liquid phases recovered from ethanol and glycerol tests, respectively.

Quantum efficiency (QE) was calculated with the following equation:

QE = 2 mol H2/absorbed photons
irradiating the sample with a 4 W Hg Pen-ray.


Development of new nanostructured catalysts for the sustainable production of H2 has become one of the most fertile fields of research in catalysis, given the central position of the H2 molecule in industry and energy-related applications. Our approach is based on the recognition that functional multicomponent materials are leveraged to the H2 production with unmatched efficiency, provided that each component is appropriately endowed with the required characteristics for the specific role. Here, we have engineered and accurately interlaced three nanoscale elements, namely (i) functionalised MWCNTs, (ii) functionalised (3–5 nm) Pd nanoparticles and (iii) crystalline TiO2 in a precise hierarchical order. The hybrid catalyst is able to drive the light-induced (both UV and solar simulated irradiation) H2 evolution from biomass-derived alcohols (ethanol and glycerol) with production rates which are amongst the best reported for carbon-supported catalysts, with the highest activity (under UV) setting at 2.4 mmol mcat−2 after 24 hours (rate per gram: 25 mmol g−1 h−1). By comparing with the state-of-the-art photocatalysts, the definite step forward arises from several advantages, such as: (1) the use of lower-powered irradiation lamps, (2) the use of Pd in lieu of the intrinsically more active, but more costly, Pt, and (3) the replacement of generally used methanol as a sacrificial electron donor with alcohols of higher value. All these advantages are realized without compromising the H2 productivity. The outstanding performance of the catalysts critically depends on the developed synthetic protocols, allowing a uniform and tight interfacing of the three components, which can therefore work in perfect synergy. The catalysts also exhibit excellent stability over at least 24 h. Finally, in the wider view, the anatomy of the synthetic photocatalyst to unveil the corresponding structure/reactivity relationships is expected to provide the key guidelines to evolve a new generation of carbon supported catalysts for artificial water splitting.


The research leading to these results has received funding from the University of Trieste (project FRA2015), INSTM, the Seventh Framework Programme [FP7/2007–2013] under grant agreement no. 310651 (SACS project). Dr A. Aneggi and Prof. A. Trovarelli (University of Udine, Italy) are kindly acknowledged for access to the XRD facility.

Notes and references

  1. N. Armaroli and V. Balzani, ChemSusChem, 2011, 4, 21 CrossRef CAS PubMed.
  2. S. Zinoviev, F. Müller-Langer, P. Das, N. Bertero, P. Fornasiero, M. Kaltschmitt, G. Centi and S. Miertus, ChemSusChem, 2010, 3, 1106 CrossRef CAS PubMed.
  3. Y. Shiraishi, Y. Sugano, S. Tanaka and T. Hirai, Angew. Chem., Int. Ed., 2010, 49, 1656 CrossRef CAS PubMed.
  4. M. Cargnello, T. Montini, S. Y. Smolin, J. B. Priebe, J. J. Delgado Jaèn, V. V. T. Doan-Nguyen, I. S. McKay, J. A. Schwalbe, M. M. Pohl, T. R. Gordon, Y. Lu, J. B. Baxter, A. Brückner, P. Fornasiero and C. B. Murray, Proc. Natl. Acad. Sci. U. S. A., 2016, 113, 3966 CrossRef CAS PubMed.
  5. K. Liu, C. Song and V. Subramani, Hydrogen and Syngas Production and Purification Technologies, Wiley, Hoboken, 2010 Search PubMed.
  6. A. Fujishima and K. Honda, Nature, 1972, 238, 37 CrossRef CAS PubMed.
  7. V. Balzani, A. Credi and M. Venturi, ChemSusChem, 2008, 1, 26 CrossRef CAS PubMed.
  8. H. B. Gray, Nat. Chem., 2009, 1, 7 CrossRef CAS PubMed.
  9. N. S. Lewis and D. G. Nocera, Proc. Natl. Acad. Sci. U. S. A., 2006, 103, 15729 CrossRef CAS PubMed.
  10. K. Jun, Y. S. Lee, T. Buonassisi and J. M. Jacobson, Angew. Chem., Int. Ed., 2012, 51, 423 CrossRef CAS PubMed.
  11. A. Gallo, T. Montini, M. Marelli, A. Minguzzi, V. Gombac, R. Psaro, P. Fornasiero and V. Dal Santo, ChemSusChem, 2012, 5, 1800 CrossRef CAS PubMed.
  12. D. Barreca, G. Carraro, V. Gombac, A. Gasparotto, C. Maccato, P. Fornasiero and E. Tondello, Adv. Funct. Mater., 2011, 21, 2611 CrossRef CAS.
  13. K. Shimura and H. Yoshida, Energy Environ. Sci., 2011, 4, 2467 CAS.
  14. X. Chen, S. Shen, L. Guo and S. S. Mao, Chem. Rev., 2010, 110, 6503 CrossRef CAS PubMed.
  15. S. Zinoviev, F. Müller-Langer, P. Das, N. Bertero, P. Fornasiero, M. Kaltschmitt, G. Centi and S. Miertus, ChemSusChem, 2010, 3, 1106 CrossRef CAS PubMed.
  16. T. Montini, M. Monai, A. Beltram, I. Romero-Ocaña and P. Fornasiero, Mater. Sci. Semicond. Process., 2016, 42, 122 CrossRef CAS.
  17. A. V. Puga, Coord. Chem. Rev., 2016, 315, 1 CrossRef CAS.
  18. M. Cargnello, A. Gasparotto, V. Gombac, T. Montini, D. Barreca and P. Fornasiero, Eur. J. Inorg. Chem., 2011, 28, 4309 CrossRef.
  19. O. Carp, C. L. Huisman and A. Reller, Prog. Solid State Chem., 2004, 32, 33 CrossRef CAS.
  20. A. Primo, A. Corma and H. Garcia, Phys. Chem. Chem. Phys., 2011, 13, 886 RSC.
  21. T. Montini, V. Gombac, L. Sordelli, J. J. Delgado, X. Chen, G. Adami and P. Fornasiero, ChemCatChem, 2011, 3, 574 CrossRef CAS.
  22. A. V. Korzhak, N. I. Ermokhina, A. L. Stroyuk, V. K. Bukhtiyarov, A. E. Raevskaya, V. I. Litvin, S. Ya. Kuchmiy, V. G. Ilyin and P. A. Manorik, J. Photochem. Photobiol., A, 2008, 198, 126 CrossRef CAS.
  23. M. Bowker, C. Morton, J. Kennedy, H. Bahru, J. Greves, W. Jones, P. R. Davies, C. Brookes, P. P. Wells and N. Dimitratos, J. Catal., 2014, 310, 10 CrossRef CAS.
  24. X. Chen, L. Liu, P. Y. Yu and S. S. Mao, Science, 2011, 331, 746 CrossRef CAS PubMed.
  25. D. Eder, Chem. Rev., 2010, 110, 1348 CrossRef CAS PubMed.
  26. M. Melchionna, M. Bonchio, F. Paolucci, M. Prato and P. Fornasiero, Top Curr. Chem., 2014, 348, 139 CrossRef PubMed.
  27. M. Melchionna, S. Marchesan, M. Prato and P. Fornasiero, Catal. Sci. Technol., 2015, 5, 3859 CAS.
  28. K. Woan, G. Pyrgiotakis and W. Sigmund, Adv. Mater., 2009, 21, 2233 CrossRef CAS.
  29. Y. Yu, J. C. Yu, J. G. Yu, Y. C. Kwok, Y. K. Che, J. C. Zhao, L. Ding, W. K. Ge and P. K. Wong, Appl. Catal., A, 2005, 289, 186 CrossRef CAS.
  30. W. Wang, P. Serp, P. Kalck and J. L. Faria, J. Mol. Catal. A: Chem., 2005, 235, 194 CrossRef CAS.
  31. W.-D. Zhang, B. Xu and L.-C. Jiang, J. Mater. Chem., 2010, 20, 6383 RSC.
  32. S. Sakthivel and H. Kisch, Angew. Chem., Int. Ed., 2003, 42, 4908 CrossRef CAS PubMed.
  33. L.-C. Chen, Y.-C. Ho, W.-S. Guo, C.-M. Huang and T.-C. Pan, Electrochim. Acta, 2009, 54, 3884 CrossRef CAS.
  34. B. Ahmmad, Y. Kusumoto, S. Somekawa and M. Ikeda, Catal. Commun., 2008, 9, 1410 CrossRef CAS.
  35. K. Dai, T. Peng, D. Ke and B. Wei, Nanotechnology, 2009, 20, 124603 Search PubMed.
  36. K. Dai, X. Zhang, K. Fan, P. Zeng and T. Peng, J. Nanomater., 2014, 2014, 694073 Search PubMed.
  37. C. G. Silva, M. J. Sampaio, R. R. N. Marques, L. A. Ferreira, P. B. Tavares, A. M. T. Silva and J. L. Faria, Appl. Catal., B, 2015, 178, 82 CrossRef CAS.
  38. M. Cargnello, M. Grzelczak, B. Rodrıguez-Gonzalez, Z. Syrgiannis, K. Bakhmutsky, V. La Parola, L. M. Liz-Marzan, R. J. Gorte, M. Prato and P. Fornasiero, J. Am. Chem. Soc., 2012, 134, 11760 CrossRef CAS PubMed.
  39. J. L. Bahr and J. M. Tour, Chem. Mater., 2001, 13, 3823 CrossRef CAS.
  40. K. Bakhmutsky, N. L. Wieder, M. Cargnello, B. Galloway, P. Fornasiero and R. J. Gorte, ChemSusChem, 2012, 5, 140 CrossRef CAS PubMed.
  41. T. Ohsaka, F. Izumi and Y. Fujiki, J. Raman Spectrosc., 1978, 7, 321 CrossRef.
  42. O. Frank, M. Zukalova, B. Laskova, J. Kürti, J. Koltai and L. Kavan, Phys. Chem. Chem. Phys., 2012, 14, 14567 RSC.
  43. K. S. W. Sing, D. H. Everett, R. A. W. Haul, L. Moscou, R. A. Pierotti, J. Rouquerol and T. Siemieniewska, Pure Appl. Chem., 1985, 57, 603 CrossRef CAS.
  44. J. H. Kang, E. W. Shin, W. J. Kim, J. D. Park and S. H. Moon, J. Catal., 2002, 208, 310 CrossRef CAS.
  45. Y. Li, B. Xu, Y. Fan, N. Feng, A. Qiu, J. M. J. He, H. Yang and Y. Chen, J. Mol. Catal. A: Chem., 2004, 216, 107 CrossRef CAS.
  46. J. M. Gatica, R. T. Baker, P. Fornasiero, S. Bernal, G. Blanco and J. Kašpar, J. Phys. Chem. B, 2000, 104, 4667 CrossRef CAS.
  47. J. M. Gatica, R. T. Baker, P. Fornasiero, S. Bernal and J. Kašpar, J. Phys. Chem. B, 2001, 105, 1191 CrossRef CAS.
  48. M. Boudart and H. S. Hwang, J. Catal., 1975, 39, 44 CrossRef CAS.
  49. D. W. Bahnemann, M. Hilgendorff and R. Memming, J. Phys. Chem. B, 1997, 101, 4265 CrossRef CAS.
  50. M. Melchionna, A. Beltram, T. Montini, M. Monai, L. Nasi, P. Fornasiero and M. Prato, Chem. Commun., 2016, 52, 764 RSC.
  51. G. L. Chiarello, M. H. Aguirre and E. Selli, J. Catal., 2010, 273, 182 CrossRef CAS.
  52. T. Montini, V. Gombac, L. Sordelli, J. J. Delgado, X. Chen, G. Adami and P. Fornasiero, ChemCatChem, 2010, 3, 574 CrossRef.
  53. C. Ampelli, R. Passalacqua, C. Genovese, S. Perathoner, G. Centi, T. Montini, V. Gombac, J. J. Delgado Jaen and P. Fornasiero, RSC Adv., 2013, 3, 21776 RSC.
  54. J. M. Kesselman, O. Weres, N. S. Lewis and M. R. Hoffmann, J. Phys. Chem. B, 1997, 101, 2637 CrossRef CAS.
  55. C. Y. Wang, R. Pagel, D. W. Bahnemann and J. K. Dohrmann, J. Phys. Chem. B, 2004, 108, 14082 CrossRef CAS.
  56. G. L. Chiarello, D. Ferri and E. Selli, J. Catal., 2011, 280, 168 CrossRef CAS.
  57. D. I. Kondarides, A. Patsoura and X. E. Verykios, J. Adv. Oxid. Technol., 2010, 13, 116 CAS.


Electronic supplementary information (ESI) available. See DOI: 10.1039/c6gc01979j

This journal is © The Royal Society of Chemistry 2017