1 Introduction

There is a growing interest in exploiting the “biorefinery” platform for the production of sustainable chemicals and fuels instead of century-old petroleum refineries [1]. However, traditional biomass biorefinery platform technologies, such as thermochemical (e.g., combustion, gasification, and pyrolysis) and bioprocesses (e.g., microbial fermentation and enzymatic digestion), are facing setbacks at current stage due to the carbon-intensive process and high capital cost [2]. The photoreforming of biomass and its derivatives into high-value chemicals and hydrogen (H2) is an emerging strategy for alleviating the present fossil fuel shortage crisis, global warming, and environmental pollution [3,4,5].

Glucose, the fundamental unit of cellulose, is the most widely investigated model substrates for biomass photoreforming in the past four decades [6]. It can be transferred to 5-hydroxymethylfural, arabinose, and formic acid during the photoreforming process [7,8,9]. Among them, arabinose has attracted much attention because of its wide applications in biological, chemical, and medical fields. Extraction of arabinose from some natural plants is a possible way, but the limitation is its high cost and the complex purification procedures [10]. Directly converting glucose to arabinose via the photoreforming by the controlled C1-C2 bonds scission is a prospective strategy using oxidizing active species such as holes (h+) [11]. Nevertheless, the low selectivity of arabinose as a result of the over-oxidation of glucose is still challenging [12]. Consequently, it is still urgent to design photocatalysts with controllable oxidation capacity and corresponding catalytic system to obtain high efficiency hydrogen (H2) and high selectivity chemicals simultaneously.

Titania (TiO2) has been widely investigated in biomass photoreforming because of its suitable conduction band and valence band position, as well as the corrosion resistance, nontoxic, and low cost [12,13,14]. However, the wide band gap and high photogenerated electron–hole pairs (e−-h+) recombination rate of conventional TiO2 usually lead to a low photoreforming efficiency [15]. Recently, Hu et al. proved that gold nanoparticles decorated three dimensionally ordered macropore TiO2 (3DOM TiO2-Au) can dramatically improve glucose conversion and arabinose selectivity [4]. And Surawut et al. also demonstrated that Ag-doped TiO2 nanofibers have excellent recycling ability with high-performance photoreforming glucose ability due to the localized surface plasmon resonance (LSPR) of Ag NPs [16]. Furthermore, Hynd et al. demonstrated that Ni−Pd bimetallic NPs deposited on TiO2 (Ni−Pd/TiO2) can minimize the cost of catalyst and increase the H2 production activity [17]. The modification of Au nanoparticles (NPs) with LSPR effect and high work function can significantly facilitate the separation of e−-h+ and visible light absorption ability in TiO2 [4, 18]. In addition, considering the potential of Ni NPs also with high work function in C–C bond cleavage (NiΦ = 5.3 eV) [19,20,21], depositing Ni-Au bimetal NPs not only improved the photoreforming activity, but also minimized the cost. This has been demonstrated to be an excellent option to increase the performance of photoreforming [17, 22]. Besides, it is worth mentioning that the effective contact between photocatalysts and glucose substrates can promote mass diffusion between these two, which is especially important for biomass photoreforming. Three-dimensional (3D) hierarchically structure favors facilitating the mass diffusion based on Murray’s law [23]. Depositing Ni-Au bimetal on 3D hierarchically urchin-like TiO2 is a promising strategy for minimizing costs as well as maximizing the photoreforming activity.

Here, Ni-Au bimetal NPs modified 3D hierarchically urchin-like TiO2 photocatalysts (NixAu0.5-x/TiO2) were designed and their application for H2 and arabinose co-generation by glucose photoreforming was demonstrated. NixAu0.5-x/TiO2 exhibited excellent H2 production accompanied by arabinose production activity. Under the optimized reaction condition, the H2 production rate of Ni0.05Au0.45/TiO2 (6391.86 μmol h−1 g−1) was 118.57, 30.78, and 1.65 times of pure TiO2 (53.91 μmol h−1 g−1), Ni0.5/TiO2 (207.56 μmol h−1 g−1), and Au0.5/TiO2 (3867.12 μmol h−1 g−1), respectively. Meanwhile, the glucose conversion rate and the corresponding arabinose selectivity over Ni0.05Au0.45/TiO2 were up to 95.00% and 36.54% after 4-h photoreforming. The photoreforming mechanism was also investigated by means of photoelectrochemistry and carrier dynamics.

2 Experimental

2.1 Reagents and instruments

The information of reagents and instruments were showed in supporting material.

2.2 Preparation of photocatalyst

2.2.1 Synthesis of 3D hierarchically urchin-like TiO2

The 3D hierarchically urchin-like TiO2 (abbreviated as TiO2) was synthesized according to our previous work through a one-step hydrothermal method. In details, acetone (3 mL) and titanium tetrachloride (0.5 mL) were added into a sample bottle with ultrasonic treatment for half an hour. Then, the sample bottle containing the above solution was placed in a Teflon-lined autoclave, and heated from an ambient temperature to 135 °C and kept for 48 h. Finally, the samples were collected by cooling the Teflon-lined autoclave to an ambient temperature, collecting the sediment by rinsing and centrifuging before drying it under 60 °C for 10 h.

2.2.2 Synthesis of AuxNi0.5−x/TiO2

A series of NixAu0.5−x/TiO2 (x = 0, 0.05, and 0.5 wt%) were synthesized via impregnation method. Firstly, 0.2 g TiO2 was suspended in 5.5 mL of Ni(NO3)2 (0.001 mol L−1) and 26.5 mL of HAuCl4 (0.001 mol L−1) solution with magnetic agitation for 1 h. Then, 0.1 mol L−1 L-Lysine was added into the above mixture. After that, 5 mL newly prepared NaBH4 solution was added dropwise into it with stirring for another 10 min. The suspension was experimentally aged under an ambient temperature for 24 h. Finally, the catalyst was gathered by centrifugation and drying. The formation process of NixAu0.5−x/TiO2 (x = 0, 0.05, 0.15, and 0.25) is schematically described in Scheme 1.

Scheme 1
scheme 1

Scheme presentation of the synthetic route of NixAu0.5-x/TiO2

Full size image

2.3 Photocatalytic activity test

Typically, 25 mg of the prepared photocatalyst, 31.8 mg of Na2CO3 and 0.05 g glucose were dispersed into 50 mL aqueous solution containing 25 mL water and 25 mL CH3CN. The suspension was vacuumed for 0.5 h to remove the existing gas before illumination. The temperature was controlled at 6 °C by circulating condensation. Then, the system was irradiated by a 300 W Xe-lamp under magnetically stirring. The evolved H2 gas was automatically quantitatively analyzed every 0.5 h during 4 h photoreforming process by gas chromatography with a thermal conductive detector. High-performance liquid chromatography was carried to determine liquid products. Glucose and arabinose were detected by using a 1260 ELSD detector with a mobile phase of CH3CN and aqueous solution containing 0.3% ammonium hydroxide in a volume ratio of 8:2 with a flow rate of 0.5 mL min−1. The column temperature was set at 40 °C.

3 Results and discussion

3.1 Photocatalyst structure characteristics

To investigate the morphology and microstructure of Ni0.05Au0.45/TiO2, FESEM and HRTEM were performed. FESEM images showed that 3D hierarchically urchin-like TiO2 microspheres with a diameter of around 6 μm were composed of one-dimensional (1D) nanorods (Fig. 1a, b). HRTEM image further verified this point (Fig. 1c). Some studies have shown that this special 3D hierarchical structure is conducive to mass transfer in the biomass reforming process [4, 23]. HRTEM image of Ni0.05Au0.45/TiO2 in Fig. 1d indicated that TiO2 still maintained complete 3D hierarchical urchin-like morphology after impregnation. HRTEM images of Ni0.05Au0.45/TiO2 exhibited that Au and Ni NPs around 5 nm were well deposited on TiO2 nanorods, and the shorter distance between Au and Ni NPs played a vital role for their coordination. Also, both Au and Ni NPs were in a tight contact with the TiO2 nanorods (Fig. 1e, f), which was favorable for the effective photogenerated carriers. The measured lattice spacings of 0.204 and 0.233 nm corresponded to Ni (111) facet and Au (111) facet, respectively [24, 25]. Furthermore, EDS mapping also presented that the Au and Ni NPs were homogeneously distributed throughout the TiO2 sample (Fig. 1g–j), and the EDS spectrum shown in Fig. 1k further verify the successful loading of Au and Ni NPs on TiO2.

Fig. 1
figure 1

a, b FESEM images of TiO2. c HRTEM image of TiO2. d, e, f TEM and HRTEM images of Ni0.05Au0.45/TiO2. g–k Elements mapping and EDS of Ti, O, Au, and Ni in Ni0.05Au0.45/TiO2

Full size image

As shown in Fig. 2a, the crystallinity and crystallographic phases of NixAu0.5−x/TiO2 were analyzed by pXRD. The diffraction peaks located at 2θ = 27.35°, 35.86°, 41.05°, 43.83°, 54.11°, 56.42°, 63.82°, and 69.57° could be attributed to (110), (101), (111), (210), (211), (220), and (301) lattice planes of rutile TiO2 [26]. Meanwhile, the observed weak diffraction peaks at 2θ = 62.63° and 68.73° were indexed to the anatase TiO2 (204) and (116) planes [15], respectively. The diffraction peaks (Fig. 2a) of Au NPs possessed four peaks at 2θ = 77.34°, 64.76°, 44.10°, and 38.18°corresponding to (311), (220), (200), and (111) planes, respectively [27]. It was worth noting that Au (200) plane was indistinguishable because of its overlapping with the (210) plane of rutile TiO2. Moreover, the diffraction peaks related to Ni NPs were not observed in the pXRD patterns of NixAu1-x/TiO2, which might be attributed to the high dispersity of Ni NPs on TiO2, as shown in EDS mapping (Fig. 1j) [28] or to the content below the detection limit [29]. Interestingly, the peak at 27.35° of TiO2 shifted to little lower angles slightly after the deposition of metals NPs (in inset in Fig. 2a). This kind of offset phenomenon could be explained by the integration of Ni-Au bimetal NPs with TiO2, particularly for TiO2 with e−-rich oxygen vacancy (Ov) sites [24]. The existence of lattice defects (Ov sites) was conductive to the interaction between the Ti lattice and metal NPs, so that the Ni-Au bimetal NPs could be better dispersed on the surface of TiO2 (Fig. 2i, j) [24, 30,31,32,33,34].

Fig. 2
figure 2

a pXRD patterns of NixAu0.5-x/TiO2 (inset displays magnified angles from 26° to 29°). b–f The high-resolution XPS spectra of survey, O 1 s, Ti 2p, Au 4f, and Ni 2p in Ni0.05Au0.45/TiO2

Full size image

XPS was performed to analyze the surface elemental compositions as well as the chemical states of Ni0.05Au0.45/TiO2 (Fig. 2c–f). Figure 2b revealed the existence of C, O, Ni, Ti, and Au elements. The C 1 s peak could be ascribed to adventitious carbon [35]. The Ti 2p image depicted in Fig. 2c exhibited four peaks of Ti4+ 2p1/2, Ti3+ 2p1/2, Ti4+ 2p3/2, and Ti3+ 2p3/2 located at 464.4, 463.5, 458.9, and 458.0 eV, respectively. As shown in Fig. 2d, the O1s peaks at 529.9, 530.4, and 531.6 eV corresponded to the O atoms in Ti–O, -OH, and Ov neighbors to Ti3+, respectively [15, 36,37,38]. The existence of Ov in TiO2 facilitated visible light absorption and charge separation simultaneously [39,40,41,42,43,44,45]. The peaks at 86.9 and 83.3 eV were assigned to Au 4f5/2 and Au 4f7/2of Au0 (Fig. 2e) [28, 46]. In the Ni 2p spectrum in Fig. 2f, a signal peak of Ni0 2p3/2 peak was identified at 852.9 eV, which revealed the existence of Ni0 [17]. The peaks at 863.4 and 855.7 eV belonged to Ni 2p1/2 and Ni 2p3/2 of Ni2+, respectively, which proved the presence of NiO in Ni0.05Au0.45/TiO2 [17]. Since the Ni species was very sensitive to oxygen, exposing Ni to air might cause the partially oxidization of Ni NPs into NiO clusters during the drying process. The above analysis showed that Au0, Ni0 and NiO coexist in Ni0.05Au0.45/TiO2, which is beneficial to the subsequent biomass photoreforming reaction.

The effective contact between photocatalysts and glucose substrates can promote mass diffusion, which was an important point in biomass photoreforming. Therefore, the N2 adsorption–desorption isotherms (at 77 K) were performed to determine the specific surface. Meanwhile, the pore-size distributions of NixAu0.5-x/TiO2 were conducted to determine the pore volumes. As displayed in Figs. S1-S2, all tested photocatalysts showed a type IV isotherms with H3 hysteresis loop, and the pore size ranged from 2 to 38 nm [15, 47]. Tables S1 and S2 show that the deposition of Ni-Au bimetals NPs had no significant effect on the specific surface area and pore volume of TiO2 itself, and they still retained perfect 3D hierarchically urchin-like morphology, as indicated by HRTEM in Fig. 1d.

3.2 Optical and photoelectrical properties

The optical characteristics of TiO2 and NixAu0.5-x/TiO2 were analyzed by UV–vis DRS (Fig. 3a, b). An absorption band edge of TiO2 appeared at about 424 nm because of the doping of Ti3+ in it. Ni0.5/TiO2 showed a flat absorption within the visible region because of the d-d transition in NiO [48]. Au0.5/TiO2 demonstrated an obvious absorption band edge redshift and an extra extended, and reinforced visible light absorption around 480–630 nm, which was caused by the localized surface plasmon resonance (LSPR) of Au NPs [17, 49]. The UV–vis DRS spectrum of Ni0.05Au0.45/TiO2 was consistent with that of Au0.5/TiO2. This indicated that no alloy was formed between the Au and Ni NPs, which were independent and relatively close to each other, as displayed in Fig. 1f. Moreover, the Kubelka–Munk method was adopted to calculate the equivalent bandgaps (Eg) by the following equations (Eqs. (1), (2), (3), and (4)) [50]. And the Eg of the Ni0.05Au0.45/TiO2 was calculated to be 2.86 eV (Fig. 3b).

Fig. 3
figure 3

a UV–vis DRS, b the corresponding bandgap energy, c PL spectra, d transient photocurrent response, and e Nyquist plot of TiO2 and NixAu0.5-x/TiO2. f Mott-Schottky plot of Ni0.05Au0.45/TiO2 (inset in Fig. 4f shows the energy band structure of Ni0.05Au0.45/TiO2)

Full size image
$$\frac{K}{S}=\frac{{\left(1-{R}_{\infty }\right)}^{2}}{2{R}_{\infty }}=F\left({R}_{\infty }\right)$$
(1)
$${R}_{\infty }=\frac{{R}_{\mathrm{sample}}}{{R}_{\mathrm{standard}}}$$
(2)
$$\alpha hv=B{\left(hv-{E}_{g}\right)}^{n}$$
(3)

Putting F(R∞) instead of α into Eq. (3) yields the form:

$${\left(F\left({R}_{\infty }\right)hv\right)}^{n}=B\left(hv-{E}_{g}\right)$$
(4)

where S is scattering coefficient, K is absorption coefficient, R∞ is the diffuse reflectance, F(R∞) is Kubelka–Munk function, α is linear absorption coefficient, v is light frequency, A is proportionality constant, and n is taken equal to the 1/2 for direct band gap materials [50].

The role of loading Au-Ni bimetal NPs in alleviating the recombination of photogenerated e−-h+ was surveyed by photoluminescence (PL) spectra. As presented in Fig. 3c and S4, all PL emission spectra of NixAu0.5-x/TiO2 exhibited similar positions to that of bare TiO2 at 469 nm, which was induced by the deep energy level emission of the typical structural defects (Ov) existing in the photocatalysts [51]. It was clearly observed that the PL emission intensity of Ni0.05Au0.45/TiO2 was dramatically decreased compared with other samples, which manifested that the introduction of Ni-Au bimetal NPs would highly improve the separation efficiency of the photoexcited charges carriers. It was speculated that the excited e− might migrate from the conduction band of TiO2 to the Ni-Au bimetal NPs due to the high work function of Ni and Au NPs [29], which could promote the transition and separation of the charge carriers, then distinctly prolonged the reactive lifetimes of the photogeneration carries. This was also beneficial to the enhancement of photoreforming activity.

To validate the influence of Ni-Au bimetal NPs on promoting the photoexcited charge carrier separation, transient photocurrent and EIS were conducted (Fig. 3d, e). All the catalysts showed reproducible and relatively stable photocurrent throughout the chopped on/off light cycles (Fig. 3d). Among them, Ni0.05Au0.45/TiO2 showed the highest photocurrent density, which proved that the deposition of Ni-Au bimetal NPs on TiO2 could significantly improve the photocurrent density of TiO2 [52,53,54,55]. The semi-circular signal of the EIS was employed to evaluate the charge transfer resistance (Fig. 3e) [56]. In normal conditions, the charge transfer resistance of the surface was equal to the diameter of the semi-circular part of the Nyquist diagram, a shorter radius corresponds to a smaller surface charge transfer resistance [57]. It was obvious that the Ni0.05Au0.45/TiO2 exhibited the lowest charge transfer resistance as well as the optimal photo-induced e−-h+ separation efficiency.

Afterwards, the Mott-Schottky (M-S) plot (Fig. 3f) was measured to determine the electronic band structure of the Ni0.05Au0.45/TiO2. The positive slope on the M-S plot of Ni0.05Au0.45/TiO2 revealed that Ni0.05Au0.45/TiO2 had a n-type semiconductor characteristic. The x-axis intercept of M-S plot revealed the flat band potential (Efb) of semiconductor electrode [58]. Figure 3f shows that the Efb of Ni0.05Au0.45/TiO2 was −0.62 eV (vs. Ag/AgCl). The flat band potential (Efb) was somewhere around 0.1 eV lower than the corresponding conduction band (ECB) for the most n-type semiconductors [59]. The corresponding ECB value of Ni0.05Au0.45/TiO2 was equivalent to −0.52 eV (vs. NHE). And as described above in Fig. 3b, the Eg value was calculated to be 2.86 eV of Ni0.05Au0.45/TiO2. Therefore, it could be determined that the EVB value Ni0.05Au0.45/TiO2 is to be +2.34 eV (vs. NHE) based on the equation of Eg = EVB – ECB. The electronic energy band structure of Ni0.05Au0.45/TiO2 was drawn in the inset of Fig. 3f.

3.3 H2 evolution from photoreforming glucose

The photoreforming H2 production performance of various photocatalysts (TiO2 and NixAu0.5-x/TiO2) was tested, with glucose as a model substrate. The effects of different conditions on the photocatalytic H2 production by Ni0.05Au0.45/TiO2 were first investigated (Fig. 4a, b and S3). It was clear that the H2 release rate was the highest when the volume ratio of CH3CN and H2O was 1:1 (CH3CN 25 mL, H2O 25 mL). On this basis, the addition of Na2CO3 further improved the H2 production (Fig. 4b). This might be because the addition of CH3CN made the resulting liquid product more conducive to separation and did not cover the catalyst active sites. Adding Na2CO3 (soluble base) highly improved the glucose photo-oxidation efficiency [60]. All subsequent photocatalytic reactions were proceeded under the optimal condition. As shown in Fig. 4c, the H2 accumulation of photocatalyst increased with time from the glucose photoreforming without any trend of slowing down after 4 h reaction. The calculated H2 release rate of TiO2 was approximately 53.91 μmol h−1 g−1 (Fig. 4d). The H2 generation efficiency of TiO2 could be attributed to the Ov introduced by Ti3+ doping and the sufficient contact between the photocatalyst and the glucose substrate due to the unique 3D hierarchical urchin-like morphology without any co-catalyst. However, the H2 production efficiency of pristine TiO2 was far lower than any of NixAu0.5-x/TiO2 (x = 0, 0.05, and 0.5), especially for Ni0.05Au0.45/TiO2. The H2 production activity of Ni0.05Au0.45/TiO2 could reach 6391.86 μmol h−1 g−1, which was 118.57, 30.78, and 1.65 times than that of pure TiO2 (53.91 μmol h−1 g−1), Ni0.5/TiO2 (207.56 μmol h−1 g−1), and Au0.5/TiO2 (3867.12 μmol h−1 g−1), respectively. Such excellent H2 evolution activity of Ni0.05Au0.45/TiO2 was not only related to the properties of TiO2 itself (the Ov and 3D hierarchical urchin-like morphology), but also more attributable to the modification of Ni and Au NPs with the synergetic effect. Firstly, Au NPs LSPR greatly promotes the light absorption by the catalyst, resulting in excited photogenerated e−-h+ for photoreforming reaction [55, 61]. Secondly, a Schottky contact could be formed between TiO2 and Ni NPs and between TiO2 and Au NPs with a high work function (AuΦ = 5.3–5.6 eV, NiΦ = 5.3 eV), where e− could be transferred from the CB of TiO2 to Ni and Au NPs for H+ reduction, but not back [17, 62].

Fig. 4
figure 4

a The H2 evolution amount of the Ni0.05Au0.45/TiO2 in solvent with different ratios of CH3CN and H2O. b The H2 evolution amount of the Ni0.05Au0.45/TiO2 in different Na2CO3 concentrations (the ratio of CH3CN and H2O was 1:1). c, d The H2 evolution from glucose photoreforming for different catalysts. e Long time recycling test of the optimal system for H2 evolution from glucose photoreforming. f The H2 evolution rate of the optimal system from different biomass derivatives. The reaction condition of c, d, e, and f: 25 mg catalyst, 25 mL H2O, 25 mL CNCH3, 6 mmol L−1 Na2CO3, 50 mg glucose, 300 W Xenon lamp

Full size image

The stability and versatility also matter much for a photoreforming system. As shown in Fig. 4c, the H2 production efficiency slightly decreased after 5 cycles, which might be owing to the Ni and Au NPs partially dislodged during the washing process and the fracture of the 1D nanorods in TiO2. The versatility of the system was carried out by photoreforming different biomass derivatives (e.g., cellobiose, glucose, α-cellulose, xylose, galactose, and fructose) under the same conditions (Fig. 4f). It could be clearly seen that the monosaccharides with simpler structures (e.g., glucose, galactose, fructose) exhibited a higher H2 production rate than other saccharides (such as saccharose, xylan, and α-Cellulose). It was worth mentioning that Ni0.05Au0.45/TiO2 showed relatively good H2 generation activity even for high crystallinity α-cellulose without any pretreatment under weak alkaline conditions (6 mmol L−1 Na2CO3), which demonstrated the potential of NixAu0.5-x/TiO2 in the photoreforming field. The H2 evolution performance of Ni0.05Au0.45/TiO2 was comparable to other TiO2-based catalysts reported in the literatures (Fig. 6a and Table S2).

3.4 Photoreforming for arabinose production

What’s charming of the photoreforming reaction is that the value-added chemicals can be generated, while producing a clean fuel H2 under the solar radiation [55]. Here, the liquid phase products of TiO2 and NixAu0.5-x/TiO2 and control system were assessed, and the dominant liquid phase product was arabinose within the detection limit of high-performance liquid chromatography (HPLC, Agilent 1290 Infinity II). The glucose conversion ratio of isolated TiO2 was determined to be around 9.1% and the arabinose selectivity was 23.46% after 4-h photoreforming reaction (Fig. 5a–c). As discussed before, the introduction of Ov triggered by Ti3+ doping and the sufficient contact between the photocatalyst and the glucose substrate because of the unique 3D hierarchical urchin-like morphology in TiO2 is pivotal in arabinose production by photoreforming [55, 61]. As expected, the Ni0.05Au0.45/TiO2 demonstrated a high glucose conversion of 95.08% after 4-h photoreforming reaction, while Au0.5/TiO2 and Ni0.5/TiO2 gave a glucose conversion of 87.15%, and 14.25%, respectively (Fig. 5a). The arabinose selectivity (Fig. 5b) after photoreforming of these catalysts also maintained the same trend, that was, Ni0.05Au0.45/TiO2 (36.54%), Au0.5/TiO2 (34.56%), Ni0.5/TiO2 (21.86%). These results further confirmed the synergistic effect of Au and Ni bimetallic NPs in the photoreforming process. In addition to the LSPR of Au NPs and the Schottky contact between TiO2 and Au/Ni NPs, we speculated that the ability of Ni NPs in C–C bonds cleavage in glucose also played a vital role in the photoreforming [21]. Moreover, the arabinose concentration for Ni0.05Au0.45/TiO2 system after 4 h photoreforming could be as high as 347.45 mg L−1, which was 1.15, 11.15, and 16.27 times that of Au0.5/TiO2 (301.25 mg L−1), Ni0.5/TiO2 (31.15 mg L−1) and TiO2 (21.35 mg L−1), respectively (Fig. 5c).

Fig. 5
figure 5

a Glucose conversion, b arabinose generation, and c arabinose selectively of different catalysts; d long time recycling test of the optimal system for glucose conversion. Reaction condition: 25 mg catalyst, 25 mL H2O, 25 mL CNCH3, 6 mmol L−1 Na2CO3, 50 mg glucose, 300 W Xenon lamp

Full size image

Furthermore, glucose conversion of Ni0.05Au0.45/TiO2 in cyclic experiments was studied (Fig. 5d). It could be found that the glucose conversion was maintained above 85% after 4 cycles of measurement, which proved the good cycling stability of photocatalyst. And the TEM images and pXRD patterns of Ni0.05Au0.45/TiO2 before and after 4 h photoreforming was shown in Fig. 6b, c, which proved the morphology and structural stability of Ni0.05Au0.45/TiO2.

Fig. 6
figure 6

a The comparison of H2 generation activity for the Ni0.05Au0.45/TiO2 and some reported TiO2-based photocatalysts (all details were described in Table S2 in the Supporting Information). b HRTEM of Ni0.05Au0.45/TiO2 after photoreforming. c PXRD of Ni0.05Au0.45/TiO2 fresh and after photoreforming

Full size image

3.5 Mechanism investigation

To better understand the glucose photoreforming mechanism for the enhanced coproduction of H2 and arabinose over Ni0.05Au0.45/TiO2, the detection of reactive species (•O2−, •OH, 1O2, and h+) was executed by electron spin resonance (ESR). 5,5′-Dimethyl-1-pyrroline N-oxide (DMPO), 2, 2, 6, 6-tetramethylpiperidine (TEMP), and 2, 2, 6, 6-tetramethylpiperidinooxy (TEMPO) were used as the spin traps for •O2− and •OH, 1O2, and h+, respectively. As shown in Fig. 7a, b, no ESR signal of •O2− and •OH was observed in the system under dark conditions. When the optimal system was under illumination, both •O2− and •OH were generated and they reacted with DMPO to present the typical ESR signals. Moreover, the ESR signals were enhanced gradually as the irradiation time increases. The change trend of 1O2 was consistent with the •O2− and •OH. Three typical signals of 1O2 with 1:1:1 relative intensity appeared with the extension of irradiation time (Fig. 7c). As for h+ detection, it was observed from Fig. 7d, the TEMPO characteristic peaks with a relative intensity of 1:1:1 were observed under dark. Once the system was exposed to light, TEMPO would combine with e− produced by the photocatalyst to form TEMPOH, the ESR signal of TEMPO was weakened. More e− were produced in the system, the lower the signal peak intensity would be. Since e− and h+ were generated in pairs during the photoreforming process, the ESR signal of the h+ was consistent with that of e−.

Fig. 7
figure 7

DMPO ESR spin-labeling for a •O2− and b •OH; c TEMP ESR for spin-labeling 1O2; d TEMPO ESR spin-labeling for h+; e glucose convention rare; f H2 production

Full size image

The main oxidative species in glucose photoreforming were studied by quenching experiments. Isopropanol (IPA), tryptophan (Trp), KI, and benzoquinone (BQ) were applied in the optimum system as sacrificial agents to consume •OH, 1O2, h+, and •O2−, respectively. As observed in Fig. 7e, the glucose conversion was inhibited in the presence of IPA (5.78%), Trp (6.68%), KI (17.50%), and BQ (60.75%), which indicated that the effect of •OH, 1O2, h+, and •O2− on facilitating the photoreforming reaction was in turn improved. According to the band structure analysis of Ni0.05Au0.45/TiO2 in Sect. 3.2, the EVB of Ni0.05Au0.45/TiO2 (+2.34 V vs. NHE) was more positive than OH−/•OH (1.90 eV vs. NHE), so the •OH was produced by this route. Meanwhile, the ECB of Ni0.05Au0.45/TiO2 (−0.52 V vs. NHE) was more negative than O2/•O2− (Eθ = −0.69 eV vs. NHE), •O2− could be obtained. Furthermore, the 1O2 could be produced by oxidizing h+. This was further validated the ESR analysis. Figure 7f reveals the H2 production activity under different trapping agents; the result was coincident with Fig. 7e.

A plausible explanation for the significant increase in H2 and arabinose co-generation and glucose conversion over Ni0.05Au0.45/TiO2 was proposed based on our results and some previous works [55, 63, 64] (Fig. 8). Under visible irradiation, the h+-e− was produced by TiO2, and then the e− was transferred from CB of TiO2 to Au and Ni NPs for the H2 production by H+ reduction. In the meanwhile, glucose was oxidized to produce gluconic acid by •O2− [4]. Gluconic acid was subsequently decarboxylated through C1-C2 α-scissions due to the promoting effect of Ni NPs on the C–C cleavage, resulting in the formation of arabinose and formic acid [61]. But no formic acid was detected, which possibly because it was dehydrated in the subsequent reaction, and CO was produced [12].

Fig. 8
figure 8

The illustration of H2 and arabinose production by glucose photoreforming over Ni0.05Au0.45/TiO2

Full size image

4 Conclusions

In summary, a high-efficiency photocatalyst NixAu1-x/TiO2 has been successfully synthesized for H2 and arabinose co-production by glucose photoreforming. Au NPs with localized surface plasmon resonance (LSPR) effect and high work function significantly facilitated the separation of photogenerated carriers; the transition metal Ni NPs with low coat and high work function (NiΦ = 5.3 eV) promoted C–C bond cleavage. The NixAu0.5-x/TiO2 obtained by one-step impregnation showed the prominent photoreforming activity by fully exerting the synergistic effect of Ni and Au NPs. Under optimal conditions, the H2 production rate of 6391.86 μmol h−1 g−1, 95.00% glucose conversion, and 36.54% arabinose selectivity could be achieved on Ni0.05Au0.45/TiO2. The enhanced photoreforming activity could be attributed to the synergistic effect of Ni and Au NPs, including the localized surface plasmon resonance (LSPR) and Schottky junction of Au nanoparticles and the promoting effect of Ni particles on C–C cleavage in glucose, as well as the glucose accessibility caused by the 3D hierarchical urchin-like TiO2. The photoreforming mechanism showed that •O2− plays an important role in the glucose oxidation. Glucose was converted to gluconic acid by •O2−. Gluconic acid was subsequently decarboxylated through C1-C2 α-scissions, resulting in the formation of arabinose and formic acid. This work demonstrates a feasible way to sustainable coproduction of H2 and chemicals, which would be a very robust alternative in biomass photoreforming systems.