Qizhao Zhang a , Hao Zhang a , Bang Gu a , Qinghu Tang b , Qiue Cao a , Wenhao Fang a,*

a School of Chemical Science and Technology, National Demonstration Center for Experimental Chemistry and Chemical Engineering Education, Yunnan University, 2  North Cuihu Road, 650091 Kunming, China

b School of Chemistry and Chemical Engineering, Collaborative Innovation Center of Henan Province for Green Manufacturing of Fine Chemicals, Key Laboratory of  Green Chemical Media and Reactions, Ministry of Education, Henan Normal University, 453007 Xinxiang, China

1. Introduction

Chemical conversion of renewable and abundant lignocellulosic  biomass sources into sustainable fuels and chemicals has created a large  consensus [1,2]. 5-Hydroxymethylfurfural (HMF) as a typical model  compound for C6 carbohydrates is recently studied for conversion of  cellulose to the upgraded production of various building-block chemicals [3,4]. Of all, oxidative transformation of HMF is an important  valorization route [5,6]. 2,5-Diformylfuran (DFF), the symmetrical  furanic dialdehyde obtained from the selective oxidation of HMF, can  serve as an essential intermediate for synthesizing diamines and Schiff  bases [7,8]. Besides, DFF is viewed as an important monomer for furyl  polymers, pharmaceutical intermediates, antimicrobial, etc [9,10]. As is  known, the product distribution during oxidation of HMF can be  complicated. Hence, the high production of DFF from HMF strongly  relies on a suitable catalyst and some mild reaction conditions.

Photocatalysis enables to drive the selective oxidation of HMF to DFF  under mild reaction conditions [10–14]. In that case, only room or  near-room temperature and atmospheric O2 are used. This is a promising  alternative to thermal catalysis thanks to the photocatalysts with photogenerated holes and reactive oxygen species generated in the presence  of O2. However, very few effective catalysts have been reported in  literature. Although MnO2 nanorods [9], C3N4-supported Ti3C2 [15],  and lead-halide perovskite [16] are described as photocatalysts to give  ≥ 90% yield of DFF using atmospheric O2 upon ultraviolet or  visible-light irradiation, the catalytic efficiency in terms of productivity  to DFF is found to reach only a moderate value (≤30 mg gcatal. − 1 h− 1 ).  Notably, the complex preparation of catalysts and the compulsory use of  organic solvents (i.e., nitriles and benzotrifluoride) can badly hamper  the green footprint of the photocatalysis. Thereby, oxide catalysts that  are low-cost, easily-prepared and readily-tunable are highly demanded.  Moreover, using water instead of organic solvent as reaction medium is  greener but can be much challenging for oxide catalysts. Water as a  byproduct tends to block the forward reaction and water may also  aggravate metal leaching. Recently, a typical Bi2WO6 photocatalyst was  shown to afford 19% yield of DFF at a conversion of 26% using air and  water and driven by visible-light [17].

Titanium dioxide (TiO2) is regarded as one of the most common  oxide photocatalysts [18–20]. TiO2 has been extensively explored for  photocatalytic degradation of organic pollutants [20–22]. However,  very few applications of TiO2 on the photocatalytic selective oxidation  of biomass-derived compounds are reported, e.g., selective oxidation of HMF to DFF in water phase [23–25]. Earlier, S. Yurdakal et al. prepared  TiO2 with different crystalline phases (i.e., anatase, rutile, and brookite)  and investigated their performances in photocatalytic oxidation of HMF  to DFF in aqueous phase [23]. It was found that TiO2 showed a strong  oxidation power to quickly drive the reaction under ultraviolet irradiation at 35 ◦C. However, the used crystalline allotropic phases seemed  not to influence the global selectivity (about 21%) of the process. The  conversion could be increased but with formation of total oxidized  products (i.e., CO2 and H2O). Thereafter, to further explore the selective  photo-oxidation property of TiO2, I. Krivtsov et al. prepared  nitrogen-doped and oxygen-enriched TiO2. This N/TiO2 catalyst may  show a higher selectivity of DFF (26%) at 40% conversion of HMF at  25 ◦C in water using visible-light irradiation [24]. It was demonstrated  that the doping of nitrogen formed Ti-O-N bonds and thus rendered the  surface devoid of hydroxyl groups. This can suppress the generation of  •OH radicals on TiO2 surface and interfered the interaction of surface  sites with products, thus reducing the decomposition of DFF. Recently,  A. Allegri et al. prepared Au-Cu alloy particles loaded on porous  TiO2-SiO2 nanocomposites but this Au3Cu1/Ti15Si85 photocatalyst  afforded an unsatisfactory performance [25]. Only 21% conversion and  34% selectivity were obtained upon simulated sunlight irradiation at  30 ◦C. It was found that photogenerated holes and •OH radicals gave rise  to the decomposition of HMF and its derived oxidation products.

It is clear to see for TiO2-based photocatalysts that simultaneously  tuning conversion of HMF and selectivity of DFF is critical to achieve a  high yield of DFF. However, the stabilization of DFF at a high conversion  of HMF remains greatly challenging. As expected, dual-functional photocatalytic systems are believed to show synergistic oxidation performance for selective formation of DFF from HMF [26]. Particularly,  incorporating the extra active sites responsible for the selective conversion of an alcohol to an aldehyde to TiO2 can be probably an effective  strategy. For instance, recombining two semiconductors with different  properties may often bring about unexpected results, such as formation  of heterojunction or Schottky junction, which would provoke unique  visible-light response or photothermal effect to a catalyst. Typically, S.  Xie et al. investigated a series of traditional semiconductor catalysts for  photocatalytic oxidative coupling of methanol to ethylene glycol.  Interestingly, cuprous oxide (Cu2O) was shown to be highly selective to  oxidize methanol to formaldehyde under visible-light irradiation [27].  Recently, D. A. Giannakoudakis attempted a CuOx nanoclusters (<4 nm)  decorated TiO2 photocatalyst which afforded 30% selectivity of DFF at  98% conversion of HMF but in acetonitrile under ultraviolet irradiation  at 30 ◦C [28]. To the best of our knowledge, no efforts have been made to  utilize the catalytic property of TiO2 and Cu2O for synergistic oxidation  catalysis in aqueous phase.

Herein, this work reports a (Cu2O)x‖TiO2 heterojunction photocatalyst that is prepared by chemical reductive precipitation method.  The strong interaction between Cu2O and TiO2 semiconductors enables  synergistic control of the conversion of HMF and the selectivity of DFF.  Hence the (Cu2O)x‖TiO2 catalyst attains the optimum yield and productivity of DFF in aqueous phase under simulated sunlight irradiation,  in comparison with a series of typical catalysts. The prime objective is to  unravel the unique photocatalytic property of the Cu2O-TiO2 heterostructure and then to discuss the reaction mechanism in light of the  structure-activity relationship and the energy-band structure of the  catalyst.

2. Experimental

2.1. Chemicals and reagents

Anatase (99.9%), rutile (99.5%), copper(II) acetate monohydrate  (≥98%), glucose (99%) and Na2CO3 (99.8%) from Alfa Aesar, and P25  from Evonik Degussa were used for catalyst preparation. HMF (98%),  DFF (98%), HMFCA (98%), FFCA (98%) and FDCA (98%) from Ark  Pharm were used for catalytic reactions and quantitative analysis.

2.2. Catalyst preparation

(Cu2O)x‖TiO2 catalysts were prepared by the chemical reductive  precipitation method. Different Cu2O/TiO2 molar ratios (x = 0.08, 0.16,  0.5 and 1.5) and TiO2 crystalline phases (anatase, rutile, as well as P25)  were investigated. Unless it was specified, anatase was used as default.  As an example for (Cu2O)0.16‖TiO2, copper acetate (1.4 mmol) and  anatase (4.2 mmol) were mixed by 20 mL of deionized water and the  mixture was ultrasonicated for 5 min. Subsequently, 20 mL of glucose  solution (0.2 mol L− 1 ) was dropwise added under a magnetic stirring at  800 rpm and the mixture was stirred for 30 min. Following that, 0.2 mol  L− 1 of Na2CO3 solution was added dropwise to adjust the pH to 9. Afterwards, the mixture was heated to 60 ◦C and aged for 2 h under stirring. The obtained solids were filtered and washed with deionized water  and ethanol till the pH of filtrate reached neutral. Finally, the  (Cu2O)0.16‖TiO2 catalyst was dried under vacuum at 60 ◦C for 12 h.  Cu2O was prepared by the same method.

2.3. Photocatalytic reactions

Typically, HMF (0.1 mmol), catalyst (30 mg) and deionized water  (100 mL) were charged into a 150 mL quartz reactor, as illustrated in  Scheme S1. The reaction mixture was first ultrasonicated (40 kHz) for  15 min and then stirred at dark under 500 rpm for another 15 min. This  allowed a desorption equilibrium between catalyst and reactant. A Xe  lamp (λ: 350–780 nm, light intensity: 0.75 W cm− 2 ) CEL-HXF300-T3  from Beijing China Education Au-light Technology Co., Ltd. was used.  The lamp was fixed vertically about 6 cm above the liquid surface, and  the reactor was connected to a circulating water of 35 ◦C and an O2 flow  of 10 mL min− 1 . Continuous sampling (0.5 mL) was done every 30 min.  The quantitative analysis was conducted on an Agilent 1260 Infinity  HPLC armed with a photodiode array detector and a Shodex SH-1011  sugar column (8 mm, 300 mm, 6 µm) using a dilute H2SO4 solution  (5 mM) as mobile phase. Specific wavelength was set for the detector, i.  e., 285 nm for analyzing HMF, 290 nm for DFF and FFCA, and 260 nm  for HMFCA and FDCA, respectively. The quantification was based on an  external standard method using the standard solutions of reactant and  products at different concentrations. Each reaction was performed at  least twice to guarantee reliable data. Conversion of HMF, selectivity of  products, yield and productivity of DFF were calculated on the basis of  carbon balance.

Conv.(%) = nHMF,initial − nHMF,final nHMF,initial (1)  

Select.(%) = nproduct nHMF,initial − nHMF,final (2)  

Yield(%) = nDFF nHMF,initial − nHMF,final (3)  

Prod. ( mgDFFg− 1 catal. h− 1) = WDFF Wcatal. × reaction time (4) 

2.4. Characterization methods

X-ray diffraction (XRD) analysis was performed on a Bruker D8  Advance diffractometer with Cu Kα radiation and a beam voltage of  40 kV. The patterns were registered in the 2θ domain of 10–90◦ at a  screening rate of 0.1◦ s − 1 . Transmission electron microscopy (TEM) was  carried out on a field-emission JEOL-2100 F system with an acceleration  voltage of 200 kV. Energy dispersive spectroscopy (EDS) elemental  mapping was conducted on an X-MaxN 80 T IE250 at 200 KV. Photoluminescence spectroscopy (PL) was recorded on an Edinburgh Instruments FSL980 with 200–1700 nm steady-state spectrum. The  excited wavelength was fixed at 480 nm. X-ray photoelectron spectroscopy (XPS) and Auger electron spectroscopy (AES) were carried out on a Thermo Scientific Kα+ system equipped with Al Kα radiation under ultrahigh vacuum. The binding energy shift due to the surface charging  was adjusted with a reference of the C 1 s line at 284.8 eV. Ultravioletvisible diffuse reflectance spectrum (UV–vis DRS) was recorded on a  Shimadzu UV3600 apparatus. Low-temperature electron paramagnetic  resonance (EPR) spectroscopy was carried out at 100 K on a Bruker  A300 EPR spectrometer with X-band frequency of 9.4 GHz. Quantitative  calculation was based on the DPPH standard. Photoelectrochemical  (PEC) performance of the catalyst was analyzed on an AutoLab electrochemical workstation model no. PGSTAT302N using a standard  three-electrode cell. Indium tin oxide (ITO) glass deposited by the  catalyst was used as the working electrode, a saturated calomel electrode (SCE) was used as the reference electrode, and a Pt wire was used  as the counter electrode. To prepare the working electrode, 5 mg of  catalyst was dispersed into 0.5 mL of ethanol, and then 10 μL of Nafion®  solution was added and the mixture was sonicated for 2 h to make it  homogeneous. Afterwards 10 μL of the mixed droplets were coated on  the ITO glass with a controlled area of 0.25 cm2 , followed by drying at  room temperature for 30 min to form a film electrode. The photoelectrochemical test was conducted under a 300 W xenon lamp, and a  Na2SO4 solution (0.5 mol L− 1 , 100 mL, pH = 6.8, 25 ◦C) was used as the  electrolyte.  

3. Results and discussion

3.1. Formation of Cu2O‖TiO2 p-n heterojunction and synergistic  photocatalytic behavior  

The actual molar ratios of Cu to Ti were precisely measured by ICPMS. The data show almost the same ratios to the theoretical values  (Table S1), which reflects the effective preparation of the (Cu2O)x‖TiO2  catalysts by the chemical reductive precipitation method. The crystalline structure of catalyst was analyzed by XRD. As shown in Fig. 1A, well  crystallized Cu2O can be prepared from copper(II) acetate by the  chemical reductive precipitation method. The characteristic diffraction  peaks due to the (110), (111), (200), (220) and (311) planes of facecentered cubic Cu2O (JCPDS #05–0667) are detected. In addition,  from AES spectrum reported in Fig. S1, Cu1+ ions with a typical kinetic  energy at 916.5 eV are found to be the dominant species on surface [29].  For the (Cu2O)x‖TiO2 catalyst, Cu2O and TiO2 (anatase, JCPDS  #21–1272) phases are clearly observed to coexist. Moreover, the intensity of Cu2O peak apparently grows higher with increasing the Cu/Ti  ratio, but the position of TiO2 peak preserves unchanged. This indicates  that Cu2O and TiO2 can be combined to form a mixed oxide [30]. The  microstructure of the representative (Cu2O)0.16‖TiO2 catalyst was  further inspected by TEM and EDS. As displayed in Fig. 1B, this catalyst  shows relatively uniform nanoparticulated morphology (ca. 20 nm).  EDS elemental maps present well distributed and evenly overlapped  signals of Cu, Ti and O. In addition, as shown in the high-resolution TEM  image (Fig. 1C), the lattice fringe with d spacing of 0.234 nm and  0.350 nm can be attributed to the (111) plane of Cu2O [31] and the  (101) plane of TiO2 [32], respectively. Notably, these crystal planes are  closely contacted, which implies the formation of p-n heterojunction  between Cu2O and TiO2 through the combination of a p-type and an  n-type semiconductors.

Photocatalytic oxidation of HMF over the (Cu2O)x‖TiO2 catalyst was  carried out under Xe lamp (λ = 350–780 nm) in aqueous phase. As  shown in Fig. 2, as reference catalysts, anatase presents 100% conversion of HMF but no oxidation products are generated, whereas Cu2O  affords 100% selectivity of DFF but it is barely active. Previously, S.  Yurdakal et al. disclosed that TiO2 can effectively catalyze oxidation of  HMF, but selectivity of DFF was low. They observed that CO2 production  would accumulate during the oxidation process over the lab-made TiO2  catalysts, i.e., brookite and P25 [23]. Later, S. Xie et al. found that Cu2O  showed superior ability to selectively convert OH group to C– –O groups  in photo-oxidation reactions, e.g., Cu2O was greatly selective for  visible-light-driven oxidation of methanol to formaldehyde [27]. Interestingly, on the (Cu2O)x‖TiO2 catalyst the conversion of HMF continuously declines from 61% to 11% while the selectivity of DFF steadily  goes up to 85% from 29% with increasing the Cu2O/TiO2 ratio from 0.08  to 1.5, as plotted in Fig. 2. Meanwhile, FFCA from over-oxidation of DFF  is obtained and stabilizes at about 15% (in select.). Other products  mainly including CO2 decreases quickly and finally become zero. These  results show that Cu2O and TiO2 can decide in synergy the performance  of the (Cu2O)x‖TiO2 photocatalyst. As expected, a volcano-like relationship can be built between productivity of DFF and molar ratio of  Cu2O/TiO2. The (Cu2O)0.16‖TiO2 catalyst achieves the optimal productivity of 64.5 mg g− 1 h− 1 .

To understand the unique photocatalytic results, further characterizations were carried out. Usually the intensity of PL spectrum can reflect the recombination efficiency of photogenerated electron-hole,  and a low peak intensity indicates a low electron-hole recombination  rate. As shown in Fig. 3A, TiO2 displays the lowest fluorescence intensity  in PL spectroscopy, which can correspond to the most effective separation of photogenerated electrons and holes [33]. In contrast, Cu2O  presents the highest fluorescence signal. Obviously, the formation of  Cu2O‖TiO2 heterostructure enables to modify the pristine photoelectric  property of TiO2. These results can explain the huge difference in photocatalytic activity between TiO2 and Cu2O for HMF conversion.  Notably, the (Cu2O)0.16‖TiO2 catalyst shows the optimum separation of  photogenerated electrons and holes among all the dual metal catalysts.  Furthermore, photocurrent response caused by the separation and  diffusion of photogenerated electron-hole pairs under light irradiation  can provide some deeper information. One can see from Fig. 3B that all  the catalysts show photocurrent responses but only the (Cu2O)0.16‖TiO2  catalyst apparently presents the highest value, which can reach about  1.3 and 6 folds over those of TiO2 and Cu2O, respectively. This unique  phenomenon confirms the best separation and mobility of photogenerated electron-hole pairs on (Cu2O)0.16‖TiO2 [34], which may  probably explain the highest productivity of DFF on this catalyst.

To investigate the light utilization on a photocatalyst, UV–vis DRS  was performed. As displayed in Fig. 3C, TiO2 is evidenced to show a  strong absorption in UV-light region but no photoresponse in visiblelight region. Interestingly, the formation of Cu2O‖TiO2 heterojunction  brings about an intense visible-light absorption. Moreover, UV-light  response can be also enhanced in comparison with TiO2 and Cu2O.  Following that, the band-gap energy (Eg) of the photocatalyst was estimated using the Tauc method by plotting (F(R) hν) 1/2 against hν [24], as  displayed in Fig. 3D and Table S2. When the energy of photons is equal  to or greater than the band gap of a photocatalyst, electrons in the  ground state can be excited and move to the high-energy level. When the  absorption rapidly increases with decreasing wavelength, the corresponding energy of photons is equal to the band gap of the photocatalyst, hence electrons can move from the top of valence band to the  bottom of conduction band. As shown in Fig. 3C, D, on the pristine TiO2  and Cu2O semiconductors, the absorption is quickly elevated from 383  and 628 nm with the corresponding band gap of 3.25 and 1.98 eV,  respectively. On the (Cu2O)x‖TiO2 photocatalyst, one can clearly see the  absorption edges of both semiconductors, in which the absorption edge  of TiO2 rises from 383 to 395 nm with increasing Cu2O content, so that  the band gap becomes relatively narrower from 3.25 to 3.15 eV. On the  contrary, Cu2O shows a decreasing absorption edge (628–610 nm) and a  wider band gap (1.98–2.04 eV). It is well known that the wider the band  gap is, the stronger its redox ability can be [35]. Hence, the narrower the  band gap is, the stronger ability to utilize light can be obtained.  Therefore, the formation of Cu2O‖TiO2 p-n heterojunction by combining  the two semiconductors is highly important for achieving a high yield of  DFF via photocatalytic oxidation of HMF. That is because p-n heterojunction can enhance the light utilization ability of TiO2 and  simultaneously improve the redox ability of Cu2O.