Pd supported on defective TiO2 polymorphic mixtures: Effect of metal-support interactions upon glycerol selective oxidation

Results in Engineering 16 (2022) 100737 Available online 31 October 2022 2590-1230/© 2022 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/bync-nd/4.0/). Pd supported on defective TiO 2 polymorphic mixtures: Effect of metal-support interactions upon glycerol selective oxidation Matías Gast´ on Rinaudo a , * , Ana María Beltr´ an b , Asunci´ on Fern´ andez c , Luis Eduardo Cadús a , Maria Roxana Morales a a Instituto de Investigaciones en Tecnología Química (INTEQUI-CONICET), Universidad Nacional de San Luis (UNSL), Facultad de Química Bioquímica y Farmacia, Almirante Brown 1455, Capital, 5700, San Luis, Argentina b Departamento de Ingeniería y Ciencia de los Materiales y del Transporte, Escuela Polit´ ecnica Superior, Universidad de Sevilla, Virgen de ´ Africa 7, 41011, Sevilla, Spain c Instituto de Ciencia de Materiales de Sevilla, (CSIC-Univ. Sevilla), Avda. Am´ erico Vespucio 49, 41092, Sevilla, Spain ARTICLE INFO Keywords: Titanium oxide polymorphs Glycerol selective oxidation Palladium-based catalyst Metal-support interaction Mechanochemical activation ABSTRACT Palladium catalysts supported on defective mixes of anatase, TiO 2 (II) and rutile crystalline phases, previously obtained by high-energy ball milling, were synthesized and tested for glycerol selective oxidation. A deep characterization of these unusual materials was carried out to elucidate catalytic and physicochemical features. Electron density transfer from support to metal or vice versa, depending on the polymorphs present, could not only alter palladium particle sizes and its surface oxidation state but also reducibility and oxygen mobility of catalysts. Furthermore, acid-base properties achieved also influenced catalytic activity under mild conditions of liquid-phase glycerol oxidation. A conversion of 94% and a selectivity to glyceric and lactic acids of 48% and 22% respectively were obtained for the Pd catalyst supported on mechanochemically activated anatase. The presence of several polymorphs in a metal oxide support could therefore benefit or handicap catalytic cycle for a particular reaction. Metal-support interactions play a key role in heterogenous catalysts and thus the rational design of supports comes on the scene. 1. Introduction With the actual trend to biomass-based fuels, valorization of byproducts obtained in these processes becomes an interesting alternative to promote green and sustainable technologies [1,2]. In case of biodiesel, direct substitute of diesel, the increasing demand observed in the last years is also accompanied by raw glycerol as a surplus by-product (obtained in order of 10 wt%). This situation leads to the need of glycerol conversion into value-added products, to increase carbon balance and improve the economy of biodiesel production [3,4]. Fortunately, the renewable nature and high-functionality of glycerol has attracted the interest of scientific community in its transformation [3,5, 6]. Among different routes, glycerol selective oxidation is an attractive alternative due to low-cost, availability and eco-friendly origin of air or oxygen used as co-reactant, giving expensive oxygenated products such as dihydroxyacetone, glyceric and lactic acids as products [4,7–9]. Glyceric acid is currently used as platform to produce amino acids such as serine but also to treat skincare disorders due to its proven biological activity [10]. In case of lactic acid, it is widely used in different areas including food, pharmaceutical and textile industries [2]. However, its most significant application is as precursor for the production of polylactic acid, which is considered as one of the most promising bioplastics to substitute petroleum-derived plastics such as polystyrene, polypropylene, and polyethylene terephthalate [11]. Liquid-phase glycerol selective oxidation under basic conditions and using supported catalysts includes a complex reaction network (Scheme 1) that involves several pathways to obtain corresponding aldehyde, ketone and carboxylic acids [8,9]. As stated in the literature, a high pH commonly achieved by NaOH addition is needed for the initial deprotonation of glycerol and further dehydrogenation to glyceraldehyde and/or dihydroxyacetone, followed by oxidation to acid products [3,8,12]. Among different steps in both solution and over catalyst’s surface, several authors agreed that oxygen is indirectly involved during oxidation, removing electrons from the surface and regenerating hydroxyl groups to close catalytic cycle [8, 12,13]. Accordingly, to achieve good conversion and selectivity, the selection of an active phase and support gains great relevance. In this matter, mono and bimetallic catalysts based on Pd, Pt and Au supported on different metal oxides and carbon materials have been extensively * Corresponding author. E-mail address: [email protected] (M.G. Rinaudo). Contents lists available at ScienceDirect Results in Engineering journal homepage: www.sciencedirect.com/journal/results-in-engineering https://doi.org/10.1016/j.rineng.2022.100737 Received 16 August 2022; Received in revised form 15 October 2022; Accepted 28 October 2022 Results in Engineering 16 (2022) 100737 2 used in liquid-phase oxidation reactions, where both dispersion and particle size of metal particles are considered to be the dominant factors for a good activity and selectivity [14,15]. However, the role of supports is not limited to act as metal particles carriers [16–18]. Several times, the interface between metal and support becomes a better active site than the metal itself and thus, an appropriate selection of a catalyst support could reduce the quantity of expensive metals needed (e.g. noble metals) [3,15,16,19–21]. In case of glycerol oxidation, metal oxides with redox properties are known to have higher mobility of adsorbed hydroxyl and oxygen species at the interface between metal and support [8]. Moreover, metal-support interaction is a well-known phenomenon attributed to electronic and geometric effects that induce charge transfer between metal particles and support [15,19,22]. These interactions together with acid-base nature of support might not only enhance the adsorption capacity of active sites but also inhibit the sintering of metal particles during oxidation [3,17,23]. A shortcoming remains in the literature when it comes to understanding support’s contribution on catalytic performance, e.g. when defective metal oxides and different polymorphs are present in a given support. A rational design including the support is therefore crucial to obtain sustainable metal-supported catalyst with high activity, reusability and selectivity to desired products [16,22,24,25]. Considering several metal oxides, TiO 2 is a recognized support for metals which shows different properties depending on the polymorphs present in its structure [16]. Since most reactions in heterogeneous catalysis take place on the surface of metal active phase and/or boundaries between metal and support, the presence of various crystalline phases could impact on catalytic properties [15,16,26]. To the best of our knowledge, catalysts with modified TiO 2 supports formed by defective structures and unusual polymorphic mixtures have not been used for glycerol selective oxidation until now. Hence, we propose the study of palladium catalysts supported on mixes of anatase, TiO 2 (II) and rutile phases obtained by high-energy ball milling in a previous work [27] to evaluate the role of their different physicochemical properties and metal-support interactions in liquid-phase glycerol selective oxidation. 2. Experimental 2.1. Catalysts preparation Wet impregnation method was used to deposit palladium on three different TiO 2 supports obtained in a previous work [27]. 2 mL of a 0.1 M aqueous solution of HCl (37.0 wt%, Anedra) with appropriate quantities of PdCl 2 (99.9%, Sigma Aldrich) was added to each titania support in order to achieve palladium loadings of 0.25 wt% on a metal basis. Recovered solids were dried in an oven under vacuum at 70 ◦C overnight. Then, samples were calcined in a muffle furnace under air at 500 ◦C for 4 h using a heating ramp of 10 ◦C min −1 to eliminate the rests of metal precursor. Final Pd supported catalysts were labeled Pd/Ti x Scheme 1. Reaction network for glycerol oxidation. M.G. Rinaudo et al. Results in Engineering 16 (2022) 100737 3 where x corresponds to the milling time of previously synthesized supports. 2.2. Characterization 2.2.1. X-ray diffraction (XRD) X-ray diffraction was performed using a Rigaku Ultima IV diffractometer operated at 20 mA and 30 kV with a Cu K α (λ =0.15405 nm) radiation lamp. Step scan data were recorded at a step width of 0.02◦ and a counting time of 2 s. 2.2.2. Raman spectroscopy Raman spectroscopy was carried out using an Ar + laser excitation (514 nm) varying the source power and a resolution of 3 cm −1 (Model RAMAN-Renishaw - Invia Raman Microscopy). 2.2.3. Scanning electron microscopy (SEM-EDX) Morphology of the catalysts was studied with a scanning electron microscope coupled to Electron Dispersive X-ray Spectroscopy system (LEO 1450 VP SEM). Samples were deposited on aluminum sample holders and sputtered with gold before analysis. 2.2.4. Transmission Electron Microscopy (TEM) Transmission Electron Microscopy (TEM) studies were performed in a FEI TALOS F200S microscope operating at an accelerating voltage of 200 kV. For TEM analyses, samples were dispersed in ethanol by ultrasound before dropping them in a holey carbon film on a copper grid. Samples were studied by High-Angle Annular Dark-Field (HAADF) imaging, Scanning Transmission Electron Microscopy (STEM) and Energydispersive X-ray Spectroscopy (EDS). 2.2.5. Textural analyses Textural properties were evaluated by Brunauer-Emmet-Teller (BET) method using nitrogen adsorption at 77 K in a Micromeritics Gemini V equipment. Catalysts were previously outgassed overnight at 250 ◦C under N 2 . 2.2.6. Inductively coupled plasma – optical emission spectroscopy (ICPOES) Pd content was checked in an ICP- Varian-Ashilet equipment after digestion of the catalysts in a mixture of HNO 3 , HCl and HF acids. 2.2.7. X-ray photoelectron spectroscopy (XPS) XPS spectra were recorded with a SPECS Multitechnique instrument equipped with a dual X-ray source of Mg/Al and a PHOIBOS 150 hemispheric analyzer. A pass energy of 30 eV and an Al anode operated at 200 W was used. The pressure was kept under 1 ×10 −9 mbar during all measurements. Spectra were obtained for the C 1s, O 1s, Ti 2p and Pd 3 d regions. C 1s binding energy (284.6 eV) was used as internal reference to correct peak positions. XPSPeak 4.1 free software was used for data processing. 2.2.8. Temperature programmed measurements H 2 -Temperature Programmed Reduction (H 2 -TPR) and O 2 -Temper- ature Programmed Desorption (O 2 -TPD) were carried out in a homemade equipment using a quartz tubular reactor and a TCD detector under the same conditions reported in a previous study [27]. 2.2.9. 2-Propanol decomposition test Acid-base properties were studied through 2-propanol decomposition test in a fixed-bed stainless-steel tubular reactor. 50 mg of each catalyst and support, with particle sizes between 300 and 600 μ m, were firstly pretreated in Ar (150 mL min −1 ) for 30 min at 290 ◦C and atmospheric pressure. Reaction tests were performed introducing 7.5 vol % isopropanol (99.5%, Sintorgan) to the reactor, using Ar as carrier and the same operational conditions for 120 min. Products were analyzed in an online gas chromatograph (Buck Scientific 910) equipped with an FID detector and a silica capillary GC column (SPB1). 2.3. Catalytic tests 2.3.1. Reaction conditions Glycerol selective oxidation was carried out in a 50-mL steel Parr type reactor equipped with a pressure gauge, oven and controllers for temperature and agitation. 20 mL of 0.15 M glycerol (99.7%, Biopack) aqueous solution, sodium hydroxide (≥97.0%, Biopack) to reach NaOH: glycerol molar ratio of 2 and 200 mg of catalyst (glycerol:Pd theoretical molar ratio of 600) were added into the reactor. Mixture was heated to 100 ◦C under 1000 rpm of agitation speed and 1 bar of pure oxygen for 4 h. 2.3.2. Products analyses Products analyses were performed in an HPLC equipped with an Aminex HPX-87H column (300 ×7.8 mm) and an UV diodes array as detector. A 0.005 M solution of H 2 SO 4 (98.0%, Anedra) was used as eluent at 30 ◦C and a flow rate of 0.6 mL min −1 . Identification and quantification of products was carried out by comparison with external calibration curves using standard solutions of each chemical: glycerol (99.7%, Biopack); lactic acid (Pharmaceutical Secondary Standard, Merck), D-glyceric acid sodium salt (≥95.0%, Merck); oxalic acid (99.5%, Ucb); formic acid (85.0%, Thermo Scientific). An aliquot of each reaction sample was firstly filtered and neutralized by adding a fixed quantity of H 2 SO 4 solution before chromatographic analysis. 3. Results and discussion Materials presented herein consist of a complex matrix that includes palladium particles and different arrangements of TiO 2 polymorphs with defective structures induced by ball-milling. An exhaustive characterization is therefore required to comprehend its properties and how they could impact on catalytic behavior. 3.1. XRD analyses It is known that the nature of supports can highly affect the shape and dispersion of active species, improving the stability of catalysts during reaction [16,28]. Hence, XRD analyses were performed to evaluate the crystalline structure of supported catalysts. We observed (Fig. 1 A) that samples were formed by different mixtures of titania polymorphs. Pd/Ti5 showed intense reflections associated to anatase phase (PDF 00-21-1272) but also two broad and low-intensity reflections at angles of 27.54◦and 31.53◦attributed to (110) rutile reflection (PDF 00-21-1276) and (111) high-pressure TiO 2 (II) reflection (PDF 00-72-0021) respectively. These observations confirmed the presence of an almost pure anatase phase in this catalyst, as expected for its support [27]. On the other hand, Pd/Ti45 exhibited an increase in both TiO 2 (II) and rutile reflections but a decrease in anatase reflections. Pd/Ti120 also showed the reflection attributed to TiO 2 (II) phase and the appearance of several high-intensity rutile reflections. Obtained results revealed the different composition of titania phases present in the three catalysts. Furthermore, broad signals observed were expected due to lattice defects and small particles present in milled supports [29]. Palladium content of catalysts (0.25 wt%) could be considered out of the detection limits of XRD equipment [26]. However, an incipient signal of (101) PdO reflection (PDF 00-06-0515) at 34.04◦was detected in Pd/Ti5 catalyst. This behavior suggested the presence of larger palladium particles deposited on this catalyst, also supported by a further TEM analysis. In addition, oxidized species of palladium were likely to be formed since all catalysts were calcined in air at 500 ◦C [30]. The absence of this signal in Pd/Ti45 and Pd/Ti120 catalysts could be associated to smaller particles finely dispersed on the supports, favored by the greater number of defects and the combination of polymorphs present in these supports M.G. Rinaudo et al. Results in Engineering 16 (2022) 100737 4 [25,27,30–32]. We also made a rough calculation of average crystallite sizes of the catalysts by means of Scherrer equation and the relative intensities of the most intense reflection of each crystalline phase, (101) anatase, (110) rutile and (111) TiO 2 (II), as suggested by Dulian et al. [33]. We observed (Table 1) a reduction trend according to supports’ milling extent and the particle refinement caused by mechanochemical process [27,34]. These results were also in line with the increase in S BET values (Table 1) observed in a subsequent analysis. We found smaller crystallite sizes for Pd/Ti5 and Pd/Ti45 compared to the values obtained for respective supports (66 and 49 nm respectively) [27]. This fact was associated to the combined effect of acid medium during preparation, palladium deposition and calcination at 500 ◦C which could modify titania structures. Being anatase the main phase present in both catalysts, less stable particles of support may be deagglomerated and re-crystalized during synthesis steps. In case of Pd/Ti120, the crystallite size increased compared to the support (23 nm), showing more intense reflections, which could be associated to the stability of rutile and crystal growth during calcination [27]. We also estimated the catalysts composition (Table 1) using the most intense Bragg reflection of each titania phase and the equations reported by Duvarci et al. [35,36]. Results showed the unusual mixes of titania phases present in the supports [27] with greater quantity of anatase phase in case of Pd/Ti45 and more evidently in Pd/Ti120. These results could be associated to palladium addition and further calcination, which is known to stabilize anatase phase, as reported by other authors [37]. Moreover, formation of stable intermetallic structures between Ti from anatase and palladium could induce a reverse transformation from TiO 2 (II) and/or rutile to anatase, increasing the quantity of the latter [37]. Even though the mixture of crystalline structures and impurities is a drawback when using mechanochemical and thermal steps of synthesis, the existence of grain boundaries between different phases of titania could not only benefit charge mobility but also serve as energetic sites for reactants and/or products, improving the catalytic cycle during reaction [38]. 3.2. Raman, morphological and textural analyses Raman spectroscopy (Fig. 1 B) was performed to sustain the results from XRD. High-intensity bands observed at 142, 397, 515 and 638 cm −1 were attributed to anatase phase while bands at 171, 315, 427 and 525 cm −1 were associated to TiO 2 (II) phase [27,39,40]. Additionally, signals observed at 430, 445 and 608 cm −1 were related to rutile phase [41,42]. Once again, Raman results also confirmed titania phases present in the catalysts, with no evidence of palladium species due to its low concentration and/or high dispersion. Fig. 1. XRD spectra (A) and Raman spectra (B) of catalysts. Table 1 Weight fraction (%) of titania polymorphs, average crystallite size and textural properties of the catalysts. Catalyst Anatase wt% TiO 2 (II) wt% Rutile wt% Average crystallite size (nm) S BET (m 2 g −1 ) Pore diameter (nm) a Total Pore Volume (cm 3 g −1 ) b Pd/Ti5 99.2 0.3 0.5 54.2 10.0 20.9 0.05 Pd/Ti45 48.4 43.3 8.3 44.1 15.3 26.0 0.10 Pd/Ti120 23.0 49.4 27.6 42.1 17.2 8.8 0.04 a BJH adsorption branch average pore diameter. b Quantity of N 2 adsorbed at relative pressure of 0.98. M.G. Rinaudo et al. Results in Engineering 16 (2022) 100737 5 Textural and morphological properties are also important to understand catalysts’ surface architecture. From textural analyses, we observed specific surface area (S BET ) values (Table 1) in accordance with the milling time of each titania support. Considering the measurement error of ±1 m 2 g −1 , values were in the order of the ones obtained for the supports [27] even after palladium addition and calcination. These results might indicate that neither the small quantity of palladium nor the synthesis conditions blocked and/or collapsed supports’ pores [15]. It is worth highlighting that surface areas obtained were lower than typical values for metals supported on oxides (at least 50 m 2 g −1 ). However, these values were adequate enough to disperse small quantities of palladium (0.25 wt%), as observed by STEM following analyses. Pore diameters varied in the range 9–21 nm (Table 1) with slightly higher values in case of Pd/Ti5 and Pd/Ti45 compared to the supports [27]. This could indicate that pores’ morphology changed at some extent during catalysts synthesis, particularly in the case of supports with greater quantity of anatase phase, and in line with previous XRD discussion [43,44]. Additionally, total pore volumes located in the range 0.04–0.1 cm 3 g −1 (Table 1) as expected for the low porosity of titania [26,40]. These values were also higher than respective ones for Pd/Ti5 Fig. 2. SEM micrographs (10 kx) and Particle Size Distribution (PSD) histograms of catalysts: Pd/Ti5 (A), Pd/Ti45 (B) and Pd/Ti120 (C). M.G. Rinaudo et al. Results in Engineering 16 (2022) 100737 6 and Pd/Ti45 supports, which could also be attributed to the catalysts preparation [44]. In spite of these changes, S BET values were not affected. It is important to notice that pore sizes obtained were higher enough to transport big reactant molecules, such as glycerol (average molecule size of 0.62 nm) and hence reaction rate decrease due to internal diffusion is improbable to happen [3,45]. Furthermore, low porosities observed may indicate that S BET values corresponded almost entirely to the external surface area [3], which could induce the preferential location of palladium species on external surface of supports, favoring the access of reactants to start the mechanism of reaction [3]. Morphology of catalysts was studied by SEM. Pd/Ti5 sample (Fig. 2 A) showed particles with spherical shape and diameters between 120 and 220 nm. Pd/Ti45 catalyst (Fig. 2 B) presented grape-like clusters formed by a couple of particles, with average diameters between 100 and 200 [46]. These agglomeration might be a consequence of electrostatic effect on fine particles (with increased surface energy) as reported by other authors [46,47]. Smaller sizes were observed in case of Pd/Ti120 (Fig. 2C) within a range of 80–180 nm. Results were in line with the ones observed for the supports, with a reduction in particle sizes when increasing milling time [27]. Moreover, the smaller particle sizes correlated well with the increase in S BET values (Table 1) from Pd/Ti5 to Pd/Ti120. 3.3. ICP-OES and STEM-EDS analyses Pd loadings, measured by ICP-OES, were found in the range 0.2–0.36 wt% (Table 2) compared to the theoretical 0.25 wt%. Differences were attributed to experimental errors during synthesis and/or digestion of the samples. STEM and EDS analyses were carried out to have an insight into the sizes of palladium species. It was difficult to discriminate between Pd and/or PdO particles and the supports’ particles due to the small metal quantity and/or its high dispersion (Fig. 3A–D), precluding the measuring of an statistical average size of metal domains. However, we could infer a heterogeneous metal distribution since some Pd particles in the range of 4–5 nm but also around 20 nm were found in Pd/Ti5 catalyst. Moreover, we could scarcely observe metal particles in the range of 4–6 and 5–10 nm for Pd/Ti45 and Pd/Ti120 respectively [48]. These observations sustained previous XRD results, where PdO reflection could only be seen in Pd/Ti5, and thus bigger metal particles could be expected in this catalyst. Furthermore, it has been found that several solids submitted to mechanochemical activation exhibited an improved sorption capacity of divalent heavy metals from solution [49,50]. This behavior could also be extrapolated to the case of Pd 2+ cations (from PdCl 2 precursor) present in the impregnation solutions used in this study to add Pd on titania supports. The quantity of defects obtained in the supports by milling process, and their distribution, could therefore act as active centers for the adsorption of metal particles and increase the dispersion of metallic domains on the surface [28,49,50]. Along with this behavior, smaller average metal domains could also be expected when using low metal amounts, as reported by other authors [51]. Particularly, the heterogenous distribution of metal particles found in Pd/Ti5 could be attributed to the slight mechanochemical activation of the support and the major presence of anatase crystal phase. Moreover, Debecker et al. [28] have demonstrated that a support composed of a mixture of TiO 2 crystal structures (anatase and rutile) dictates the morphology and dispersion of Ru species owing to lattice matching and enhanced stability (impeding sintering) upon annealing. Hence, different metal-support interactions could be achieved according to the crystalline phases present in the catalysts, changing both dispersion and oxidation state of palladium particles [28]. Subsequently, an XPS study was performed to elucidate surface chemistry of each catalyst. 3.4. XPS analyses At the Ti 2p level, Pd/Ti5 presented Ti 2p 3/2 and Ti 2p 1/2 doublets (Fig. 4 A and Table 3) with binding energies (BE) values close to the ones reported for Ti 4+ in anatase (458.59 eV and 464.31 eV respectively) [52]. These signals were shifted to higher BE compared to the supports (458.37 and 464.04 eV respectively), indicating certain degree of metal-support interaction after palladium addition [27]. Ti 2p 1/2 could also be deconvoluted into a low-intensity signal attributed to Ti 3+ . In case of Pd/Ti45 and Pd/Ti120, doublets presented BE with values close to the ones reported for Ti 4+ in rutile (458.46 eV and 464.23 eV respectively) [52]. For both cases these signals could be mainly assigned to Ti 4+ from TiO 2 (II) phase since the two catalysts contained between 40 and 50% of this crystalline phase. Moreover, signals shifted to lower BE after palladium addition when compared to the supports. This could be attributed to a partial reduction of Ti 4+ to Ti 3+ generating oxygen vacancies, as reported by other authors [15]. In line with this observation, both catalysts also presented a low-intensity signal located around 460.3 eV and assigned to Ti 3+ at Ti 2p 1/2 level. At the Pd 3 d level, obtained signals were deconvoluted into four peaks (Fig. 4 B and Table 3) corresponding to Pd 0 (reference values of 334.7 eV for Pd 3d 5/2 and 340.4 for Pd 3d 3/2 ) and PdO (reference values of 336.2 eV for Pd 3d 5/2 and 341.4 for Pd 3d 3/2 ). We selected Pd 3d 5/2 signals for interpretation of palladium oxidation state (Table 3). In case of Pd/Ti5, signals shifted to lower BE than the values reported for Pd 0 and PdO, indicating metal-support interaction even in non-reduced catalysts. For Pd/Ti45 and Pd/Ti120, these signals shifted to the opposite direction (higher BE) also indicating a metal-support interaction, but different from Pd/Ti5. Shifts to higher BE for Pd 0 have been usually associated to smaller particles of Pd [22]. This is in line with our STEM observations, were Pd/Ti45 and Pd/Ti120 showed smaller metallic domains than Pd/Ti5. For Pd/Ti5, the shifts to higher BE for Ti 2p and lower BE for Pd 3 d could indicate an electronic transfer from support to metal, as observed by Namdeo et al. [3] for Pd/TiO 2 catalysts. This behavior was attributed to a possible formation of TiPd x O structures where palladium oxidation state could be more negative than zero [3,37]. Due to the high electronegativity of Ti, a charge transfer from support to metal could be expected, generating an electron cloud towards palladium [3]. In line with this explanation, Pd/Ti5 catalyst presented the highest content of Pd 0 (34.7%). According to Murata et al. [53], Pd 0 –PdO interfaces are more active in C–H dissociation of methane than PdO or metallic Pd. In our case, the presence of Pd 0 /PdO interfaces in the catalysts could be beneficial for glycerol oxidation in liquid phase in terms of OH − adsorption and glycerol dehydration to glyceraldehyde intermediate [8, 12]. The opposite shifts observed for Pd/Ti45 and Pd/Ti120 to lower BE in Ti 2p and higher BE in Pd 3 d could be associated to an electronic transfer from small metal particles to support, generating oxidized species (Pd δ+ with δ ≥2) and/or Pd/PdO x interfaces [3,54]. Dodson et al. [15] reported a higher content of these species in rutile phase. In our case, both catalysts share a great quantity of TiO 2 (II) phase (43.3 and 49.4 wt% respectively) and an upward trend in rutile (8.3 and 27.6 wt% respectively) which might be the cause of the opposite electron transfer compared to the almost pure anatase present in Pd/Ti5. These results shed a light on the importance of grain boundaries in polymorphic transformations, which could change the direction of electron density transfer between metal and support, affecting speciation and catalytic properties of materials. Increased number of defects, and content of rutile and TiO 2 (II) phases in the supports milled for longer Table 2 XPS and ICP-OES analyses of catalysts. Catalyst at%, theoretical at%, XPS wt%, theoretical wt%, ICP-OES Pd/Ti Pd/Ti Pd Pd Pd/Ti5 0.19 0.39 0.25 0.20 Pd/Ti45 0.51 0.29 Pd/Ti120 0.64 0.36 M.G. Rinaudo et al. Results in Engineering 16 (2022) 100737 7 Fig. 3. STEM images and EDS analyses: Pd/Ti5 (A* and B), Pd/Ti45 (C) and Pd/Ti120 (D). *Reproduced with permission from ref. [43]. M.G. Rinaudo et al. Results in Engineering 16 (2022) 100737 8 time would lead to more electron-deficient palladium species (oxidation states higher than zero) due to electron migration from palladium to titanium, as can be seen in Pd 0 /(Pd 0 +Pd δ+ ) trend (Table 3) [14,15,17]. Cationic species of palladium and/or Pd/PdO boundaries obtained in the three catalysts are expected to accelerate the mechanism of glycerol oxidation [3,8]. Additionally, from the point of view of transfer-induced electrostatic interactions, the sintering of positively charged palladium particles demands more energy than zero valence palladium particles [17]. Therefore, the availability of cationic species on the surface may also stabilize the catalysts under reaction cycles, preventing a high rate of sintering. Pd/Ti at% (Table 2) were in all cases higher than the theoretical calculation for the bulk (0.19 at%) but similar when affected by S BET , which sustained the dispersion of Pd on the surface of supports [16]. O 1s level was deconvoluted into two signals (Fig. 5 and Table 3) associated to lattice oxygen (O lat ) (ca. 529.9 and 529.7 eV for anatase and rutile respectively) and oxygen adsorbed on oxygen vacancies (O ads ) (ca. 531.09 and 531.22 eV for anatase and rutile respectively) [52]. In case of Pd/Ti5, O lat showed a BE close to the value reported for anatase while in Pd/Ti45 and Pd/Ti120 this value was closer to the BE for rutile, in accordance with the crystalline phases coexisting in the catalysts. Also, Ti/O lat atomic ratios (Table 3) were in all cases lower that the theoretical (0.5) which may give evidence of certain number of oxygen vacancies present in the catalysts. In line with these values, Pd/Ti5 showed the higher O ads quantity, 16.26%, while Pd/Ti45 and Pd/Ti120 showed values in the order of 10%. These contents were in the same order as observed in the supports [27], with a slight variation ascribed to palladium addition and further calcination. These low values of O ads achieved after milling were previously explained in terms of oxygen vacancies migration from the surface to the bulk in order to induce polymorphic transformation during the synthesis of supports [27]. Distinctive properties observed in the three catalysts might have a decisive role on activity and selectivity during glycerol oxidation, where charge and oxygen mobility are key factors. In order to emphasize support’s milling effect, we decided to prepare and analyze a catalyst composed of 0.25 wt% palladium using pristine TiO 2 support (anatase, S BET of 8 m 2 g −1 ) [27] and labeled it Pd/Ti. At the Ti 2p level we obtained signals in the order of reported ones for anatase phase with no substantial shifts in binding energies. In case of O 1s we observed an O ads quantity of 23%, displaying some degree of surface Fig. 4. XPS spectra of Ti 2p (A) and Pd 3 d (B) of catalysts. Table 3 XPS analyses of catalysts. Catalyst Binding Energy (eV) at%, XPS at. ratio, XPS Ti 2p 3/2 (Ti 4+ ) Ti 2p 1/2 (Ti 4+ ) Ti 2p 1/2 (Ti 3+ ) O 1s (O lat ) O 1s (O ads ) Pd 3d 5/2 (Pd 0 ) Pd 3d 5/2 (Pd δ+ ) Pd 0 /(Pd 0 + Pd δ+ ) O ads /(O lat + O ads ) Ti/O lat Pd/Ti5 458.79 464.48 460.56 529.73 531.06 334.37 336.04 34.70 16.26 0.45 Pd/Ti45 458.38 464.06 460.33 529.65 531.22 334.32 336.21 30.33 10.10 0.42 Pd/ Ti120 458.33 464.02 460.40 529.58 531.07 334.42 336.18 20.04 10.42 0.42 M.G. Rinaudo et al. Results in Engineering 16 (2022) 100737 9 oxygen vacancies, in agreement with its support [27]. More interestingly, deconvoluted Pd 3d 5/2 signals gave values of 338.23 and 339.46 eV, indicating the only presence of cationic palladium species [55]. Since no substantial shifts in Ti 2p signals were observed, we could deduce a negligible metal-support interaction in this catalyst, where Pd δ+ species would be a consequence of surface oxygen in pristine titania. These results allowed us to remark that even a short milling time could provoke the support’s ‘activation’ and benefit metal-support interaction, inducing electron migration according to the defects introduced and the TiO 2 polymorphs achieved. To deepen in the study of oxygen mobility and oxygen species present in the catalysts, of vital importance in oxidation, we carried out O 2 -Temperature Programmed Desorption (O 2 -TPD) experiments. 3.5. O 2 -TPD analyses Ouyang et al. [56] reported different oxygen species for O ads signals including oxygen from the surface of palladium particles and oxygen from the surface and subsurface of supports. Desorption signals (Fig. 6) found below 100 ◦C were related to oxygen species weakly bonded to the surface, as reported by other authors [57,58]. On the other hand, desorption signals located around 450 ◦C have been associated to surface lattice oxygen from the support [27]. These signals were observed at higher temperatures when compared to the supports, which could be explained by the metal-support interaction experienced after palladium addition, making oxygen release more difficult. Signals in the range 174–220 ◦C were attributed to oxygen species adsorbed on oxygen vacancies [59]. These oxygen species at the surface could be involved in the formation of oxygenated palladium species, with stronger bonds, that therefore desorb at higher temperatures. This fact could explain the disappearance of several low-temperature signals that were observed in the supports [27]. Also, the higher intensity and lower temperature of the desorption peak at 174 ◦C observed in Pd/Ti5 could be associated to the greater quantity of oxygen vacancies exhibited by XPS (Table 3). Furthermore, according to Ouyang et al. [56] signals located close to 200 ◦C could be attributed to molecular oxygen adsorbed on palladium particles, since they observed them in Pd black reference. Signals near 350 ◦C could be associated to the desorption of oxygen from finely dispersed PdO while the signal at 575 ◦C, only observed in Pd/Ti5, might be related to oxygen from larger PdO particles, as reported by Huang et al. [58]. These observations also have a good correlation with the palladium particles inferred by STEM, with larger sizes for Pd/Ti5 and smaller sizes for Pd/Ti45 and Pd/Ti120, leading to more oxygen-active species in the former. Catalysts with higher oxygen mobility are then expected to promote glycerol selective oxidation, in terms of oxygen supply and/or regeneration of the active surface [12,30,60]. 3.6. H 2 -TPR analyses Grain boundaries and interfaces between titania phases and/or palladium could increase charge mobility and thus reducibility of the catalysts [32]. H 2 -TPR measurements were performed to further study metal-support interactions, that play a major role in catalytic reactions that proceed via redox cycle [53]. Strong signals found in the range 533–578 ◦C (Fig. 7) were associated to the reduction of surface Ti 4+ Fig. 5. XPS spectra of O 1s of catalysts. Fig. 6. O 2 -TPD of catalysts. M.G. Rinaudo et al.