**E**ff**ect of Metal Loading in Unpromoted and Promoted CoMo**/**Al2O3–TiO2 Catalysts for the Hydrodeoxygenation of Phenol**

**J. Andrés Tavizón-Pozos 1,2,\*, Carlos E. Santolalla-Vargas <sup>3</sup> , Omar U. Valdés-Martínez <sup>1</sup> and José Antonio de los Reyes Heredia 1,\***


Received: 29 May 2019; Accepted: 14 June 2019; Published: 19 June 2019

**Abstract:** This paper reports the effects of changes in the supported active phase concentration over titania containing mixed oxides catalysts for hydrodeoxygenation (HDO). Mo and CoMo supported on sol–gel Al2O3–TiO2 (Al/Ti = 2) were synthetized and tested for the HDO of phenol in a batch reactor at 5.5 MPa, 593 K, and 100 ppm S. Characterization results showed that the increase in Mo loading led to an increase in the amount of oxide Mo species with octahedral coordination (MoOh), which produced more active sites and augmented the catalytic activity. The study of the change of Co concentration allowed prototypes of the oxide species and their relationship with the CoMo/AT2 activity to be described. Catalysts were tested at four different Co/(Co + Mo) ratios. The results presented a correlation between the available fraction of CoOh and the catalytic performance. At low CoOh fractions (Co/(Co + Mo) = 0.1), Co could not promote all MoS2 slabs and metallic sites from this latter phase performed the reaction. Also, at high Co/(Co + Mo) ratios (0.3 and 0.4), there was a loss of Co species. The Co/(Co + Mo) = 0.2 ratio presented an optimum amount of available CoOh and catalytic activity since the XPS results indicated a higher concentration of the CoMoS phase than at a higher ratio.

**Keywords:** hydrodeoxygenation; phenol; Al2O3-TiO2; CoMo; CoMoS; biofuels; MoS2

#### **1. Introduction**

Studies on the transformation of fast pyrolysis oils from lignocellulosic biomass as an alternative to produce clean and renewable transportation fuels have been increasing for the last years [1–3]. Bio-oil, rich in oxygenated compounds (30%–40%), needs to be upgraded to enhance its heating value, chemical and thermal stability, and miscibility with fossil fuels [4]. To achieve this, catalytic hydrodeoxygenation (HDO) can be used to eliminate oxygen and to hydrogenate instaurations for hydrocarbons chains. HDO can proceed in a wide range of temperatures (200–400 ◦C) and pressure (1–7 MPa) [5,6]. Also, as some oxygenated molecules present in pyrolysis oils are soluble in water, aqueous phase reactions can be carried out in the presence of a catalyst [7,8]. However, the aqueous hydrodeoxygenation may only be effective for some fractions of the bio-oil. For the more refractory molecules, this process could take advantage of the current technology of typical hydrotreatment

processes used for petroleum feeds since upgraded bio-oil can be blended with them and be transformed in the same step [5]. Several works have been focused on the improvement of their catalytic properties, for instance selectivity, activity, and active phase dispersion [2,9–12]. These studies have used resulting lignin representative probe molecules, for example, guaiacol, eugenol, furans, cresol, catechol, anisole, and phenol, to understand catalytic performance and functionalities [7,13–15]. Particularly, phenol has been used as a probe molecule since its reactions could expose information about the reactivity of the CAR–OH bond and hydrogenation capacity of the catalyst. Additionally, it is formed as a remnant of more complicated oxygen containing molecules [10,14,16,17]. HDO of phenol can proceed by two pathways: Hydrogenolysis or direct dehydrogenation (DDO) and hydrogenation (HYD) [10,17]. The DDO route consists in the cleavage of the CAR–OH bond (414 kJ/mol) to form benzene. For its part, the HYD route proceeds by hydrogenation of the π bonds of the aromatic ring to generate an oxygenated intermediate (O–I, cyclohexanol, and cyclohexanone). Therefore, the C–OH bond splits to produce cyclohexene and subsequently cyclohexane [10,17].

Typical hydrotreatment CoMo/Al2O3 sulfided catalysts have been used in HDO with probe molecules with promising results [10,13,18,19]. It is widely accepted that the active sites are located at the edges of the formed sulfided slabs. The Mo-edge has been attributed the role of hydrogenation due to its metallic character, whereas the active sites for hydrogenolysis path (C–O scission) have been identified as coordinatively unsaturated sites (CUSs) of sulfur at the S-edge on the MoS2 phase [20,21]. These sulfided vacancies possess an electrophilic character, where oxygen from phenol can adsorb and SH groups provide hydrogen to carry out the C–O bond cleavage [5,22]. When Co is added as a promoter atom, hydrotreatment activity increases significantly by the formation of the CoMoS II phase as proposed by Topsoe et al. [21]. In this highly active phase, it has been proposed that Co has a preference to be located at the S-edge on the MoS2 phase, which may promote the hydrogenolysis pathway [22,23]. Nonetheless, in HDO reactions, there is still debate concerning the way that the oxygenated functional groups of the molecules react on the different sulfide phases. On this regard, to achieve improvements of the catalysts by increasing the concentration of the CoMoS phase, the synergic effect of the Co concentration in the active sites for these reactions needs to be understood [24–26].

To develop highly active CoMoS catalysts supported on alumina, metal–support interaction (MSI) between oxide precursors for both metal sulfides is an important factor to tune for the supported active phase structure, dispersion, reducibility, and promotion [20–22]. One possible way is to modify Al-based mixed oxides, such as Al2O3–TiO2 (AT). This mixed oxide has been proposed as an alternative support to take advantage of the combined properties of alumina and titania [27–33]. It had displayed notorious improvements in the textural and physicochemical properties in comparison to alumina or titania supports [30–34]. Particularly, CoMoS supported over sol–gel synthetized Al2O3–TiO2 catalysts with an Al/Ti = 2 ratio (called "AT2") showed that the interactions between the metal oxide phases and the support enhance the formation of the sulfided active phases compared to alumina, leading to higher activities [32,35–37]. In a previous work of the HDO of phenol, it was shown that CoMo/AT2 was twice as active and was more selective to the DDO route than CoMo/Al2O3 catalyst [34]. These results were attributed to a decrease in the MSI due to the presence of titania in the support. The differences in activity and selectivity were attributed to a higher fraction of supported MoOx species with octahedral coordination (MoOh) on AT2 than on alumina. These MoOh species are easier to reduce than MoOx with tetrahedral coordination (MoTh), which were more abundant on alumina [34]. As it has been reported, variations on metal loading may be induced in the formation of oxide species and therefore in the active phase formation [26]. Additionally, the Co/(Co + Mo) ratio must be optimized in order to avoid promoter loss into the support and segregation to complete the understanding of the synergic effect of the promoter on the MoS2 in CoMo/AT2 catalysts. However, there is no information about the effects of metal loadings on the formation of different species over AT supports, and their impact on the activity and selectivity of HDO reactions. Therefore, the objective of this work was gain further

insight on the Co/(Co + Mo) ratio and its effect on C–O bond cleavage of CoMoS/AT catalyst for the HDO process using phenol as the model molecule.

#### **2. Results and Discussion**

#### *2.1. E*ff*ect of Mo Loading*

#### 2.1.1. Catalytic Activity

The catalytic hydrodeoxygenation (HDO) experiments results, for non-promoted Mo catalysts are shown in Figure 1.

**Figure 1.** Normalized initial reaction rate of HDO of phenol for unpromoted Mo/AT2 catalysts at different Mo loadings at 5.5 MPa, 593 K, and 100 ppm S, (-) by gram of catalyst, () by gram of Mo.

These results showed that the AT2 support was active and the initial reaction rate increased linearly with the Mo loading. Specifically, 5 wt.% Mo exhibited twice the activity of the support, while the activity for the 20 wt.% Mo catalyst was also five times higher. When the initial reaction rates were quantified by gram of supported Mo, a decrease was observed when the metal loading increased. However, at loadings higher than 10 wt.% Mo, activity remained constant. This may indicate that at this Mo loading, a monolayer was achieved. This result seems to be adequate to the system, since the Mo/Al2O3 monolayer coverage was near 10 wt.% Mo and 6.6 wt.% for Mo/TiO2 [38–40]. Selectivity data at 20% of phenol conversion are presented in Figure 2.

It was observed that the AT2 support alone mainly generated incomplete hydrogenation products, i.e., oxygenated intermediates (O–I, cyclohexanone, and cyclohexanol) and cyclohexene. Therefore, AT2 support presented a higher hydrogenation functionality than hydrogenolysis. However, it is possible that there were not enough sites to achieve cyclohexene hydrogenation. When Mo was added, even at low loadings, the product yields changed to an increase in benzene and cyclohexene production compared with the AT2 support. Also, as all Mo catalyst yields did not present significant changes, it is possible that the active sites' nature was the same.

It has been suggested that S-edge sites are responsible for the hydrogenolysis pathway, while Mo-edge sites are responsible for hydrogenation reactions [5]. On this basis, it is possible to suggest that S-edge sites predominate in all Mo/AT2 catalysts since the hydrogenation of cyclohexene to cyclohexane was limited. This resulted in the capability to cleave the CAR–OH and the C–OH (C = O) bonds from the O–I to produce benzene and cyclohexene, respectively. However, the O–I yield decreased with the Mo loading. Therefore, it is possible to propose that both hydrogenolysis processes may occur in the same type of sites and their abundance increased with the Mo loading.

**Figure 2.** Products yields of HDO of phenol for unpromoted Mo/AT2 catalysts at different Mo loadings at 5.5 MPa, 593 K, and 100 ppm S.

2.1.2. Diffuse Reflectance Spectroscopy UV-Vis

Figure 3 gives the DRS UV-vis spectra for the calcined MoOx catalysts supported on AT2. For all studied samples, DRS UV-vis spectra showed a single signal located from 200 to 600 nm. For the AT2 sample, a simple band was visible between 200 and 350 nm and corresponds to the metal–ligand charge transfer (MLCT) for O2<sup>−</sup> <sup>→</sup> Ti4<sup>+</sup> [41].

**Figure 3.** Diffuse reflectance UV-vis spectra of the unpromoted Mo/AT2 catalysts calcined at 673 K with different Mo loadings: **a**) AT2, (**b**) 5 wt.%, (**c**) 10 wt.%, (**d**) 15 wt.%, and (**e**) 20 wt.%.

For the supported Mo catalysts, it could be considered that the band between 200 and 400 nm included the signals of Mo and titania [36]. The MLCT band for O2<sup>−</sup> <sup>→</sup> Mo6<sup>+</sup> was located between 200 and 400 nm. In this band, Mo with tetrahedral coordination (MoTh) (MoO4 <sup>2</sup><sup>−</sup>, Mo2O7 <sup>2</sup>−) were located in the 200–300 nm range. Besides, Mo with octahedral coordination (MoOh) from heptamolybdates and octamolybdates were included between 300 and 400 nm [41–43]. Even when a clear band assignation is complex, in a comparison between AT2 support with the different catalysts, it is possible to observe a shift in the reflectance bands to longer wavelengths with increments of the Mo loading. This could indicate a higher concentration of MoOh compared with MoTh at high Mo loadings.

#### 2.1.3. Laser Raman Spectroscopy

Figure 4 displays the laser Raman spectra for 10 and 15 wt.% Mo calcined catalysts.

**Figure 4.** Laser Raman spectra of Mo/AT2 catalyst calcined at 673 K with different Mo loadings: (**a**) 10 wt.%, (**b**) 15 wt.%.

For the 10 wt.% catalyst, four peaks were displayed at 664, 815, 950, and 992 cm<sup>−</sup>1. These signals presented a slight shift to a higher stretching frequency for 15 wt.% Mo due to the increment of metal content. The band located at 664 cm−<sup>1</sup> was attributed to symmetric Mo-O-Mo deformations. Besides, 815 and 992 cm−<sup>1</sup> frequencies were related to antisymmetric Mo-O-Mo stretching for bulk-like MoO3 crystallites [26,44]. The band at 950 cm−<sup>1</sup> was assigned to terminal Mo = O symmetric and asymmetric stretching bonds for coexisting polymolybdates (Mo7O246- and Mo8O36<sup>4</sup>−, MoOh). The intensity of this signal increased for the 15 wt.% Mo loading, suggesting that more octahedral species could be present compared with the 10 wt.% Mo sample, in agreement with UV-vis results. This signal involved heptamolybdates and octamolybdates species, which were normally found around 920, 945, and 965 cm−<sup>1</sup> [45,46]. Hence, this could indicate that MoO3 structures are formed before the total coverage of the monolayer.

#### 2.1.4. Temperature Programed Reduction

Unpromoted oxide Mo catalysts and AT2 support TPR profiles are shown in Figure 5. The AT2 TPR profile (Figure 5a) presented two reduction peaks attributed to the reduction of surface Ti species at 900 K and bulk Ti species (Ti4<sup>+</sup> <sup>→</sup> Ti3<sup>+</sup>) at 1050 K [31,47].

**Figure 5.** TPR profiles for unpromoted Mo/AT2 catalysts calcined at 673 K at different Mo loadings. (**a**) AT2, (**b**) 5 wt.%, (**c**) 10 wt.%, (**d**) 15 wt.%, and (**e**) 20 wt.%.

However, according to Platanitis et al. [47], only approximately 24% of anatase could be reduced on pure TiO2. Hence, it is expected that less than this percentage of titania could be reduced in the mixed oxide support. In the Mo containing samples, all catalysts showed reduction peaks corresponding to Mo6<sup>+</sup> <sup>→</sup> Mo4<sup>+</sup> and Mo4<sup>+</sup> <sup>→</sup> Mo0 at low temperatures (400–800 K) and high temperatures (>800 K), respectively [48–50]. Particularly, for the 5 wt.% Mo catalysts, a peak centered at 740 K corresponded to an easily reducible Mo species, possibly in octahedral coordination [50]. The signal located at 940 K was attributed to Mo strongly interacting with the support. Furthermore, the 10 wt.% Mo catalysts also showed two signals at 740 and 940 K. However, it is possible that the 940 K peak could not only be assigned for the MoTh species reduction, but also to a contribution from the MoO3 species, as LRS results showed. The TPR profile of 15 and 20 wt.% Mo catalysts presented four signals at 780, 850, 880, and 1000 K. The peaks located between 780 and 900 K could be attributed to the reduction of MoOh and a mixture of MoOh and MoTh, respectively. The high temperature peaks (880 and 1000 K) could be caused by the presence of MoTh and bulk MoO3 [47,49,50]. The absence of an 850 K peak at 5 and 10 wt.% Mo would indicate that at low Mo loadings, there may not be a notorious mixture of MoOh and MoTh and more MoOh was formed when Mo loading was augmented.

From the impregnation of Mo, the ammonium heptamolybdate solution for the 5 wt.% Mo catalyst's pH value was 5.2 and decreased to 4.5 for the 20 wt.% Mo catalyst. Consequently, ionic polymolybdates complexes, including [H3MoO24] <sup>3</sup><sup>−</sup>, [H2Mo7O24] <sup>4</sup><sup>−</sup>, [Mo7O24] <sup>6</sup><sup>−</sup> [Mo8O26] <sup>4</sup><sup>−</sup>, and [HMo7O24] <sup>5</sup><sup>−</sup> species, were predominant in the solution as has been mentioned by the literature [46,51]. Additionally, since the AT2 isoelectric point value was 7.6, acid terminal OH groups were predominant on the surface at the impregnation pH. Hence, anionic polymolybdates species could anchor to the support by electrostatic interactions [52]. On this basis, at low metal loadings (5 wt.% Mo), the anionic Mo species could be well dispersed on the surface. As the metal loading increased, the anionic species were closer to each other and conglomerated, generating large Mo species that would have a weaker metal–support interaction than those present at low metal loadings. Characterization analysis showed that MoOh oxide species increased with the metal loading. However, as MoTh species are not clearly visible in the DRS results, the Raman and TPR profiles confirmed the presence of a fraction of them and MoO3 crystallites. These results may indicate that Mo dispersion decreased at high Mo loadings.

The presence of MoO3 at the 10 wt.% Mo catalyst, and the catalytic results shown in Figure 1, indicated that the monolayer coverage would be complete at 20 wt.% Mo and that MoO3 can be formed before it [42,53]. This behavior seems to be quite similar to the Mo/Al2O3 catalyst, which due to the parallel configuration of the hydroxyl groups of the support, leads to a rearrangement of the Mo species during calcination. In contrast, Mo/TiO2 tends to form MoO3 after total coverage of the monolayer, caused by the homogeneity of its hydroxyl groups [26]. Therefore, supported polymolybdates on AT2 may present a configuration more alike to alumina than titania. However, due to the presence of the latter, the generation of MoO3 was delayed. On this basis, it can be suggested that this could occur just before the total formation of the monolayer.

Catalytic evaluation indicated that the AT2 support was active and presented a selectivity to partial hydrogenation. In contrast, when Mo was supported, even at low metal loadings, the selectivity changed to cyclohexane and benzene production. Nevertheless, since all unpromoted catalysts' selectivity did not present significant changes, it is possible to suggest that the generation of products was due to a high number of active sites with the same nature. Consequently, the non-covered parts of the support could contribute slightly to the production of the O–I and cyclohexene, whereas the MoS2 phase contributed to the generation of DDO and HYD sites. Considering that MoOh were easier to reduce than MoTh, more MoS2 would be present at high Mo loadings (>10 wt.% Mo). In this sense, MoTh species may not change the selectivity; however, some of them would not be completely sulfided, i.e., only the number of active sites changed. This is deduced by the fact that even in the presence of CS2, benzene was still produced. Due to the presence of a sulfiding agent (CS2), the production of benzene was limited by the competition of CS2 for the electrophilic sites located at the S-edge [5,10,34,54]. Then, as MoOh increased, more S-edge sites were formed, and the resistance to inhibition was improved.

#### *2.2. E*ff*ect of Co Loading*

#### 2.2.1. Catalytic Activity

The initial reaction rates for Co/AT2 and AT2 are presented in Table 1. The Co/AT2 catalyst was 1.3 times more active than the AT2 support. This means that Co containing sample presented active sites that improved the catalytic activity. Nevertheless, this activity was 0.3 times lower than that for the 5 wt.% Mo/AT2 sample (see Figure 1). The selectivity showed that cyclohexane was the main product on Co/AT2 in contrast with cyclohexene on AT2.

**Table 1.** Initial reaction rate of the HDO of phenol at 5.5 MPa and 593 K and product yields at 15% of the conversion for AT2 support and Co/AT2 sulfided catalyst.


The O–I yield decreased when Co was present on the catalyst, while benzene production remained the same. On this basis, the Co sulfide phase did not provide enough sites to cleave the CAR–OH bond (468 kJ/mol). However, the high production of cyclohexane indicated that O–I could transform (339 kJ/mol) into cyclohexene to hydrogenate in a further step [17]. Nevertheless, these active sites presented a more hydrogenating character than hydrogenolysis, leading to cyclohexane. Furthermore, since Co did not cover the entire surface of the support, the AT2 support may have had a role. Since the AT2 support showed selectivity to O–I and cyclohexene, it is possible that these products were generated by the support in Co/AT2. In this sense, the AT2 support could have provided Bronsted acid sites, whereas the Co9S8 phase provided metallic sites, which have a hydrogenating character.

The synergic effect of Co for the HDO of phenol activity is presented in Figure 6.

**Figure 6.** Initial reaction rates of HDO of phenol at different Co/(Co + Mo) ratios in the promoted CoMo/AT2 catalyst (-) at 10 wt.% Mo and (-) 15 wt.% Mo.

For both series, the dependence of the activity by the concentration of Co on the catalyst presented parallel volcano type curves with a maximum at Co/(Co + Mo) = 0.2. Since the concentration of Mo was higher at the promoted 15 wt.% Mo catalysts than the 10 wt.% Mo catalysts, the initial reaction rate increased 1.3 times. Table 2 presents a synergic factor to compare the activity of the promoted with the unpromoted catalysts, at different Co/(Co + Mo) ratios.


**Table 2.** Synergic effect in the HDO of phenol with CoMo/AT2 at different Co concentrations.

Following the volcano like curve, the Co/(Co + Mo) = 0.1 ratio presented a lower promoting factor (1.6) compared with the other samples. At an atomic ratio of 0.2, a maximum synergic factor was found (2.8) and this decreased by 20% when Co was loaded at higher atomic ratios. This indicated that Co loading played an important role in the interaction with MoS2 slabs and subsequently the promotion of active sites. The selectivity changed for the different atomic ratios as illustrated in Table 3.

**Table 3.** Direct deoxygenation and hydrogenation route ratio of HDO of phenol at 20% of the phenol conversion for the CoMo/AT2 catalyst.


In all catalysts, phenol was the main product, following the direct deoxygenation (DDO) route by direct incision of the CAR–OH bond. However, hydrogenated products (HYD), including O–I, cyclohexene, and cyclohexane, increased with the Co loading. As Co/AT2 showed, Co9S8 functionalities promoted the HYD route, thus the increment of the Co concentration led to the formation of this sulfide. At low Co loadings (Co/(Co + Mo) = 0.1), a fraction of DDO sites were promoted by this element at the edges of the MoS2 phase [55]. At the Co/(Co + Mo) = 0.2 ratio, a high concentration of the CoMoS phase could be formed due to this ratio presenting the higher activity. Since not all MoS2 would be promoted, the presence of Co9S8 could increase, leading to an increase in HYD selectivity. In this sense, at the Co/(Co + Mo) = 0.3 and 0.4 ratios, Co not only promoted the DDO route at the edges of the CoMoS phase but also contributed to the HYD route due to the presence of the Co9S8 segregated phase.

#### 2.2.2. Diffuse Reflectance Spectroscopy UV-Vis

The Co/AT2 calcined catalyst spectra are presented in Figure 7.

**Figure 7.** DR UV-vis spectra of the Co/AT2 calcined at 673 K.

*Catalysts* **2019**, *9*, 550

It is possible to observe a shoulder centered at 260 nm corresponding to the MLCT O2<sup>−</sup> <sup>→</sup> TiO4<sup>+</sup> bands. Then, the Co transition bands were in a wide signal between 400 and 1000 nm. The first signal at 400 nm was attributed to the Co2<sup>+</sup> species with octahedral coordination (CoOh) [56]. However, like the Mo results, this band was overlaid with the titania MLCT and the analysis was difficult. The wide signal involves a triplet centered at 500, 580, and 630 nm, corresponding to d–d transitions of the Co2<sup>+</sup> species with tetrahedral coordination (CoTh) and strongly interacting with alumina (CoAl2O4) [57]. Finally, a band near 750 nm was assigned to Co3<sup>+</sup> and Co2<sup>+</sup> species with octahedral coordination in Co3O4 [57]. On this basis, CoTh species were more abundant than CoOh species since the respective were more intense. This was due to the absence of Mo and the low Co loading (2 wt.%) on the support. Co could be well dispersed on the support in small particles, which interacted strongly with the alumina present in the support. Nonetheless, as titania was highly dispersed in the alumina matrix [58], it could avoid the migration of Co into the support. This could promote the formation of CoOh, which is a Co9S8 precursor. The promoted supported catalysts' DR spectra are presented in Figure 8, showing two main bands, one located at 200–400 nm, and the other at 500 – 800 nm.

**Figure 8.** DR UV-vis spectra of the CoMo supported catalysts calcined at 673 K at different Co/(Co + Mo) ratios: (**a**) 0.1, (**b**) 0.2, (**c**) 0.3, and (**d**) 0.4.

Considering the previous unpromoted Mo and Co supported catalyst spectra, the fist signal (200–400 nm) involved the titania, and Mo and Co bands. However, the second signal only involved the Co species. In comparison to Figure 7, it is possible to observe an increase in the intensity of these bands with the Co/(Co + Mo) ratio. Also, a shift to near infrared was detected as the Co loading was increased. Additionally, the shoulder located at 400 nm (CoOh) increased its intensity at high atomic ratios. This may indicate that more Co with octahedral coordination was present at Co/(Co + Mo) = 0.4 compared with the other catalysts. Nevertheless, the CoTh transition bands increased as well. Since Co may find its migration into the titania support difficult, these Co species could be part of CoMoO4 in which Co was in octahedral coordination [59,60]. To achieve a proper comparison between the CoOh and CoTh species, Gaussian deconvolution (not showed) was carried out considering the representative signals reported in the literature [56,57,61]. Since CoMoO4 is considered a bad precursor for the CoMoS phase [62], the CoOh present in this oxide complex was considered as CoTh. It is possible to calculate the ratio between CoOh and CoTh with the area under each peak and using the F(R∞) Oh/(F(R∞) Oh <sup>+</sup> F(R∞) Th) equation [63]. The results are shown in Figure 9.

**Figure 9.** Correlation between CoOh and Co loading of the promoted CoMo catalyst with 10 wt.% and 15 wt.% Mo calcined at 673 K.

The correlation between CoOh and Co loading presented a volcano type curve with a maximum at the Co/(Co + Mo) = 0.2 ratio. At low Co/(Co + Mo) ratios, the concentration of the promoting atom was inadequate to interact with Mo. Hence, it resulted in small particles that interacted strongly with the support, leading to an incomplete promotion. In contrast, at high Co/(Co + Mo) ratios, the promoting atom was in excess. Thus, it resulted in the formation of CoMoO4 and highly dispersed particles, which did not interact with Mo but with the support. In this sense, at the maximum CoOh ratio, there was enough CoOh to interact properly with the MoOX species, leading to the formation of the CoMoS phase. The amount of CoOh and CoTh obtained from DRS UV-vis analysis was correlated with the activity, as presented in Figure 10.

**Figure 10.** Correlation between the initial reaction rate of the HDO of phenol and octahedral Co of 10 wt.% and 15 wt.% Mo supported catalysts.

This correlation showed that as CoOh was increased, the initial reaction rate also increased. This is due to the proper promotion of Mo with CoOh to form the CoMoS phase. At low Co loadings, more CoTh species were present on the support and their sulfidation was less than the CoOh species, as has been reported before [34,64]. At the optimum Co/(Co + Mo) ratio, the amount of CoOh promoted the proper formation of the CoMoS phase and a fraction of Co could be CoTh. When Co was in excess, a fraction of it promoted the CoMoS phase, but also a fraction interacted strongly with Mo, giving place to CoMoO4. However, the Co3O4 phase was also be present and could be transformed into Co9S8.

#### 2.2.3. Temperature Programmed Reduction

In Figure 11, the Co/AT2 calcined catalyst's TPR profile is presented. It is possible to observe four signals centered at 650, 780, 1030, and 1098 K.

**Figure 11.** TPR profile of Co/AT2 catalyst calcined at 673 K.

By Gaussian deconvolution, an extra peak was found in the 1030 K peak. The first signal (650 K) was assigned to the reduction of Co3O4 crystals to CoO, i.e., the CoOh, Co3<sup>+</sup> <sup>→</sup> Co2<sup>+</sup> [65,66]. The signal at 780 K corresponded to the reduction of superficial well dispersed Co3<sup>+</sup> species. Also, the deconvolution resulting peak at 923 K could be attributable to Co2<sup>+</sup> <sup>→</sup> Co0, whereas, the second resulting peak at 1040 K may be referred to the CoAl2O4 species [50]. However, these peaks may have contributed to the initial titania ions' reduction to Ti3<sup>+</sup>. Finally, the last peak at 1090 K was the result of the Ti4<sup>+</sup> <sup>→</sup> Ti3<sup>+</sup> reduction. Note that the CoTh species, with strong interaction with the support, were more abundant than CoOh. This result confirms the DR UV-vis results since the Co metal–support interaction is strong enough to generate CoTh species that are difficult to reduce. Therefore, their sulfidation was incomplete and less active phase was formed. As was previously seen in the literature [67], Co metal–support interactions decrease when Mo is present on the surface of the support. According to these results, the Co/AT2 catalyst presented a higher amount of CoTh than CoOh due to the interaction with alumina. However, both species could be sulfided and form Co9S8, which had metallic sites with selectivity to the HYD route. The benzene production was inhibited by competition for the hydrogenolysis active sites by the presence of H2S in the reactor. Additionally, the amount of these sites would be less than the hydrogenation sites. Since Co did not totally cover the support, the latter could contribute to the production of the O–I and cyclohexene.

The CoMo/AT2 catalyst's TPR profiles are presented in Figure 12. In general, all TPR profiles were similar to each other and showed three main peaks at 750, 950, and 1080 K.

The Gaussian deconvolution of these TPR profiles developed two additional peaks at 640 K and 820 K. The peak at 640 K corresponded to CoOh, Co3<sup>+</sup> <sup>→</sup> Co2<sup>+</sup> as shown in Figure 11 as well. The second peak, at 750 K, was assigned to the Mo6<sup>+</sup> <sup>→</sup> Mo4<sup>+</sup> reduction [51]. The third peak, at 820 K, could be assigned to a mixture of Co2<sup>+</sup> <sup>→</sup> Co<sup>0</sup> and Mo4<sup>+</sup> <sup>→</sup> Mo<sup>0</sup> [68]. The fourth peak near 940 K was caused by Mo strongly interacting with the support and involved CoMoO4 and CoAl4O3 species [63]. Finally, the last high temperature peak belonged to the support.

**Figure 12.** TPR profiles of CoMo/AT2 calcined catalyst at 15 wt.% Mo with different Co/(Co + Mo) ratios: (**a**) 0.1, (**b**) 0.2, (**c**) 0.3, and (**d**) 0.4.

In general, the reduction peaks did not present significant shifts to low temperatures with the increase in the Co loading. The Co/(Co + Mo) = 0.1 ratio catalyst showed a slight shift of the second main peak to low temperatures due to the low Co concentration, i.e., the contribution by the Mo reduction was predominant. By its part, the Co/(Co + Mo) = 0.2 catalyst consumed more H2 in the first peak than in the second one, meaning that more octahedral Co and Mo species were present in this catalyst. This could lead to a better promotion of the active phase and therefore an improvement of the activity. At Co/(Co + Mo) = 0.4, the peak located at 640 K was more intense than in the other samples. This signal could be the result of the high concentration of segregated Co3O4. Additionally, since more segregated Co was present on these catalysts, CoTh was present in a minor fraction.

#### 2.2.4. X-Ray Photoelectron Spectroscopy

Figures 13 and 14 exhibit the XPS spectra for Mo3d and Co2p core levels for sulfided CoMo/AT2 catalysts at the Co/(Co + Mo) = 0.2 ratio with 15 wt.% Mo.

**Figure 13.** XPS deconvolution of Mo3d core level for sulfided 15 wt.% Mo CoMo/AT2 catalysts at the Co/(Co + Mo) = 0.2 ratio.

**Figure 14.** XPS deconvolution of Co 2p core level for sulfided 15 wt.% Mo CoMo/AT2 catalysts at the Co/(Co + Mo) = 0.2 ratio.

The Mo3d spectra displayed a doublet for the two spin orbit components, 3d5/<sup>2</sup> and 3d3/2, located at 228.76 and 231.8 eV, respectively. Additionally, the S2s core levels' band was detected at 226.3 eV. The XPS for the Mo3d5/2 decomposition showed peaks attributed to Mo in sulfide (Mo4<sup>+</sup>), oxysulfide (Mo5<sup>+</sup>), and oxidic (Mo6<sup>+</sup>) species. These peaks were located at 228.79 eV, 230.4 eV, and 232.80 eV, respectively [38,69,70].

For the Co2p3/<sup>2</sup> core levels displayed in Figure 14, the Co9S8, CoMoS, and Co2<sup>+</sup> oxide species were found by decomposition of the main signal. The binding energies for Co species were identified at 778.16 eV for Co9S8, at 778.76 eV for CoMoS, and at 780.7 eV for oxidic Co2<sup>+</sup> [38,69,71]. The CoMo/AT2 catalyst at the Co/(Co + Mo) = 0.4 ratio is not shown since it presented similar signals. In these samples, sulfidation of Co and Mo was incomplete since the presence of oxysulfide and oxidic species was detected. However, the presence of oxysulfide molybdenum species indicated the transition of Mo6<sup>+</sup> to Mo5<sup>+</sup> during the sulfidation process. Despite this, sulfided phases, such as MoS2, Co9S8, and CoMoS, were predominant on 10 wt.% and 15 wt.% Mo supported catalysts. The relative contributions of Mo and Co species from the data obtained from sulfide CoMo/AT2 catalysts at Co/(Co + Mo) = 0.2 and 0.4 ratios are presented in Tables 4 and 5.



**Table 5.** Binding energies of the Co 2p3/<sup>2</sup> contributions obtained for sulfided 15 wt.% Mo CoMo/AT2 catalysts at Co/(Co + Mo) = 0.2 and 0.4.


Table 4 shows that MoS2 species were the main phase on both catalysts (69%–59%), while oxysulfide and oxide species represented about 30% to 19% and 12% to 10%, respectively. It is possible to observe that a fraction of the MoS2 phase decreased by 14% at high Co loadings, indicating that the sulfidation degree was higher for the Co/(Co + Mo) = 0.2 ratio than the Co/(Co + Mo) = 0.4. However, since the oxide Mo6<sup>+</sup> species were the same, a partial sulfidation of Mo species occurred at high Co loadings due to the presence of 60% more oxysulfide species than at low Co concentrations. This could indicate that the excess of Co caused a decrease in the capacity of sulfidation of the Mo species due to the formation of the CoMoO4 phase [72]. The Co species' contributions shown in Table 5 indicated that the mixed phase CoMoS represented 51% to 44%.

Besides, the Co9S8 and oxide phase contributed to 19% to 13% and about 37% of the Co supported on both catalysts. At Co/(Co + Mo) = 0.4, the amount of Co9S8 phase increased by 20%, whereas the CoMoS phase decreased by 15% and the oxide form was essentially the same. A high concentration of Co induced the formation of Co9S8 over the CoMoS phase due to the presence of CoMoO4 species, which were difficult to sulfide. Therefore, the segregated CoOh was easily reduced and formed its sulfided phase. At low Co concentrations, Co was capable of occupying the octahedral sites of the MoS2 phase. Thus, as this latter phase was more abundant, it led to an increase of the CoMoS phase concentration. Therefore, the Co/(Co + Mo) = 0.2 ratio was adequate for generation of the CoMoS phase, which is widely accepted to be the most active phase. Since the Co loading was different, a normalization of the fraction of each phase was carried out. Results of the concentration and mass fraction of the sulfided species are displayed in Table 6.

**Table 6.** Concentrations and atomic ratios of the Co and Mo species of the 15 wt.% Mo CoMo/AT2 catalyst at Co/(Co + Mo) = 0.2 and 0.4.


It is possible to observe that the concentration of sulfided and oxide Co species increased with the Co loading, whereas the MoS2 concentration dropped. Nevertheless, the CoMoS/Co9S8 and CoMoS/Co2<sup>+</sup> fractions decreased by 40% and 16%, respectively at Co/(Co + Mo) = 0.4. This means that as more Co was interacting with Mo at high Co concentrations, less MoS2 could be formed. In other words, the formation of the CoMoS phase was limited due to the formation of CoMoO4 and the sulfidation of the active phase was not complete. At low Co loadings, less Co9S8 and Co2<sup>+</sup> were present on the catalyst than at high Co loadings. As Figure 10 indicates, more CoOh was present at the Co/(Co + Mo) = 0.2 ratio than at 0.4. Hence, these CoOh species interacted with Mo species to form the CoMoS phase. However, the Co concentration was not enough to promote all Mo species, and more MoS2 was generated, i.e., a higher promotion of the active sites was achieved (see Table 2). In contrast, the excess Co provoked segregated species that could either form Co9S8 or Co oxide. Since, at high Co loadings, segregated Co oxide species in octahedral coordination were present on the surface, the sulfided Co phase tended to form. In comparison, the Co oxide species with tetrahedral coordination were difficult to sulfide and were more abundant at Co/(Co + Mo) = 0.4 than at Co/(Co + Mo) = 0.2.

The formation of the CoMoS phase provided electrophilic active sites that were selective to the hydrogenolysis route [5,73]. However, when the Co loading was increased, the HYD route did as well, due to the presence of more CoOh species. At Co/(Co + Mo) = 0.1, the production of benzene was higher than at the other ratios, indicating the presence of the CoMoS phase with hydrogenolysis sites. Nevertheless, the Co concentration was not enough to promote all MoS2 slabs. Hence, a fraction of Co was segregated in the Co9S8 phase and a minor fraction interacted with the support. In this sense, the HYD route was mainly caused by metallic sites from unpromoted MoS2 and Co9S8 phases than by CoMoS. The Co/(Co + Mo) = 0.2 ratio presented a higher catalytic activity and amount of CoMoS. This means that at this ratio, the amount of Co was adequate for direct interaction with Mo during the sulfidation process. The CoOh was available to promote the MoS2 phase and generate the CoMoS

phase over the other sulfided phases. At Co/(Co + Mo) > 0.2, activity decreased; however, the HYD route was enhanced. In these cases, there was enough Co to promote the MoS2 phase. Nonetheless, the excess Co led to the production of a more segregated Co9S8 phase than at lower ratios. Hence, the metallic sites responsible for hydrogenation reactions increased their number. As more Co was added, more CoTh and CoMoO4 phase could have been formed, which were difficult to sulfide and consequently, sulfidation was not optimal.

#### **3. Materials and Methods**

#### *3.1. Support and Catalysts Synthesis*

#### 3.1.1. Support Synthesis

Mixed oxide Al2O3–TiO2, (Al/Ti = 2 labelled as AT2) support was synthesized by the sol–gel method as described in a previous work [34]. As organic precursors, tri-sec-butoxide (Al (OCH(CH3) C2H5)3; Aldrich 99.9%, St. Louis, MI, USA) and titanium isopropoxide (Ti (OCH3H7)4; Aldrich 98%, St. Louis, MI, USA) were employed. As a solvent, 2-propanol (CH3)2CHOH; Baker 99.5%, Ecatepec, Estado de Mexico, Mexico), and, as hydrolysis catalyst, nitric acid (HNO3) were used. The nominal molar ratio used in all supports was 2-proponol:H2O:alkoxide:HNO3 = 325:100:5:1 [30]. 2-Propanol was cooled to 0 ◦C and under vigorous stirring, the theoretical amounts of Al and Ti were added. Then, HNO3 aqueous solution was added dropwise. The obtained gel was aged for 24 h at 273 K. Subsequently, it was dried at 333 K. Finally, the dried gel was calcined for 3 h at 773 K with a rate of 3 K min<sup>−</sup>1. AT2 support textural properties were: SBET = 359 m2 g<sup>−</sup>1, Vp = 1.1 cm<sup>3</sup> g<sup>−</sup>1, Dp = 7.7 nm, and PZC = 7.6, as previously reported by [32,34].

#### 3.1.2. Catalyst Synthesis

The Al2O3–TiO2 support was impregnated by the successive wetness impregnation method using an aqueous solution of ammonium heptamolybdate ((NH4)6Mo7O24·4H2O; Aldrich 99.9%, St. Louis, MI, USA) and cobalt nitrate ((Co(NO3)2·6H2O; Aldrich 99%, St. Louis, MI, USA). The non-promoted Mo series were loaded at 5, 10, 15, and 20 wt.% Mo, while only Co catalyst was loaded at 2 wt.%. The promoted CoMo catalyst series were impregnated following four different molar Co/(Co + Mo) ratios: 0.1, 0.2, 0.3, and 0.4. For monometallic catalysts, the Mo (or Co) solution was impregnated on AT2 support and was macerated at room conditions for 12 h. After that, it was dried at 393 K and calcinated at 673 K for 5 h. For the promoted catalyst, the Mo calcined materials were impregnated with the cobalt solution and the heat treatment was repeated. Before XPS analysis and the HDO of phenol tests, calcined CoMo samples were sulfided ex-situ in a glass tube reactor with a 10 vol.-% H2S/H2 mixture at 673 K for 2 h. After this, sulfided catalysts were immediately immersed in dodecane to avoid oxidation from air.

#### *3.2. Materials Characterization*

#### 3.2.1. Diffuse Reflectance Ultraviolet-Visible Spectroscopy

The diffuse reflectance UV-Vis (DRS UV-Vis) spectra of the synthesized support and promoted and unpromoted oxide catalyst series were recorded with a Lambda 35 spectrometer equipped with an integration sphere (Labsphere RSA-PE-20, North Sutton, NH, USA). The data acquisition was in the 200–1000 nm range with an interval of 0.5 nm and a scan speed of 240 nm<sup>−</sup>1. The spectra were recorded in the reflectance mode for infinitely thick samples (R∞) using the reflectance of MgO as a reference.

#### 3.2.2. Laser Raman Spectroscopy

Laser Raman spectroscopy (LRS) of unpromoted Mo catalyst at 10 and 15 wt.% were analyzed with a Perkin Elmer GX Raman FT-IR (Waltham, MA, USA), equipped with an Nd: YAG (1064 nm) laser and InGaAs detector. The data acquisition was carried out with a laser power of 40 to 300 mW at the 3600 to 100 cm−<sup>1</sup> Raman shift range with a resolution of 2 to 4 cm<sup>−</sup>1.

#### 3.2.3. Temperature Programmed Reduction

Temperature programmed reduction (TPR) experiments of the promoted and unpromoted oxide catalysts series were carried out with in an Altamira Instruments AMI-80 (Pittsburgh, PA, USA) apparatus provided with a thermal conductivity detector (TCD) interfaced to a data station. For each TPR test, 50 mg of catalyst precursor were set into a U-shaped quartz cell and pretreated in situ at 523 K for 1 h under 35 mL min−<sup>1</sup> He flow. After this, the catalyst precursor was cooled to room temperature. TPR analysis was performed under a stream of 10 vol% of H2/Ar, with a heating rate of 15 ◦C min−<sup>1</sup> up to 1100 K. A moisture trap was used to avoid measurement interference.

#### 3.2.4. X-Ray Photoelectron Spectroscopy (XPS)

The sulfided catalyst were analyzed in a K-alpha Thermo Fischer Scientific spectrometer equipped (Waltham, MA, USA) with a hemispherical electron analyzer and an Al Kα (hν = 1486.6 eV) X-ray source. The residual pressure was kept below 7 <sup>×</sup> 10−<sup>7</sup> Pa during data acquisition. The binding energies (BEs) were referenced to the C 1s peak (284.9 eV) to account for the charging effects. The areas of the peaks were computed after fitting the experimental spectra to Gaussian/Lorentzian curves and removing the background (Shirley function). After that, surface atomic ratios were calculated from the peak area ratios normalized by the corresponding atomic sensitivity factors. The spectra were analyzed from the 222 to 244 eV region where the Mo 3d levels were located. The surface relative abundance percentages of the Mo and Co species were calculated according to the methodology proposed by Chen et al. [38]. Moreover, we estimated the relative amount of the CoMoS mixed phase, considering the Co species in this phase:

$$\left[\mathrm{Mo^{4+}}\right](\%) = \frac{\mathrm{A\_{Mo^{4+}}}}{\mathrm{A\_{Mo^{4+}}} + \mathrm{A\_{Mo^{5+}}} + \mathrm{A\_{Mo^{6+}}}} \times 100,\tag{1}$$

$$\left[\text{CoMoS}\right]\left(\%\right) = \frac{\text{A}\_{\text{CoMoS}}}{\text{A}\_{\text{CoMoS}} + \text{A}\_{\text{CoPoS}} + \text{A}\_{\text{Co}^{2+}}} \times 100\,,\tag{2}$$

$$\mathbf{C}(\text{MoS}\_2) = \mathbf{C}(\text{Mo}) \left[ \text{Mo}^{4+} \right] / 100 \,, \tag{3}$$

$$\text{C}(\text{CoMoS}) = \text{C}(\text{Co})[\text{CoMoS}]/100,\tag{4}$$

$$f\_{\frac{\text{CoMoS}}{\text{MoS}\_2}} = \frac{\text{C}(\text{CoMoS})}{\text{C}(\text{MoS}\_2)},\tag{5}$$

where AMo4+, AMo5+, AMo6+, ACoMoS, ACo9S8, and ACo2<sup>+</sup> are the area of each species fitted from the Mo3d and Co2p XPS spectra; C(Mo) and C(Co) are the theoretical mass concentrations of Mo and Co per gram of oxidic catalyst (gmetal/gcatalyts); C(MoS2) and C(CoMoS) are the mass concentrations of MoS2 and CoMoS species per gram of catalyst; and fCoMoS/MoS2 is the mass ratio of the CoMoS and MoS2 species. The C(Co9S8) and C(Co<sup>2</sup>+) concentrations and mass ratios were quantified in the same way.

#### 3.2.5. Catalytic Performance

To evaluate the catalytic performance of the Mo, Co, and CoMo sulfided catalyst, hydrodeoxygenation of phenol was carried out. HDO of phenol facilitates an understanding of the functionality and reaction mechanisms in catalytic tests since it is a relatively simple molecule and it constitutes the main component in the remnant of the HDO of guaiacol [40]. The reaction took place in a Parr Series 4540 high-pressure batch reactor (Parr Instrument Co., Moline, IL, USA) equipped with a Parr 4842 controller, mechanic impeller, wall-baffles, liquid and gas inlet/outlet gas valves, internal

thermocouple, and pressure gauge. The reaction mixture consisted of phenol (500 ppm of oxygen) and CS2 (100 ppm of sulfur) dissolved in 100 mL of n-dodecane and 0.1 g of freshly sulfide catalysts with a particle size between 150 and 180 μm. Catalyst sulfidation was carried out as was described in the XPS methodology. The reactor was pressurized up to 1.4 MPa with N2 to flush air and to prevent oxidation of the sulfide catalysts. After this, the reactor was heated up to 593 K and kept an isothermal operation mode during the reaction. Then, N2 was vented slowly, and hydrogen was introduced up to 5.5 MPa. The reactor operated in an isobaric mode during the reaction time with manual addition of H2 and vigorous agitation of 1000 rpm. The reaction time started from the incorporation of H2. Small samples were collected (0, 10, 20, 30, 45, 60, 90, 120, 180, 240, and 300 min), ensuring that the sum of the volume samples was less than 5% of the initial volume. A gas chromatograph (Agilent 7820A, Santa Clara, CA, USA) equipped with a CP Sil-5 CB capillary column (100% dimethylpolysiloxane, 60 m × 0.32 mm) and a flame ionized detector (FID) were used for the quantification of the products' concentrations. The initial reaction rates, reagent mol transformed per time of reaction, and mass of a sulfided catalyst (mol g<sup>−</sup>1s−1) were compared.

#### **4. Conclusions**

In the unpromoted catalysts at high Mo loadings, more easily reduced MoOh species were formed, and more active sites were present on these catalysts than at low Mo loadings. Nevertheless, a hardly reduced MoO3 presence was detected at a high Mo content. On the other hand, the functionalities to produce benzene and cyclohexene did not change with the Mo content since all catalysts showed the same selectivity. This indicated that the active sites could cleave the CAR–OH bond but could not properly hydrogenate the π-bonds of the cyclic compounds, i.e., S-edges sites played a more major role than Mo-edges.

Catalytic activity and selectivity were related to the CoOh content. A Co/(Co + Mo) = 0.2 ratio presented a maximum activity for these catalysts. At this Co concentration, CoOh was the main Co oxide coordination species. These CoOh species properly promoted the Mo oxide species and subsequently formed more CoMoS phase than at higher ratios. At a low Co content, it was insufficient to totally promote the Mo oxide species. Therefore, CoMoS phase formation was limited and the MoS2 phase was predominant since benzene was the main product. Plus, Co may have generated CoTh species that were difficult to reduce. At high ratios, the excess Co concentration led to the CoMoO4 phase, and the formation of the CoMoS phase was inadequate. Hence, activity dropped compared with the Co/(Co + Mo) = 0.2 ratio. Moreover, the surplus of Co was segregated and formed CoOh, which transformed into the Co9S8 phase, leading to an enhancement of the HYD route. Finally, the segregated Co oxide species may have formed CoTh as well; thus, catalytic activity decreased.

**Author Contributions:** C.E.S.-V. and O.U.V.-M. performed and analyzed the characterization experiments and data; J.A.T.-P. conceived, designed, and performed experiments, analyzed the data, and wrote the manuscript; J.A.d.l.R.H. contributed to writing—review of the manuscript, funding acquisition, and was the project administrator and laboratory chief.

**Funding:** This research was funded by Consejo Nacional de Ciencia y Tenología (CONACYT—Mexico), grant number 237857.

**Acknowledgments:** The authors are grateful to CONACYT for the financial support 237857 and for the scholarship of J.A. Tavizón-Pozos with number 221991. The authors acknowledge financial support from Instituto Politénico Nacional (Proyecto SIP 20196722) and CONACYT for the projects CB-2017-2018 #A1-S-32418 and the Cátedras-CONACYT number 216.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Synthesis and Regeneration of Nickel-Based Catalysts for Hydrodeoxygenation of Beech Wood Fast Pyrolysis Bio-Oil**

**Caroline Carriel Schmitt 1,2,\*, María Belén Gagliardi Reolon 1,3, Michael Zimmermann 1, Klaus Raffelt 1, Jan-Dierk Grunwaldt 1,4 and Nicolaus Dahmen <sup>1</sup>**


Received: 17 August 2018; Accepted: 8 October 2018; Published: 12 October 2018

**Abstract:** Four nickel-based catalysts are synthesized by wet impregnation and evaluated for the hydrotreatment/hydrodeoxygenation of beech wood fast-pyrolysis bio-oil. Parameters such as elemental analysis, pH value, and water content, as well as the heating value of the upgraded bio-oils are considered for the evaluation of the catalysts' activity and catalyst reuse in cycles of hydrodeoxygenation after regeneration. The reduction temperature, selectivity and hydrogen consumption are distinct among them, although all catalysts tested produce upgraded bio-oils with reduced oxygen concentration, lower water content and higher energy density. Ni/SiO2, in particular, can remove more than 50% of the oxygen content and reduce the water content by more than 80%, with low coke and gas formation. The evaluation over four consecutive hydrotreatment reactions and catalyst regeneration shows a slightly reduced hydrodeoxygenation activity of Ni/SiO2, mainly due to deactivation caused by sintering and adsorption of poisoning substances, such as sulfur. Following the fourth catalyst reuse, the upgraded bio-oil shows 43% less oxygen in comparison to the feedstock and properties comparable to the upgraded bio-oil obtained with the fresh catalyst. Hence, nickel-based catalysts are promising for improving hardwood fast-pyrolysis bio-oil properties, especially monometallic nickel catalysts supported on silica.

**Keywords:** hydrodeoxygenation; fast-pyrolysis bio-oil; nickel catalyst; upgrading

#### **1. Introduction**

The increase in global energy demand, depletion of fossil fuel reserves and climate change issues have drawn attention to renewable alternatives, particularly to biomass [1–3], considering its CO2 neutrality for fuel applications and widespread availability [1,4]. Products such as heat, power, biomaterials, chemical compounds, and transportation fuels can be obtained from biomass [5]. Considering this purpose, thermochemical, chemical-catalytical or biological processes are used. Regarding the first category, combustion, gasification and pyrolysis are most common. Although considered the simplest way to convert biomass in either power or heat, combustion shows high emissions and ash generation [5]. Gasification is considered a very efficient method to obtain fuels, but it requires a high investment due to large-scale installations, storage and transportation [6]. Therewith, pyrolysis has been considered a promising process as it balances simple operation techniques with reasonable costs.

Pyrolysis is a thermochemical process in which the biomass is heated and converted in an inert atmosphere into a liquid fraction called bio-oil, a carbon-rich solid (biochar) and a mixture of non-condensable gases [7]. The bio-oil obtained has poorer physical and chemical characteristics if compared to liquid fossil fuels. The heating value is usually lower, only 40–50% compared to conventional fossil fuels (42–45 MJ/Kg), mainly due to the high oxygen and water content. Additionally, it shows high viscosity, low chemical stability and solid particles [8–10] due to incomplete solid separation or polymerization reactions during storage, for example. Carboxylic acids present in the bio-oil composition lead to high acidity (pH value around 2–3.7), resulting in a bio-oil with potentially corrosive properties. Furthermore, it is highly unstable during storage due to ongoing chemical reactions, resulting in larger molecules by polymerization, etherification and esterification [9], for example. Additionally, it is immiscible with fossil fuels and tends to undergo phase separation when stored for a long time. Considering these poor fuel properties, the direct application of bio-oil is limited to furnaces and boilers, being unsuitable for application in gas turbines, diesel engines and other applications without further treatment [11]. Concerning bio-oil production today, wood with low ash content is used, leading to relatively "well-natured" bio-oils. Using ash-rich feedstocks, bio-oil yield and quality is decreased, while the tendency for phase separation increases.

To improve these properties and obtain a product resembling diesel fuel, bio-oil requires an additional upgrading treatment. Upgraded bio-oil can then be used as feedstock for producing chemicals, such as phenols for resin production, additives for fertilizers and pharmaceutical industries, as well as flavoring agents in the food industry [12]. Regarding terms of energetic use, upgraded bio-oil might be used as feedstock in oil refineries and fuels in engines [13].

A variety of upgrading techniques already have been proposed, such as catalytic cracking, hydrodeoxygenation (HDO) and esterification in supercritical fluids [14]. Among them, HDO appears to be a propitious route, due to its flexibility with respect to the biomass feed, the good economy of the input materials, and its compatibility with refinery infrastructures [15]. HDO is a high-pressure catalytic treatment in which oxygen is removed by hydrogen resulting in water, which is environmentally benign [16]. Usually, sulfides, noble metals and transition metal catalysts are used [17]. Noble metals such as Pt, Pd and Ru have been evaluated widely for HDO and are often the first choice in hydrogenation reactions. Additionally, they have a low tendency to be poisoned by the sulfur present in the bio-oil [18]. Their relatively high costs, however, prevent them from being widely used. Recently, nickel-based catalysts have become more attractive, considering their lower price, availability, activity and reduced hydrogen consumption [10,19]. Jin et al. [20] evaluated a series of nickel-based catalysts on different supports (SiO2, Al2O3, AC and SBA–15 mesoporous silica) for the HDO of anisole, used as a model compound. Boscagli et al. [21] investigated the HDO of the bio-oil light phase over a variety of nickel-based catalysts (NiCu/Al2O3, Ni/SiO2, Ni/ZrO2, Ni/TiO2 and NiW/AC). Dongil et al. [22] studied the HDO of guaiacol over nickel-based catalysts, using different carbon-based supports.

The combination of nickel in bimetallic catalysts has also attracted attention for HDO, especially in combination with copper. Ardiyanti et al. [23] evaluated the application of NiCu at different loadings supported in δ–Al2O3 for upgrading of model compounds and fast-pyrolysis bio-oil. Dongil et al. [24] investigated the effect of Cu loading on nickel catalysts supported in carbon nanotubes over the HDO of guaiacol. Mortensen et al. [25] screened different catalysts, including NiCu/SiO2 for phenol HDO and, more recently, Boscagli et al. [26] tested NiCu/Al2O3 for the HDO of phenol and bio-oils reusing the catalyst after a regeneration step.

The investigation of nickel and nickel–copper catalysts on SiO2 and ZrO2 supports with real feedstock (fast-pyrolysis bio-oil) is of interest, especially with respect to the lower acidity in comparison to Al2O3, a commonly studied support for HDO catalysts [27]. Many studies are focused on alumina-supported catalysts (Al2O3) [23,26] and upgrading applying model compounds [20,22] but supports with higher stability are required. The alumina-support is well known for its acidity, tendency for increased coke formation, low water tolerance, and conversion to boehmite, resulting in the oxidation and deactivation of the active metal [15,28,29]. According to He et al. [28], the selection of the support for HDO of bio-oils, must consider the resistance to the water content, the acidity of the supports to reduce coke formation, the porosity and its ability to keep the active metal dispersed for the activation of hydrogen. Hence, the investigation of different supports, such as SiO2 and ZrO2, appears interesting, especially when including catalyst regeneration in consecutive cycles of HDO-regeneration, evaluating the thermal stability of the catalyst [30]. Presently, only a few works have investigated the regeneration and evaluation of the reuse of the catalyst [26]. Most works, in fact, only consider one regeneration step and do not contemplate Ni catalysts [26,31,32]. It is an essential step to reduce costs, minimizing the waste generation at the same time helping to increase the reusability and recyclability of the catalysts, extending its lifetime [33]. Additionally, previous studies considered the HDO of model compounds whereas others considered the application of fast pyrolysis bio-oil. Usually, different temperature, pressure and reactor designs are used, which makes the comparison of the performance of different nickel-based catalysts difficult.

The current work synthesizes, characterizes and evaluate four nickel-based catalysts for a multi-phase fast-pyrolysis bio-oil upgrading. Supports with higher stability (SiO2 and ZrO2) are selected. The catalyst with the best performance is then reused in subsequent HDO-regeneration steps, resulting in four consecutive reactions. Finally, the performance and the catalytic activity along the HDO-regeneration steps are assessed and discussed.

#### **2. Results**

#### *2.1. Characterization of the Synthesized Catalysts*

The results obtained from the temperature programmed reduction (H2–TPR), in Figure 1, were useful to identify the catalysts' reduction temperatures before the hydrodeoxygenation (HDO) reactions. The H2–TPR profile for Ni/SiO2 showed a clear peak at 350 ◦C, while the peak for Ni/ZrO2 is found between 350–400 ◦C, in agreement with literature [21,34,35]. Since the reduction of bulk Ni oxide occurs around 400–450 ◦C [36,37], the reduction temperature of Ni/SiO2 and Ni/ZrO2 was set to 500 ◦C to ensure a full reduction before hydrodeoxygenation (HDO) reactions. The addition of Cu to Ni [36,37], as well as the higher loading of Ni [38,39] seems to promote the reduction of Ni oxide, as for both bimetallic NiCu catalysts the temperature of reduction is lower compared to monometallic Ni catalysts. Concerning NiCu/SiO2, it occurs at 300 ◦C, while for NiCu/ZrO2 it was closer to 200 ◦C. Different peaks are present in the H2–TPR profiles, attributed to bulk nickel oxide reduction, reduction of Cu(II) to Cu(0) (below 250 ◦C), as well as reduction of bimetallic NiCu species which, according to Ardiyanti et al. [40], should occur approximately in the range of 290–390 ◦C. Regarding the current catalysts, most of the reduction was observed at lower temperatures. Consequently, the reduction of NiCu catalyst was set at 350 ◦C.

Powder X-ray diffraction (XRD), provided information about the crystalline structure of the catalyst (Figure 2). Both SiO2-supported catalysts show similar XRD patterns. The reflections located at 37.25◦ and 43.29◦ indicate the presence of Ni oxide (NiO) in the calcined catalysts. Additionally, metallic Ni was identified due to the reflections at 44.49◦, 51.85◦, 76.38◦, 92.93◦ and 98.44◦. Following the reduction, reflections attributed to NiO disappeared, remaining just metallic Ni reflections. A similar behavior can be seen for Ni/ZrO2 and NiCu/ZrO2 (Figure 2). Both NiO and metallic Ni are present in the calcined catalyst. Subsequently, the reduction reflections of NiO are no longer observed for Ni/ZrO2 and showed a reduced intensity for NiCu/ZrO2. Reflections attributed to copper were not observed in the bimetallic catalysts, which can be a result of high dispersion of the metal, as well as low concentration [21,23].

**Figure 1.** Temperature programmed reduction profile for the nickel-based catalysts. TCD: Thermal conductivity detector; a.u.: arbitrary units.

**Figure 2.** X-ray powder diffraction of the freshly synthesized catalysts, Ni/ZrO2, Ni/SiO2, NiCu/ZrO2 and NiCu/SiO2. a.u.: arbitrary units.

The crystallite sizes were estimated using the Scherrer equation. The crystallite size was 17.7 nm for Ni/SiO2, whereas for NiCu/SiO2, the value was estimated to be 21.4 nm. NiCu/ZrO2 showed a crystallite size of 43.3 nm and the crystallite size of Ni/ZrO2 was estimated at 9.7 nm.

The metal concentration, as well as the specific surface area, is compiled in Table 1. The Ni/SiO2 catalyst displays the highest specific surface area (215 m2/g), while NiCu/ZrO2 shows the lowest (50 m2/g). Usually, SiO2-supported catalysts show higher surface areas in comparison to ZrO2-supported catalysts [41]. The BET surface area of the catalysts, as well as the micropore area and volume, is reduced with the addition of Cu and higher nickel loading. This behavior was also observed by Dongil et al. [24] and Zhang et al. [42]. Furthermore, no micropores were observed in the zirconia-supported catalysts.

**Table 1.** BET surface area, pore area, volume, diameter, and metal content in the freshly synthesized catalysts.


#### *2.2. Hydrotreatment Reactions*

#### 2.2.1. Upgraded Bio-Oil Yields and Properties

The HDO reactions with Ni/SiO2 showed the highest yield of upgraded bio-oil (49.36 wt.%), and the lowest yields of the aqueous phase (35.57 wt.%) and solids (Table 2). These tendencies were not clear for the remaining catalysts. The lowest yield of upgraded bio-oil (45.32 wt.%) was obtained with Ni/ZrO2, while NiCu/ZrO2 showed the highest production of solids (0.32 wt.%) and aqueous phase (43.52 wt.%). The highest production of gas was obtained with Ni/ZrO2 (4.54 wt.%) whereas the lowest was obtained with NiCu/SiO2 (3.56 wt.%).

**Table 2.** Mass balance, elemental analysis and physicochemical properties of the upgraded bio-oils obtained with fresh Ni-based catalysts.


The elemental composition of the upgraded bio-oils by the different nickel-based catalysts is presented in Table 2. The concentration of carbon increases in all upgraded bio-oil in comparison to the concentration in the feed (57.31 wt.% dry basis) [43]. The highest carbon content (73.15 wt.% dry basis) was obtained applying Ni/SiO2. The remaining oils upgraded with the other catalysts displaying a

concentration of around 72.25 wt.% in dry and 67.38 wt.% in wet basis, respectively. The hydrogen concentration was slightly higher in the upgraded bio-oils in comparison to the initial feed, while the oxygen content was reduced in comparison to the original beech wood bio-oil (35.84 wt.% in dry basis). The upgraded bio-oil over Ni/SiO2 shows the lowest oxygen concentration (17.86 wt.% dry basis), followed by NiCu/SiO2 (19.00 wt.% dry basis), Ni/ZrO2 (19.46 wt.% dry basis) and NiCu/ZrO2 (20.35 wt.% dry basis).

The water content in the upgraded bio-oils was reduced significantly from 26.77 wt.% in the original feed to values between 4.85 wt.% and 7.30 wt.% in the upgraded bio-oils. More than 70% of the water content in the original oil was removed by the hydrotreatment. The upgraded bio-oil over Ni/SiO2 showed the lowest water content (4.85 wt.%), while NiCu/SiO2 had the highest value (7.30 wt.%). The opposite was observed for the upgraded aqueous phase. While the upgraded aqueous phase obtained with Ni/SiO2 showed 74.35 wt.% of water, the aqueous phase obtained with NiCu/SiO2 showed 67.15 wt.% of H2O (Table S1). This is because the water removed from the upgraded bio-oils mainly was concentrated in the aqueous phase.

The higher heating value (HHV) also changed by the reactions. In the feedstock, this value was 24.33 MJ/kg, whereas, for the upgraded bio-oils, the HHV increased to values ranging from 29.86 MJ/kg to 31.18 MJ/kg in the case of SiO2 supported catalysts. A slight increase in the pH value was observed, except for the upgraded bio-oil with Ni/ZrO2. The density of the upgraded bio-oils decreased after the hydrotreatment. Comparing the density of the upgraded bio-oils with the density of the heavy phase, for example, the value was reduced from 1.19 g/cm<sup>3</sup> to values between 1.09 g/cm3 and 1.12 g/cm3. The upgraded bio-oil over NiCu/SiO2 displayed the lowest density (1.09 g/cm3), while the highest corresponded to Ni/ZrO2 (1.12 g/cm3).

Further information regarding the upgraded bio-oils composition was obtained by proton nuclear magnetic resonance (1H-NMR), depicted in Figure 3. Table 3 shows the assignment of chemical groups to the integration range of the spectra [44]. The aqueous phases are compared to the light phase (LP) of the feedstock, whereas the upgraded bio-oil is compared to the heavy phase (HP) of the feedstock.

**Figure 3.** 1H-NMR spectra integrals of the bio-oil components, HP (**top**) and LP (**bottom**), in contrast to the products (upgraded bio-oil and aqueous phase) obtained by different catalysts.


**Table 3.** Integration ranges of 1H-NMR spectra and their corresponding proton assignment [44].

The signals for aldehydes (10.1–9.5 ppm) are very small in the feed, below 0.1 mmol/g sample in both LP and HP. They were not observed in the upgraded products as they are very reactive even at mild conditions and, therefore, are hydrogenated to alcohols [45]. The main signals for the upgraded bio-oil were found in the region of α-protons to carboxylic acid or keto-groups and α-protons to unsaturated groups (3.0–1.5 ppm).

The aromatic (8.5–6.0 ppm) were concentrated mostly in the upgraded bio-oil and almost absent in the aqueous phase (7.0 versus 0.2 mmol/g sample on average). No significant differences among catalysts were observed. A slight increasing tendency could be observed for the concentration of alcohols, ethers, and alkenes (4.3–3.0 ppm), although it was not significantly different among all tested catalysts. The concentration of protons in this region was more abundant in the upgraded bio-oil (14.5 mmol/g sample on average) in comparison to the aqueous phase (8.2 mmol/g sample on average). The abundance of alkanes (1.5–0.5 ppm) almost triples in comparison to the feed (heavy phase) and was significantly higher in the upgraded bio-oil than in the aqueous phase (23.4 versus 2.2 mmol/g sample).

The accumulation of water in the aqueous phase is also confirmed by the 1H-NMR measurements, considering that the main signal obtained for the aqueous phase was found in the carbohydrates, water, and O-H exchanging groups (6.0–4.3 ppm). The high concentration of protons in this region is attributed to the removal of water from the bio-oil [46,47], in agreement with Karl–Fisher results. On the other hand, the proton in this region decreases in all upgraded bio-oils, especially for Ni/SiO2, which produced the upgraded bio-oil with the lowest water concentration (Table 3).

To identify the main compounds in the upgraded bio-oils, as well as to investigate differences in selectivity among the catalysts tested, the upgraded bio-oils were analyzed qualitatively by gas chromatography-mass spectrometer (GC-MS), depicted in Figure 4 and Table 4. The chromatograms of the upgraded bio-oils are discussed in comparison to the chromatograms of the heavy phase (feedstock), available in the Supplementary Material (Figure S3). The chromatograms of the upgraded aqueous phases, as well as the light phase of the feedstock, are also available as Supplementary Material (Figures S3 and S4).

Typical compounds were observed in the upgraded bio-oils, such as carboxylic acids, ketones, phenolic compounds and others [48]. The main reaction pathways identified and later discussed are available in Figure S6.

A variety of ketones, especially but not limited to cyclic forms, were identified in all the upgraded bio-oils, such as 2-pentanone, cyclopentanone, 3-methyl-cyclopentanone, 2-ethyl-cyclohexanone, cycloheptanone and others, in agreement with Boscagli et al. [46] and Ardiyanti et al. [23]. They are attributed as products of sugar conversion and its derivatives.

Aromatic compounds initially present in the feedstock also were observed in the upgrade-oil, such as guaiacol (2-methoxy-phenol), phenol and 4-ethyl-2-methoxy-phenol. Acetic acid also was present in the feedstock as well as in the upgraded products.

**Figure 4.** Chromatograms of the upgraded bio-oils over different nickel-based catalysts.

**Table 4.** Retention time of the main compounds in the upgraded bio-oil identified by GC-MS.


After the upgrading reaction, compounds initially present in the bio-oil mixtures, such as furfural, were completely converted for all the catalysts tested. The same was observed for eugenol and compounds with GC retention times higher than 30 min, such as vanillin. Although very similar compositions of all upgraded bio-oils, some differences in selectivity were observed for the catalysts. 1-propanol (5.82 min) was observed only in the bio-oils upgraded with bimetallic catalysts, as well as the peak at 16.04 min, attributed to tetrahydro-2-furanmethanol (tetrahydrofurfuryl alcohol), also observed in the bio-oils upgraded with NiCu catalysts. It shows that furfural is completely hydrogenated to tetrahydrofurfuryl alcohol over these catalysts [49]. Earlier studies observed the selectivity of copper-containing catalysts for furfural hydrogenation, particularly at high temperatures [50]. Furfural also can be converted to cyclopentanone [48,49] and 2-pentanone [8,51], both identified in the products, as well as other cyclopentanones [52].Compounds such as methane, identified in the gas phase and later discussed, also can be derived from furfural conversion; the conversion of furfural to furfuryl alcohol and later to furan leads to methane formation [53]. Furthermore, the peak at 17.57 min, attributed to propylene glycol, was observed only in the upgraded bio-oils as well as in the aqueous phases (Figure S5) upgraded with NiCu catalysts. Propylene glycol can be obtained from hydrogenation of hydroxyacetone [54], which was converted completely after the HDO reactions (see peak at 12.46 min Figure S3). While copper-containing catalysts seem to favor the production of propylene glycol in comparison to other catalysts [55], nickel catalysts seems to follow a different pathway. Resulting from the C-O bond cleavage of propylene glycol [54], 1-propanol was observed in the oils upgraded with bimetallic catalysts. Furthermore, compounds initially absent in the feedstock, such as 2-methoxy-4-propyl-phenol, resulting from the hydrogenation of the double bond of eugenol, were identified in the upgraded bio-oil [47]. A peak at a retention time of 26.9 min was observed in the upgraded aqueous phases obtained with bimetallic catalysts, although the identification of the compound was not possible.

Differences among the feed (LP) and upgraded products (aqueous phases), as well as among the products obtained with different catalysts, were observed. For example, 2-methyl-propanol (13.08 min) was observed in all upgraded aqueous phases, except for NiCu/SiO2. The peak attributed to tetrahydro-2-furanmethanol increased significantly, mainly in the aqueous phases obtained by bimetallic catalysts, in the same way as observed in the upgraded bio-oils. Propanoic acid (16.77 min) was observed for all aqueous phases.

#### 2.2.2. Hydrogen Consumption and Gaseous Products

The hydrogen consumption was considered for the evaluation of the different catalysts. NiCu/SiO2 presented the highest consumption (239.3 NL/kg bio-oil), followed by NiCu/ZrO2 (201.6 NL/kg bio-oil), Ni/SiO2 (186.2 NL/kg bio-oil) and Ni/ZrO2 (181.9 NL/kg bio-oil). Yin et al. [35] reported, higher hydrogenation activity can be seen when adding copper to Ni catalysts, due to changes in the catalytic activity and selectivity, favoring some hydrogenation reactions [56]. The hydrogen consumption can be correlated with the H/C molar ratio. A higher H/C molar ratio and lower O/C molar ratio indicates an upgraded bio-oil with improved properties [15]. The catalysts evaluated resulted in H/C ratios between 1.37 (Ni/ZrO2) to 1.53 (NiCu/SiO2), showing a tendency between the hydrogen uptake and the H/C molar ratio. The same tendency was not observed for the O/C ratio. The lowest ratio was observed using Ni/SiO2 (0.18) whereas the highest was for NiCu/ZrO2 (0.28). This indicates that a higher consumption of H2 does not reflect in the HDO of the feedstock.

Although a very similar total gas production (Figure 5) was obtained for all tested catalysts, Ni/ZrO2 showed the highest total gas production (1.06 mol/kg bio-oil), followed by Ni/SiO2 (0.99 mol/kg bio-oil), NiCu/ZrO2 (0.90 mol/kg bio-oil) and NiCu/SiO2 (0.86 mol/kg bio-oil). The total gas production was determined mainly by the production of CO2, the most abundant gas product for all four catalysts, in agreement with other studies [35,40,41,46]. Decarboxylation of carboxylic

acids can result in CO2 formation [46]. Carbon monoxide, methane, and other gases such as propane, propene, ethane and ethene were formed in smaller amounts.

**Figure 5.** Production and composition of the gas phase for the nickel-based catalysts tested.

Monometallic Ni catalysts produced the highest amount of CO2, followed by NiCu/ZrO2 and, later, by NiCu/SiO2. According to Gallakota et al. [14], in an ideal scenario all C atoms should be converted to hydrocarbons without CO2 formation. Nevertheless, since the total gas production of the four catalysts represented around 4 wt.% (see mass balance Table 2), the produced amounts of CO2 are, in general, quite reduced in these four cases. The formation of CO2 might indicate lower hydrogen consumption [57]. This effect was observed when the CO2 production was compared to the hydrogen consumption—both variables were inversely proportional (Figure S2), as reported by Boscagli et al. [46]. This tendency only was noticed in the CO2 production, but not for the remaining produced gases.

The formation of methane was observed mainly for NiCu/SiO2. The smallest concentration of methane was obtained for Ni/SiO2. Methane resulted from the hydrogenation of carbohydrates, acetic acid decomposition, cleavage of C-C bonds of alcohols or even from methoxy groups demethylation [57]. It is important to highlight that the higher methane production was observed with NiCu/SiO2, the catalyst that showed the highest H2 consumption. This is an indication of hydrocracking of the molecules and excessive hydrogen consumption [35,58].

CO formation was more abundant in the SiO2-supported catalysts. NiCu/SiO2 displayed the highest amount, followed by Ni/SiO2, Ni/ZrO2 and NiCu/ZrO2. The formation of CO can be a result of the C-O cleavage of different groups, such as carboxylic acids, aldehydes, and alcohols. The loss of C via CO (decarbonylation), is less advantageous than the loss of CO2 (decarboxylation), considering more O is removed per mole of lost carbon [57].

#### 2.2.3. Catalysts Characterization

The spent catalysts were evaluated mainly in terms of metal leaching, XRD, sintering, carbon deposition as well as surface area and composition. Inductively Coupled Plasma Emission Spectroscopy (ICP-OES) was further used to evaluate the metal content present in the upgraded aqueous phases and determine the metal leaching. Ni/SiO2 presented 0.8% of Ni leached while using Ni/ZrO2 0.43% of Ni was washed into the aqueous phase. Both NiCu catalysts showed 0.16% of Ni leached while NiCu/SiO2 and NiCu/ZrO2 displayed 0.08% and 0.04% of Cu leached, respectively. Similar leaching levels also were found by Boscagli et al. [26] for a reaction at 240 ◦C compared

to a NiCu/Al2O3 catalyst. The extent of leaching seems to correlate with the support. The lowest proportions of metals leached were observed in ZrO2-supported catalysts. Reported previously, ZrO2 seems to be stable in the harsh reaction conditions [28,40,42,59]. Due to the severity of the HDO reactions, the structure, morphology and texture of the catalysts might be affected [40], therefore, the spent catalysts were analyzed by XRD after the reactions. No significant differences were observed between the X-ray diffraction patterns of the fresh and spent catalysts, except for NiCu/ZrO2. The small diffractions attributed to NiO disappeared in the spent catalysts, indicating further nickel reduction. The XRD patterns are available in Figure S1. The crystallite sizes were calculated for the spent catalysts. The crystallite size of NiCu/SiO2 increased from 21.4 nm to 43.0 nm after the reaction. Such an increase in the crystallite size was previously observed [60] and is an indication of sintering, which might result in loss of the catalyst activity [61]. Conversely, the crystallite size of Ni/ZrO2, Ni/SiO2 and NiCu/ZrO2 remained in the same range observed for the fresh catalysts. The composition of selected particles was analyzed by EDX (Energy Dispersive X-ray spectroscopy) and the results obtained for the fresh and spent catalysts are presented in Table 5.

**Table 5.** Composition of selected particles of different Ni-based catalysts (fresh and spent) obtained by EDX.


Carbon deposition was observed for all catalysts after HDO, with lower concentration observed in Ni/SiO2. Poisoning substances, such as sulfur and calcium, were observed near the detection limit in the spent forms. Sulfur appeared in all spent catalysts, whereas calcium and iron were observed only in zirconia-supported catalysts. The Ni proportion over the catalysts' surfaces decreased, attributed mainly to the carbon deposits of higher molecular weight polymerization products [35]. The specific surface areas of the spent catalysts were reduced in comparison to the fresh ones, especially for the silica-supported catalysts. The specific surface area of Ni/SiO2 was reduced to 46 m2/g, a sharp reduction in comparison to the original surface area (215 m2/g). NiCu/SiO2 showed a reduction from 156 m2/g to 36 m2/g. It can be attributed to the fact that the pores probably were blocked by carbonaceous deposition. The micropore area of Ni/SiO2 was reduced to 8 m2/g, whereas the micropore area of NiCu/SiO2 was reduced to 3 m2/g. The reduction in the surface area of the zirconia-supported catalyst was less significant. Ni/ZrO2 showed a reduction to 57 m2/g (original: 65 m2/g), whereas NiCu/ZrO2 was reduced from 50 to 36 m2/g. The reduction of the surface area is attributed to deposition of carbon in all cases, as already documented by many authors [41,62–64].

Considering that SEM-EDX (Scanning Electron Microscopy/Energy Dispersive X-ray spectroscopy) provides the elemental mapping of only selected regions [65], further composition analysis was performed measuring the active metals and poisoning substances in the bulk catalyst (Table 6).

Agreeing with EDX results, an increase in the calcium concentration over both zirconia-supported catalysts was observed. Sulfur, considered a very persistent poisoning substance of nickel [26,66], increased in all catalysts tested after the reaction, with slightly lower concentrations at Ni/SiO2. Interestingly, no differences were observed in the XRD spectra, such as nickel sulfide formation [66]. The concentration of active metals (Ni and Cu) was reduced in the spent forms in comparison to the fresh catalyst. This behavior correlates to the carbon deposition. Accompanying a higher concentration of carbon, which initially is absent in the fresh catalysts, the average concentration of nickel and copper

in the spent catalyst decreases. Furthermore, leaching also can play a role in the reduction of active metal in the catalyst, as discussed previously. The carbon concentration obtained by elemental analysis shows the same tendencies as observed with the EDX measurements.


**Table 6.** Metal content and poisoning substances on the catalyst before and after the reaction.

\* Results obtained by ICP-OES. \*\* Results obtained by elemental analysis.

#### *2.3. Cycles of HDO and Regeneration: Catalyst and Product Behavior*

Ni/SiO2 was selected for further consecutive HDO reactions as soon as the original reactions with the fresh catalyst were carried out. Key parameters were considered for selection of this catalyst for consecutive HDO-regeneration cycles: HDO activity, water and carbon concentration in the upgraded bio-oil, hydrogen consumption as well as solid formation. The upgraded bio-oil obtained with Ni/SiO2 showed the highest carbon content and the lowest oxygen content in comparison to other catalysts, an indication of the improvement of the oil quality (higher energy density and better chemical stability). Additionally, upgrading with Ni/SiO2 resulted in an upgraded bio-oil with the lowest water concentration, low hydrogen consumption, lowest amount of solids formed, highest HHV and pH (Table 2). Finally, the smallest production of methane also was obtained for this catalyst (Figure 5). Therefore, Ni/SiO2 was selected to carry out the following regeneration steps.

Along the cycles of HDO-regeneration, the catalyst was characterized in the intermediate steps of the process: spent, calcined, and reduced form. Only a small amount of sample was available for characterization, considering that the catalyst needed to be reused in the following experiments. Due to that reason, two techniques which require low sample amounts showing meaningful results were selected to monitor the catalyst along the cycles—SEM-EDX and XRD. The upgraded products were characterized using the same techniques as described previously.

Looking at the SEM-EDX results, it was possible to follow the main changes occurring on the catalyst surface (Figure 6). The fresh catalyst showed a good dispersion of nickel particles over the support surface (Figure S7 fresh a). Along the cycles, sintering of the nickel particles was observed. Specific regions were analyzed by EDX to estimate the composition. While the fresh catalyst mainly was composed of nickel, silica and a low concentration of carbon, the spent catalysts clearly showed a small concentration of sulfur (near the detection limit) located around the nickel particles. Interestingly, in the regions with very low nickel concentration or even absent of nickel, no sulfur was identified. Al and Fe were also observed in the EDX spectrum, although in very small concentrations (below 0.1 wt.%). Aluminum was observed in small concentrations (support composition) but the authors cannot discard the possibility of contamination during the removal of the catalyst from the autoclave (aluminum paste is used to seal the reactor).

Considering that EDX measurements can be made either as element mapping or point analysis, both methods were used. Further compilation of images, as well as the mapping of the catalyst surface along the cycles, is available in the Supplementary Material (Figures S7 and S8 and Table S3).

**Figure 6.** *Cont.*

**Figure 6.** SEM-EDX images and spectra of Ni/SiO2 along the consecutive cycles.

Following the fourth reaction, the catalyst was analyzed by SEM-EDX in three different situations—spent, after the calcination and, finally, after the reduction step. Although punctual analyses were performed (Table 7), some trends could be observed. Regarding carbon, an expected lower concentration was observed after the calcination step (spectrum 14–16), in comparison to the spent catalyst (spectrum 9–13). Furthermore, the sulfur remained over the reduced catalyst, mainly in the regions with higher nickel concentration, indicating a persistent adsorption. Further structural investigation along the cycles was obtained by XRD and, in this case, the catalyst was analyzed along each new regeneration cycle, in the spent, calcined, and reduced forms (Figure 7). Metallic Ni reflection was observed in the fresh, spent, and reduced catalysts along the consecutive reactions. Due to the calcination step, reflections attributed to NiO were observed which disappeared after the reduction. Considering the low concentration of sulfur over the spent catalyst surface, reflections attributed to Ni3S2 as found by other authors [26,67,68] were not observed, even after the fourth reaction.


**Table 7.** SEM-EDX scan composition of selected regions of Ni/SiO2 over consecutive cycles.

**Figure 7.** XRD patterns of Ni/SiO2 catalyst along consecutive reactions and regeneration steps. Dashed blue lines refer to Ni and dashed red lines refer to NiO.

An increase in the crystallite sizes could be observed, as also identified by other authors [26]. Ni/SiO2 showed initially a crystallite size of 17.7 nm. Following subsequent regenerations, the crystallite size reached 37.3 nm (fourth reaction), due to particle sintering [45], in agreement with the SEM-EDX observations.

Due to the small amount of sample available, the BET surface area was measured only after the fourth reaction (spent catalyst). The original catalyst had a surface area of 215 m2/g of catalyst, reduced to 46 m2/g after the first reaction and later to 39 m2/g after the fourth reaction (spent catalyst), slightly lower compared to the first use. Furthermore, the amount of nickel leached to the light phase was calculated. It could be noticed that along the cycles the amount of nickel leached was being reduced as follows: during the 1st reaction 0.8% of nickel was leached, followed by 0.6% in the 2nd, 0.12% in the 3rd reaction and, finally, in the 4th only 0.10% of nickel was leached.

The physicochemical properties of the upgraded bio-oils over consecutive reactions also were determined (Table 8). The carbon content remained above 72 wt.%, although a slight decrease was observed. The water concentration slightly increased after the second reaction, remaining in the same range over the following reactions (around 5.4 wt.%). The HHV and pH remained in the same range as observed in the reaction performed with the original catalyst. The oxygen concentration increased after the third reaction, reaching 20.35 wt.% after the fourth reuse of the catalyst. Similar tendencies were found by Boscagli et al. [26] evaluating the performance of regenerated catalysts, although, in their study, only one regeneration was considered and a different catalyst (NiCu/Al2O3) was used. As the number of consecutive reactions increased, the hydrogen consumption decreased, mainly between the first and second reuse. A reduction of 17.18% in the consumption of hydrogen between the original reaction and the first regeneration, 6.28% between the first and the second, and just 0.66% between the second and third regenerations was observed. The lower consumption of hydrogen indicates a decline in the hydrotreatment activity, which can be a result of the sintering, leaching and due to the poisoning substances [61] observed by XRD, ICP-OES and EDX measurements.

**Table 8.** Physicochemical properties, elemental analysis (in dry basis) of the upgraded bio-oil, hydrogen consumption and total gas production over consecutive HDO reactions and catalyst regeneration.


Furthermore, as the number of consecutive reactions increased, higher amounts of CO2 were produced (Figure 8). Discussed previously, the consumption of hydrogen seems to be inversely proportional to the amount of CO2 produced; Additionally, as the number of consecutive reactions increased, the higher the amounts of CO, propene, ethane and ethene were produced, although less significant in comparison to CO2. Methane remained constant along the cycles.

The changes over consecutive reactions also were monitored by 1H-NMR (Figure 9). Some trends observed for the original reaction also were observed along the cycles. The signal for aldehydes (10.1–9.5 ppm), already absent in the first reaction due to its high reactivity at low temperatures [45], was not observed in any of the phases for all regeneration stages. The highest signal for the upgraded bio-oil was found in the α protons to unsaturated, carboxylic acids and keto-groups (3.0–1.5 ppm) while, for the aqueous phase, it was in the water, O-H exchanging and carbohydrate groups (6.0–4.3 ppm). Aromatics (8.5–6.0 ppm) mostly were concentrated in the upgraded bio-oil and almost absent in the aqueous phase (8.3 mmol/g sample versus 0.2 mmol/g sample on average). Alkenes, alcohols, and ethers (4.3–3.0 ppm) were concentrated in the upgraded bio-oil (10.6 mmol/g sample versus 6.3 mmol/g in the aqueous phase on average). The same was observed for alkanes (1.5–0.5 ppm), 22.3 mmol/g bio-oil versus 1.1 mmol/g aqueous phase on average. It could be seen that the concentration of aromatics remained similar along the reactions. The values also were similar as observed for the feedstock (HP). A similar trend was observed for the carbohydrates, water, and O-H exchanging groups. The increase in this region can be attributed to the higher water concentration in

the upgraded bio-oils, as well as a lower conversion of compounds belonging to this region. As the number of reactions increased, the concentration of alcohols, ether and alkenes in the upgraded bio-oils decreased. The α proton to unsaturated groups in the upgraded bio-oils was much higher in comparison to the feedstock. An increase of protons was observed in the first reaction, dropping in the following reactions. Alkanes were also in higher concentration compared to the feedstock and the values were quite similar among all the reactions. Conversely, within the aqueous phase, the opposite tendency was observed for the carbohydrates, water and O-H exchanging groups. A decline along the regenerations was seen, probably due to the increase of the water content in the upgraded bio-oil (Figure 9). The signal for alcohols, ethers and alkenes also dropped, from 8.7 mmol/g after the first reaction to 5.2 mmol/g after the fourth regeneration. The signals for α proton to carboxylic acid or keto-groups, α proton to unsaturated groups, and for alkanes displayed a very small reduction after each reuse of the catalyst.

**Figure 8.** Production and composition of the gas phase obtained for the cycles of hydrotreatmentregeneration.

Further qualitative investigation was performed with GC-MS (Figure 10). The composition of the upgraded bio-oils along the cycles was very similar, although some changes could be observed. The intensity of the peaks attributed to cycloheptanone (12.23 min) and 2-methyl-2-propanol (13.08 min) was reduced along the cycles. The same was observed for two more substances: dihydro-5-methyl-2-(3*H*)-furanone was being reduced until disappeared in the fourth reaction while a visible reduction in the γ-butyrolactone peak (18.4 min) was visible.

The dehydration of C6 sugars, such as glucose, results in compounds such as hydroxymethylfurfural [69]. The following conversion of hydroxymethylfurfural resulted in 5-(Hydroxymethyl)dihydro-2(3*H*)-furanone which can be further converted to γ-valerolactone (Dihydro-5-methyl-2(3*H*)-furanone) through hydrogenation and direct deoxygenation [70]. Conversely, γ-butyrolactone is obtained through hydrogenation of 2(5*H*)-furanone [70]. These sugar derivative compounds then can be further converted to ketones and alcohols [28,69]. Based on these results, a correlation between 1H-NMR and GC-MS could be proposed to explain these findings: Considering the higher signal obtained for the protons belonging to the carbohydrates, water and O-H exchanging groups, it can be assumed that the upgraded bio-oils along the cycles showed higher concentration of not just of water, but possibly also sugars, in comparison to the first reaction. Furthermore, based on the lower intensity of the GC-MS peaks of sugar derivative compounds, such as γ-valerolactone, γ-butyrolactone as well as 2-methyl-2-propanol, it is assumed that the conversion of sugars along the upgrading cycles is reduced. The peak intensity of 2-ethyl-cyclopentanone in the light phases (Figure S5) is also reduced along the cycles. It also is evidence of lower conversion of sugars along the

cycles, considering that this compound can be obtained from a compound with similar structure to furfural [71].

**Figure 9.** 1H-NMR spectra integrals of upgraded bio-oil and HP (**top**) and aqueous phase and LP (**bottom**) along the HDO-regeneration cycles.

**Figure 10.** GC-MS upgraded bio-oil along the regeneration cycles.

#### **3. Discussion**

The synthesis and evaluation of nickel-based catalysts showed differences among the catalysts evaluated. The temperature programmed reduction (H2-TPR) showed that the catalysts, with the addition of copper as a promoter and higher loading of nickel, had lowered the reduction temperature of nickel oxides, which is in agreement with previous studies [30,38].

The evaluation of the catalysts for hydrodeoxygenation (HDO) reactions showed that upgraded bio-oil obtained with Ni/SiO2 showed the best properties in terms of low oxygen concentration, low water concentration and high HHV. The higher HDO activity for this catalyst can be correlated with its higher surface area in comparison to the other catalysts tested in this study, which could be beneficial to increase the dispersion of the active components, resulting in a more active catalyst [20,64]. Coincidently, the ZrO2-supported catalysts showed a lower HDO activity, which might be attributed to the lower surface area of this support [42]. Similar to the current study's findings, Dongil et al. [24] observed lower HDO activity for NiCu catalysts in comparison to monometallic Ni catalysts. The authors attributed the lower guaiacol HDO to the larger particle size, the presence of NiO particles as well as to copper particles located at the nickel active sites. Furthermore, a higher crystallite size of bimetallic catalysts, as observed for both NiCu catalysts, also might play a role in the lower HDO activity; the higher crystallite size decreases the number of step/corner sites, which, according to Mortensen et al. [15], are more active for breaking C-O bonds.

Coke formation is another important parameter for the selection of the catalyst. Observed for both Ni/SiO2 and NiCu/SiO2, the volume of the micropores was reduced significantly after the reactions. Coke is known for blocking the pores, covering the catalyst active sites, resulting in partial or even complete loss of activity [61], therefore, it is considered one of the main causes of deactivation in HDO reactions [8,15,28]. Considering that the catalyst acidity is connected directly to coke formation [8], the lower coke formation observed with Ni/SiO2 could be related to the low acidity of silica [28]. Due to the amphoteric nature of ZrO2, reduced coke formation would be expected [28]; However, SiO2 was more resistant to coke formation. The addition of Cu also seems to contribute to the slightly higher amount of coke deposited on the bimetallic catalysts. Stated by Zhang et al. [42], the addition of copper increases the acidity of the catalyst, which can result in higher coke deposition [27], as observed. Hence, the lowest coke deposition observed for Ni/SiO2 can be related to the lower acidity compared to NiCu/SiO2 and higher resistance of SiO2 in comparison to ZrO2.

Reactive compounds, such as aldehydes, were converted completely after the upgrading. The reduction of these very reactive compounds results in a more stable oil [15,29]. The lower conversion of aromatic compounds over all the catalysts evaluated also is interesting; if further upgrading is intended, aiming at fuel production, the presence of aromatic compounds could result in a high octane number gasoline [29].

Different selectivity among mono and bimetallic catalysts was attributed mostly to the addition of copper [69], with a minor contribution of nickel loading [22]. Furfural, also a very reactive compound [72], mainly seemed to be hydrogenated to tetrahydrofurfuryl alcohol over NiCu catalysts. Additionally, the presence of propylene glycol leads to the conclusion that hydroxyacetone was mainly hydrogenated to this compound over bimetallic catalysts. Concurring with other authors, the addition of copper seems to increase the hydrogenation [24], increasing the hydrogen consumption. Furthermore, the higher hydrogen consumption observed for NiCu catalysts is in agreement with literature [21,23] giving evidence that the addition of a second metal can increase the hydrogenation activity of the catalyst [69]. Stated by Mortensen et al. [15], the H/C and O/C is used to evaluate the quality of the upgraded product. A higher H/C ratio is intended, whereas a lower O/C ratio is desired. In this case, it was observed that the higher H/C ratio for NiCu/SiO2 agrees with the higher consumption of H2 observed for this catalyst, but it was not reflected in the O/C ratio. The lowest O/C was obtained for Ni/SiO2 (0.18), being much lower in comparison to the O/C ratio of the feed (0.47). It is important to note higher hydrogenation results in higher consumption of hydrogen [57,73], but it is not necessarily reflected in the reduction of the oxygen content. Moreover, the higher consumption of

hydrogen can also result in higher methane formation [35], considering that during the hydrocracking the consumption of hydrogen is higher compared to hydrotreating [74]. This behavior was observed for NiCu/SiO2. The catalyst showed not just the higher hydrogen consumption, but also a higher methane formation. The desired catalyst should be able to remove the larger amount of oxygen with minimal hydrogen consumption [24] as hydrogen consumption, bio-oil yield and catalyst deactivation are among the most important parameters to be considered in the HDO process [75]. Ni/SiO2, therefore, was considered the more appropriate catalyst in terms of H2 consumption and HDO activity.

After the reaction, compounds such as calcium (only over ZrO2) and sulfur were observed on the spent catalysts by two different analytical techniques, Inductively Coupled Plasma Emission Spectroscopy (ICP-OES) and Scanning Electron Microscopy/Energy Dispersive X-ray spectroscopy (SEM-EDX). Calcium, observed in higher concentrations in comparison to sulfur over the zirconia-supported catalysts, on the one hand acts as a poisoning agent, reducing the mobility and re-dispersion of the active metal centers over the support and, on the other hand, can reduce sintering, due to the reduction in the atom mobility, resistance to dissociation and migration [61].

Despite the detection of sulfur on the catalyst's surface, structural changes were not observed after the reactions by Powder XRD, as observed in previous investigations. Mortensen et al. [66] observed a reflection at approximately 2θ = 45.2◦, attributed to NiS during the evaluation of the influence of sulfur over the conversion of guaiacol with Ni/ZrO2. Boscagli et al. [26] observed the formation of Ni3S2, which was persistent to regeneration and changed the catalyst structure. The concentration of sulfur in the feed, in this case, varies significantly in both works. While Mortensen et al. used a model mixture containing 0.05 wt.% of sulfur, Boscagli et al. used feedstocks with a higher concentration of sulfur (light bio-oil phase with 0.05 wt.% of sulfur and straw bio-oil obtained at 450 ◦C with 0.3 wt.% of sulfur). The high concentration of sulfur resulted in a spent catalyst with a much higher concentration of sulfur on the catalyst (0.6–2.0 wt.%, SEM-EDX), in comparison to the current study's findings (feed sulfur concentration = 0.012 wt.% and spent catalysts ≤0.1–0.4 wt.%, SEM-EDX). Even without structural changes, sulfur is one of the most persistent poisons for nickel catalysts. It is irreversibly chemisorbed and responsible for blocking the reaction-adsorption active site, modifying electronically the neighbor atom of metals and, thereby, reducing the ability to adsorb and dissociate H2, at the same time influencing the diffusion or reactants, blocking their contact with the active site [61].

Along with the consecutive reactions with Ni/SiO2, comparable carbon and hydrogen concentrations and a slightly higher oxygen concentration was observed in the upgraded bio-oils. The oxygen concentration reached20.34 wt.% after the fourth reaction. It was higher compared to the first reaction (17.86 wt.%) but still much lower compared to the feed (35.84 wt.%, dry basis). It gives evidence of a low rate of deactivation and the possibility to reuse the catalyst. The increase in the oxygen concentration can be correlated with the reduction in the H2 uptake, in agreement with lower HDO activity of Ni/SiO2 in comparison to the fresh catalyst [76]. The current authors assumed that the reduced activity can be correlated mainly to sintering, poisoning and coke deposition (although easily removed during the calcination step), as the leaching was negligible. The crystallite size increased over the cycles in the same proportion that the HDO activity was reduced along the consecutive reactions. The fresh catalyst showed a crystallite size of 17.7 nm, reaching 37.3 nm after the fourth reuse. Thus, the number of active surfaces available were reduced with the increase of the crystallite size [61]. Furthermore, the lower H2 uptake can also result in a higher amount of CO2 formation, as observed along the cycles, considering the higher the hydrogen uptake, the lower the CO2 production [46]. Curiously, the sulfur concentration remained approximately constant along the cycles. Furthermore, no differences were observed in the XRD diffractions. The current authors concluded that sulfur was strongly adsorbed on the catalyst, affecting surface-sensitive reactions, resulting only in partial loss of activity of the catalytic surface [61].

The results obtained by 1H-NMR showed that the concentration of protons in the region related to water and carbohydrates increased along the cycles in the upgraded bio-oils. Moreover, the gas chromatography-mass spectrometer (GC-MS), showed that the peak of some sugar derivatives became smaller over the consecutive cycles. It was assumed that the catalytic conversion of sugars was possibly affected by the lower activity of Ni/SiO2 along the cycles. This assumption corroborates the results of GC-MS, from which the intensity of sugar derivative compounds (γ-valerolactone γ-butyrolactone 2-ethyl-cyclopentanone 2-methyl-2-propanol) have been reduced as the number of cycles increased. Considering the assumption that the conversion of sugars through hydrogenation is reduced with the increase in the number of consecutive reactions, higher amounts of coke formation could be expected, considering that the thermal polymerization of the sugar fraction can lead to increased char deposition [35,46,77] over the cycles. The results of SEM-EDX showed a slightly higher carbon deposition over the spent catalyst after the fourth reaction.

Since poisoning substances might affect some specific reactions [61], the investigation of model compound conversions as well as the effect of poisonings, over single compounds, could contribute to the understanding of the selectivity changes observed along the cycles. Considering the difficulty for regeneration of sulfur-poisoned catalysts, due to the harsh conditions required (700 ◦C in steam) [61], its influence over the conversion of model compounds should be investigated in detail. Since deactivation mechanisms are difficult to be monitored in batch experiments [66], continuously operated reactors are more appropriate for this investigation.

#### **4. Experimental**

#### *4.1. Catalyst Syntheses*

The four nickel-based catalysts, Ni/SiO2, Ni/ZrO2, NiCu/SiO2 and NiCu/ZrO2, were synthesized by wet impregnation as follows. Both supports (silica and zirconia from Alfa Aesar, Haverhill, MA, USA) were milled to 0.125–0.250 mm and added to the metal solution in a ratio of 1:10. The metal precursors were Ni(NO3)2.6H2O (Sigma–Aldrich, St. Louis, MI, USA) and Cu(NO3)2·2.5H2O (Alfa Aesar, Haverhill, MA, USA). The water was evaporated at 35 ◦C, 45 mbar, 100 rpm in a rotary evaporator (Hei-VAP Advantage ML/G3) and the formed catalyst was dried for 12 h at 105 ◦C. The catalysts were then calcined at 450 ◦C for 4 h after reaching the set point, with a heating ramp of 10 ◦C/min (Thermolyne F6010) and later reduced in a 25% H2/N2 flow of 3 L/min, heating ramp of 5 K/min, during 4 h. The reduction temperatures were defined by temperature programmed reduction (H2-TPR) experiments (see Sections 2.1 and 4.5). Regarding the monometallic Ni catalyst, the metal concentration in the catalyst was defined in 8.6 wt.% [78] while, for the bimetallic catalysts, the metal concentration was defined in 28 wt.% of Ni and 3.5 wt.% Cu [23].

#### *4.2. Beech Wood Fast-Pyrolysis Bio-Oil*

The experiments were carried out with a beech wood fast-pyrolysis bio-oil (BTG Group, Enschede, The Netherlands). Following intentional aging (24 h, 80 ◦C) two phases were observed. The final oil was then composed by 41 wt.% of heavy phase (HP) and 59 wt.% of light phase (LP), respectively. Both phases were separated, and previously characterized separately, as presented elsewhere [43]. The main physicochemical properties and elemental analyses are presented (Table 9).

**Table 9.** Physicochemical properties and elemental analysis of the HP and LP of aged beech wood fast pyrolysis bio-oil [43].


#### *4.3. HDO Reactions*

The HDO (hydrodeoxygenation) experiments were conducted in a 200 mL volume batch reactor. More details about the reactor are given in the Supplementary Material. The conditions selected for this study were fixed to 325 ◦C and 80 bar of H2 based on the current authors' previous investigation [47]. Fifty grams of bio-oil (composed of the mixture of the LP and HP, in the proportion advised in Section 4.2) and 5 wt.% of catalyst in relation to the amount of the bio-oil were added to the autoclave. The reactor was closed and purged with N2 for 5 min and pressurized with H2 (Air liquid Alphagaz 2, purity 6.0) at ambient temperature. The stirrer was switched to 1000 rpm and the heating program was started at a rate of 5 ◦C/min. The global reaction time, including the heating ramp, was 120 min. When this time was reached, the reaction was quenched first using a flow of compressed air and, later, in a cold water bath with ice, until reaching ambient temperature. Two experiments were performed for each set of conditions and are presented as an average.

Following the reaction, the final pressure was recorded for hydrogen consumption determination and the gas phase was collected for gas chromatography analysis. The mixture of spent catalyst, coke/char, upgraded bio-oil and aqueous phase was collected, centrifuged for 40 min at 7000 rpm in a Heraeus Biofuge Stratos centrifuge (Thermo Fischer Scientific, Waltham, MA, USA) and then separated for further characterization. The H2 consumption was calculated using the ideal gas law, as the difference of the moles of hydrogen loaded to the reactor and the remaining moles after the hydrotreatment, from the pressure before and after the reaction and the gas composition determined by gas chromatography [40].

#### *4.4. Products Characterization*

The gas composition was determined by collecting and analyzing the gaseous fraction by gas chromatography. A 100 μL sample was injected at 250 ◦C (split 28:1) in an Agilent 7890A (two detectors: thermal conductivity and flame ionization detector, TCD and FID respectively) equipped with two columns: Restek 79,096 Hayesep Q and Restek Molsieve 5A. The oven temperature was programmed as follows: the initial temperature was set to 50 ◦C maintained for 10 min; increased to 90 ◦C at a heating rate of 3 ◦C per minute and then increased to 150 ◦C at a heating rate of 20 ◦C per minute, maintained at this temperature for 16 min and finally heated to 230 ◦C at 50 ◦C per minute and kept for 10 min.

The liquid products (aqueous phase and upgraded bio-oil) were characterized using the same methodology, except the higher heating value (HHV), which was not determined for the aqueous phases. The pH values were measured with a pH-meter from Metrohm. The HHV was determined using a calorimeter IKA C5000 control and the water content using a volumetric Karl–Fischer titrator from Metrohm (Titrando 841, titration reagents Composite 5 and dry Methanol). Carbon, hydrogen, and nitrogen content were measured using a micro-elemental analyzer Elementar Vario el Cube. The content of oxygen was estimated by the difference.

Quantitative 1H-NMR (proton nuclear magnetic resonance) was employed to characterize the functional groups in the product molecules, based on the number of protons in the corresponding 1H-shift range [44]. 1H-NMR spectra were recorded at 25 ◦C on a Bruker Biospin spectrometer, equipped with a 5.47 T magnet (1H frequency 250 Hz). Sample preparation consisted of 0.1 g of either upgraded bio-oil or aqueous phase and their dilution in 0.7 g of deuterated methanol (CD3OD) containing TMSP-d4 [3-(trimethylsilyl)-2,2,3,3-tetradeuteropropionic acid sodium salt] as an internal standard (0.1 g TMSP in 50 mL CD3OD). Subsequently, the samples were centrifuged (removal of particles not solubilized) and placed in NMR tubes.

The liquid samples also were analyzed qualitatively using a gas chromatography mass spectrometer (GC-MS) HPG1800A, with a Restek stabilwax column. Prior to measurement, the samples were diluted in methanol and filtrated using a 0.25 μm filter. A 1 μL sample was injected at 250 ◦C, with a split of 1:20. The oven temperature was programmed as follows: the initial temperature was

set to 40 ◦C, maintained for 5 min; increased to 300 ◦C with a heating rate of 20 ◦C per minute and maintained for 20 min at this condition.

#### *4.5. Catalysts Characterization*

The concentration of metal in the catalysts was analyzed by ICP-OES (Inductively Coupled Plasma Emission Spectroscopy Agilent 725 Spectrometer). The sample was dissolved using a mixture of HNO3 (2 mL), HCl (6 mL) and H2O2 (0.5 mL) and digested in a microwave oven for 45 min at 240 ◦C.

To determine the reduction temperature profile of the active metal, an Autochem HP 2950 (Micrometrics, Ottawa, ON, Canada) was used for the temperature programmed reduction by hydrogen (H2-TPR). The measurements were performed at a heating rate of 1 K/min until 500 ◦C and 5% H2 in Ar at 30 mL/min. The samples were dried in a 30 mL/min flow of Ar at 300 ◦C for 3 h before the measurement.

The total specific surface area of the catalyst was determined by nitrogen physisorption with a Belsorp Mini II at 77 K and calculated by applying the BET theory in the fitting rate between 0.05–0.30 p/p0 (12 points). Powder XRD was measured using an X'Pert PRO MPD instrument (PANalytical GmbH, Nuremberg, BY, Germany) equipped with a Cu anode (Cu Kα 1.54060 Å). The XRD patterns were recorded in a 2θ range between 5–120◦ (1 h, step size 0.017◦). The average crystallite size was estimated using the Scherrer equation (shape factor K = 0.9) after correcting the instrumental line broadening.

Leaching of the catalyst was monitored by analyzing the aqueous phase by ICP-OES. Sample preparation involved the filtration with a 0.2 μm polytetrafluoroetylen (PTFE) filter membrane of the produced aqueous phase after each reaction.

To identify the elements present on the particles of fresh and spent catalysts, SEM/EDX (Scanning Electron Microscopy/Energy Dispersive X-ray spectroscopy) was applied. The equipment for this technique was a GeminiSEM 500, Zeiss, software SmartSEM Version 6.01, with a thermal Schottky field-emitter cathode. An energy dispersive X-ray spectrometer X-MaxN from Oxford with a silicon drift detector (80 mm2 and resolution of 127 eV) was employed for the quantitative analysis of micro areas and the distribution of the elements, in addition to the software Aztec 3.3. The C, H, N and S of the spent catalysts were measured by a micro-elemental analyzer Elementar Vario el Cube. The solid over the spent catalysts (coke) was calculated by the carbon concentration over the spent catalyst (determined by elemental analysis), considering oxygen concentration negligible [46]. More detail is given in the Supplementary Materials (Equation (S1)).

#### *4.6. Catalyst Regeneration*

Based on the experimental results, such as oxygen content, water concentration, solid and gas production as well as upgraded bio-oil yield, one of the catalysts was selected for further application in cycles of HDO and regeneration. The cycles each consisted of an HDO reaction, calcination of the spent catalyst (as described in Section 4.1), followed by a reduction and a subsequent HDO reaction. Altogether, the catalyst was used four times. To evaluate the behavior of the catalyst over the consecutive uses, the spent catalyst was analyzed through SEM-EDX and XRD between the regenerations: EDX was performed for the spent and reduced catalysts, whereas XRD was performed for the spent, calcined, and reduced catalysts. Furthermore, the upgraded products also were characterized along the cycles, as described in Section 4.4.

#### **5. Conclusions**

Four nickel-based catalysts on different supports were synthesized and evaluated for the HDO of a multi-phase beech wood fast-pyrolysis bio-oil. The bimetallic catalysts showed lower reduction temperature, attributed to the addition of copper, and higher metal loading. Furthermore, NiCu catalysts presented higher consumption of hydrogen and different selectivity toward the conversion of compounds such as furfural, compared to monometallic catalysts. Upgraded bio-oils

with reduced concentration of oxygen, lower water concentration and higher carbon content were obtained after the HDO reactions. Ni/SiO2, in particular, showed the highest HDO activity, reducing more than 50% of the oxygen content and more than 80% of the water content, thus selected for application in cycles of regeneration-reaction. During the consecutive reactions, the activity of the catalyst decreased, attributed mainly to sintering and poisoning by sulfur, as coke was removed easily during the regeneration steps. Lower hydrogen consumption and higher carbon dioxide production were observed in comparison to the reactions when applying the fresh catalyst. Correlating the results obtained by 1H-NMR and GC-MS, it was possible to observe that compounds known as sugar derivatives were being reduced along the consecutive reactions while the concentration of protons in the region attributed to water, carbohydrates and O-H exchange increased. The partial loss of activity seemed to lower the conversion of sugars.

Accompanying that, the catalysts evaluated seemed to be suitable for HDO of hardwood fast-pyrolysis bio-oil, especially Ni/SiO2. Further investigation will be addressed, considering a detailed investigation of model compound conversions and the effect of sulfur and calcium over HDO of single compounds. The influence of sintering over the activity and selectivity will also be investigated. Due to the limitation of the batch reactor for deactivation mechanism studies, the application of a trickle bed reactor in future studies is intended.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4344/8/10/449/ s1.

**Author Contributions:** C.C.S. design the experiments, performed part of the experiments, analyzed the data and wrote the paper; M.B.G.R. performed most of the experiments and analyzed the data; M.Z. contributed with the microscopy experiments; J.-D.G., K.R. and N.D. contributed with the discussion, reviewed and supervised the work.

**Funding:** This research received no external funding.

**Acknowledgments:** The authors are grateful to the Bioeconomy Graduate Program—BBWForwerts, Brazilian National Council for Science and Technology (CNPQ) and BeMundus for the financial support. The authors are also grateful to Herman Köhler, Pia Griesheimer, Petra Janke, Jessica Heinrich, and Simon Wodarz for the support with analytical techniques, Oliver Schade for the support with the catalyst reduction oven and to the assistant bachelor student (Hiwi student) Mouhannad Moulla, for the activities developed during his internship which supported the development of this work.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
