University of Birmingham Reaction kinetics of vanillin hydrodeoxygenation in acidic and nonacidic environments using bimetallic PdRh/Al2O3 catalyst

: The kinetics of hydrodeoxygenation (HDO) reaction in literature are mostly reported for single model compounds in bio-oil. However, these kinetic models may become invalid in the real bio-oil environment where other model compounds are present. This study investigates the e ﬀ ect of acetic acid, which is a major compound in bio-oil, on the liquid-phase HDO reaction kinetics of vanillin (VL). A synthesized bimetallic catalyst (PdRh/Al 2 O 3 ) was utilized in a batch reactor at 308 − 328 K, 1 − 4 MPa H 2 gas partial pressure ( P H ), 263 − 526 mM initial VL concentration ( C VL0 ), and 1.9 − 4.6 kg/m 3 catalyst loading with ethyl acetate as the reaction solvent. N 2 adsorption − desorption, scanning electron microscopy (SEM), X-ray di ﬀ raction (XRD), X-ray ﬂ uorescence (XRF), CO chemisorption, and mercury porosimetry methods were used to determine the physicochemical properties of the catalyst. Transport limitations in the system were ruled out via the Madon − Boudart test, Weisz − Prater criterion, agitation, and particle size test. In vanillin-only and vanillin − acetic acid environments, non ﬁ rst-order dependence on P H and C VL0 was found. Notably, lower reaction rates were observed in the presence of acetic acid compared to vanillin-only environment. Kinetic data for the vanillin-only and vanillin − acetic acid environments were successfully modeled using derived expressions from Langmuir − Hinshelwood − Hougen − Watson approach under the assumption of competitive dissociative hydrogen adsorption. The estimated activation energy for VL HDO reaction was 24.1 kJ/mol in vanillin-only environment and 51.0 kJ/mol in the presence of acetic acid.


INTRODUCTION
The world energy consumption is expected to grow by 20− 30% from 2019 to 2040 as a result of the increase in population and rapid development in countries like China and India. 1 Conversely, the volume of fossil fuels accessible from reserves around the world has steeply declined over the past decade. 2,3 This disturbing trend in capacity of world fossil reserves and growing concern on the impact fossil fuels consumption has on the environment motivates the development of alternative renewable and sustainable energy sources. 4−6 Renewable sources of energy such as wind, solar, and hydrogen fuel cell are attractive substitutes for fossil resources. However, most of the existing infrastructure in the world works on liquid hydrocarbon fuels. 7,8 Hence, to utilize the above listed renewable energy sources, drastic changes will be required, necessitating a long duration until they are fully realized. At the moment, plant biomass is the only renewable and sustainable carbon source for producing liquid hydrocarbon fuel, otherwise known as bio-oil. 9−12 Consequently, production of bio-oil from plant biomass has increasingly become the subject of interest in energy research studies. Fast pyrolysis (FP) is the preferred technology for producing bio-oil because it requires lower costs compared to competing technology like liquefaction. 9,13−15 FP involves rapid heating of biomass to a temperature around 923 K in the absence of air. 16−19 Bio-oils produced via FP usually contain a significant amount of highly unstable oxygenated compounds such as aldehydes, ketones, carboxylic acids, guaiacols, cresols, syringols, and anisoles. 20−23 The fast processing conditions used in the production of bio-oil via FP are primarily responsible for the thermodynamic instability of these oxygenated compounds, which adversely affects the fuel quality. For instance, bio-oils have low pH, poor volatility, lower energy density, and high viscosity. 13,24−27 Fortunately, the fuel quality of bio-oil can be improved through a hydrotreating process known as hydrodeoxygenation (HDO). 19,28−35 However, the complexity of HDO processes and bio-oils makes it difficult to collect kinetic data for reactor sizing and overall process-optimization purposes. Most research studies on HDO are based on individual model compounds found in bio-oil to simplify the reaction and gain better understanding of the steps involved in the transformation process. 36−40 However, in the presence of other model compounds, the established kinetic relationship from single model compound studies may become invalid. Ligninderived phenolics containing multiple functional groups such as cresols, guaiacols, and anisoles are commonly used as model compounds in HDO reactions. 37,41 In this work vanillin, a lignin-derived phenolic compound with three functional groups (ether, hydroxyl, and aldehyde), is selected because it has been reported in more than four different analytical studies as a component of real bio-oil but is less reported as a model compound for HDO. 42−46 Sulfided transition metals, which include cobalt molybdenum (CoMo) and nickel molybdenum (NiMo) on conventional support such as alumina (Al 2 O 3 ), were traditionally used to investigate HDO reaction. However, cofeeding with H 2 S is required to maintain activity of these catalysts. 47,48 As a result, research attention drifted toward the use of noble metals, which includes ruthenium (Ru), platinum (Pt), palladium (Pd), and rhodium (Rh) as HDO catalysts. 49,50 These catalysts are active at moderate operating conditions and do not require sulfur addition to remain active. However, the ease with which they are poisoned and the availability of these metals pose economic concern. 51,52 Transition metal carbides, nitrides, oxides, and phosphides recently gained attention as low-cost and readily available alternative catalysts for HDO reaction. For instance, Shit et al. 53 reported the use of cobalt phosphides on porous organic polymers (POPs) as catalyst for vanillin HDO reaction. However, concerns on the stability of these catalysts under typical conditions for HDO limit their application. Remarkably, the majority of the previous works on kinetics of vanillin HDO reaction utilized monometallic catalysts such as Ru/C by Bindwal and Vaidya 54 and Pt/C by Sulman et al. 55 However, as previously highlighted, these catalysts are prone to poisoning. Fortunately, their resistance against poisoning can be improved when combined with another metal. In addition, bimetallic catalysts have demonstrated superior HDO performances in terms of selectivity toward deoxygenated products and conversion in past works. 56 Nonetheless, information on the reaction kinetics of vanillin over bimetallic catalysts is still missing in the literature. In this work a bimetallic catalyst comprising palladium and rhodium on alumina support (PdRh/Al 2 O 3 ) was synthesized, characterized, and used to investigate the kinetics of vanillin HDO reaction in the liquid phase. The interaction between Pd and Rh is expected to improve the selectivity toward creosol, a potential future fuel product. Remarkably, the reaction kinetics of vanillin HDO in the presence of other major lignin-derived compounds has not been reported anywhere. The aim of this work is to fill the identified gap in kinetic studies of HDO reaction. On the basis that competition for finite catalyst active sites increases with the addition of another model compound, the energy barrier to vanillin HDO reaction will most likely be higher in the presence of acetic acid. To test these assumptions, the effects of reaction temperature, catalyst loading, hydrogen gas pressure, and initial VL concentration were studied under a kinetically controlled regime in vanillin− acetic acid mixture and vanillin-only environments.

EXPERIMENTAL SECTION
2.1. Materials and Catalyst Preparation. Vanillin (purity 99%), vanillyl alcohol (4-hydroxy-3-methoxybenzyl alcohol, purity 99%), creosol (2-methoxy-4-methylphenol, purity 99%), ethyl acetate (purity 99%), and PdCl 2 solution (5 wt % in 10 wt % HCl) were purchased from Sigma-Aldrich, United Kingdom. Methanol (purity 99.9%) was obtained from Alfa Aesar, United Kingdom. Hydrogen (H 2 ) and nitrogen (N 2 ) gas cylinders (purity 99.9%) were supplied by BOC, United Kingdom. Commercial 5 wt % Rh/Al 2 O 3 catalyst was purchased from Johnson−Matthey PLC Catalyst, United Kingdom. The catalyst used to conduct this study was prepared via wet incipient impregnation method. This method involves the addition of 10.0 g of PdCl 2 solution dropwise to a slurry containing 10.0 g of Rh/Al 2 O 3 and 30 mL of methanol under slow stirring to achieve the desired Pd loading of 2.7 wt %. The solution was stirred at room temperature for an additional 4 h. Subsequently, the resultant mixture was filtered, washed, and dried in a fume cupboard overnight. Further drying in an oven set at 353 K was carried out for 2 h prior to calcination at 773 K for 4 h.
2.2. Catalyst Characterization. The characteristic length and apparent density of the catalyst were obtained from Micrometrics Autopore IV 9500 mercury porosimeter using low-(0−345 kPa) and high-pressure (up to 228 MPa) analysis. Crystallinity of fresh catalysts were examined by XRD using a Bruker AXS GmbH (D8 Advanced XRD, Karlsruhe, Germany). An HITACHI TM3030 Plus Tabletop Scanning Electron Microscope (SEM) equipped with Energy Dispersive X-ray (EDX) analyzer was used to examine the surface morphology of the catalyst before and after the reaction. The characteristics of surface metallic atoms such as active particle diameter and dispersion were determined through CO chemisorption using a Quantachrome ChemBet Pulsar instrument. In addition, the Brunauer−Emmett−Teller (BET) specific surface area, micropore volume, and average pore diameter of the fresh catalyst were determined according to ASTM C1274 using Micromeritics Analytical Instrument ASAP 2010.
2.3. Reactor System and Procedure. The experimental setup and schematic of reactor system used in this work has been described comprehensively by Aliu et al. 57 In a typical experiment, the autoclave was loaded with the appropriate amount of vanillin (2.2−4.4 g), PdRh/Al 2 O 3 catalyst (0.10−0.25 g), acetic acid (3.85−7.70 g), and 48.6 g of ethyl acetate as solvent. To create an inert atmosphere for impending reaction, the autoclave was purged with N 2 gas. After the purging process, the autoclave was heated to the desired reaction temperature (308−328 K). Once the autoclave attained the desired reaction temperature, the stirring speed was increased to 1000 rpm and H 2 gas was introduced until the desired reaction pressure (1.0− 4.0 MPa) was achieved. Timing of the reaction commenced from the point at which H 2 gas was charged. Further details of the experimental procedure were reported by Aliu et al. 57 2.4. Product Analysis. Reaction samples were initially analyzed via a gas chromatography unit equipped with mass spectrometer (GC-MS) to establish the products formed. Subsequently, multipoint calibration curves were developed to quantify the products and reactants. An automated trace gas chromatography unit equipped with flame ionization detector (GC-FID) containing a mild polar capillary column (i.e., ZB wax with specification of 250 μm internal diameter, 0.25 μm film thickness, and 30 m column length) was deemed suitable for the quantification process. Details on the temperature program applied are reported elsewhere. 57 3. THEORETICAL BASIS 3.1. Mass and Heat Transfer Effect. The collection of reliable quantitative kinetic data requires the absence of external and internal transport limitations; this includes heat and mass transfer effects. 58 In a typical heterogeneous catalytic reaction such as VL HDO, the following events occur: (i) diffusion of VL and H 2 through the boundary layer surrounding the catalyst particle; (ii) intraparticle diffusion of VL and H 2 into the catalyst pores to the active sites; (iii) adsorption of VL and H 2 onto active sites of the catalyst; (iv) surface reactions to convert adsorbed VL and H 2 into intermediates and possibly surface diffusion steps; (v) desorption of products (i.e., vanillyl alcohol and creosol) from active sites of the catalyst; (vi) intraparticle diffusion of vanillyl alcohol and creosol through pores of the catalyst; (vii) diffusion of vanillyl alcohol and creosol across the boundary surrounding the catalyst. 57 However, the scope of the present work necessitates consideration of only conditions within the intrinsic reaction kinetics control regime (i.e., steps 3−5). Transport limitations as a result of thermal gradients are insignificant at low temperatures. 59 Hence, it was decided to study only the temperature range of 308−328 K. The conventional approaches used to eliminate mass transfer limitations involve agitation speed and catalyst particle size tests to respectively determine the minimum stirring speed and the maximum particle size at which external and internal resistances become insignificant. These tests were performed at a reaction temperature of 318 K. In addition to these tests, the Weisz−Prater criterion was used to verify the absence of

Energy & Fuels
Article internal diffusion limitations. According to the Weisz−Prater criterion, the absence of significant internal diffusion is confirmed if the dimensionless parameter known as observable modulus (ηφ 2 ) has a value of <0.3 for reaction orders ≤2. 60 From the definition of observable modulus (ηφ 2 ) in eq 1, values significantly <1 indicate faster internal mass transfer rate compared to reaction rate. 61 ηϕ Herein, r 0 represents the initial reaction rate, η is the effectiveness factor, φ it the Thiele modulus, ω is the catalyst loading, and L is the characteristic length of the spherical catalyst particle. The correlation suggested by Pintar et al. 62 was used to estimate C H (kmol/m 3 ), while the Wilke−Chang equation 63 was used to estimate the liquid-phase effective diffusivities of VL and H 2 . The initial reaction rate representative of activity was estimated from the conversion−time data using the procedure described by Chen et al. 64 Following a review of the literature 58,60 on mass and heat transfer verification in laboratory-scale reactors, the Madon− Boudart test is the most definitive approach for establishing transport limitation significance. According to this test, in the kinetically controlled regime, the catalyst activity is directly proportional to the number of active sites. Hence, the In−In plot of the catalyst activity against surface metal atom concentration should be linear with a slope of unity to confirm absence of transport limitation. 60 This test was carried out at 318 and 328 K to confirm the absence of both mass and heat transfer resistances in the system.  Table 1. All the catalysts exhibit mesoporous and type IV isotherm characteristics. Notably, post impregnation of the Pd metal into Rh/Al 2 O 3 catalyst, the specific surface area decreased from 164 to 145.5 m 2 /g. On the basis of the CO chemisorption analysis, 12% of the metals in the prepared bimetallic PdRh/Al 2 O 3 catalyst are exposed on the catalyst surface as well as accessible to the reactants. Figure 1 shows the scanning electron microscopy with energy-dispersive X-ray spectroscopy (SEM-EDX) of the monometallic Rh/Al 2 O 3 and the prepared bimetallic PdRh/ Al 2 O 3 catalysts. The EDX images in parts b and f of Figure 1 confirmed the monometallic and bimetallic catalyst compositions, respectively. It can be seen from Figure 1h that Pd metals are well-dispersed in the prepared bimetallic PdRh/ Al 2 O 3 catalyst.
XRD is a common analytical technique used to identify phases and crystallinity of a sample material such as the prepared PdRh/Al 2 O 3 catalyst. The XRD patterns of the commercial monometallic Pd/Al 2 O 3 , unmodified commercial Rh/Al 2 O 3 , and Pd-modified Rh/Al 2 O 3 catalyst to produce bimetallic PdRh/Al 2 O 3 catalyst are shown in Figure 2 The SEM images and particle size distribution of PdRh/ Al 2 O 3 catalyst before and after reaction presented in Figure 3 show no significant difference in packing of the particles. In addition, the energy dispersive x-ray (EDX) analysis showed no carbon deposition on the used catalyst surface (Figure 3a and c). This thereby indicates that the reaction did not induce any observable change in the structure of the catalyst particles. This observation is in agreement with reports in the literature that catalyst deactivation as a result of carbon deposition is less likely at mild operating temperatures. 57,65,66 However, the particle size distribution before and after reaction (Figures 3b  and d) shows that 95% of the fresh catalyst particles have sizes within the range of 200−400 nm. Likewise the sizes of 93% of reused catalyst particles fall within the interval of 200−400 nm. Therefore, this indicates a slight drop in the proportion of catalyst particle sizes within the range of 200 nm and an increase in particle size of 300 nm. This may be as a result of little agglomeration of the particles of 200 nm into 300 nm.

Verification of Kinetically Controlled Reaction Regime.
To establish the influence of external and internal transfer limitations, the effect of agitation speed and particle size on initial reaction rates (r o ) was examined. Figure 4a depicts the effect of agitation speed, and Figure 4b represents the influence of particle size. Interestingly, Figure 4a shows that the initial reaction rate is independent of stirring speed in the range of 900−1300 rpm. Therefore, external diffusion limitation is absent when the stirring speed falls within this range. All further experiments in this work were carried out at a stirring speed of 1000 rpm. Figure 4b shows that intraparticle diffusion limitation is negligible for particles smaller than 250 μm. As a result, subsequent experiments were carried out with particles smaller than 250 μm.
To reaffirm the absence of internal mass transfer limitation, eq 1 was used to compute the observable modulus at 318 K. The results are presented in Table 2; because the computed observable moduli for both VL and H 2 were significantly <1, it can be concluded that internal mass transfer resistance is negligible. In addition, the observable modulus of H 2 found is 10 times bigger than that of VL, therefore confirming H 2 as the   Clearly, the result in Table 3 provides further justification for the suitability of PdRh/Al 2 O 3 for VL HDO reaction as it shows significantly improved selectivity toward CR (99.5%) and high conversion (99.7%).
In this study, the typical concentration−time profile observed is presented in Figure 5. It clearly confirmed VA as the intermediate product and CR as the final product, thereby supporting the proposed mechanism in the literature for VL HDO reaction. 54,55,57 4.4. Effect of Catalyst Loading. The influence of catalyst loading (ω) on the initial rates of VL disappearance (r 0 ) was studied over the range of 1.86−4.64 kg/m 3 at T = 318 and 328     Effect of H 2 Gas Partial Pressure. The effect of hydrogen gas partial pressures (P H ) on initial rates of VL disappearance (r 0 ) was investigated in vanillin−acetic acid and vanillin-only environments at 318 K. Figure 7 represents observed variation in r 0 as P H changes from 1.0 to 4.0 MPa. It shows that, in vanillin-only environment, r 0 increased from 9.6 × 10 −4 to 17.9 × 10 −4 kmol/kg cat min as P H increases from     This conclusion complements the initial finding in section 4.2 that H 2 gas is the limiting reactant in the system. The order of reaction with respect to P H found from the slope of In (r 0 ) against In (P H ) in Figure S1 is 0.44 in the vanillin-only environment and 0.63 in vanillin−acetic acid mixture. On the basis of the obtained values, it can be concluded that VL HDO reaction exhibits nonfirst-order reaction kinetics with respect to hydrogen gas partial pressure in vanillin-only and vanillin− acetic acid environments. Again, this is in agreement with the earlier claim in section 4.2 that the reaction order with respect to (wrt) P H is <2. Interestingly, the nonfirst-order dependence on P H reported in this study is in agreement with previous work on VL HDO reaction over monometallic Pd/C catalyst. 57 Notably, a similar observation of nonfirst-order dependence on P H was reported in the work of Wan et al. 67 on hydrogenation of 2-butanone over Ru/C catalyst using water as the reaction solvent. 4.6. Effect of Initial VL Concentration. The effect of starting VL concentration (C VL0 ) on the initial rate of disappearance (r 0 ) was investigated at 318 K in vanillin-only and vanillin−acetic acid environments, with the values of P H and ω kept at 2.0 MPa and 4.64 kg/m 3 , respectively. The results are summarized in Table 4; it shows that r 0 decreased from 13.9 × 10 −4 to 7.6 × 10 −4 kmol/kg cat min as C VL0 increased from 263 to 526 mM in the vanillin-only environment. Similarly, in the vanillin−acetic acid environment, r 0 decreased from 9.2 × 10 −4 to 4.2 × 10 −4 kmol/kg cat min as C VL0 increased from 263 to 526 mM. The observed trend is unusual but in agreement with findings from Aliu et al. 57 on VL hydrodeoxygenation over Pd/C catalyst. Interestingly, the work of Maccarrone et al. 68 on hydrogenation of 1-heptyne over alumina-supported Pd and Ni showed the same trend of initial reaction rates decreasing as starting concentration increases. This unusual trend suggests decreased catalytic activity as C VL0 increases, possibly because of increased competition for a finite number of active sites. In addition, it indicates nonfirst-order dependence on VL concentration. In fact, from the plot of In (r 0 ) against In (C VL0 ) in Figure S2, the order of reaction wrt VL concentration found is −0.8 (in vanillin-only environment) and −1.1 (in vanillin−acetic acid environment). These values reaffirm the conclusion in section 4.2 regarding the reaction order with respect to VL from the estimated VL observable modulus. In comparison to the value of −0.7 reported as the order of reaction wrt VL concentration in the work of Aliu et al., 57 the values found in this study are in the same range. Hence, the change in catalyst from monometallic Pd/C to PdRh/Al 2 O 3 had little effect on the reaction dependence on initial VL concentration. Additionally, observed initial rates of VL disappearance were lower in the presence of acetic acid. However, the estimated orders of reaction wrt VL in vanillin− acetic acid and vanillin-only environments are identical. Table 5 shows the distribution of products in vanillin-only and vanillin−acetic acid environments; the same trend can be observed in the intermediate and final product selectivities as the starting concentration of vanillin increases. Notably the selectivity toward the desired product creosol differs between the vanillin-only and vanillin−acetic acid environments. However, the change cannot be attributed to active site modification. Rather, increased competition for a finite number of sites on the catalyst surface is a more plausible reason for the observed change in reaction rate, activation energy, and product distribution.   To examine the stability and reusability of PdRh/Al 2 O 3 catalyst, it was recycled thrice under the conditions of 338 K, 1000 rpm, 2.0 MPa, and 1 h reaction time using a catalyst loading of 4.67 kg/ m 3 and initial VL concentration of 263 mM. Prior to reusing the catalyst, it was dried overnight at room temperature. As shown in Figure 8, conversion, which is indicative of the catalyst activity, remained unchanged throughout the three cycles. However, slight changes can be observed in the product distribution between the first and second cycles. This indicates a small change in metal sites due to the absence of regeneration. Because the rate calculations were based on conversion, which remained the same for the three cycles, it follows that modeling of the reaction kinetics does not require a deactivation term.
4.8. Proposed Scheme for Vanillin HDO. On the basis of the products observed in this study, Scheme 1 represents the proposed steps in vanillin (VL) HDO reaction. Interestingly, this scheme is in agreement with Table 3 in section 4.3 and literature. 54,55,57 The first step involves hydrogenation of VL to vanillyl alcohol (VA), while in the last step VA is further hydrogenated to the product of interest, creosol (CR). At temperatures higher than the range considered in this study, other products like guaiacol and aryl group hydrocarbons may appear as a result of side reactions such as demethylation and decarbonylation. 54,57 4.9. Modeling of VL HDO Reaction Kinetics. Most of the previous attempts in the literature to model the HDO reaction of VL and other related bio-oil model compounds favored the Langmuir−Hinshelwood−Hougen−Watson (LHHW) approach for reasons elucidated in Aliu et al. 53,57 Hence, the LHHW approach is used in this work to model the reaction kinetics of VL over PdRh/Al 2 O 3 catalyst. The majority of fluid solid catalytic reactions modeling via the LHHW method assumed that hydrogen adsorbed dissociatively onto the catalyst surface. 69−71 Therefore, the derived rate expressions presented in Table 6 are based on dissociatively chemisorbed hydrogen atoms. The elementary steps associated with competitive adsorption of dissociatively chemisorbed H 2 and VL are represented by eqs 2−4 (see model derivation in section S1).
In the presence of acetic acid (AA), the elementary steps for competitive adsorption of dissociatively chemisorbed H 2 , VL, and AA are represented by eqs 5−8 (see derivation of model for the binary mixture in section S1).
where H 2 is molecular hydrogen, H is atomic hydrogen, AAS is adsorbed acetic acid intermediate, HS is adsorbed atomic hydrogen, VLS is the adsorbed vanillin intermediate, P is the final product, and S is the catalyst active site. It is worth mentioning that previous work on the VL HDO reaction over Pd/C confirmed the suitability of competitive adsorption models over noncompetitive models, 57 hence the decision in this work to focus only on competitive dissociative adsorption models. In addition, other possibilities of H 2 and VL adsorption steps as the rate-determining step (RDS) have been ruled out in previous work. 57 To estimate the kinetic  Figure 9. Parity plots comparing predicted and experimental initial reaction rates of VL in (a) nonacidic and (b) acidic environments. Lines represent best fit to data.

Energy & Fuels
Article parameters (i.e., surface reaction rate constant (k s ), hydrogen adsorption constant (K H ), VL adsorption constant (K VL ), and acetic acid adsorption constant (K A ) in the rate expressions for vanillin−acetic acid and vanillin-only environments (see Table  6), a nonlinear generalized reduced gradient (GRG) solver in Microsoft Excel was used to minimize the objective function defined as residual sum of squares (RSS) in eq 9.
Herein, r Exp denotes experimental reaction rates, and r Mod represents predicted reaction rates from the models. The estimated values for surface reaction rate constant, adsorption constants, heat of adsorption, and activation energy are provided in Table 7. These values were estimated using the Arrhenius expression for surface reaction rate constant and Van't Hoff isochore for the heats of adsorption. Interestingly, Table 7 shows that the VL HDO reaction activation energy is ∼2 times higher in the presence of acetic acid. This explains the observed lower initial VL reaction rates in the presence of acetic acid compared to the vanillin-only environment in section 4.6. In comparison to values of 50.6 kJ/mol reported in a previous work by Aliu et al. 57 and 41.2 kJ/mol in the work of Bindwal and Vaidya, 54 the synergy between Pd and Rh metals significantly reduced the VL HDO reaction activation energy; this is consistent with the TOF comparison in section 4.4. However, this benefit diminished when the reaction was carried out in the presence of acetic acid. As shown in Table 7, the estimated activation energy for VL HDO reaction in the presence of acetic acid is the same as the value reported by Aliu   57 On the basis of the enthalpy of adsorption estimated from the models, it can be concluded that the strengths of adsorptions of both VL and H 2 are significantly smaller compared to that for acetic acid. The strong interaction between the catalyst surface and the acetic acid increased competition for available active sites. As a result, the VL HDO reaction requires higher activation energy in the presence of acetic acid. The parity plots in parts a and b of Figure 9 illustrate the level of agreement between predicted and observed initial reaction rates of VL in vanillin−acetic acid and vanillin-only environments. Clearly, with R 2 values of 0.99 (vanillin-only environment) and 0.98 (vanillin−acetic acid environment) for the parity plots, the proposed models reliably describe the VL HDO reaction in a batch reactor. Parts a and b of Figure 10 further reaffirm the fit of the proposed models as they indicate strong agreement between predicted and experimental VL concentrations at three different reaction temperatures.
The fractional coverage of H 2 , VL, and acetic acid as a function of time can be seen in Table 8. It shows that the fractional coverage of VL decreases over time, while those of H 2 and acetic acid were almost constant. Notably, the mild temperature used in this study accounts for the negligible change in acetic acid coverage. In addition, H 2 coverage did not change over time because of the fixed external H 2 gas supply applied to maintain a constant reaction pressure.

CONCLUSIONS
In summary, the specific objective of this work is to examine any differences between the hydrodeoxygenation (HDO) reaction kinetics of vanillin, a typical model compound of bio-oil in pure and acidic environments. In addition, the roles of noble metals such as Pd and Rh on alumina were investigated to obtain a combination of metals that maximizes selectivity toward the desired fuel product creosol. The use of Pd/Al 2 O 3 favors the conversion of vanillin (99%) but resulted in low selectivity toward creosol (4%). Conversely, in the presence of Rh/Al 2 O 3 , low conversion (25%) and high selectivity toward creosol (49%) were achieved. This stimulated the use of PdRh/Al 2 O 3 to further improve selectivity toward creosol while maintaining high conversion. Interestingly, selectivity toward creosol increased significantly to 99.5% with conversion still hovering above 99% in the presence of PdRh/Al 2 O 3 . Consequently, all other experiments in this work were conducted using the prepared bimetallic PdRh/Al 2 O 3 catalyst. It was found that stirring speed in the range of 900−1300 rpm and particle size less than 250 μm ensure the absence of transport limitations within the system. In the vanillin-only environment, the order of reaction with respect to VL concentration was −0.8, while with respect to hydrogen gas partial pressure (P H ) the order was 0.44. Conversely, in the presence of acetic acid, the estimated order of reaction with respect to VL concentration was −1.1 and and that with respect to P H was 0.63. The initial rates of VL disappearance during the reactions in vanillin−acetic acid and vanillin-only environments were modeled through derived rate expressions from the Langmuir−Hinshelwood−Hougen− Watson (LHHW) method. The estimated values from the model for the vanillin-only environment were 24.1 kJ/mol for reaction activation energy, 13.4 kJ/mol for VL adsorption enthalpy, and 7.8 kJ/mol for H 2 adsorption enthalpy. The corresponding values from the modified model for VL HDO reaction in the vanillin−acetic acid environment include 51.0 kJ/mol for the activation energy, 13.8 kJ/mol for VL adsorption enthalpy, and 30.8 kJ/mol for H 2 adsorption enthalpy. On the basis of the results obtained from this study, the presence of acetic acid significantly increased the energy barrier for vanillin HDO reaction.

* S Supporting Information
The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.energyfuels.9b02993.

■ ABBREVIATIONS
C H = concentration of hydrogen in the liquid phase, kmol m −3 C VL0 = initial concentration of vanillin in the liquid phase, kmol m −3 CR = creosol D e i = effective diffusivity of species i in eq 1 , m 2 /s E A = activation energy of the reaction, kJ/mol ΔH H = enthalpy of adsorption for hydrogen, kJ/mol ΔH VL = enthalpy of adsorption for vanillin, kJ/mol ΔH A = enthalpy of adsorption for acetic acid, kJ/mol HDO = hydrodeoxygenation K H = adsorption equilibrium constant for hydrogen gas, m 3 / kmol K S = surface reaction rate constant, kmol/(kg cat min) K VL = adsorption equilibrium constant for vanillin, m 3 /kmol K A = adsorption equilibrium constant for acetic acid, m 3 / kmol L = characteristic length of catalyst particle LHHW = Langmuir−Hinshelwood−Hougen−Watson P H = hydrogen gas partial pressure, MPa r o = initial reaction rate, kmol/(kg cat min) rpm = revolutions per min r Exp = experimental values of reaction rates, kmol/(kg cat min) r Mod = reaction rates predicted from LHHW models, kmol/ (kg cat min)

Energy & Fuels
Article R 2 = coefficient of determination RSS = residual sum of squares S i = active site i on the catalyst surface T = temperature, K t = time, min VA = vanillyl alcohol VL = vanillin ω = catalyst loading, kg m −3 Φ = Thiele modulus η = effectiveness factor ηΦ i 2 = observable modulus for species i in eq 1 θ V = vacant site fractional coverage θ VL = vanillin fractional coverage θ A = acetic acid fractional coverage θ H = hydrogen fractional coverage TOF = turnover frequency, 1/min