Continuous hydrogen production from liquid-phase formic acid dehydrogenation over Pd/AC catalysts: A kinetic study

Hydrogen production using formic acid (FA) as renewable carrier has been investigated in a fixed bed reactor packed with a commercial Pd/AC catalyst. For the first time, both FA disappearance and evolved gas flow rate have been monitored upon space-time, enabling the elucidation of the FA reaction pathway and the development of a kinetic model that accounts for catalyst deactivation. Nearly complete FA conversion and a production of 10 mL min (cid:0) 1 of hydrogen gas were achieved under the following operating conditions: C FA,0 = 1 M, T = 45 º C and τ = 66.7 g CAT h L (cid:0) 1 . The reaction was found not to be controlled the mass transfer limitations. The kinetic model reveals a first order with respect to FA concentration, with FA disappearing through dehydrogenation into hydrogen and CO 2 (E a = 53.6 kJ mol (cid:0) 1 ) as well as sorption onto the catalyst surface without reaction (E a = 36.7 kJ mol (cid:0) 1 ). The catalyst deactivation is attributed to the accumulation of reaction species, including FA/ HCOO - (reversibly sorbed) and CO 2 (irreversibly chemisorbed), on the Pd active sites and the progressive decrease in the Pd 2 + /Pd 0 ratio.


Introduction
Due to climate change caused by carbon dioxide emissions and the decreasing supply of fossil fuels, the initiative to develop alternative fuels has gained momentum.Hydrogen is one of the most promising candidates since it is considered an efficient and clean energy carrier that can be used as a carbon-emissions free fuel for various devices, such as portable and stationary electronic devices and fuel cell vehicles, among others [1,2].In fact, hydrogen has almost three times more energy content than gasoline [3].
Despite this, hydrogen implementation and launch as an energy vector is delayed due to some challenges that must be faced in the four stages of the hydrogen commercialization: production, storage, transport and end-use.Regarding to the hydrogen storage, the barriers are the high energy requirement in compressed hydrogen in tanks at 350 -700 bar, due to its low specific gravity, the contamination that can suffer the hydrogen when is bulk storage at geographic features, or the low durability of solid materials used for storage (i.e.fiber, metals, polymers etc.).However, the chemical storage, using inorganic and organic liquid compounds, can offer a high storage performance, it is a safer alternative to the physical hydrogen storage and, in case of liquid organic chemicals, commonly known as LOHCs (Liquid Organic Hydrogen Carriers), the hydrogen storage and transportation are a low investment risk due to their compatibility with the existing fuel infrastructure [4].From LOHCs, the hydrogen is extracted on demand by a catalytic dehydrogenation reaction that it must be reversible to recover the LOHC in a cyclic process.
The research focused on LOHCs proposes the use of cycloalkanes, such as cyclohexane, methylcyclohexane and decalin [5], N-heterocycles such as dodecahydro-N-ethylcarbazole [6], formic acid (FA) [7][8][9][10][11][12][13][14][15][16] and carbohydrates [5].The last two are very promising due to their low cost, renewable source and high purity hydrogen generated, only accompanied by CO 2 , which can be easily removed.The advantage of FA on the use of carbohydrates is that the co-product CO 2 can be hydrogenated back to FA in a second step.
The present work deals with the production of hydrogen from FA using a commercial palladium (10 wt%) on activated carbon catalyst (Pd/AC).In our previous work [17], the suitability of this Pd/AC catalyst to produce CO-free hydrogen from FA decomposition in liquid phase was demonstrated by investigating its activity, stability and recyclability in a batch-wise slurry reactor.Also, the kinetic model, including the catalytic deactivation in absence of mass transfer limitation, was elucidated.It was reported that the reaction follows a first order kinetic for FA while the activity showed an exponential decay with the initial FA concentration and reaction temperature.Both variables affected the amount of CO 2 produced, which is the main responsible species for the Pd/AC deactivation due to its interaction with Pd active sites.Interestingly, it was possible the recovery of the initial activity by drying the catalyst at 60 ºC, though a slight progressive and irreversible catalyst deactivation occurred upon the cycles use.
For the practical application of the FA as LOHC, and more considering the desirable hydrogen production on-demand, the FA dehydrogenation reaction should be operated in continuous mode.In this sense, the Bellers' group has been working on the development of a continuous stirred tank reactor set-up using homogeneous catalysts [18][19][20][21].Homogenous catalysts such as RuH 2 (ddppe) 2 are very active, producing unprecedent hydrogen flows of 47 L h − 1 at 25 ºC under optimized conditions [21].Unfortunately, these Ru catalysts require the use of an amine, to form a reactive FA-amine adduct and the recycling of the amine still represents a challenge.The use of solid catalysts, such as the above mentioned Pd/AC, would allow the continuous production of hydrogen in fixed-bed reactors with the consequent increase in reaction rates.In spite of the benefits of this type of catalytic reactor (i.e.ease to build and catalyst immobilization, low cost of construction, operation and maintenance, the use of more reactant concentrations and greater separation between the products/reactants and catalyst), few works deal with the FA decomposition in aqueous phase [22,23], conditions under which the total selectivity to the dehydrogenation reaction is usually assured, and CO-free hydrogen is produced.Hu et al. [22] and O'Neill et al. [24] studied the kinetic of the FA decomposition reaction over a 10 wt% Pd on carbon (Pd/C) and PtRuBiO x supported on carbon (PtRuBiO x /C) catalyst, respectively.In both studies, the deactivation was reversible, and it was caused by the reactive species or products accumulation.The deactivation kinetic model was not considered in the FA reaction rate, but the species accumulation was considered by including the adsorption term in the kinetic equation, such as the adsorption of CO, H 2 and FA on the Pd/AC catalyst [22] and the adsorption of FA and formate ion (HCOO -) over PtRuBiO x /C [24].In addition, the apparent activation energy values reported were very different (i.e.37 kJ mol − 1 for PtRuBiO x /C and 67 kJ mol − 1 for Pd/C) indicating that the PtRuBiO x /C exhibits a higher catalytic activity than Pd/C with the increase in temperature, as expected in a trimetallic catalyst.Lately, Caiti et al. [23] investigated the FA decomposition over a 5 wt% Pd/C catalyst in a fixed-bed reactor and found that the HCOO - species were the main responsible species of the catalyst deactivation.This deactivation was more pronounced in a fixed-bed than in batch-wise slurry reactor (by 35-fold), due to the exposition of the catalyst to higher amounts of FA/HCOO -species caused by the continuous FA feeding.Consequently, the conversion of FA was unpredictably more favourable in the slurry than in the fixed-bed reactor.The activity could be restored by washing the catalyst with water.In none of the continuous studies [22,24], the evolved gas phase was monitored upon reaction, nor the hydrogen production quantified.However, this is a key aspect for comprehending and modelling the chemical process, as will be demonstrated in this work.
The motivation of the present work is to progress in the implementation of hydrogen technology using FA as sustainable LOHC.To achieve this, the continuous hydrogen production in a commercial 10 wt % Pd/AC fixed-bed reactor has been modelled assuming a plug-flow reactor.For the first time, FA conversion and evolved gas flow rate have been monitored upon space-time, and the kinetic equation for FA disappearance, including the deactivation kinetic, has been reported in a flow reactor.Furthermore, the influence of the external and internal mass transport of FA and hydrogen has been examined.The obtained results have been compared with those from our previous work carried out in a batch-wise slurry reactor [17].

Pd/AC catalyst and characterization
A commercial Pd (10 wt%) supported on activated carbon catalyst (10 wt% Pd/AC, henceforth Pd/AC) was supplied by Evonik (Noblyst ® P1070) in powder form.The characterization of the fresh Pd/AC was carried out elsewhere [17], being the average Pd particle size (d p ) = 1.6 ± 0.2 nm and Pd 2+ /Pd 0 ratio = 2.4.Moreover, the BET surface area (S BET ) and the external area (A ext ) measured are 887 and 404 m 2 g − 1 , respectively, which in accordance with the IUPAC classification, Pd/AC is a microporous material.Herein, different techniques were used to characterize the use Pd/AC catalysts once dried overnight at 60 ºC.The d p was examined by the transmission electron microscopy operating at 200 kV (TEM, JEOL 2100 F).The exposed Pd surface species (Pd 0, Pd 2+ ) were quantified by X-ray photoelectron spectroscopy (XPS).Each Pd species displays two peaks due to the 3d 5/2 and 3d 3/2 transitions.Peaks at 335.5 and 340.9 eV are ascribed to Pd 0 while peaks at 337.5 and 342.6 eV to Pd 2+ , with PdO being the most likely species [25][26][27][28].The spectra measurements were performed with the PHI5000 VersaProbe II using a monochromatic Al-Ka X-ray beam (1486.6 eV).Charge referencing was measured against adventitious carbon (C 1 s at 284.8 eV).All XPS spectra were analysed using the CasaXPS software.Baseline corrections were made using a Shirley type background correction and the peak shapes was approximated as a Gaussian line shape.
The S BET and A ext of the used catalysts were measured at 77 K using a Micromeritics Tristar 3000 apparatus nitrogen using the Brunauer-Emmett-Teller (BET) method.Before the measurement, the samples were outgassed at 120 • C overnight.The elemental composition of the catalysts was measured using a LECO CHNS-932 analyzer.The overall Pd loading in the catalysts after reaction was measured by total reflection X-ray fluorescence (TXRF) using a benchtop S2 PICOFOX TXRF spectrometer (Bruker Nano).Attenuated total reflection (ATR) spectroscopy was performed to detect species adsorbed on Pd nanoparticles, and the formation of Pd hydride.A Bruker vector 22 FTIR spectrometer equipped with a MCT detector and Harrick diffuse reflectance accessory was used.Spectra was obtained with apparatus loading on ~ 0.2 g of sample mixed with KBr scanning at 4 cm -1 resolutions with 20 scans from 4000 to 450 cm -1 .

Catalytic dehydrogenation of FA in a fixed bed reactor
Dehydrogenation experiments were conducted in the set-up provided in Fig. 1.The main part is the tubular reactor, a glass tube of 16 mm of internal diameter with a water jacket for thermostating (GE Healthcare, XK16/20 mm).The catalytic bed consisted of the appropriate mass of powdered Pd/AC catalyst placed inside a cylindrical β-SiC  foam, supplied by SICAT, with a diameter of 15 mm, height of 22 mm and pores per linear inch (PPI) equal to 30, to avoid the catalyst caking upon reaction.The foam was held between two small beds of spherical glass beads of 2 mm diameter to help to the flow distribution, at the entrance, and to retain the powdered catalyst, at the exit.
Constant flow rates of FA solution and Helium gas were obtained by using a piston pump (Gilson 307 HPLC) and a mass flow controller (Bronkhorst High Technology), respectively.Both streams, liquid and gas, were mixed, preheated and fed to the reactor in concurrent up-flow.The reactor effluent was cooled down by means of a Peltier cell placed in a steel phase separator.
For the kinetic studies, a FA aqueous solution, at a given initial FA concentration (C FA,0 = 0.25 -1 M), was continuously fed to the reactor at different flow rates (Q L = 0.125-1 mL min -1 ) to cover the experimental range of space-time values (τ = W Q L -1 = 8.3-66.7 g CAT h L -1 ).A 17 NmL min -1 pure Helium flow was continuously feeding in all the experiments to assure a constant gas flow rate.The dehydrogenation runs were performed at a constant temperature, ranged from 25 to 55 ºC, atmospheric pressure and using a mass of catalyst (W) equal to 0.5 g.
Long-term experiments to study the catalyst deactivation and recyclability in continuous mode operation were carried out at C FA,0 = 1 M, Q L = 0.25 mL min -1 , Q He =17 NmL min -1 , T = 25-55 ºC, W = 0.5 g and τ = 33.3g CAT h L -1 .After each experiment, the catalyst was recovered from the foam and dried overnight at 60 ºC, as regeneration treatment, before next use.Additionally, two control experiments were conducted to further understand the causes of catalyst deactivation.In one experiment, the catalytic bed was washed with water for several hours before the reaction.In the other experiment, the commercial Pd/AC underwent a reduction treatment at 250 ºC under a H 2 /N 2 stream (50/100 NmL min -1 ) for 2 h at a heating rate of 10 ºC min -1 , prior to the reaction.Both experiments were conducted under the following operating conditions: Also, the external mass transfer was experimentally evaluated by studying the effect of the liquid flow rate (Q L =0.05-0.4mL min − 1 ) on FA conversion in a reactor loaded with 0.2 g of catalyst.The liquid flow rate was proportionally changed to the mass of catalyst to operate in the same range of space-time as in the reactor loaded with 0.5 g.
The FA conversion (X FA ) and the evolve gas flow rate (Q GAS ) were used for evaluating the catalytic performance: ⋅dilution factor (2) where C FA,exit (mol L -1 ) is the FA concentration in the reactor liquid effluent, at the reactor exit, and the dilution factor is calculated according to the gas composition measured in the gas effluent.The detailed calculations are provided in the Supporting Information.The evolve hydrogen flow rate (Q H2,GC ) was obtained from Q GAS and the hydrogen concentration measured by GC, in percentage, in the gas effluent.
Furthermore, the theoretical hydrogen flow rate (Q H2,cal in L min -1 ) was calculated from the theoretical hydrogen molar flow rate (F H2,cal in mol H2 min -1 ) using the ideal gas law: where R is the ideal gas constant (0.082 atm L K -1 mol -1 ), T (in K) is the reaction temperature and P (in atm) is the atmospheric pressure.F H2,cal is calculated from the amount of FA converted as: where F FA,0 is the FA molar flow rate fed to the reactor, calculated as Q L C FA0 .

Analytical methods
The progress of the reaction was followed by analysing the liquid and gas samples collected from the separator.The FA concentration was determined in a Cary 60 UV− vis spectrophotometer at a wavelength of 210 nm.The total inorganic carbon (TIC) was measured using a TOC analyser (Shimadzu, mod.TOC-Vsch) to quantify the content of dissolved CO 2 in the liquid effluent.The composition of the evolve gas was analyzed by a gas chromatograph (Agilent 6890) equipped with a thermal conductivity detector using a Varian select permanent gases/ CO 2 column.The H 2 , CO, CO 2 , O 2 and N 2 gases were calibrated using two commercial standards.Helium was used as carrier gas.Finally, the content of Pd in solution was measured by TXRF.

Kinetic model
The mass balance of FA in the fixed-bed reactor, assuming isothermal plug-flow and absence of reaction in the liquid phase, can be expressed as: where Considering that Q L remains constant, and the definition of τ, the reaction rate can be expressed as: The FA reaction rate depends not only on the temperature and concentration of FA, but also on the activity of the catalyst (a), as shown in the following equation: where ( − r FA ) w,0 is the initial reaction rate, without catalyst deactivation, in mol g CAT -1 h − 1 .
The activity is calculated as the ratio between the X FA at the exit of the reactor at a given time on stream and the initial conversion, without catalyst deactivation: The activity decay with time on stream (t) always attends to a potential model: where k d is the deactivation kinetic constant (in h -1 ) and m is the empirical exponent, which takes values of 0, 1 and 2 for linear, exponential, or hyperbolic decay curves, respectively.It is expected that the kinetic equation proposed for (− r FA ) w,0 to describe the influence of temperature and concentration is a first order reaction, as it was elucidated in our previous work by performing the hydrogen production in presence of the Pd/AC catalyst in a batch-wise slurry reactor [17], and also that the activity decreases exponentially with the time on stream.Therefore, the activity decay is also described by a first-order equation, m=1 in Eq. (8).Consequently, the FA dehydrogenation reaction rate should be expressed as: (10) where k FA (in L g CAT -1 h -1 ) is the initial kinetic rate constant of FA dehydrogenation reaction.OriginLab 2017 program was used for the non-linear regressions aimed at calculating the kinetic parameters.It is based on the Levenberg-Marquardt algorithm to minimize the chi-square (χ 2 ) C. Martin et al. function, defined as the the residual sum of squares (RSS) by the degrees of freedom.The initial conditions considered are C FA = C FA,0 at τ = 0 g CAT h L -1 and t = 0 h.The model discrimination was based on statistical analysis, considering the coefficient of determination (R 2 ) closer to one, and the physical meaning of the estimated parameters.

Catalytic activity
The influence of the reaction temperature on both the X FA and evolve Q H2,GC was systematically investigated by performing the dehydrogenation reaction from 25 to 55 ºC at the following standard conditions: C FA,0 = 1 M, W = 0.5 g and Q He = 17 NmL min -1 in a wide range of space-time values.The results are depicted in Fig. 2. They are values monitored at stationary state of the reaction system, before the X FA began to decrease because of the deactivation.
As can be seen in Fig. 2a, the X FA increases with the space-time and the rise of temperature.Almost complete conversion (X FA = 96 %) is achieved at T = 45 ºC, C FA,0 = 1 M and τ = 66.7 g CAT h L -1 (equivalent to 1.5 g CAT h g FA -1 ).It is difficult to compare this activity with that of other catalysts employed in the liquid-phase FA dehydrogenation in continuous reactors due to the varying operating conditions used.For instance, the 5 wt% Pd/AC catalyst studied by Caiti et al. [23] exhibited a X FA = 82 % at T = 50 ºC and 0.44 g CAT h g FA -1 and a X FA = 100 % at T = 90 ºC and 0.44 g CAT h g FA -1 , while the PtRuBiO x /C catalyst provided very low FA conversions, X FA = 1-3 % at T = 60 ºC, because the space-times selected were also very low, from 0.007 to 0.016 g CAT h g FA -1 [22].
Regarding to hydrogen production (Fig. 2b), the Q H2,GC is favored by the increasing temperature (viz. up to 4, 6, 10 and 11 mL min -1 are detected at 25, 35, 45 and 55 ºC, respectively) though the temperature influence seems to be less significant above 45 ºC.Noticeable, at low reaction temperatures, 25 and 35 ºC, there is a maximum noticed at the low space-time of 16.7 g CAT h L -1 , while at 45 and 55 ºC, the hydrogen flow rate profiles diminish in the range of the space-time used.Initially, these profiles may seem contradictory because a lower hydrogen flow rate is obtained at higher FA conversion.However, this behaviour is expected considering that the increase in space-time is achieved by decreasing the liquid flow rate, and consequently the molar flow rate of FA fed in the reactor.In fact, the same trends are obtained when hydrogen flow rates are calculated from the FA conversion and liquid flow rate (Eqs.3 and 4), curves provided in Figure S1 of the Supporting Information.It is noteworthy that the Q H2,GC values measured by GC (Eq.2) are lower than the calculated ones, Q H2,cal , (Eqs. 3 and 4), especially noticeable at high liquid flow rates (or low space-time values) and lower reaction temperatures (Figure S1 of the Supporting Information).This behavior was not observed in a batch-wise stirred tank reactor operated at the same range of reaction temperature and with the same catalyst [17].In this catalytic system, the measured hydrogen production coincided with the value calculated from the FA conversion.
A straightforward reason for the mismatch observed in fixed bed reactors would be the presence of mass transfer limitation, which could also explain the poor influence of the temperature on the hydrogen production above 45 ºC.However, the mass transfer resistance increases at low liquid flow rates (or high space-time values), conditions to which the calculated and the experimental values are mostly coincident (Figure S1 of the Supporting Information).Therefore, though the mass transfer will be further analyzed, a different explanation must be provided.In this line, the retention of hydrogen and/or FA/HCOO -species on the Pd/AC catalyst may be the reasons behind the mismatch between FA conversion and hydrogen production.For instance, Pd has the property of absorbing hydrogen in the crystal matrix to form Pd hydrides even at low hydrogen pressures and room temperatures [29,30].Also, FA/HCOO -species could remain adsorbed on the Pd sites, reducing the number of active sites available [22][23][24].A deep characterization of the used catalysts will be conducted to clarify this point.
Regarding the catalyst selectivity, only hydrogen and CO 2 were detected in the gas effluent; CO was never observed.Therefore, FA decomposition occurs exclusively through the dehydrogenation of FA, and the catalyst demonstrates 100 % selectivity towards hydrogen.
Finally, the H 2 /CO 2 molar ratio measured in all the experiments in Fig. 2 was slightly above 1 (see Figure S2 of the Supporting Information).These findings suggest that CO 2 may either be retained in the liquid effluent [23] or on the catalyst surface.However, the acidic pH of the liquid effluent, with the lowest measured value being 2.4 in the reaction carried out at T = 25 ºC and 2.9 at 55 ºC, indicates that the concentration of CO 2 in the liquid phase can be considered negligible.In fact, TIC analyses have been conducted on the liquid effluents, with their value consistently null in all cases.Therefore, it can be concluded that CO 2 is not present in the liquid reactor effluent, and most probably it is retained on the catalyst surface.

Mass transfer analysis
The external mass transfer effect on the catalytic performance has been experimentally evaluated by using the same fixed bed reactor as in Fig. 2, with the same configuration, but with different amount of catalyst (W = 0.2 g instead of 0.5 g) at the following selected operating conditions: C FA,0 = 1 M, T = 35ºC and Q He = 17 NmL min -1 .Thus, the liquid flow rate was proportionally changed to the mass of catalyst to operate in the same range of space-time.
Fig. 3a shows the X FA obtained in both reactors.The temporal profile for X FA is the same in both cases, except at very low flow rates.Specifically, at Q L = 0.05 mL min -1 (W = 0.2 g), the X FA is lower than the value C. Martin et al. obtained at Q L = 0.125 mL min -1 (W = 0.5 g).This indicates that at FA flow rates below 0.125 mL min -1 , the diffusion of FA from the liquid to the catalyst surface limit the chemical reaction rate, with FA mass transfer rate controlling the overall chemical process.
Regarding to hydrogen production (Fig. 3b), the calculated Q H2,GC values are lower in the fixed bed reactor loaded with W = 0.2 g compared to 0.5 g, as expected.This is because the same X FA values are achieved in both cases (Fig. 3a) but the liquid flow rates, or the FA molar flow rates fed to the reactor, are lower when 0.2 g of catalyst is used.Again, the Q H2 values measured by the GC are below the calculated ones.If the reason were the control of the process by the hydrogen mass transfer to the gas phase, then the difference between the Q H2,GC and Q H2,cal would be more evident at higher space-times or lower liquid flow rate, but this is not the case.Hence, the lower production of hydrogen is likely attributable to the accumulation of reaction species on the catalyst surface rather than external hydrogen mass transfer limitations.
On the other hand, the effect of internal mass transfer limitation on the catalytic rate is unlikely considering that powdered Pd/AC catalyst is used (grain particle size of 25 μm).To prove it, the Thiele modulus has been calculated for FA at the highest reaction temperature.The calculation is provided in the Supporting Information.The FA diffusion coefficient was estimated by the Wilke-Chang equation [31].The apparent rate constants (k FA,app ) were calculated from the FA concentration vs. τ profiles assuming first order for FA.The values are summarized in Table 1 at each temperature.As expected, the k FA,app values rise with the temperature, that allows to achieve a higher FA conversion and hydrogen and CO 2 production at a given space-time value (Table 1 and Fig. 2).The Thiele modulus value at 55 ºC was 0.005, and then the internal effectiveness factor equal to unity.Therefore, the control of internal diffusion of FA can be neglected, as well as for hydrogen since the hydrogen molecular diffusion coefficient is higher than FA.

Catalyst stability and characterization
To study the Pd/AC stability, X FA as a function of time-on-stream was monitored until the complete loss of activity, at different reaction temperatures and τ = 33.3g CAT h L -1 .In no case, Pd was leached out from the catalyst and detected in the liquid effluent.The profiles are provided in Fig. 4. The initial X FA values are those already included in Fig. 2a at τ = 33.3g CAT h L -1 .They have been also collected in Table 1.As can be seen, from that initial value, the conversion exponentially decreases with time-on-stream (Fig. 4a) as well as the hydrogen produced (Fig. 4b), which points out that the activity evolution attends to a firstorder equation, m = 1 in Eq. 9. From the activity vs. time-on-stream profiles shown in Fig. 4c, the deactivation rate constant (k d ) values were estimated at each temperature and summarized in Table 1.The k d values also rise with temperature.Therefore, a higher reaction temperature implies a faster deactivation, coinciding with a higher production of hydrogen and CO 2 at a given space-time value (Fig. 2b).Similar behavior was observed by performing the reaction in a slurry reactor [17].
After each use, the catalysts were dried overnight at 60 ºC and then characterized.The TEM images, including the particle size distribution histograms, and the XPS spectra of the used Pd/AC catalysts are displayed in Figures S3 and S4 of the Supporting Information, respectively.These results along with the textural properties (S BET and A ext ) and the bulk C and H content are summarized in Table 2.
As can be seen, a slight sintering of Pd particle size takes place upon the reaction, and it becomes more significant with increasing the reaction temperature, i.e. from 1.6 nm in the fresh Pd/AC to 3.4 nm in the 1st used Pd/AC at 55 ºC, while is not affected by the consecutive uses.Also, a significant reduction of Pd 2+ to Pd 0 species occurs upon reaction and it is favored by both the rise of the reaction temperature and time-onstream, i.e. from 2.4 in the fresh Pd/AC up to 0.32 in the 3rd used Pd/ AC at 55 ºC.This points out that the number of active sites, which are the electrodeficient Pd species where the HCOO -species interact [22,25,[32][33][34], decreases upon time-on-stream, and they are not recovered after the dried treatment.Besides, the S BET and A ext diminish somehow upon reaction from temperatures above 35 ºC, though the augment of the C and the decrease of H content are not evidenced (Table 2).This could be because the molecules present on the catalyst surface, such as FA/HCOO -species, hydrogen or CO 2 are small molecules, and they scarcely contribute to the content of C and H in a carbon-based catalyst.
With the aim of gaining insight into this all the 1st used Pd/AC catalysts at the different reaction temperatures were analyzed by ATR.The absence of the characteristic band associated with palladium hydride (at 2150 cm -1 ) [35] suggests that either palladium hydride was decomposed and the hydrogen release during the storing and handling of the catalysts expose to the air atmosphere [29,30], or hydrogen absorption did not occur on the very small Pd nanoparticles during the reaction.In fact, previous studies have demonstrated that for very small Pd nanoparticles (smaller than 3 nm) hydrides are not formed, and  hydrogen is instead adsorbed on the Pd nanoparticles [36].Therefore, the formation of palladium hydride is most likely occurring at the higher reaction temperature tested, from 45 ºC.
On the other hand, the presence of CO 2 chemisorbed on the Pd nanoparticles was detected when the reaction was performed at 45 and 55 ºC, peaks at 2337 and 2375 cm -1 (Fig. 5b), corresponding to temperatures associated to a higher production of CO 2 .The presence of this species can contribute to the decrease of the S BET and A ext values, particularly at 45 and 55 ºC.
The retention of FA or HCOO -species on the Pd/AC catalyst causing deactivation cannot be discarded, about all when the catalyst activity is recovered by drying at 60 ºC overnight, as it will be further demonstrated.Unfortunately, the interaction between FA or HCOO -species and Pd nanoparticles, typically 1720-1739 cm -1 for CO, 2938 cm -1 for CH and 1578 cm -1 for OCO vibrations from FA molecules [37,38] could not be exclusively identified in the used catalysts since these bands are also present in the fresh Pd/AC catalyst, due to the presence of those groups in the activated carbon support.However, note that these species may be responsible for the slight decrease observed in the S BET and A ext values with the catalyst used (Table 2), and for the observed mismatch between the calculated and experimental hydrogen flow rates.Finally, no evidence of CO on Pd nanoparticles was discerned.The absence of bands in the 1975-2252 cm -1 range in the ATR spectra of any of the used catalysts [34,35] confirms that the FA decomposition upon the Pd/AC catalyst occurs exclusively through the dehydrogenation reaction, yielding H 2 and CO 2 .
According to the above results, the decline in Pd 2+ species can be attributed to various factors, including nanoparticle sintering and occupation of these sites by CO 2 molecules.Also, the retention of FA/ HCOO -species should be considered, and even the in-situ reduction of Pd 2+ species by the hydrogen produced upon reaction.The Pd sintering and the presence of FA/HCOO -and CO 2 can provoke the slight diminish observed in the S BET and A ext .
Similar causes of deactivation were identified in Pd/AC catalysts operated in the batch-wise slurry reactor [17], though in the latter case, the catalysts were characterized after three cycles of use (with a total of 12 h instead of 24 h on-stream as in this work) with regeneration in between by drying at 60 ºC.The characterization results of the Pd/AC catalyst used at 55 ºC in the slurry reactor are also collected in Table 2.As can be deduced, the Pd sintering is similar in both reactors, while the decreasing of the surface areas (S BET and A ext ) is more pronounced in the fixed bed reactor, as well as the loss of Pd 2+ species.This is reasonable considering that continuous mode operation in fixed bed reactors  subjects the catalyst to higher amounts of FA reactants and products.For instance, CO 2 molecules are now detected on the Pd sites during reaction conducted at 45 ºC, in contrast to the conditions in batch-wise slurry reactors where this detection occurred at 85 ºC [17].

Catalyst recyclability
Fig. 6 shows the temporal X FA and Q H2,GC profiles in consecutive reaction cycles at 25 ºC (Fig. 6a and b) and 55 ºC (Fig. 6c and d) at τ= 33.3 g CAT h L -1 with regeneration of the Pd/AC catalyst between cycles.
The deactivation rate constant values estimated by Eq. 9 have been included in Table 1 and the characterization results for the 3rd use Pd/ AC catalysts are collected in Table 2.The TEM images and XPS spectra of the 3rd use Pd/AC catalysts are provided in Figure S3 and S4 of the Supporting Information, respectively.
As can be seen in Fig. 6a, the initial X FA can be fully restored after regenerating the catalyst by drying it at 60 ºC overnight.These results suggest that the reversible deactivation observed is due to the accumulation of reaction species that can be removed during a dry treatment, such as water and FA/HCOO -.To further investigate this, a control experiment was conducted by washing the catalytic bed with water and then performing the FA decomposition reaction at T = 25 ºC, C FA,0 = 1 M and τ = 33.3g CAT h L -1 .The obtained X FA and Q H2,GC profiles upon reaction were identical to those observed in the absence of the washing step (see Figure S5 of the Supporting Information).This suggests that the retention of FA/HCOO -species on the Pd active sites is the primary cause of catalyst deactivation over time on stream.Also, the FA adsorption on  C. Martin et al. the AC support may not be discarded.This FA/HCOO -species accumulated on the Pd active sites and AC support that remain unreacted can explain the observed discrepancy between FA conversion and hydrogen production.
On the other hand, the initial Q H2,GC rate exhibits a distinct trend compared to the initial X FA in each cycle, as shown in Fig. 6b.Unlike X FA , the initial Q H2,GC values are not restored but rather decrease in each cycle, with a more pronounced decline observed at 55 ºC compared to 25 ºC.This suggests that hydrogen production is dependent on the presence of specific Pd species, whether electrodeficient or metallic Pd, while the retention of FA/HCOO -species is not.To confirm this, the FA dehydrogenation reaction has been conducted over a pre-reduced Pd/AC catalyst upon time on stream.This catalyst presented a lower Pd 2+ /Pd 0 ratio than the as-received one (Pd 2+ /Pd 0 = 0.6 vs. 2.4) under the same operating conditions (T = 25 ºC, C FA,0 = 1 M and τ = 33.3g CAT h L -1 ).
The X FA profiles are the same in both catalysts whereas the hydrogen production was lower over the pre-reduced Pd/AC catalyst.The XPS spectra and X FA and Q H2,GC profiles are shown in Figure S6 of the Supporting Information.These results confirm that hydrogen production requires the presence of Pd 2+ sites and put in relevance the significant contribution of the adsorption phenomenon to the observed FA conversion in fixed bed reactors, since this is independent of the nature of the Pd sites whereas the dehydrogenation reaction requires Pd 2+ sites.
Regarding to the deactivation observed with time-on-stream over consecutive cycles, the progressive decline in X FA and Q H2,GC occurs more rapidly with each cycle, as evidenced by the increase in k d values with catalyst use (Table 1).This can be attributed to the lower initial number of active sites in each cycle, as indicated by the lower Pd 2+ /Pd 0 ratio (Table 2).In fact, the pre-reduced Pd/AC catalyst has a k d = 0.43± 0.10 h -1 , higher than the as-received one (k d = 0.22 ± 0.18 h -1 , in Table 1) under the same reaction conditions (T = 25 ºC, C FA,0 = 1 M and τ = 33.3g CAT h L -1 ).Supporting this, there is a linear relationship (see Fig. 7) between the measured Pd 2+ /Pd 0 ratio at the end of the reaction (after the 25 h on time-on-stream, data collected in Table 2) and the estimated k d values (Table 1).As can be seen, the lower the Pd 2+ /Pd 0 ratio, the higher the k d value, and this proportional incremental change is valid for Pd 2+ /Pd 0 values above 0.37.Finally, along with the Pd 2+ /Pd 0 ratio, the S BET and A ext also decrease in each cycle (see Table 2) and particularly at 55 ºC.

Kinetic modelling
As demonstrated, the FA dehydrogenation reaction is not affected by FA mass transfer in the fixed bed reactor at flow rates higher than 0.25 mL min -1 (Fig. 3).To obtain the kinetic equation for the initial reaction rate, (− r FA ) w,0 , a first order reaction is assumed for FA concentration.This assumption is supported by our previous work, where hydrogen production with the Pd/AC catalyst was conducted in a batchwise slurry reactor without mass transfer limitations [17].As shown in Fig. 8, the FA disappearance rate can indeed be described by a first-order rate equation (Fig. 8a) with a pre-exponential kinetic factor and activation energy values of 6.13 10 4 L g CAT -1 h -1 and 36.7 kJ mol -1 , respectively (Fig. 8b).The apparent kinetic rate constant values at each temperature are provided in Table S2 of the Supporting Information.It is worth noting that this activation energy value is significantly lower than the one estimated in the slurry reactor with the same Pd/AC catalyst, which was 53.6 kJ mol -1 [17].This disparity suggests that FA disappears faster in continuous fixed bed than in batch-wise slurry reactors.This can be explained by considering a parallel reaction for FA disappearance, likely due to the retention of FA or HCOO -species on the Pd/AC, as before mentioned.Based on this, the following FA reaction route is proposed: where the corresponding reaction rates follow a first-order kinetic equation: According to this, the initial FA disappearance rate, in absence of deactivation, can be expressed as: and considering the Arrhenius equation, as: ) ) C FA (16) where the values of k 1 ʹ ,0 (which is k 1 • C CAT in L g CAT -1 h -1 ) and Ea 1 (in kJ mol -1 ) correspond to those estimated in the slurry reactor for the FA dehydrogenation reaction [17], while the values of k 2 ʹ ,0 (which is ) and Ea 2 (in kJ mol -1 ) are estimated considering the apparent values of the global reaction (i.e.6.13 10 4 L g CAT -1 h -1 and 36.7 kJ mol -1 ) and k 1,0 and Ea 1 .Thus, the following kinetic model equation describes the FA reaction at initial condition in absence of deactivation: C FA (17) Regarding to the modelling of catalyst deactivation, the Pd/AC catalyst deactivates according to a first-order model (see the curves and fitting in Fig. 4c).The estimated k d values increase with the temperature at the given initial FA concentration of 1 M (Table 1), following the Arrhenius equation with the low activation energy value of 18.8 kJ mol - 1 (see Figure S7a of the Supporting Information).On the other hand, the influence of initial FA concentration on the Pd/AC catalyst at a given reaction temperature (T = 35 ºC) has been studied since higher concentrations of FA suggest the accumulation of adsorbed FA/HCOO - molecules on the active sites and faster accumulation of CO 2 .Fig. 9 provides X FA and activity profiles at three different initial FA concentration values.While, the initial X FA values are similar, they decrease C. Martin et al. faster with increasing initial FA concentration (see Fig. 9a).Fig. 9b displays the mathematical fit of the activity curves.Note that the initial FA concentration has a higher impact on deactivation compared to the reaction temperature (see Fig. 4c).This finding supports that the retention of FA/HCOO -species on the Pd active sites is the primary cause of catalyst deactivation, over the presence of reaction products.
The deactivation rate constant calculated at each initial concentration (Eq.9) follows a linear relationship with the initial FA concentration, see Figure S7b of the Supporting Information.Therefore, the deactivation rate constant can be expressed as: where k 3 (in L mol -1 h -1 ) is 0.32 at T=35ºC.According to above results, the FA dehydrogenation reaction rate can be expressed as: Thus, the following kinetic model equation for the FA dehydrogenation reaction over Pd/AC catalyst for CO-free hydrogen production is elucidated: C FA (20)

Conclusions
Liquid phase catalytic FA dehydrogenation is a promising way to produce CO-free hydrogen.The scalability of this process has been investigated in a fixed bed reactor using a commercial powdered Pd/AC catalyst.The continuous hydrogen production was conducted at various operating conditions (C FA,0 = 0.25-1 M, T = 25-55 ºC, W = 0.2-0.5 g), and the FA dehydrogenation reaction was kinetically modelling.The reaction was found not to be controlled the mass transfer limitations.
The elucidated kinetic equation is a first order with respect to FA concentration and accounts for two pathways of FA (or HCOO -species) disappearance.Firstly, FA decomposes into hydrogen and CO 2 with an activation energy of 53.6 kJ mol -1 .Secondly, FA is sorbed onto the Pd/ AC catalyst without reacting, with an activation energy of 36.7 kJ mol -1 .In addition, the catalyst activity exhibits exponential decay with both initial FA concentration and reaction temperature.This is attributed to deactivation caused by the accumulation of reaction species, including FA/HCOO -(reversibly sorbed) and CO 2 (irreversibly chemisorbed) on the Pd active sites.Deactivation occurs more rapidly during continuous operation compared to a batch-wise reactor, due to the intensified process employing a continuous flow of FA.The Pd/AC catalyst can be nearly fully regenerated by drying at 60 ºC during 24 h in an air atmosphere, though deactivation takes place faster with catalyst recycling, as the amount of Pd 2+ species (identified as the main active sites) gradually diminishes over time-on-stream.Along with the Pd 2+ /Pd 0 ratio, the S BET and A ext decrease upon time-on-stream, while the sintering of Pd nanoparticles is mainly influenced by the reaction temperature.All these factors contribute to a slow but irreversible deactivation of the Pd/AC catalyst.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Fig. 1 .
Fig. 1.Schematic diagram of the experimental setup for FA dehydrogenation.

C
.Martin et al.

Fig. 3 .
Fig. 3. Study of the influence of liquid flow rate on (a) FA conversion and (b) hydrogen flow rate produced.Operating conditions: C FA,0 = 1 M, W = 0.2 and 0.5 g, T = 35 ºC and Q He = 17 NmL min -1 .

Fig. 5 .
Fig. 5. (a) ATR spectra and (b) zoom in the window from 2100 to 2600 cm -1 of the 1st use Pd/AC catalysts in a fixed-bed reactor at different reaction temperatures.

Fig. 9 .
Fig. 9. (a) FA conversion and (b) catalytic activity upon time-on-stream at different initial FA concentrations.Operating conditions: W = 0.5 g, T = 35 ºC, τ = 33.3g CAT h L -1 and Q He = 17 NmL min -1 .Lines in Fig. 9b are the best fit to estimate the k d values.

Table 1
Initial FA conversions, apparent kinetic rate constants and deactivation rate constants of the FA dehydrogenation reaction at different temperatures.

Table 2
Pd particle size (d Pd ), atomic surface ratio Pd 2+ /Pd 0 , BET surface area (S BET ) , external surface area (A ext ) , and bulk C and H content of different Pd/AC catalyst.
C.Martin et al.