Hydrodeoxygenation of cracked vegetable oil using CoMo/Al2O3 and Pt/C catalysts

During the continuous catalytic hydrodeoxygenation (HDO) of cracked vegetable oil (CVO), CO2 and CO occur as the main reaction gases, in addition to hydrocarbon gases such as CH4 and C2H6. The catalysts used were cobalt-molybdenum (CoMo) on an Al2O3 support and platinum (Pt) on an active carbon. All named gas components can result directly from the decomposition of CVO. The results of batch experiments for gas phase reactions (GPRs) under the same 50 bar H2 atmosphere using the same catalysts (CoMo, Pt) indicate that CO and CH4 can also be formed by GPRs. CO can result from the reverse water–gas shift reaction (RWGS), and CH4 from CO- or CO2-methanation. The found CO-yields from GPRs are within the theoretical thermodynamic limits based on equilibrium. An unexpected inhomogeneity of the gas component concentrations in the reactor during batch investigations was observed despite the elevated temperature (380 °C) and high RPM (1100) due to the high density difference compared to H2, especially in the case of CO2.


Introduction
The rapid growth of world's population and rising desire for prosperity are driving the demand for energy and transportation fuels. According to the IEA scenarios [1,2], this goes along with increasing importance of biofuels. To overcome limitations in blending potential of biofuels, drop-in quality is favourable. One established pathway in this direction is the production of hydroprocessed vegetable oil [3].
A comparable approach is a two-step upgrading of vegetable oils via thermal deoxygenation (DO) in the first step producing CVO and subsequent hydroprocessing to obtain hydroprocessed CVO [4,5]. The CVO production step was investigated in detail before; during this step, a decrease of the vegetable oils' initial oxygen content (approximately 10 wt %) to 5 wt % was reached [4,6]. It was shown that the CVO's oxygen is mainly (87 wt %) bound in free fatty acids [5]. Batch studies on HDO of CVO gave the first indications of catalytic influences on oxygen and energy content of hydroprocessed CVO [5]. Semi-continuous studies were performed to approach suitable feed rates (CVO, H 2 ) for the continuous HDO of CVO [7]. Optimal reaction conditions for all investigations of this work were found to be 380°C, 50 bar H 2 pressure and a gas entrainment impeller rotation speed of 1100 RPM [5,7]. DO and HDO of fatty acids is widely discussed in literature [8][9][10][11][12][13][14][15]. HDO implies the direct participation of H 2 in a reaction, while DO in general does not necessarily require H 2 . According to the aforementioned literature, the main gaseous products from DO/HDO of fatty acids are CO 2 from decarboxylation and CO from decarbonylation reactions for both DO and HDO, while reaction water may result from at least three pathways: first, along with decarbonylation; second, from dehydration (during DO/ HDO) and third, via hydrogenation (reduction) in case of HDO. Further gaseous products may be short-chain hydrocarbons such as CH 4 and C 2 H 6 . One possibility for the formation of hydrocarbon gases is cracking [8,16]. CH 4 can also result from GPRs such as methanation of CO and CO 2 . Furthermore, CO can also result from CO 2 and H 2 via RWGS. These GPRs as shown in Eqs. (1-3) are reported by Gusmão et al. [12], Wagman et al. [17], Snåre et al. [18], Lestari et al. [19] and Madsen et al. [11].
The present work focuses on three main questions: Which are the main gaseous and liquid products occurring during HDO of CVO? Which interrelationship between gas and liquid-phase composition can be identified? Which GPRs can take place in parallel of HDO of CVO? Based on these questions, the objective of this work is to study the influence of two different catalysts: A typical hydrotreatment catalyst CoMo/Al 2 O 3 and Pt/C as a noble metal catalyst [20]. For the HDO investigations of CVO, continuous experiments are performed. GPRs are studied using a batch configuration of the setup to allow an overall mass balance of the experiments. Thermodynamics of GPRs are worked out to support the findings from the experimental work.

Thermodynamic section of GPRs
The thermodynamic section shall give some insights regarding Gibbs free energy and reaction equilibrium data for the considered GPRs. The main gaseous products observed during HDO of CVO are CO 2 , CO and CH 4 . According to the literature mentioned above, RWGS and methanation of CO and CO 2 are the expected GPRs under H 2 atmosphere (1-3).
CO þ 3H 2 $ T;p Equation (3) may be represented by the combination of Eqs. (1) and (2). Therefore, reactions (1, 2) will be mainly considered in the following. To compare experimental results with thermodynamic equilibrium, data for reactions (1, 2) were calculated for the applied reaction temperature. First, it was proved if the GPRs are exergonic or endergonic (Eq. 4). The calculation of reaction enthalpy and reaction entropy according to (5,6) is carried out with data from Basu [21] and National Institute of Standards and Technology [22].
where DG R Gibbs free energy (kJ/mol), DH R reaction enthalpy (kJ/mol), T temperature (K), DS R reaction entropy (kJ/mol K), DH R ðT 0 Þ reaction enthalpy at T 0 (1 bar) using data from National Institute of Standards It was found that the RWGS is endergonic in the considered temperature range, whereas CO-methanation is exergonic, as can be seen from Gibbs free energy for each reaction. Detailed results are shown in Table 1.
Due to these thermodynamic results, a low equilibrium constant (K p,(1) ) is expected for RWGS, while equilibrium constant for CO-methanation should be high. The thermodynamic equilibrium constant for the RWGS reaction (1) and CO-methanation (2) are calculated according to Eq. (7) based on the data from Elvers [23].

Methods
The experiments were all performed using an autoclave (type 4576A, Parr Instruments, Moline, IL, USA) at 380°C, 50 bar H 2 pressure and with the gas entrainment impeller running at 1100 RPM. In previous investigations, these process conditions were found to be optimal for the catalytic HDO of CVO [5,7]. The piping and instrumentation diagram of the test plant is presented in Fig. 1.
11 experiments were performed in total ( Table 2). The experiments included the following: For the continuous catalytic HDO of CVO, H 2 and CVO were fed continuously into the process. The operation mode of the pilot plant is termed reactive stripping and was derived from the reactive distillation mode in the production process of CVO [4]. The reactor was mounted, rendered inert and finally CVO was provided in the reactor to give a filling level of two-thirds at the reaction temperature relative to the overall reactor volume. H 2 acts as a stripping and hydrogenation gas. The H 2 flow was set to 2.5 g/h by means of a mass flow counter (type M 12202139A, Bronkhorst High-Tech B.V., Ruurlo, Netherlands). Comparable investigations used a similar flow for H 2 [13,30]. The liquid feed rate was 3.5 g/h and was chosen according to the preliminary investigations in semi-continuous mode [7]. Condensate samples were released continuously every 30 min at the beginning and then every 60 min after approximately 3 h time on stream (condensate fraction). The sump fraction was collected in the reactor. The catalytic HDO of CVO was run continuously investigating time on stream between 7 and 12 h. The preparation of all continuous catalytic HDO experiments was identical to assure reproducibility.
Continuous discharge of gas (CO and CO 2, CO x in the following) was calculated according to Eq. (8) based on measurement data of the nondispersive infrared photometer.
Volume fractions (v j þ v jþ1 ) measured by a nondispersive infrared photometer in each time interval (t jþ1 À t j Þ, the total gas volume flow ( _ V ges;Dt;j ) and density of CO x (q CO x ;Dt;j ) were required parameters for the calculation.
The net molar production of the respective alkanes was taken from GC/MS/FID analysis. The selection for alkanes (C 8 , C 9 , C 10 , C 15, C 16 , C 17 and C 18 ) is related to chain length of the respective main fatty acids (C 9 , C 10 , C 16 , C 18 ) in the CVO. These four selected fatty acids were found in concentrations C 1.5 wt % in CVO and decomposed totally during HDO. The decomposition of these fatty acids can result in unchanged chain length alkanes, so called C n -alkanes (C 9 , C 10 , C 16 , C 18 ) via hydrogenation (reduction). The second possibility of decomposition is the chain length reduction by one carbon, resulting in the so-called C n-1 -alkanes (C 8 , C 9 , C 15 and C 17 ) via decarbonylation and decarboxylation. The procedure of using molar product ratios for the characterisation of possible reactions was used in other studies before [13,[31][32][33].
For the continuous gas residence time studies, the same constant flow of H 2 was used as in continuous HDO experiments for the purpose of comparison (2.5 g/h). CO 2 was injected for pulsed tagging, once the reaction temperature was reached by means of a Teledyne Isco highpressure pump (type 500 HP, Lincoln, NE, USA) via the liquid feed inlet. Two setups were investigated: one without liquid filling the reactor and one with filling the reactor with an inert liquid medium (BP Energol CS-HB 220). The liquid level of the reactor was (according to the experiments for the catalytic HDO of CVO) two-thirds at the reaction temperature relative to the overall reactor volume. Regarding the gas residence time, the CO 2 pulse passed the test plant ( Fig. 1) in the same way as the reaction gases formed in an experiment for the catalytic HDO of CVO Fig. 1 Piping and instrumentation diagram of the test plant. Acronyms according to [29]: A alarm, C control, F flow, I indication, P pressure, R record, S speed, T temperature Batch investigation of GPRs (Pt catalyst) 10, 11 Batch investigation of inhomogeneity effect (experiment 1-5) as follows: it was injected in the reactor and passed the condenser and the settling vessel of the liquid samples before the gas concentration was measured by the nondispersive infrared photometer. For both experiments on the gas residence time, approximately the same amount of CO 2 was injected. The mean gas residence time was calculated considering CO 2 concentration over time using linear interpolation between the measurement points. The gas residence time investigations were performed without catalysts to avoid GPRs. For the batch investigations of the catalyst's selectivity for GPRs, the reactor was operated in batch mode. The procedure of preparation was the same as for a continuous HDO experiment, but without liquid filling. The catalyst amount was the same as for an experiment for the catalytic HDO of CVO (0.91 g for CoMo/Al 2 O 3, 0.48 g for Pt/C, Table 3). After filling with H 2 , the reactor was heated to the reaction temperature (380°C). The first sample of each experiment was taken before the CO 2 injection to obtain a comparison value. Afterwards, the reactor pressure was adjusted to 50 bar H 2 , and CO 2 was injected for pulsed tagging. The gas entrainment impeller was run at 1100 RPM. The experiments with both the catalysts were performed with similar initial concentration of CO 2 to ensure the comparability of the results. Gas sampling intervals were 5 min after injection. Each sample was drawn directly from the reactor after cooling and expansion to ambient conditions. Six samples could be taken from the overall reactor volume. Samples were then prepared for the GC/ FID analysis.
The inhomogeneity of the gas phase in the reactor was investigated in batch mode considering two different sampling locations, using the CO 2 concentration as an indicator. The CO 2 injection was performed once the reactor reached 380°C. Sampling was performed approximately 5 min after the CO 2 injection. The first sampling location was on the top of the reactor (approximately 86 mm from the bottom), and the second sample was drawn via the dip tube (approximately 25 mm from the bottom). Both samples were taken simultaneously.

Liquid samples
For the liquid sample analysis, a G1530A GC (Agilent Technologies, Santa Clara, CA, USA) coupled with a quadrupole MS (5972A HP 6890, Hewlett Packard, Palo Alto, CA, USA) and FID was used. The database of the National Institute of Standard and Technology was used for the component identification. A VF5-ms capillary column from Agilent Technologies was installed (60 m 9 0.25 mm 9 0.25 lm). Perylene (C 20 H 12 ) and used as an internal standard to avoid overlapping with component peaks. The split ratio was set to 15:1. The initial temperature of the column was 45°C (holding time 4 min), followed by a heating ramp of 3°C/min until 320°C was reached. The final column temperature was held for 20 min. The GC was run in a constant flow mode with helium as the carrier gas at 2 mL/min. The average flow velocity was 31 cm/s. The FID was run at 350°C. Nitrogen was used as the makeup gas with a flow of 45 mL/min; the synthetic air flow was 450 mL/min, and the H 2 flow was 40 mL/min. MS and FID detector was used in tandem for best identification and quantification of the components. The process of sample preparation was reproducible. Linear fitted curves were used for the calculation of the quantity of each component in agreement with the procedure applied by Artigues et al. [34].

Continuous gas composition
A calibrated Infralyt 50 nondispersive infrared photometer (Saxon Corporation, Dessau-Roßlau, Germany) was used for the continuous gas analysis. The nondispersive infrared

Batch gas samples
The compositions of the batch gas samples were quantified with GC/FID on an Agilent Technologies 6890 N (G1530 N) machine equipped with a CoraPlot Q capillary column (CP7554, 25 m length, 0.53 mm inner diameter and 20 lm film thickness). The split was set to 10:1. The carrier gas was helium, with a flow of 20 mL/min. The initial oven temperature was 50°C. The first heating ramp was 20°C/min until 95°C was reached and held for 4 min. The final temperature was 100°C, reached with a second heating ramp of 12°C/min. The FID was run at 220°C. The H 2 flow was 40 mL/min, the air flow was 400 mL/min and the nitrogen makeup flow was 12 mL/min. The components were assigned according to their retention time similarity. Calibration runs with different test gases were performed as repeat determinations. Linear fitted curves were used to calculate the quantity of each component. The gas samples were additionally analysed on a calibrated micro gas chromatograph (GCM Micro Box I, SLS Micro Technology Corporation, Mainz-Kastel, Germany) to confirm the gas composition results. A HayeSepA packed column was used for the measurement by means of a thermal conductivity detector. Argon was used as a carrier gas at a pressure of 3.5 bar.

Results and discussion
Results and discussion are given in two main chapters in the following order: Continuous processing • Liquid-phase DO via catalytic HDO of CVO for upgrading raw material CVO and to identify possible reactions. 11 experiments were performed in total. An overview of these experiments is given in Table 2.

Continuous processing
The results of the HDO experiments as well as the gas residence time behaviour of the test plant are given in this chapter.
Liquid-phase DO via Catalytic HDO of CVO Table 3 gives the overall mass balances and oxygen content of CVO and main products (sump fraction) for five experiments of the continuous catalytic HDO of CVO with 3.5 g/h feed and 2.5 g/h H 2 mass flow. The rotational speed of the gas entrainment impeller was 1100 RPM.
As can be seen from Table 3, sump fraction is the main liquid product. This is due to the initial filling of the reactor with CVO. The highest DO of the sump fraction was observed with 12 h time on stream in experiment 5. The lowest oxygen concentration is confirmed by the highest mass of m CO CO was detected slightly before CO 2 in every experiment. As the decrease in CO concentration in experiment 5 from 0.5 to 0.4 vol % took place very slowly (approximately within 5 h), it may be expected that the process was close to steady state at point C. It can be seen in Figs. 2 and 3 that the main DO of the liquid phase takes place rapidly at the beginning of the experiment due to steep inclination of the gas concentrations. Main product's (sump fraction) oxygen content confirms high DO after time on stream with both the catalysts (Table 3). Possibly the CO and CO 2 concentrations result from the DO/HDO reactions decarboxylation and decarbonylation. Considering this assumption, it can be seen that the Pt catalyst was more selective for these DO/HDO reactions compared to the CoMo catalyst. This is in agreement with literature, as can be seen in Table 4.
A dec/hyd ratio of 1 indicates the presence of both decarbonylation/decarboxylation and hydrogenation to the same extent. This is the case for both experiments with the CoMo catalyst (experiment 1 and 2). The same dec/ hyd ratios of these experiments confirm reproducibility. Pt catalysts give clearly higher dec/hyd ratios than CoMo catalysts using vegetable oils as feedstocks. This effect is still visible in the HDO of CVO in this work. The lower dec/hyd ratio for CVO compared to the other results with Pt catalyst is due to the strongly reduced oxygen content of the CVO [36,37]. This means, that the main part of decarbonylation/decarboxylation already occurred in the previous CVO production step [4]. In general, higher CO x -yields should go along with increasing dec/hyd ratios. This is confirmed by the data of this work given in Table 4.
Due to the higher concentration of CO compared to CO 2 , both catalysts seem to be more selective for decarbonylation over decarboxylation. However, it is possible that the CO 2 is continuously decomposed to CO via GPRs during the HDO of CVO as well. CH 4 and C 2 H 6 were measured only batch-wise every 30 min by micro gas chromatography. The maximum values for the CoMo catalyst (experiment 1) were 0.13 vol % for CH 4 and C 2 H 6 , and for the Pt catalyst (experiment 3) 0.16 vol % for CH 4 and 0.30 vol % for C 2 H 6 were reached.
Due to the gas phase composition observed (mainly H 2 , CO and CO 2 ), GPRs may occur (Eqs. 1-3). The gas concentrations (Figs. 2, 3) are compared to gas residence time behaviour of the plant to identify reaction time of main DO/HDO.

Gas residence time behaviour of the test plant
The gas residence time behaviour of the plant was investigated with and without filling the reactor with an inert liquid medium. Two experiments (6, 7) were performed. CO 2 pulsed tagging was applied for these investigations. Figure 4 shows the results of the CO 2 concentration plotted over time measured via a nondispersive infrared photometer.
Mean gas residence time for experiment 6 (B) and for experiment 7 (C) is shown. Downtime (A) of both the experiments was 26 min. As can be seen in Fig. 4, the downtime of both experiments was 26 min. It is defined as the time period between the pulse injection and the first detection of the gas by the nondispersive infrared photometer. Taking the downtime into account, the starting time of the DO during the catalytic HDO of CVO  [20]. Partly HDO of rapeseed oil with CoMO catalyst at 200°C was reported from Pinto et al. [38]. HDO processes in this temperature range (B250°C) are often called mild HDO [20,39]. Experiment 7, with an inert liquid filling, showed a higher maximum CO 2 concentration but the same downtime as experiment 6, without liquid filling. This may be due to both a higher concentration of CO 2 , caused by the smaller gas volume in the reactor due to the liquid filling and the higher amount of injected CO 2 , caused by the limited setting accuracy of the high-pressure pump. According to Fig. 4, the injected quantity of CO 2 and/or the filling of the reactor with an inert liquid media influenced the mean gas residence time slightly. Mean gas residence time was evaluated by linear interpolation between the measurement points. The mean gas residence time was 42 min for experiment 6 and 45 min for experiment 7.

Characterisation of gas phase reactions
GPRs (1-3) may be responsible for the rapidly decreasing CO 2 concentration in the gas phase (Figs. 2, 3). For this reason, the selectivity of these chemical reactions was investigated with both the catalysts in experiments 8 and 9. After the CO 2 injection, six gas samples were taken from the reactor in each experiment (Fig. 1). The development of the reactor pressure during sampling in experiments 8 and 9 is shown in Fig. 5.
As can be seen in Fig. 5, the reactor pressure was constant between each gas sampling in the experiment with the CoMo catalyst (experiment 8), but it increased in the experiment with the Pt catalyst. It is assumed that this finding is due to the adsorption/desorption effects on the support of the catalyst (active carbon). Each stage in the diagram indicates a gas sample. Gas samples were removed until the reactor was evacuated. The adsorption/desorption of H 2 on Pt catalyst on active carbon (5 wt % Pt, BET = 650 m 2 /g) was observed as well by Wang and Yang [40] (6 wt % Pt, BET = 3126 m 2 /g). These authors showed a higher adsorption capacity of metal-doped carbon than that of plain carbon. At 50 bar and 298 K, the amount of H 2 adsorbed on Pt/carbon was found to be approximately 0.75 wt %, related to the mass of the catalyst [40]. The adsorbed amount depends on the reaction time, temperature and BET surface area. Results from GC/FID analysis of the gas sample composition in experiments 8 and 9 are shown in Figs. 6 and 7. Experiments 8 and 9 showed a remarkable CO 2 gas phase conversion due to CO and CH 4 generation. CO 2 conversion was obviously supported to a higher extend with the Pt catalyst compared to the CoMo catalyst. The detection of CO was unexpected, as the equilibrium constant for the CO-methanation reaction (2) according to Eq. (7) is 6 9 10 4 bar -2 , indicating the complete reaction of CO to CH 4 . A possible explanation is the kinetic inhibition of the CO-methanation (2). A further unexpected observation in Fig. 7 is the initial concentration of CH 4 above zero (0.22 vol %) prior to the injection of CO 2 . A reason could be that the CH 4 was generated by Pt supported reaction of the active carbon carrier material and the H 2 atmosphere during the heating phase (approximately 50 min).
Based on the results shown in Figs. 6 and 7, the quantity of each component was calculated according to ideal gas law, and the results are summarised in Table 5.
Higher hydrocarbons were not considered due to their very low concentrations (B0.11 vol %) during the entire experiments. The total carbon discharge in the collected gas samples was used to calculate the initial concentration of the CO 2 in the reactor, as given in Table 5 and marked  Fig. 6. In the two experiments, the injected quantity of CO 2 was nearly the same. The experimental procedure was reproducible and the initial CO 2 concentrations were comparable in both experiments (8 and 9).
With these experimental data and the equilibrium constant calculated before the theoretical maximum partial pressure of CO (p CO ) at RWGS's equilibrium without subsequent methanation was calculated according to Eq. (10).
where K p;ð1Þ equilibrium constant for RWGS (1), p CO equilibrium partial pressure CO (bar), p H 2 O equilibrium partial pressure H 2 O (bar), p CO 2 equilibrium partial pressure CO 2 (bar), p H 2 equilibrium partial pressure H 2 (bar), p 0;CO 2 initial partial pressure CO 2 (bar). Disregarding subsequent reactions of CO, partial pressures for equilibrium of RWGS were defined as follows: The initial partial pressure of CO 2 p 0;CO 2 À Á in Eq. (10) was calculated from the total initial reactor pressure multiplied by the experimental initial CO 2 concentration, as given in Table 5. The equilibrium partial pressure of H 2 p H 2 ð Þ was set equal to the difference between the measured initial reactor pressure after injection of CO 2 and the calculated p 0;CO 2 by approximation. Based on these preconditions, the equilibrium partial pressure of CO (p CO ) was calculated from Eq. (10). Results are given in Table 6.
The results from Table 6 show that p CO 2 must be lower at RWGS's equilibrium compared to p CO . Based on the   Table 7.
The experimental CO-yields of experiments 8 and 9 according to Table 6 are approximately 50 % below the theoretical maximum yields as shown in Table 7 and therefore within the expected theoretical limits. Presuming RWGS to occur, the formation of H 2 O occurs in the same quantity compared to CO. The calculation of the H 2 Oyields based on the experimental CO-yields for experiment 8 and 9 gave the following results: These low mass yields of H 2 O could not be detected. They probably disappeared by condensation in the gas samples. As expected, the theoretical maximum yields of H 2 O were found to be higher than the calculated yields based on measurements. This can be seen when comparing the values from the experiments (8,9) to the values given in Table 7.

Inhomogeneity effect
During the experimental investigations of the GPRs, an inhomogeneity of the gas phase was observed in the reactor. The first indicators could be detected in experiments 8 and 9, and the phenomenon is proven in experiments 10 and 11. This inhomogeneity effect made further kinetic investigations of GPRs impossible. However, the balance of the detected gas components could be analysed ( Table 5).
The steady increase of CO and CO 2 shown in Fig. 6 gives the first indication of this inhomogeneity, which was caused due to density differences of the gases compared to H 2 (CO and CO 2 by factors 22 and 14, respectively). In the first gas sample, H 2 was predominantly removed, but in the last sample, the CO and CO 2 could expand more in the reactor, thus gradually raising the local concentration at the measuring point for each sample. The indication of inhomogeneity in experiments 8 and 9 can be derived from the fact that the measured CO 2 concentrations in the gas samples exceed the initial CO 2 concentration at the end of each investigation (Fig. 6), despite the CO 2 decomposition. Remarkably, this inhomogeneity effect occurred even though the gas entrainment impeller was run at 1100 RPM and the reactor temperature was 380°C.
The measuring points for the gas samples are indicated in Fig. 8 in a simplified true-scale drawing (experiments 8 and 9: sample dip tube, experiments 10 and 11: sample dip tube and on the top of the reactor).
For the practical validation of the inhomogeneity effect, experiments 10 and 11 were carried out without catalyst and liquid filling in the reactor, considering the two different measuring points in the reactor. The CO 2 concentrations of the samples were analysed via micro gas chromatography, and the results are given in Fig. 9.
Sampling was performed 5 min after the CO 2 injection in both experiments 10 and 11 (sample 1) as before (in experiment 8 and 9). Sample 2 in experiment 11 was drawn immediately after sample 1. The quantitative reproducibility is limited (comparing experiment 10 and experiment 11, sample 1), but the inhomogeneity effect is obvious in all samples: The CO 2 concentration is higher at the dip tube measuring point, close to the bottom of the reactor, despite simultaneous sampling. The increase in the CO 2 concentration with the decreasing reactor pressure (comparing experiment 11, sample 1 and sample 2) is due to the expansion of the heavy gas phase.

Conclusions
Continuous experiments for the HDO of CVO were performed using an autoclave in a reactive stripping process. Used catalysts (Pt, CoMo) were suitable for this process due to high DO of the main product (sump fraction, Table 3). The main reaction gases were found to be CO 2 and CO, while the latter showed a higher concentration. Dec/hyd ratios were determined based on liquid phase composition. These calculated ratios correlate with the CO x -yields calculated from the gas phase composition. The selectivity for DO/HDO via decarboxylation/decarbonylation is indicated by the dec/hyd ratio and was higher with the Pt catalyst compared to the CoMo catalyst.
To investigate whether the RWGS reaction (1) occurs, batch experiments for GPRs (CO 2 ? H 2 ) were performed without CVO, using the same catalysts as in the continuous HDO experiments (CoMo, Pt). In these batch experiments, CO was found to be a product gas, indicating that the RWGS reaction occurs under the applied experimental conditions (380°C, ca. 50 bar H 2 and ca. 2 bar CO 2 ). Additionally, CH 4 was found, indicating that methanation (2-3) occurs as well.
Thermodynamic data showed that RWGS is endergonic and CO-methanation is strongly exergonic under the applied conditions. Due to that, low RWGS yields and high CO-methanation yields are expected. Nevertheless, calculated RWGS yields based on equilibrium were found to be still significant due to H 2 excess. This was confirmed by GPR investigations with both catalysts. The experimental determined CO-yields were found to be within the theoretical limit. However, COmethanation yields were found to be much lower than the expected according to low CH 4 -concentrations from GPR investigations. This indicates kinetic inhibition. Further kinetic evaluations of RWGS and CO-methanation reactions (1-2) were not possible due to an unexpected inhomogeneity of the gas component concentrations in the reactor.
The enrichment in CO 2 and CO at the bottom of the reactor was clearly caused by the high density differences of the gas components CO 2 , CO and CH 4 compared to H 2 (factors 22, 14 and 8, respectively), despite the elevated temperature and high RPM (380°C, 1100 RPM).
Such inhomogeneity found in a common type reactor widely applied in research may cause similar problems in comparable investigations of other research groups. Nevertheless, this issue was not observed in the considered literature. CFD tools are to be applied as a tool for further clarification of this issue and to allow the confirmation of the unexpected experimental results shown in the present investigation.
Acknowledgments The present work is supported by a scholarship of the Hanns-Seidel-Stiftung by funds from the Federal Ministry of Education and Research, Germany. GC/MS/FID analyses were performed at the Thünen Institute of Wood Technology and Wood Biology, Hamburg, Germany.
Author contributions The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript and contributed equally.

Compliance with ethical standards
Conflict of interest The authors declare that they have no conflict of interest.
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://crea tivecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.