Catalytic gasification of glycerol in supercritical water

The conversion of glycerol in supercritical water (SCW) was studied at 510 – 550°C and a pressure of 350 bars using both a bed of inert and non-porous ZrO2 particles (hydrothermal experiments), and a bed of 1 % Ru/ZrO2 catalyst particles. Experiments were conducted with a glycerol concentration of 5 wt% in a continuous isothermal fixed-bed reactor at a residence time between 2 and 10 s. Hydrothermolysis of glycerol formed water-soluble products such as acetaldehyde, acetic acid, hydroxyacetone and acrolein, and also gases like H2, CO and CO2. The catalyst enhanced the formation of acetic acid, inhibited the formation of acrolein, and promoted the gasification of the glycerol decomposition products. Hydrogen and carbon oxides were the main gases produced in the catalytic experiments, with only minor amounts of methane and ethylene. Complete glycerol conversion was achieved at a residence time of 8.5 s at 510 °C, and at around 5 s at 550 °C with a 1 wt% Ru/ZrO2 catalyst. The catalyst was not active enough to achieve complete gasification, since high yields of primary products like acetic acid and acetaldehyde were still present. Carbon balances were between 80 and 60 % in the catalytic experiments, decreasing continuously as the residence time was increased. This was attributed partially to the formation of methanol and acetaldehyde, which were not recovered and analyzed efficiently in our set-up, but also to the formation of carbon deposits. Carbon deposition was not observed on the catalyst particles but on the surface of the inert zirconia particles, especially at high residence time. This was related to the higher concentration of acetic acid and other acidic species in the catalytic experiments, which may polymerize to form tar-like carbon precursors. Because of carbon deposition, hydrogen yields were significantly lower than expected; for instance at 550 oC the hydrogen yield potential was only 50 % of the stoichiometric value.


Introduction
Glycerol is a polyalcohol with several commercial applications in food and cosmetics that may be synthesized chemically from epichlorohydrin derived from propylene [1], or biochemically by microbial fermentation [2]. Nowadays, however, massive amounts of low-purity glycerol are being obtained as byproduct in the manufacture of fatty acids, and mostly in biodiesel production where glycerol represents around 10 wt% of the plant product [3]. The surplus amount generated from biodiesel is so largemore than 600 000 tones in Europe in 2006 [4] that glycerol market price has decreased to less that 0.05 €/kg [5,6]. Due to its low purity, crude glycerol from biodiesel cannot be used in cosmetics or food unless a costly refining process is undertaken, and that is why it is usually considered a refuse product. Still, its wide availability and cheap price offer new opportunities for chemistry and energy [3].
Conventional gas-phase catalytic processes have been studied for the conversion of glycerol into acrolein [7,8] and a wide variety of glycerol-derived chemicals, as reviewed recently by Zhou and co-workers in a comprehensive paper [9]. Hydrothermal chemistry in sub-or supercritical water (SCW), either non-catalytic or with the addition of homogeneous and heterogeneous catalysts, has received considerable attention in recent years. Non-catalytic hydrothermal processing yields a variety of aldehydes, alcohols and gases [10], but adequate catalysts and control of reaction conditions maximize the yield of specific products like acrolein [11,12].
Gasification of glycerol for the production of hydrogen and syngas is the main application that has been explored for energy purposes. Hydrogen, methane and syngas have been produced by hydrothermal gasification of biomass substrates and model compounds such as glucose, cellulose and lignin in supercritical water [13][14][15][16][17][18], and glycerol gasification has been investigated as well [10,19]. Supercritical water has particular properties that provide a highly reactive and homogeneous medium for the conversion of organic molecules. SCW is miscible with organic compounds and gases due to its low dielectric constant and weaker hydrogen bonds than liquid water, but simultaneously facilitates the occurrence of ionic chemistry due to its relatively high ion product. Mass transfer limitations and coke formation on catalyst surfaces are also reduced because of a low viscosity and high diffusivity [20]. Two approaches have been undertaken for the SCW gasification of glycerol: high-temperature SCW gasification at reaction temperatures ranging from 550 to 800 °Cwith the occasional addition of non-metallic catalyst [21] and low-temperature catalytic SCW gasification using a metal-based catalyst, usually below 550 °C.
Non-catalytic glycerol decomposition in SCW proceeds through a complex reaction mechanism that is summarized in two competing pathways, ionic and free-radical, whose predominance depends on water density and acidity [10,22]. At low temperature and high pressure (i.e., high water density) a set of ionic reactions forms acetaldehyde, formaldehyde and acrolein as main products. At low water density a free-radical pathway leads to the preferential formation of allyl alcohol and methanol. Gases are typical products of the free-radical reactions, and their yield is favored by low water density. Most studies on hydrogen production through non-catalytic processing of glycerol, glucose and other biomass model compounds have been conducted at high-temperature since complete gasification is achieved only above 700 °C. However, hydrogen is not favored as the main product and even if gasification efficiency is high with diluted glycerol solutions, it decreases dramatically when the substrate concentration is above 20 wt% because of the dominance of condensation reactions between glycerol decomposition products that lead to the formation of tarry materials and char [23,24].
Solid catalysts have been studied to promote gas formation at lower temperature, maximize hydrogen selectivity and decrease the formation of tars and char [23]. Carbon was tested as catalyst on the SCW gasification of several organic feedstocks [24], but the gas yield was still low even at 600 °C. Supported-metal catalysts are the most appropriate for SCW gasification and, among them, those based on noble metals like Ru, Rh or Pt. They do not oxidize at SCW conditions and remain active for long periods of time. Support materials should be stable in the harsh SCW environment. Carbon, zirconia (ZrO 2 ), titania (TiO 2 ) and α-alumina (α-Al 2 O 3 ) have been usually employed [14]. Ruthenium-based catalysts have been shown to offer the best results for SCW gasification. Complete gasification of lignin and cellulose was achieved to produce methane as the main reaction product [25]. Hydrogen production from cellulose and sawdust was investigated with Ru/C, Pd/C, CeO 2 , nano-CeO 2 and (CeZr) x O 2 catalysts [26].
Ru/C was the most active catalyst. Glycerol was gasified completely and the stoichiometric yield of hydrogen (7mol H 2 /mol glycerol) achieved with a Ru/Al 2 O 3 catalyst, although a high reaction temperature (700 -800 °C) was needed [27]. In fact, no catalyst capable of reaching complete conversion of glycerol and a hydrogen yield close to the stoichiometric value at a temperature below 550 °C has been found yet, partially because of the formation of methane as competing final product.
In this paper we focused on the SCW gasification of glycerol over a Ru/ZrO 2 catalyst in the low temperature range. Experiments at incomplete conversion of glycerol were performed both with and without a Ru/ZrO 2 catalyst, and the reaction products were determined and quantified at increasing residence time. The influence of the catalyst on the pathways of glycerol conversion is discussed on the basis of the data gathered.

Materials
Glycerol (99.5 %, Sharlau) was dissolved in Mili-Q water at the adequate weight ratio and the solution was filtered (0.2 μm, Whatman) and degassed under vacuum before use. The solution was stored at 5 °C and used in less than 48 h to prevent the growth of microorganisms.

Preparation and characterization of the Ru/ZrO 2 catalyst
The catalyst was prepared by incipient wetness impregnation, following literature procedures [28][29][30][31][32]. The support was impregnated with the adequate amount of the precursor aqueous solution and dried at 110 °C for 24 h. Then the sample was crushed and oxidized at 500 °C for 4 h, reduced with hydrogen at 400 °C for 5 h, cooled to room temperature under nitrogen and stored in caped vials until use. The catalyst was characterized by x-ray diffraction in a Bruker-AXS D8-Discover diffractometer. The angular 2θ diffraction range was between 5 and 70° and the data was collected with an angular step of 0.05° at 3 s per step. Surface area analysis was performed by nitrogen adsorption in a Micromeritics ASAP 2020. SEM analysis was performed on a scanning electronic microscope JEOL model JSM-6400. The samples were covered with gold before analyzing. Analysis was at 20 kV, in high vacuum and at 16 mm distance between lens and sample. ESEM analysis of variable pressure was joined with a XR microanalysis in order to determine and quantify the elements with a current of 20 kV, spot 4 (electrons ray) and at 10 mm distance between lens and sample.

SCW reactor setup
The SCW gasification experiments were performed in a tubular fixed bed reactor ( Figure 1). The feed stocks, water and glycerol solution (V-101, V-102), were pumped through the system using two HPLC pumps (GILSON model 307), (P-101 and P-102) at a flow rate between 0.2 and 5 mL/min. The ratio of glycerol solution (10 -25 wt%) to water flow rate was varied from 1:2 to 1:4. The mass flow rates fed to the reactor were calculated from the variation of mass of the liquid bottles by means of two scales (Scalter SBA 52, max. weight 2200 g ± 0.01 g). The signals from the scales, and other signals of the system like temperatures, were recorded in a computer with a data acquisition system (Compact Field-Point and LabView software, National Instruments). Water was preheated (HC-101) before entering the reactor. Glycerol at room temperature was injected into the preheated water stream just at the top of the catalyst bed to minimize degradation during preheating. The tubular reactor (R-101) had a diameter of 0.635 cm and a length of 8.68 cm (total volume 2.75 mL). It was packed with 1-3 g of catalyst previously diluted with 3-5 g of inert particles (zirconia support) to improve heat transfer and decrease temperature gradients. The particles were supported onto a stainless steel frit (10 μm, VICI) and a thermocouple (TI-103) was inserted to record the temperature of the catalyst bed.
The reactor and the water preheater were mounted inside an electric furnace. After leaving the reactor, the fluid was cooled in a heat exchanger (HC-102) refrigerated with ethylene glycol at 2 °C. The fluid passed then through a metal filter (10 μm, F-101) to retain possible particles entrained from the bed or formed during reaction. Pressure was maintained with a mechanical backpressure regulator (P-101), placed after the filter. After the expansion through the backpressure regulator, the gas and the liquid were separated in a separator (S-101) that was refrigerated by ethylene glycol at 2 °C to minimize the loss of volatile compounds through the gas stream. The gas stream was analyzed on-line with a microGC (Agilent 3000). The liquid stream flowed into a flask (V-103) that rested on a scale (Acculab Atilon 2202-I), used to determine its flow rate. Samples of the outlet liquid stream were taken for HPLC and pH analysis.

Analytical methods
The composition of the gas was determined by on-line gas chromatography (Agilent 3000A Micro GC) every four minutes. The molar flow rates of the gas products were calculated based on the molar gas flow of nitrogen that was added to the gas-liquid separator as an internal

Hydrothermal treatment of Glycerol in SCW
Residence time of the reacting mixture inside the reactor, , was estimated with equation 6, assuming that the density of the reacting mixture approached that of pure water at the temperature and pressure of the experiment. This is a coarse approximation in our case since glycerol concentration was relatively high (5 wt% in the feed), and therefore density of the reaction mixture should be calculated accounting for the change of composition due to the conversion of glycerol and the formation of products. However, this calculation is not straightforward at supercritical conditions and we used this simplified procedure to have an approximate estimation of the residence time. In equation 6, V R is the volume of the catalyst bed,  is the void fraction of the bed (which is 1 for the empty-tube experiments), m 0 the mass flow rate of water and glycerol fed to the reactor, and  m,R is the density of pure water at the reactor conditions of P and T. The latter was calculated according to Wagner and Kruse [33].
Initial experiments showed that the zirconia particles could be considered chemically inert since there was no significant difference in conversion and yield of products between the experiments performed with the empty tube and those with the zirconia particles. In consequence, all the experiments were conducted using a packed bed of ZrO 2 particles, since the hydrodynamics of the system was then equivalent to that of the catalyzed experiments. Product yields were expressed as mole of product formed per mole of glycerol fed to the reactor, according to equation 6. Hydrothermal processing at 510 °C on a bed of inert ZrO 2 particles gave a conversion of glycerol of around 22 % after 8.5 s. The main reaction products were acetaldehyde, hydroxyacetone and acetic acid, together with minor amounts of allyl alcohol, propionaldehyde and acrolein. Gaseous products comprised hydrogen, carbon oxides and methane. Trace amounts of acrylic acid and ethylene were also detected. This wide variety of products reflects the complexity of the reaction mechanisms involved in the hydrothermal decomposition of glycerol, which can be summarized into the coexistence of competing ionic and free radical pathways [10]. The ionic path dominates at high-water densitylow temperature and high pressurea situation in which the ion product of water is high enough to allow the existence of free protons and hydroxyl ions that may catalyze ionic reactions, and the static dielectric constant of water is high enough to stabilize ionic species and reaction intermediates. The free radical decomposition route dominates at low water densityhigh temperature and low pressurewhere ionic chemistry is not favored due to the low dielectric constant of the media and low water ion product. Table I shows relevant properties for the conditions we have covered in our study, which were calculated according to Wagner and Kruse [33]. At 510 °C and 350 bar the ion product of water is six orders of magnitude lower than for water at room temperature and the dielectric constant is close to that of steam, which indicates that the ionic route will play a minor role in our case and that the free-radical pathway should dominate. The latter is summarized in Figure 6, where a simplified schematic of a more rigorous mechanism [10] is given. Hydrogen transfer reactions form unstable •CH 2 CHOHCH 2 OH (Int1) and •COHCHOHCH 2 OH (Int2) radicals. The first evolves into propionaldehyde and allyl alcohol, whereas the second is converted into acrolein, acrylic acid and formaldehyde. Radical isomerization also forms •COH(CH 2 OH) 2 radicals (Int3) from Int2, which then evolve into acetaldehyde and acetic acid. Formaldehyde is only an intermediate that leads rapidly to the formation of carbon monoxide, which is transformed to carbon dioxide by the water-gas shift reaction. Methane is formed from CO and CO 2 as a result of methanation reactions. The carbon balance was between 90 to 100 % and most of reaction products were identified, which was supported by values of HYP between 6 and 7 mol H 2 /mol glycerol. The carbon unaccounted for was attributed to the formation of known products like methanol, formaldehyde [10] and 3hydroxypropanal [22], which could not be measured accurately due to the limitations of our analytical system. No carbon deposits were visible on the ZrO 2 particles.
At 550 °C the rate of glycerol decomposition was increased significantly, and almost complete conversion was achieved with only 8 s. Significant increases in the yields of all reaction products were observed. The yields of acetic acid, acetaldehyde and hydroxyacetone reached a maximum and started to decrease at high glycerol conversion, showing that they were unstable under hydrothermal conditions. The yield of gases increased continuously, especially carbon oxides since they evolved from the decomposition of acetic acid and acetaldehyde through decarboxylation and decarbonilation. The overall reaction rate for glycerol conversion was well described by pseudo-first order kinetics as seen in Figure 2

Catalyst characterization
Low porosity CaO-stabilized ZrO 2 particles were used as support to prepare the Ru catalyst.
The support particles were screened and the fraction between 100 and 200 m was selected.
The material had a low surface area of 0.1 m 2 /g. Figure 7 shows the XRD spectra of the fresh support particles, which reveals two major crystalline phasestetragonal zirconia (54.81 %) and cubic calcium zirconium oxide (Ca 0.2 Zr 0. 8 Figure 8 shows that the surface of the catalyst was rougher and appeared to be more porous than that of the support. The XRD spectra of the 1 % Ru/ZrO 2 catalyst in Figure 7 confirmed the presence of metallic Ru. The Ru peaks were low and poorly defined which pointed to a high metal dispersion. ESEM microanalysis of the surface supported the formation of a catalyst particle with a shell-type morphology. The 1 % Ru/ZrO 2 -CaO catalyst had an average surface content of 12.4 ± 1.7 % of Ru as measured by ESEM, which indicates that all Ru was distributed in a thin layer on the outer surface and did not penetrate deep inside the particle's core.

Catalytic gasification in SCW
A preliminary series of experiments was developed to assess the stability of the catalyst.
Glycerol conversion and the composition of the gas products were monitored for extended periods to determine changes in the activity of the catalyst and product selectivity. For instance,  The catalyst had a notable influence on the selectivity towards the different reaction products.
Acetic acid became the main condensable product instead of acetaldehyde. For instance, at 8.5 s residence timecomplete glycerol conversionthe yield of acetic acid grew from 0.015 mol/mol Gly in the non-catalyzed experiments to 0.25 mol/mol Gly , while that of acetaldehyde went from 0.045 to 0.15 mol/mol Gly . Hydroxyacetone yield grew from 0.025 to 0.18 mol/mol Gly , also below the yield of acetic acid. Concerning minor products, Figure 4 shows that the yields of allyl alcohol and propionaldehyde also increased significantly, but that acrolein was reduced.
Permanent gasesshown in Figure 3 were the major products, and the yields of hydrogen, carbon oxides and methane all increased when the catalyst was used. The catalyst favored both C-C cleavage reactions that formed acetic acid and acetaldehyde, and dehydration reactions that formed C3 products, mainly hydroxyacetone. The catalyst also promoted reforming to form hydrogen, carbon oxides and methane (secondary products), although glycerol was preferentially converted into primary products (i.e. acetic acid, acetaldehyde, hydroxyacetone, etc.), which were then converted into gases at a slower rate. This may be observed in the experiments at 550 °C (Figures 2 to 4). A glycerol conversion above 95 % was reached in 4 s, when a maximum in acetic acid yield was attained. Notably, the yields of acetaldehyde, hydroxyacetone, and other condensable primary products decreased with residence time in the entire interval covered by the experiment, showing that they were converted into secondary products. The yield of hydrogen was lower than at 510 °C because more carbon monoxide and methane were formed, thus reducing hydrogen yield. However, gases were not the sole secondary products. Figure 5 shows the carbon balances for the non- surface of the bed particles could be observed, even at the highest temperatures. The hydrogen yield potential (HYP, the yield of hydrogen that could be obtained if all identified products were completely converted into hydrogen and carbon dioxide) is also shown in Figure 5, and it was close to the theoretical value of 7 molH 2 /mol Gly at 510 ºC. At 550 ºC it was below 6 molH 2 /mol Gly , but if the missing carbon was assumed to be in the form of acetaldehyde then the HYP value would be close to the theoretical.
The carbon balance closure was significantly lower for the catalyzed experiments. Values of only 80-60 % were obtained, decreasing continuously as the residence time was increased.
Although the formation of methanol and formaldehyde represented a fraction of the carbon unaccounted for, the main reason for the poor carbon balance closure was the formation of carbon deposits on the bed. In fact carbon deposits were not observed on the particles of Ru/ZrO 2 catalyst but on the inert ZrO 2 particles that were used to dilute the catalyst bed, and eventually caused partial plugging of the bed and pressure build-up during extended experiments when temperature was below 510 °C. This implies that carbon deposits evolved from the primary reaction products through condensation reactions in the aqueous phase that formed tar-like carbon precursors. The formation of tar at temperatures below 600 °C has been reported for the hydrothermal gasification of glucose. In our case a proton-catalyzed ionic reaction pathway may be considered due to the relative high concentration of acetic acid and other acidic species. However, the low ion product (kW) and static dielectric constant () of water at the reaction conditions we usedfor instance, kW and  were 1.67x10 -20 (mol/kg) 2 and 1.984 at 510 °C and 350 bar, respectively [33] point to a more likely free-radical pathway as dominant. The formation of carbon deposits contributed to a lower relative yield of hydrogen.
The HYP was from 4.5 to 5.5 molH 2 /mol Gly at 510 ºC, and from only 3.5 to around 5 at 550 ºC.

Conclusions
The hydrothermal and catalytic conversion of glycerol has been studied in supercritical water at formation. The catalyst also converted the primary products into gases through reforming, although the rate of reforming was slower than that of the formation of primary products. When the catalyst was used, the higher concentration of acidic species in the fluid prompted the formation of tar-like adducts that lead to the deposition of carbon on the bedparticularly on the surface of the inert particles of ZrO 2 that were used to dilute the catalyst particlesand a poor carbon balance. In conclusion, although the Ru/ZrO 2 catalyst presented good stability and overall activity, its selectivity towards reforming reactions was not high enough in the intermediate temperature range we studied. From a practical standpoint, low temperature and pressure are desirable in the hydrothermal processing of biomass in supercritical water to reduce the requirements of construction materials and operation costs. Optimization of the properties of the catalyst to enhance the reforming activity at low temperature and the selectivity towards gas products, and the capacity for processing high concentrations of biomass, are key aspects that need to be solved for the implementation of the supercritical water gasification technology.