Hydrogenation of 9-ethylcarbazole as a prototype of a liquid hydrogen carrier

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 5 ( 2 0 1 0 ) 1 1 6 0 9 e1 1 6 2 1 Available at www.sciencedirect.com journal homepage: www.elsevier.com/locate/he Hydrogenation of 9-ethylcarbazole as a prototype of a liquid hydrogen carrier Katarzyna Morawa Eblagon a, Daniel Rentsch b, Oliver Friedrichs c, Arndt Remhof c, Andreas Zuettel c, A.J. Ramirez-Cuesta d, Shik Chi Tsang a,* a Wolfson Catalysis Centre, Inorganic Chemistry Laboratory, University of Oxford, Oxford OX1 3QR, UK Laboratory for Functional Polymers, Swiss Federal Laboratories for Materials Testing and Research, Überlandstrasse 129,CH-8600 Dübendorf, Switzerland c Hydrogen & Energy Empa e Swiss Federal Laboratories for Materials Testing and Research, Überlandstrasse 129, CH-8600 Dübendorf, Switzerland d ISIS facility, Rutherford Appleton Laboratory, Chilton, Didcot OX11 0QX, UK b article info abstract Article history: Liquid organic hydrides, e.g. 9-ethylcarbazole, are potentially interesting hydrogen storage Received 10 December 2009 materials because of their reversible hydrogen sorption properties. In the present work, Received in revised form hydrogenation reaction of 9-ethylcarbazole in the molten form was investigated over 8 March 2010 a wide variety of noble metal and nickel supported catalysts. The catalytic activity of Accepted 12 March 2010 8.2  10 Available online 21 April 2010 loaded product were recorded over 5 wt% ruthenium on alumina which was prepared by 6 mol-ethylcarbazole/g of metal/s and selectivity of 98% towards a fully hydrogen mild chemical reduction of ruthenium salt. Using this catalyst the theoretical capacity of Keywords: hydrogen uptake (5.7 wt%) was obtained and the rate of the reaction and activation energy 9-Ethylcarbazole were estimated. Due to its potential high hydrogen storage capacity, this system could be Catalytic hydrogenation a promising on-board storage candidate for mobile applications. The structures of reaction Hydrogen storage products and intermediates were identified using 2D NMR techniques. These structures 2D NMR were also predicted to be thermodynamically stable using density functional theory (DFT), DFT calculations matching well with the experimental observations. ª 2010 Professor T. Nejat Veziroglu. Published by Elsevier Ltd. All rights reserved. 1. Introduction Hydrogen is proposed as a clean energy carrier that can be used in transportation without emissions of greenhouse gases [1]. Storing hydrogen in the form of liquid organic hydrides is a relatively new approach, which offers some unique advantages. Liquid storage of hydrogen can be pumped easily not only during delivery and storage, but also within the vehicle operations. Thus, a small aliquot can be heated for dehydrogenation reaction to provide hydrogen for a fuel cell device in the catalytic chamber on board of the vehicle. This can offer higher safety and security in case of collision without the spillage of a large mass of hot reactive material present on board [2]. Other advantages of hydrogen storage in form of liquid organic hydrides are a) no CO, CO2 or other by-products, b) total reversible reaction and recyclability of reactants and products, c) that they can attain high gravimetric hydrogen storage value (5e8 wt%), d) possibility to integrate to the present infrastructure including oil tankers and gasoline distributors [3]. In the past, the use of liquid organic hydrides did not receive much attention due to the fact that the hydrogen liberation step (dehydrogenation) from organic molecule is strongly * Corresponding author. Tel./fax: þ44 1865282610. E-mail address: edman.tsang@chem.ox.ac.uk (S.C. Tsang). 0360-3199/$ e see front matter ª 2010 Professor T. Nejat Veziroglu. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.ijhydene.2010.03.068 11610 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 5 ( 2 0 1 0 ) 1 1 6 0 9 e1 1 6 2 1 endothermic [4], thereby requiring prohibitively higher reaction temperatures (600e700 K) to supply hydrogen [5] than the operating conditions of fuel cells. For traditional PEM fuel cell the optimum working temperature is in the range of 343e353 K [6], however for the novel high temperature PEM fuel cells, the working temperatures can be as high as 393e453 K, but this is still considerably lower than the required temperatures for hydrogen release from typical saturated organic molecules through catalytic dehydrogenation. On the other hand, the use of aromatic molecules with heteroatom(s) for hydrogen storage was first studied by Pez, Scott and Cheng from Air Products in their series of patents [7,8]. One of the prototypes for liquid organic hydrides studied in their research was 9-ethylcarbazole [9]. This heterocyclic compound can theoretically take up 6 moles of equivalent hydrogen (5.7 wt%). This capacity is higher than the revised storage targets (4.5 wt% by 2010) set by DOE in USA in early 2009 [10]. It has been found that the incorporation of nitrogen [11] and fused rings into an organic structure can drastically decrease the enthalpy of dehydrogenation, enabling fine tuning of the reversible thermodynamics of hydrogenation [12]. Furthermore, the presence of the ethyl group could lower the melting point of this molecule from 243  C for carbazole to 60  C for 9-ethylcarbazole allowing a possible reversible hydrogenation to take place without the use of solvent, while keeping the high theoretical hydrogen uptake capacity of the material. Thus, hydrogen can be extracted via catalytic dehydrogenation of 9-ethyl-perhydroethylcarbazole (hydrogen loaded state) and stored by the hydrogenation of 9-ethylcarbazole (hydrogen unloaded state) according to Eq. (1): C14H25N $ C14H13N þ 6H2 (1) Undoubtedly, efficient hydrogenation of 9-ethylcarbazole in the molten state for producing 9-ethyl-perhydrocarbazole is a crucial step in developing reversible hydrogen storage material based on 9-ethylcarbazole. In the literature, the dehydrogenation of 9-ethylcarbazole over heterogeneous catalysts has been studied [7e9,13e15]. There is also a study by Jensen et al., for the dehydrogenation of 9-perhydroethylcarbazole using homogeneous PCP pincer iridium based complexes [16] which were previously reported to be effective for the homogeneous dehydrogenation of cycloalkanes to arenes [17]. However, there is no in-depth study concerning the mechanism of catalytic hydrogenation of 9-ethylcarbazole for hydrogen storage purpose. Only very recently, an investigation on hydrogenation of 9-ethylcarbazole in decalin solution using a commercially available ruthenium catalyst has been initiated by Sotoodeh et al. [13,14]. On the other hand, hydrogenation of aromatic molecules in general has received a lot of attention for the production of high performance diesel fuels [18]. However, it is known that more energy input is required to hydrogenate heteroaromatic compounds (10e20 kcal/mol per double bond) than for hydrogenation of olefins, ketones and imines, etc [19]. In addition, the heteroatom can strongly bind to metal catalytic centres leading to catalyst deactivation [20]. Furthermore, it is accepted that many of them are difficult to be hydrogenated, and that in several cases hydrogenations are complicated by side reactions [21]. Conventional catalytic systems for hydrogenation of aromatics, such as Pd, Ni [21,22], Pt, Ru and Co supported on TiO2, SiO2, Al2O3, SiO2eAl2O3, zeolite, ZnO, MgO, graphite, activated carbon, charcoal and others [22] in solvents such as nhexane, methanol [23], ethanol [24] isooctane, n-heptane and cyclohexane [25] have been studied previously. It is worth emphasizing that all the studied hydrogenation reactions were conducted in solution phase but no attempt has been reported using the molten compound. We believe that it is very important to demonstrate the direct applications of hydrogenation and dehydrogenation in the molten reactant without the unnecessary use of solvent. First of all, removal of the solvent from the system would allow preserving the theoretical capacity of the hydrogen uptake of the material. This work also aims at studying whether problems like severe deactivation due to extensive adsorption of pure reactant molecules or any insurmountable diffusion problems, etc are not encountered in the pure molten reactant. It should be borne in mind that the artefacts due to the solvent dilution and the intrinsic weight of solvent could prohibit the practical validity of a reactantesolvent system. In this present work, we report for the first time the hydrogenation of molten 9-ethylcarbazole over various supported noble metal catalysts. It was found that the theoretical hydrogen uptake capacity of 9-ethylcarbazole was actually achieved over the chemically reduced (CR), 5 wt% Ru/Al2O3 catalyst, giving high catalytic activity per gram of metal and high selectivity towards fully hydrogenated product amongst all materials tested. Additionally, this catalyst was found to give much better performance than its commercial counterparts [13], due to its much higher atom efficiency. The reaction products were identified by GCeMS and the stereochemical structures of the intermediates and products were investigated using 2D NMR techniques, shedding new light to the reaction mechanism. In the previously reported work [13,14], the authors relied only on GCeMS analysis to obtain the structures of the intermediates and products and they reported only one stereoisomer of the fully loaded 9-ethyl-perhydrocarbazole, whereas in this work we have found four isomers of 9-ethyl-perhydrocarbazole. Moreover, we have optimized the total energies of the reactants and products using density functional theory calculations to gain insight into their stability as well as the probability of the formation of defined stereoisomers. The calculation results agreed well with the experimental results. From all the data, obtained in the present work, the merits of developing the melt based liquid organic hydride system for hydrogen storage are clearly demonstrated. 2. Experimental 2.1. Materials 9-Ethylcarbazole (97%) and the commercial catalysts used in this work are listed in Table 1, sodium borohydride (99%) and ruthenium chloride hydrate (38e40%) were purchased from SigmaeAldrich. Graphite, activated carbon, alumina, silicaealumina and zeolite were purchased from Johnson Matthey. Hydrogen gas of technical grade was supplied by Air Products. The HPLC grade solvents were supplied by Fisher-Scientific. i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 5 ( 2 0 1 0 ) 1 1 6 0 9 e1 1 6 2 1 Table 1 e Catalytic activity and selectivity of screened catalytic systems for hydrogenation of molten 9ethylcarbazole. COM stands for commercial catalyst, CR represents chemically reduced catalyst, using sodium borohydride solution. Sample (based on recipe value) 5 wt% Ru/Al2O3 (CR) 5 wt% Ru/TiO2 (CR) 5 wt% Ru/Al2O3 (COM) 5 wt% Ru/SiO2-Al2O3 (CR) 5 wt% Ru on zeolite (CR) 5 wt% Ru on graphite (CR) 5 wt% Ru/AC (CR) 5 wt% Pd/AC (CR) Ru black (COM) 65 wt% Ni/SiO2-Al2O3 (COM) 5 wt% Pt/AC (CR) 65 wt% Ni/AC (CR) 2.2. Catalytic activity  10 6 (mol ethc/g of metal/s) Catalytic selectivity towards sum of þ12H stereoisomers (%) 8.2 7.9 7.8 7.5 7.2 7.2 6.8 1.5 0.4 0.3 0.1 0.05 98 96 92 96 78 64 53 2 77 93 0.6 33 Catalyst preparation and characterisation Mild chemical reduction was used in the preparation of supported catalysts. An appropriate amount of metal chloride salt was dissolved in 10 mL of distilled water. This solution was then added dropwise to the aqueous slurry of the support. After 48 h of stirring, the mixture was reduced by solution of sodium borohydride in water at 90  C. Subsequently, the catalysts were washed five times with distilled water and acetone and centrifuged at 5000 rpm/min for 10 min each time. They were then left to dry overnight at room temperature under ambient atmosphere. The BrunauereEmmeteTeller (BET) specific surface areas of catalysts were determined by adsorptionedesorption of nitrogen at liquid nitrogen temperature, using Micromeritics TriStar 3000 equipment. High resolution transmission electron microscopy (HRTEM) micrographs were taken using an HRTEM JEOL 2010. The quantitative analysis of the metal content in the ruthenium based catalysts was carried out using Energy Dispersive X-ray analysis (EDAX). Representative samples were examined by High Vacuum Cambridge 360 Stereoscan Scanning Microscope fitted with an Oxford Instruments INCA for the X-ray analysis. The samples were focused with the magnification of about 600k and four different areas of each of the samples were examined by the EDX analysis with working distance set at 25 mm, dead time 50% and process time of 6 min. The influence of signal from the carbon disc sample holder was taken into consideration for the calculation of the final metal loading. 2.3. Catalytic hydrogenation and products analysis The catalytic hydrogenation reactions were performed in a 10 mL glass insert placed in a stainless steel 300 mL Parr batch reactor with magnetic stirring. The glass insert was used due to the small volume of the molten reaction mixture as compared to the total volume of the reactor. The glass insert assured the effective and homogeneous stirring of the whole molten reaction mixture. One gram of 9-ethylcarbazole 11611 was placed in the glass, followed by 0.2 g of catalyst. The reactor was then sealed, flushed with hydrogen and then heated to 130  C. When the desired temperature was reached, 70 bar of pure hydrogen was charged into the reactor and the reaction time was measured. The reaction was stirred using a PTFE encapsulated bar and a magnetic stirrer produced by Ika-Werke at the speed of 600 rpm/min, which was placed under the heating mantle of the autoclave. After a set period of time, the autoclave was cooled down to room temperature using a water bath, the hydrogen pressure was released and the reaction mixture was analysed using a GCeMS (Agilent 6890GC-MS) equipped with a non-polar capillary column (Agilent 19091s-433) and an auto-sampler. For the NMR study the intermediates and products were analysed versus time: several batches were made in parallel, with varied reaction times. Hydrogen absorption measurements were done using a Sieverts apparatus with an “in house” set-up. This set-up allowed the monitoring of the hydrogen absorption at different stable temperatures by measuring the pressure decrease in the sample chamber [26]. For pressure-drop measurements, the reaction conditions were modified using 0.05 g of 5 wt% Ru on alumina (CR) catalyst mixed with 1 g of 9-ethylcarbazole in a 15 mL sample chamber. The hydrogenation reaction was stirred using a magnetic stirrer, and to avoid the evaporation and condensation of the products outside the sample chamber, a cooling trap was used. The sample chamber was heated up using an oil bath equipped with thermocouples. A lower catalyst to reactant ratio was used in order to slow down the kinetics as compared to the previous testing protocols to follow the rates of hydrogen uptakes at various temperatures for comparison. 1 H and 13C NMR spectra were obtained at 400.13 (100.61) MHz on a Bruker Avance-400 NMR spectrometer. The 1H and 13 C NMR spectra, the 1H,1H-DQFeCOSY (double quantum coherence correlation spectroscopy), eJRES (J resolved) and eNOESY (nuclear overhauser effect spectroscopy), and the 1H, 13 C-HSQC (heteronuclear single quantum coherence), eHMBC (heteronuclear multiple bond correlation) and eHSQCeTOCSY (total correlation spectroscopy) experiments were performed at 298 K using a 5 mm broadband inverse probe with z-gradient and 90 pulse lengths of 6.8 ms (1H) and 14.9 ms (13C). All spectra were recorded with Bruker standard pulse programs and parameter sets and the 1H/13C chemical shifts were referenced internally using the resonance signals of cyclohexane-d12 at 1.38/26.43 ppm. 2.4. Theoretical calculations DFT calculations were performed using Gaussian 03 software [27] with 6-311G basis set and B3LYP hybrid functional for electron correlation method [28,29]. The calculations were performed for all the reactants and products in the gas phase only, to simplify the problem and shed light on the stability of the intermediates in this reaction. The geometries of all possible intermediates and products were optimized and the thermal energies were calculated. Thermodynamic parameters of products or intermediates involving transitional, rotational and translational modes were also taken into account [30]. Subsequently, the calculated total energy (the sum of 11612 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 5 ( 2 0 1 0 ) 1 1 6 0 9 e1 1 6 2 1 electronic energies found from geometry calculations and thermal energies) was compared with the sum of total energies of the starting material (9-ethylcarbazole) together with the total energy of moles of hydrogen added to create defined intermediate according to the formula given below in Eq. (2): (2) EC14 H13 N þ nEH2 /EC14 H13þ2n N where: EC14 H13 N is the total energy of 9-ethylcarbazole, EH2 is the total energy of added moles of hydrogen and EC14 H13þ2n N is the total energy of intermediate products. 3. Results 3.1. Catalyst screening The hydrogen saturation of 9-ethylcarbazole was investigated over a number of supported and unsupported noble metal catalysts. The catalytic activity (Ac) per gram of metal present and selectivities (Sel) were calculated according to Eqs. (3) and (4): Ac ¼ 9-ethylcarbazole converted ½molesŠ total metal content ½gŠ  time ½sŠ (3) Sel ¼ product ½molesŠ  100 sum of products ½molesŠ (4) surface area per gram shows rather poor catalytic activity, due to the poor metal utilization. Other metal catalysts prepared by the same chemical reduction method on the same support (activated carbon, AC) arranged by decreasing catalytic activity are as follows: Ru > Pd > Pt > Ni. The catalytic systems based on 5 wt% Ni were found to show no activity at all under comparable reaction conditions and thus higher metal loadings of nickel were used. Similarly, supported platinum was found to give very low catalytic activity for this reaction. Generally, platinum is well known to display high activity in hydrogenation reactions [12], however, the reasons of the poor activity observed in the present work are not yet clear. Further study will be required. Overall, the best combination of selectivity and activity were clearly displayed by 5 nominal wt% ruthenium on alumina (CR- actual loading 4.8 wt%). This catalyst proved to be superior to its commercial counterpart on a gram basis (see Tables 1 and 2). The findings also suggest that the performance of the catalyst in this reaction is not only dependent on its composition but also on the method of synthesis. The superior performance of the 5 wt% ruthenium on alumina (CR) is attributed to the better utilization of the metal atoms present in this catalyst as compared to the commercial 5 wt% of Ru on alumina. 3.2. The catalyst was compared based on the catalytic activity per gram of metal, since this value is important in terms of metal utilization and the cost of the storage system. All screened catalytic systems are depicted in Table 1. Since for hydrogen storage, it is very important to fully hydrogenate the unloaded compound without producing side products, therefore the catalytic activity and the selectivity towards the sum of stereoisomers of fully hydrogenated product (9-ethyl-perhydrocarbazole) for all of the catalytic systems studied have been compared. Careful inspection of Table 1 reveals that the catalytic activity and selectivity critically depend on the types of metal and support used. In general, the most active catalytic systems are ruthenium based supported catalysts, whereas unsupported ruthenium (ruthenium black) with expected low Catalyst characterisation The physical properties such as metal particle size, metal loading obtained by EDAX analysis and total specific surface area of the selected catalysts are shown in Table 2. All catalytic systems were mesoporous with a wide range of pore sizes, however the pore size did not seem to affect much the activity of the systems. A similar observation was previously made by Shinohara et al. in the dehydrogenation of decalin [31]. There was also no direct correlation between the specific surface area of support and activity of the catalysts. To appreciate the influence of the support, ruthenium was deposited onto various solid carriers via chemical reduction using sodium borohydride. As seen in Table 2, the metal loadings achieved for all of the synthesized catalysts were very similar to the expected 5 wt%. The typical EDAX spectrum of 5 wt% Ru on alumina (CR) with the inset of SEM micrograph of the same sample, are shown in Fig. 1. It was Table 2 e Physical properties of the catalytic systems investigated arranged in the decreasing order of catalytic activity. CR describes chemically reduced samples and COM describes commercial samples, sd is standard deviation of size measurement. Catalyst (based on recipe value) Synthesis method Particle size [nm] (s.d) BET specific surface area [m2/g] Average pore diameter [nm] Metal loading % by EDAX 5 wt% Ru/Al2O3 5 wt% Ru/TiO2 5 wt% Ru/Al2O3 5 wt% Ru/SiO2eAl2O3 5 wt% Ru/zeolite Ru/graphite 5 wt% Ru/activated carbon 65 wt% Ni on SiO2eAl2O3 Ru black CR CR COM CR CR CR CR COM COM 1.61 (0.38) 2.11 (0.55) 9.08 (1.20) 2.37 (0.61) 2.31 (0.62) 2.30 (0.58) 2.99 (1.04) 7.50 (1.54) Aggregated 152 191 83 483 343 7 538 95 22 11.7 14.6 12.0 5.4 3.2 35.3 5.2 10.0 24.4 4.8 4.8 e 4.8 4.7 4.9 4.8 e e i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 5 ( 2 0 1 0 ) 1 1 6 0 9 e1 1 6 2 1 observed that the catalytic activity is clearly dependent on the type of support used and decreases in order as follows: Al2O3 > TiO2 > SiO2eAl2O3 > zeolite  graphite > activated carbon, which is shown in Table 1, for the set of Ru catalyst on various carriers. The ruthenium supported on zeolite can be considered more active then ruthenium supported on graphite, due to the fact that the former contains lower amount of ruthenium metal (see Table 2). The rates of the reaction are apparently higher on acidic oxide catalyst than on neutral high surface area carbon or neutral low surface area graphite. Careful analysis of the ruthenium catalysts prepared by chemical reduction also reveals that the obtained activity is closely related to the inverse of metal particle size, as it can be seen in Fig. 2. With regards to selectivity to a fully hydrogenated product, when comparing Ru on silicaealumina, zeolite and graphite, all having similar particle size and metal loading, the selectivity tends to decrease with decreasing acidity of the support. Moreover, it is worth noting that the mild chemical reduction method used in the present work, which results in well dispersed small nanoparticles, is evidenced from the HRTEM images of the best catalyst, 5 wt% Ru on alumina (CR) in Fig. 3. In these images, it can be seen that the small metal nanoparticles are finely dispersed on the alumina carrier. The lattice spacings of the metal nanoparticles measured in the image on the left and were: 2.1  A and 1.9  A, respectively. These numbers correspond well with those of elemental cubic ruthenium published in the XRD database (ICSD 01-088-2333): namely, 2.21  A (111) and 1.95  A (200) crystallographic planes. Therefore, the produced nanocatalyst contained crystalline structure of elemental Ru. 3.3. Hydrogen uptake measurements A series of hydrogen absorption curves were recorded for the best catalyst 5 wt% ruthenium on alumina (CR) in order to obtain information about the uptake capabilities of 9-ethylcarbazole. The absorption isotherms were measured to observe the influence of the reaction temperature on the rate of the reaction. Therefore, the isotherms were measured at 11613 120  C, 130  C, 140  C, 150  C and 170  C where an increment of the reaction rate was observed with increased temperature (Fig. 4). Conversely, there is no significant difference between 150  C and 170  C and under both conditions, the hydrogen uptake reaches the theoretical hydrogen uptake value for 9-ethylcarbazole (5.7 wt%). As also seen from Fig. 4 that the uptake for 130  C is slightly higher than 120  C but their initial rates are very similar. Higher rates of the reaction between hydrogen and molten reactant account for higher uptakes when high temperatures are used. However, at 170  C the uptake rate might be limited by mass transfer of substrates since there is no evidence of further increase in reaction rate from 150  C (which is commonly encountered in hydrogenation). However, no such mass transfer limitation was reported in solution phase hydrogenation of the same reactant in Ref. [13]. We therefore envisage that the hydrogen solubility or the transfer of hydrogen in molten state was limited in this system. Further kinetic study should be performed. Based on these measurements, first order rate constants for each temperature were derived. The rate constants are listed in Table 3 together with catalytic activity calculated for each of the temperatures according to Eq. (3). Subsequently, the logarithm of the kinetic rate constants was plotted versus the reciprocal temperature to obtain the Arrhenius plot, as shown in Fig. 5. The apparent activation energy was estimated to be 58 kJ/mol based on the consumption of 9-ethylcarabazole. This value is lower than the apparent activation energy of 99.5 kJ/mol estimated by Sotoodeh et al. [13], indicative of a degree of mass transfer limitations in our case (molten reactant). We envisage that the hydrogen solubility or transfer of hydrogen in molten state was limited in this system. However, one must be cautious that the apparent activation energies were estimated using different experimental conditions (absence of solvent, smaller reactor, different stirring speed, different catalyst and catalyst to reactant ratio, etc.). Further detailed kinetic measurements with model fittings should be carried out to explain the difference in the measured activation energy. Fig. 1 e An SEM micrograph (inset) and EDAX spectrum showing the chemical composition of 5 wt% Ru on alumina (CR). The scale bar in the SEM micrograph is 4 microns. The highest peak in the EDAX spectrum belongs to carbon disc mounted on an aluminium SEM specimen stub (Agar Scientific). 11614 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 5 ( 2 0 1 0 ) 1 1 6 0 9 e1 1 6 2 1 Fig. 2 e Catalytic activity versus particle size for synthesized (CR) ruthenium based catalysts on different supports. There is a clear inverse relationship between catalytic activity and particle size. 3.4. The study of the reaction products and intermediates It is important to identify the structures of the reaction intermediates and products which would lead to a better understanding of the mechanism of this reversible reaction and allow us to identify the challenges of this system. The stereoisomers of the products and intermediates of the catalytic hydrogenation of 9-ethylcarbazole have not been studied previously [13,14]. It is believed that their identification is a crucial step in defining the actual reaction pathway as well as in analysis of the potential reaction reversibility. Two partially hydrogenated intermediates (9-ethyl-tetrahydrocarbazole and 9-ethyl-octahydrocarbazole) and a set of three different fractions of the fully hydrogenated compound 9-ethyl-perhydrocarbazole were separated, quantified and identified by means of GCeMS. The product distribution versus time is shown in Fig. 6, for the initial 8 h of the reaction time using 5% Ru on alumina (CR). This profile was obtained running each time span as a separate reaction due to the technical difficulties with continuous sampling of the high viscosity melt. The reaction products and intermediates identified only using GCeMS analysis together with a simplified version of the possible reaction pathway are depicted in Fig. 7. A variety of 1D and 2D NMR (2 dimensional nuclear magnetic resonance) techniques were then used to further elucidate the structures of products and related intermediates that were not provided by means of GCeMS, in order to gain an insight into the reaction pathway. In Fig. 8 the 2D-correlated 1H, 13C-HSQC (correlations of Hs to directly bound Cs, spectra at bottom) and eHMBC (correlation of Hs to Cs 2e3 bonds apart, top spectrum) NMR spectra are shown. The red and blue coloured correlation signals belong to 9-ethylcarbazole (starting material) and to 9-ethyl-tetrahydrocarbazole (þ4H), respectively. All assigned correlation signals unambiguously verified the structure of the latter compound shown in Fig. 8. It must be mentioned that in the 2D and 1D NMR spectra a total of 14 carbons with the corresponding protons were assigned (for chemical shifts, see Appendix). It is evident from Fig. 6 that 9-ethyl-tetrahydrocarbazole (þ4H) must be formed in an early stage of the reaction as it is shown in Fig. 7. The structure of the following intermediate, 9-ethyl-octahydrocarbazole (þ8H), was assigned following the same procedure: a) identification of carbon atoms not yet assigned to already known structure over the HSQC cross signals, b) identification of carbon atoms near these protons over the HMBC correlated spectra and c) additional assignments of the 1H spin systems over 1 H,1H-DQFeCOSY experiments. After longer reaction periods a lot of additional cross signals appeared in the 2D NMR spectra (data not shown), and the assignments to the structures shown in Fig. 9 became more and more complicated, but with additional 2D NMR experiments (see below) we were able to identify also some of the 9-ethyl-perhydrocarbazoles. Fully hydrogenated (þ12H) products of 9-ethylcarbazole possess 4 stereocentres at positions 4a, 4b, 8a and 9a of the molecule (Fig. 9). Therefore, 24 ¼ 16 stereoisomers are theoretically possible. Owing to mirror planes and/or C2 axes transforming one structure into another, this number is reduced to 6 molecules distinguishable by means of NMR. Owing to mirror planes or to C2 axes within the structure itself, only 8 carbon resonances are expected for the 4 symmetric products AeD and for the two asymmetric molecules E and F a total of 14 carbon resonances are expected in the 13C NMR spectra. Fig. 3 e On the right, 5 wt% ruthenium on alumina (CR) showing evenly dispersed metal nanoparticles on the alumina support. On the left, HRTEM image of the same sample showing ruthenium metal lattice fringes. 11615 6% 0.0022 0.00225 0.0023 0.00235 0.0024 0.00245 0.0025 0.00255 0.0026 0 5% -1 -2 y = -6976.9x + 11.061 R² = 0.9204 4% -3 3 2 4 3% ln k Hydrogen uptake % i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 5 ( 2 0 1 0 ) 1 1 6 0 9 e1 1 6 2 1 5 1 -4 -5 2% -6 -7 1% -8 1/T 0% 0 2 4 6 8 10 12 14 16 Time [h] Fig. 4 e Hydrogen absorption in 9-ethylcarbazole measurements at 1 e 170  C, 2 e 150  C, 3 e 140  C, 4 e 130  C and 5 e 120  C at 70 bar H2 over 5 wt% ruthenium on alumina (CR). In the aliphatic part of a 1D 13C NMR spectrum 14 carbon resonances of equal signal intensities are indeed observed (Fig. 10). When the hydrogenation reaction was quenched much earlier, 13C NMR signals were equal, but when compared to the spectrum discussed above considerably decreased relative signal intensities were observed at the same chemical shifts. With the aid of a 1H, 13C-HSQCeTOCSY experiment we showed that the magnetization of the proton at 0.9 ppm (outstanding from others) is transferred over the conjunct spin system of 6 attached carbons (Fig. 11, left) and with a lower threshold of the same spectral region it is evident that 12 carbons are attached to the selected proton (Fig. 11, right, no correlation observable over quaternary carbons or nitrogen to the ethyl group). Thus, one of the fully hydrogenated products must have one of the structures E or F shown in Fig. 9 with 14 different carbon atoms but its quantification was proved difficult. The 1H and 13C chemical shifts of (þ12H) E (see Fig. 9) were fully assigned over the DQFeCOSY and HMBC NMR experiments (see Fig. 11). Owing to 3J(H-4b/H-8a) z 10 Hz a trans diaxial configuration of these two protons can be envisioned, whereas 3J(H-4a/H-9a) is rather small. From chemical shift arguments with Dd13C of 8.7 ppm for the carbon pair C-8a/C-9a and of 4.5 ppm for C-4a/C-4b, we concluded that product (þ12H) E is consistent with the configuration Fig. 5 e An Arrhenius plot using experimental rate constants from the reactions at 120e170  C. shown in Fig. 9 with H-8a and H-9a in trans positions, whereas for the minor asymmetric and fully hydrogenated product (þ12H) F values Dd13C of 2.1 (for C-8a/C-9a) and 9.5 ppm (for C4a/C-4b) were also observed. This observation supports the presumption that H-4a and H-4b are trans to each other in this conformer. However, species F was found to be in much lower concentrations in the analysed mixture, as compared to E. Using the above mentioned NMR techniques, we have also identified two of the 4 symmetric structures AeD, depicted in Fig. 9. Owing to six distinct carbon resonances found in the HSQCeTOCSY spectra, it must be assumed that the two additional species must be symmetric and that their chemical structures belong to the four shown in the same figure (structures AeD). It is clear that from the analysis of an HSQCeTOCSY spectrum the attached N-ethyl groups are not detectable, but they can be assigned over the 1H, 13C-HMBC experiments (correlation over 2e3 bonds). Although the 1D 13C NMR spectra were not recorded using relaxation delays Table 3 e Rate constants and catalytic activity for 9ethylcarbazole hydrogenation over 5 wt.% Ru/Al2O3 catalyst. These values were calculated based on rate constants derived from measurements shown in Fig. 4. Temperature  C 120 130 140 150 170 Rate constant, k  10 2 (min 1) Catalytic activity  10 1 (mM ethc/g of metal/s) 0.13 0.15 0.26 0.63 0.81 0.40 0.46 0.67 1.12 1.23 Fig. 6 e The reaction products distribution versus time for the initial 8 h of the reaction obtained using 5% Ru on alumina (CR) as identified by GCeMS and NMR techniques. The (D4H) stands for 9-ethyl-tetrahydrocarbazole, (D6H) is 9-ethyl-hexahydrocarbazole and so on. The letters E, B and F correspond to the 9-ethyl-perhydrocarbazole isomers depicted in Fig. 9. 11616 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 5 ( 2 0 1 0 ) 1 1 6 0 9 e1 1 6 2 1 Fig. 7 e A simplified scheme containing intermediates and products of catalytic hydrogenation of 9-ethylcarbazole (D4H and D8H correspond to four and eight hydrogen atoms added to the starting material, respectively). The D6H intermediate was detected only by mass spectrometry (MS), all of the other compounds were detected by both MS and NMR. ppm 110 N 115 120 9-ethyl-tetrahydrocarbazole (+4H) 125 130 N 135 140 9-ethylcarbazole (+0H) 4.0 3.5 3.0 2.5 2.0 1.5 ppm ppm ppm 15 110 20 115 25 8.0 7.8 7.6 7.4 7.2 7.0 6.8 120 30 125 35 ppm 4.0 3.5 3.0 2.5 2.0 1.5 ppm Fig. 8 e Regions of interest of 1H, 13C-HSQC (bottom) and eHMBC (top) 2D NMR spectra of a mixture of 9-ethylcarbazole (starting material, in red) and 9-ethyl-tetrahydrocarbazole (D4H, in blue) quenched during an early stage (1 h) of hydrogenation. i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 5 ( 2 0 1 0 ) 1 1 6 0 9 e1 1 6 2 1 11617 Fig. 9 e Chemical structures of all six 9-ethyl-perhydrocarbazole stereoisomers distinguishable by means of NMR (absolute configuration shown arbitrarily) with partial numbering of carbon atoms used for signal assignments (see text and Appendix). Upper row are symmetric compounds for which 8 different carbon resonances were observed, and the bottom line are asymmetric compounds for which 14 different carbon resonances were observed. sufficient for full relaxation of the carbon nuclei, similar relative signal intensities were observed for each species, notably, the signals of C-10 and C-11 of the symmetric products each exhibit half the intensity compared to the resonances of C-1e4, C-4a and C-9a. The chemical shifts of individual positions of the core structures in the molecules were assigned over DQFeCOSY and HMBC correlations and with help of chemical shift arguments. Fig. 10 e Aliphatic region of the 1D 13C{1H} NMR spectrum of a 9-ethylcarbazole batch hydrogenated during a longer period (22 h). Totally 14 carbon signals with relative high signal intensities of D12H E product (signals of C6D12 around 26 ppm) were observed. 11618 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 5 ( 2 0 1 0 ) 1 1 6 0 9 e1 1 6 2 1 Fig. 11 e Regions of interest of 1H, 13C-HSQCeTOCSY spectra showing the correlations of H-6 of 9-ethyl-perhydrocarbazole (structure D12H E in Fig. 9) to the 6 carbons C-(1e4, 4a and 9a; left), and to all 12 carbons C-(1e8, 4a, 4b, 8a, 9a; right). The spectrum on the right is shown with a much lower threshold. 3.5. DFT theoretical calculations The relative stabilities of different candidate intermediate compounds and products were calculated using DFT methods. All the calculations have been done in a gas phase approximation. In order to introduce thermal effects, the total energy was calculated including thermal contributions to the free energies of the molecule at 423 K, in order to determine the theoretical enthalpies of reaction for a gas phase reaction [27]. In Fig. 12 the energy differences are plotted between the reactants and product according to Eq. (2). A total of 96 different possible products were modelled. It can also be seen that the free energy of one of the (þ8H) isomers gives a considerably lower energy than all the (þ10H) compounds. In the hydrogenation reaction in the molten form with the absence of the solvent, the (þ8H) intermediate was observed to be the main product formed. The calculated energetics of the system (with no consideration of activation energy barriers) suggest that it may favour the direct formation of the (þ12H) compound, as seen in Fig. 12. Based on these theoretical considerations only extremely low amounts of (þ10H) if any, would be expected to be produced in this reaction. In our experiments, the production of (þ10H) products was not detected. However, only traces of (þ10H) intermediate were previously observed as a minor product in the same hydrogenation reaction but using decalin as a solvent [13,14]. This observation clearly agrees well with our calculation results. As can be seen in Fig. 12, the intermediates that are favourable to be created in the reaction are: 9-ethyl-tetrahydrocarbazole (þ4H), 9-ethyl-hexahydrocarbazole (þ6H), 9-ethyl-octahydrocarbazole (þ8H) and five isomers of 9-ethylperhydrocarbazole (þ12H). The more negative the calculated values, the more stable the compounds are in the gas phase. It can be seen that there is an energy penalty for most of the isomers considered: there are no products with 9-ethyl-dihydrocarbazole (þ2H) that are more stable than the reactants. A general trend is that only few of the calculated structures are energetically favourable under these conditions: the (þ8H) isomers, the 9-ethyldodecahydrocarbazole (þ10H) isomers and (þ12H) isomers, where these hydrogenated products appear to be more stable than the reactants. Table 4 summarises the energy differences for these more stable compounds. The numbers in brackets correspond to the numbers of carbon to which the hydrogen atom was attached to. The numbering of the carbon atoms starting from the carbon next to nitrogen and going clockwise is shown in the scheme of 9-ethylcarbazole that is in the Fig. 12 e DFT calculated free energy differences between 9-ethylcarbazole and the most stable reaction intermediates/products (labelled) in the gas phase. i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 5 ( 2 0 1 0 ) 1 1 6 0 9 e1 1 6 2 1 11619 Table 4 e Calculated energy difference in [kJ/mol] of hydrogenated intermediate with respect to 9-ethylcarbazole as a function of hydrogen loading. Numbers between brackets refer to the carbon position at which the hydrogen atoms were added (also see Fig. 7). (þ4H) (3456) 12.98420 (1234) 9.75843 (þ6H) (þ8H) (þ10H) (þ12H) (123456) 78.93812 (23451011) 48.46200 (2345891011) 143.71344 (345678910) 59.12408 (3456891011) 57.15482 (56789101112) 50.44525 (34569101112) 33.98009 (1456891011) 28.05494 (12369101112) 27.73268 (1256891011) 24.68955 (12349101112) 23.63593 (3456781112) 21.46309 (34789101112) 14.06269 (34569101112) 7.25857 (12369101112) 5.34527 (12789101112) 3.20234 (345678912) 3.19751 (12569101112) 0.41199 (3456789101112) 125.48033 (2345789101112) 121.16840 (1234789101112) 95.10115 (1236789101112) 81.95844 (1456789101112) 74.15556 (1456789101112) 74.13903 (1256789101112) 67.49901 (1 up, 6 7 12 down) 182.25727 (1 6 up, 7 12 down) 170.54110 (all up) 166.09700 (6 up, 1 7 12 down) 160.25676 (1 12 up, 6 7 down) 157.70859 bottom of the table. It is interesting to note that the lowest energy structure corresponds to the asymmetrical fully hydrogenated product E rather than the symmetrical products, as shown in Fig. 9. Experimentally, we have not observed (þ6H) in NMR analysis, probably because the concentration observed on GCeMS was in a trace level. However, we have observed high concentrations of (þ8H) with a selectivity of up to 35e40% using ruthenium black commercial catalyst. It can be seen from Fig. 12 and Table 4 that this intermediate seems very favourable to be created due to large negative total energy difference that is almost as low as the energy of the fully hydrogenated stereoisomers of 9-ethyl-perhydrocarbazole, in fact the calculations suggest that some of the (þ8H) compounds are more stable than the (þ10H) ones. Moreover, the structures having the lowest total energy difference agreed well with the structures assigned based on NMR experiments. The theoretical calculations together with NMR analysis of the reaction intermediates both showed clearly that several different stereoisomers of intermediates are possible to arise in this hydrogenation reaction, therefore the reaction pathway is expected to be more complicated than the simplified version as shown in Fig. 7. Thus, different parallel reactions together with the isomerisation steps are likely to take place which can in turn influence the final stereochemistry of the fully loaded products. It is important to underline that the calculations do not take into consideration the effects of the catalyst in the reaction or the cohesive energy of the molecules in the melt. Nonetheless, the general trends are in agreement with 11620 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 5 ( 2 0 1 0 ) 1 1 6 0 9 e1 1 6 2 1 experiment, a further and more complete theoretical study of these systems is underway and will be reported soon. 4. Conclusions The conversion and selectivity of the catalytic hydrogenation reaction of molten 9-ethylcarbazole was found to be critically dependent on the type of metal and support used. A very active catalyst, 5 nominal wt% Ru on alumina, was synthesized by means of mild chemical reduction and has shown to give better performance than its commercially available counterparts with similar composition, mainly because of its much higher atom efficiency. Most of the chemical structures of the partially stable intermediates and final fully hydrogenated products predicted by DFT calculations were identified in the reaction mixtures using NMR techniques. The wide range of structures of the intermediates found using combined methods of NMR and theoretical calculations suggest that the catalytic reaction is not taking place as a simple consecutive hydrogenation. Apart from the identification of two symmetric hydrogenation products from possible AeD structures (þ12H) the observation of asymmetrically hydrogenated compounds E and F as products under the chosen reaction conditions is rather surprising with regard to the expected concerted attack of two protons from one and the same side of a molecule to a double bond. Their formations may likely involve intermediate isomerisation steps on different metal surface sites, or several parallel reactions which give different intermediates and products. This can give implications on the reversibility of the hydrogenation process, taking into consideration the variety of stereoisomers of the intermediates and products possible. The geometry of the intermediates will have an influence on the compound’s adsorption mode on the metal’s catalytically active sites. To derive the true reaction mechanism, it is extremely important to assign the structures of all of the intermediates and products. The next step will be to quantify their concentration and derive the reaction pathway depending on the type of catalyst used, which is currently under study and will be reported soon. In summary, the theoretical hydrogen uptake of 9-ethylcarbazole in molten state was achieved using our best synthesized catalyst with favourable kinetics in spite of a range of fast elementary surface reactions envisaged, giving various intermediates. The measured activation energy is lower than those reported in solvent, indicative of a small degree of mass transfer limitations. However, the catalytic reaction achieves the theoretical hydrogen uptake value at fast kinetics without encountering insurmountable diffusion problems or severe deactivation in the melt. This demonstrates that 9-ethylcarbazole could be a promising candidate for hydrogen storage. Acknowledgements We would like to kindly acknowledge Mr Fernando Eblagon and Dr William Oduro who made a major insightful contribution to this work. Dr K.M. Kerry Yu is acknowledged for the technical support of this work. We would also like to acknowledge Prof Heyong He and Shanghai Key Laboratory of Molecular Catalysis and Innovative Materials at Fudan University, China for TEM images; Dr Sam French of Johnson Matthey, Billingham, UK for fruitful discussion. Finally, we would like to thank University of Oxford, and STFC for financial support, computing resources provided by the U.K. e-science Centre, STFC (SCARF). Appendix. Supplementary data Supplementary data associated with this article can be found in the online version, at doi:10.1016/j.ijhydene.2010.03.068. references [1] Hiyoshi N, Miura R, Rode ChV, Sato O, Shirai M. Enhanced selectivity to decalin in naphthalene hydrogenation under supercritical carbon dioxide. Chem Lett 2005;34: 424e5. [2] Crabtree RH. Hydrogen storage in liquid organic heterocycles. Energy Environ Sci 2008;1:134e7. [3] Kariya N, Fukuoka A, Utagawa T, Sakuramoto M, Goto Y, Ichikawa M. Efficient hydrogen production using cyclohexane and decalin by pulse spray mode reactor with Pt catalyst. Appl Catal A 2003;247:247e59. [4] Moores A, Poyatos M, Yi L, Crabtree RH. Catalysed low temperature H2 release from nitrogen heterocycles. New J Chem 2006;30:1675e8. [5] Coughlan B, Keane MA. The catalytic hydrogenation of cyclohexane and methylcyclohexane over nickel loaded Y zeolites. Catal Lett 1990;5:89e100. [6] http://www.geifuelcells.com/innovation/ht-pem. [7] Pez GP, Scott AR, Cooper AC, Cheng H. US Pat, 7,101,530; 2006. [8] Pez GP, Scott AR, Cooper AC, Cheng H, Wilhel FC, Abdourazak A. US Pat 7,351,395; 2008. [9] Cooper A, Pez GP, Abdourazak A, Scott A, Fowler D, Wilhelm F, et al. Annual progress report DOE hydrogen program 2006, http://www.hydrogen.energy.gov/annual_ progress06.html; 2006. [10] Yang J, Sudik A, Wolverton Ch, Siegel DJ. High capacity hydrogen storage materials: attributes for automotive applications and techniques for materials discovery. Chem Soc Rev; 2010. [11] Clot E, Eisenstein O, Crabtree RH. Computational structureeactivity relationships in H2 storage: how placement of N atoms affects release temperatures in organic liquid storage materials. Chem Commun 2007;22: 2231e3. [12] Cooper H, Donnis BBL. Aromatic saturation of distillates: an overview. Appl Catal A 1996;137:203e23. [13] Sotoodeh F, Zhao L, Smith KJ. Kinetics of H2 recovery from dodecahydro-N-ethylcarbazole over a supported Pd catalyst. Appl Catal A 2009;362:155e62. [14] Sotoodeh F, Smith KJ. Kinetics of hydrogen uptake and release from heteroaromatic compounds for hydrogen storage. Ind Eng Chem Res 2010;49:1018e26. [15] Cooper AC, Campbell KM, Pez G. Conference material PWHEC 2006 Lyon France http://www.cder.dz/A2H2/Medias/ Download/Proc%20PDF/PARALLEL%20SESSIONS/%5BS15% 5D%20Advanced%20Materials%20Storage/15-06-06/ 461.pdf. i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 5 ( 2 0 1 0 ) 1 1 6 0 9 e1 1 6 2 1 [16] Wang Z, Tonks I, Belli J, Jensen CM. Dehydrogenation of Nethyl perhydrocarbazole catalyzed by PCP pincer iridium complexes: evaluation of a homogenous hydrogen storage system. J Organomet Chem 2009;694:2854e7. [17] Gupta M, Hagen C, Kaska WC, Cramer RE, Jensen CM. Catalytic dehydrogenation of cycloalkanes to arenes by a dihydrido iridium PeCeP pincer. J Am Chem Soc 1997;119:840e1. [18] Ho TC. Deep HDS of diesel fuel: chemistry and catalysis. Catal Today 2004;98:3e18. [19] Bird CV. Heteroaromaticity: a unified aromaticity index. Tetrahedron 1992;48:335e40. [20] Ponomarev AA, Chegolya AS, Dyukareva VN. Liquid phase hydrogenation of some heterocyclic nitrogen compounds over ruthenium catalysts. Chem Heterocycl 1966;2:239e42. [21] Toppinen S, Rantakyla TK, Salmi T, Aittamaa J. Kinetics of the liquid phase hydrogenation of di- and trisubstituted alkylbenzenes over a nickel catalyst. Ind Eng Chem Res 1996; 35:4424e33. [22] Liu Y, Chen J, Zhang J. Effects of the supports on activity of supported nickel catalysts for hydrogenation of mdinitrobenzene to m-phenylenediamine. J Chem Eng 2007;15: 63e7. [23] László H, Tabor M. Hydrogenation of pyrrole derivatives: part V. Poisoning effect of nitrogen on precious metal on carbon catalyst. Appl Catal A 2002;226:319e22. 11621 [24] Zhen G, Hue F, Heng-chang M, Qiao-Xiang K, ZhiWang Y. Hydrogenation of five-membered heterocycles of polymer-supported palladium catalyst at normal temperature and pressure. J Polym Adv Technol 2004;15: 100e4. [25] Rautanen PA, Aittamaa JR, Outi A, Krause I. Solvent effect in liquid-phase hydrogenation of toluene. Ind Eng Chem Res 2000;39:4032e9. [26] Friedrichs O, Borgschulte A, Zuettel A. Controlled mechanically activated hydrogen sorption. Int J Hydrogen Energy 2008;33:5606e10. [27] Gaussian 03, revision C.02. Wallingford, CT: Gaussian, Inc.; 2004. [28] Salter C, Foresman JB. Napthalene and azulene I: semimicro bomb calorimetry and quantum mechanical calculations. J Chem Educ 1998;75:1341e5. [29] Foresman JB, Frisch A. Exploring chemistry with electronic structure methods. 2nd ed. Gaussian, Inc.; 1998. [30] Koch W, Holthausen MC. A chemist’s guide to density functional theory. Wenheim: Wiley-VCH; 2000. [31] Shinohara C, Kawakami S, Moriga T, Hayashi H, Hodoshima S, Saito Y, et al. Local structure around platinum in Pt/C catalysts employed for liquid-phase dehydrogenation of decalin in the liquid-film state under reactive distillation conditions. Appl Catal A 2004;266:251e5.