Hairy foam: carbon nanofibers on solid foam as catalyst support : synthesis, mass transfer, and reactor modeling

The chemical reactor is at the heart of many chemical processes. The chemical industry strives for the most efficient, most compact, and safest chemical reactor. The efficiency of a chemical reactor is determined by the delicate balance of catalyst performance (i.e. selectivity and activity) and the mass transfer and hydrodynamic characteristics of the chemical reactor. The mass transfer and hydrodynamic characteristics can be optimized by the use of suitable reactor packings. The classical way employs so-called random packings (e.g. porous spheres, Raschig rings, Berl Saddles) which are dumped into the reactor. However, the use of these random packings poses challenges in terms of pressure drop and fluid maldistribution. The fluid maldistribution results in a non-uniform access of reactants to the catalyst reducing the effectiveness of the process. Also, fluid maldistribution can result in the formation of hot spots resulting in an even further reduction of process efficiency. Structured packings are promising packings which exhibit a low pressure drop and also reduce fluid maldistribution. Another benefit is (in general) the lower solids holdup of these structured packings while maintaining a high external surface to volume ratio in order to accommodate high mass transfer rates towards the catalyst. In this thesis the use of solid foam packings as a catalyst support for gas-liquid-solid reactions is discussed. Solid foams are highly porous, open celled materials consisting of a reticulated network of struts. The cell size, expressed in pores per linear inch (PPI), ranges from 5 to 100 PPI, with porosities ranging from 80% to 97%. Solid foams are available in a wide range of materials (e.g. ceramics, metals, plastics and carbon). The focus in this study is on aluminum foam and carbon foam, also called reticulated vitreous carbon. These solid foam materials exhibit high gas-liquid mass transfer rates due to their high geometric specific surface area, values of kglagl up to 0.6 m3l m-3 r s-1 have been reported. However, the absence of micropores results in a low surface area for catalyst deposition. In this study the surface area has been increased by the growth of carbon nanofibers (CNFs) on the surface of the carbon foam, called Hairy Foam. The combination of solid foam and CNFs resembles the visual inverse of a porous packed bed. Instead of depositing the catalyst in the pores of the porous particle, it can now be deposited on the outside of the CNFs. The open structure of the CNFs reduces diffusion limitation and allows for an intimate contact between the liquid flowing along the fibers and the catalyst deposited on the fibers. Hairy Foam has been produced by growing carbon nanofibers on 20 PPI, 45 PPI, and 60 PPI reticulated vitreous carbon foam (Chapters 2 & 3). The carbon nanofibers have been synthesized by the nickel catalyzed decomposition of ethylene. The effect of nickel loading on fiber diameter and morphology, carbon nanofiber coverage, and fiber layer thickness has been studied using 60 PPI reticulated vitreous carbon foam (Chapter 2). The foam is completely covered with CNFs using nickel deposition times = 4 hrs and a nickel concentration of 2.5 wt% (gNi g-1 RVC). The surface area increased from 0.12 m2s g-1s for reticulated vitreous carbon foam to 146 m2s g-1s for Hairy Foam. CNFs of 30 to 1100 nm in diameter have been observed. The CNF layer thickness ranges from 13 µm to 25 µm. Model calculations in the case of a the first-order palladium catalyzed gas-liquid-solid oxidation of glucose show that application of the Hairy Foam catalyst support will lead to mass transfer limited reaction rates at surface areas larger than 0.3 m2s g-1s . The liquid-solid mass transfer characteristics of Hairy Foam have been studied using the palladium catalyzed liquid phase oxidation of sodium formate (Chapter 3). CNFs have been grown on 20 PPI, 45 PPI, and 60 PPI reticulated vitreous carbon foam at temperatures of 773 K, 873 K, and 973 K. The effects of liquid velocity, u l = 4.7 · 10-3 to 77 · 10-3 m3l m-2 r s-1, and reaction temperature, T = 323 K and 343 K, have been studied. For all cases both liquid-solid mass transfer and chemical kinetics contribute to the overall reaction rate coefficient. For 45 PPI and 60 PPI the overall reaction rate coefficient is approximately the same as the liquid-solid mass transfer coefficient of random packings, higher than the liquidsolid mass transfer coefficient for regular solid foams, ranging from 0.075 to 0.36 m3l m-3rs-1. The overall rate coefficient for 20 PPI Hairy Foam is significantly larger than for regular solid foam packings and random packings, ranging from 0.15 to 0.90 m 3l m-3r s-1 for liquidvelocities of 4.7 · 10-3 and 64 · 10-3 m3lm-2r s-1, respectively. The overall reaction ratecoefficient is approximately the same for CNFs grown at 773 K and 873 K, and higher for CNFs grown at 973 K. The pressure drop of the Hairy Foam was within the experimental error of the pressure transducer and is at least below 4000 Pa m-1, which is much lower than for random packings, 5.2 · 103 to 1.2 · 106 Pa m-1. This shows that higher or at least equivalent mass transfer rates can be obtained, but at a considerably lower pressure drop than for random packings. The hydrodynamic accessibility of the CNFs supported on solid foams has been studied using 2-dimensional simulations (Chapter 4). The effect of solid foam PPI number, liquid velocity, the CNF layer porosity, permeability, and layer thickness on the average velocity of the liquid in the CNF layer was studied. The simulations show that the liquid velocity in the CNF layer is considerably lower than the velocity of the liquid flowing past the solid foam structure, maximum 10% of the total average liquid velocity. A model was derived on basis of Darcy’s law, accurately describing the average velocity of the liquid in the CNF layer as a function of pressure drop, foam strut diameter, CNF layer permeability, and CNF layer thickness, for Darcy numbers below 0.1 (Da =permeability/(layer thickness) 2). Only for CNF layers with a permeability larger than 10-10 m2, an enhancement in liquid-solid mass transfer due to the hydrodynamic accessibility of the CNFs is calculated. For CNF layers with a lower permeability the liquid-solid mass transfer rate is significantly larger than the flow entering the CNF layer. The liquid-solid mass transfer characteristics have been studied for cocurrent upflow operated gas-liquid aluminum foam packings (Chapter 5). The effect of aluminum foam PPI number (10 PPI, 20 PPI, and 40 PPI), gas velocity (u g = 0.1-0.8 m3g m-2r s-1), and liquid velocity (ul = 0.02 and 0.04 m3lm-2r s-1) on the liquid-solid mass transfer coefficient was studied using the Bi/Pd catalyzed oxidation of glucose. For 10 and 20 PPI the volumetric liquid-solid mass transfer coefficient is approximately the same, ranging from 2 · 10 -2 to 9 · 10-2 m3lm-3r s-1. For 40 PPI lower values for the volumetric liquid-solid mass transfer coefficient were found, ranging from 6 · 10-3 to 4 · 10-2 m3l m-3 r s-1. Both the volumetric and the intrinsic liquid-solid mass transfer coefficient show a minimum as a function of gas-velocity, indicating that multiple steps for the liquid-solid mass transfer take place. The intrinsic liquid-solid mass transfer coefficient was found to be proportional to u 0.98 l and ranges from 5.5·10-6 to 8·10-4 m3l m-2 i s-1, which is in the same range as found for random packings and corrugated sheet packings. Structured packings can be used for gas-liquid operations under dispersed flow. The gas bubbles move through the liquid, leaving a thin liquid film on the surface of the packing when a gas bubble passes by. The mass transfer through these thin liquid films can be modeled using e.g. Higbie’s penetration theory. Typically these models incorporate the contact time between the two phases amongst which mass transfer takes place. However, due to the geometry of the gas bubbles there is not one specific contact time but a distribution of contact times. (Chapter 6) presents a model for the gas-liquid mass transfer through these thin liquid films present on the packing when gas-bubbles passes, incorporating the distribution of contact times. The model has been derived for two cases: the absorption/desorption of a gaseous component into the liquid film and the transfer of the gaseous component through the liquid film to the packing surface where an infinitely fast reaction takes place. The cases have been solved for three gas-bubble geometries: rectangular, cylindrical, and spherical. For Fourier numbers below 0.3 the derived models correspond to Higbie’s penetration theory. The Sherwood numbers for cylindrical and spherical bubbles are 20% and 35% higher, pectively,than for rectangular bubbles. In case of absorption and Fourier numbers exceeding 3, the effect of bubble geometry becomes more pronounced. For Fourier numbers exceeding 10, the Sherwood numbers for cylindrical and spherical bubbles are 55% and 100% higher, respectively, than for rectangular bubbles. For the case of an infinitely fast reaction at the packing surface and Fourier numbers larger than 3, the Sherwood numbers correspond to Whitman’s film theory (Sh = 1). The solutions of the derived models consist of infinite series, these series have been approximated with engineering correlations, which give the Sherwood number as a function of gas bubble diameter, gas bubble rise velocity, liquid layer thickness, and diffusion coefficient. The engineering correlations describe all the model data within 4% accuracy. The model has been demonstrated for the three examples (1) gas-liquid mass transfer for structured packings operated in dispersed flow; (2) gas-liquid mass transfer in a microchannel operated with annular flow; (3) gas-liquid mass transfer in a microchannel with Taylor flow. The performance of Hairy Foam and washcoated solid foam is compared with that of a packed bed of porous particles using a modelling study (Chapter 7). Gas-liquid mass transfer, liquid-solid mass transfer, and diffusion with simultaneous reaction have been incorporated in the modelling. The comparison of the packings was made on the basis of two gas-liquid operated, heterogeneously catalyzed hydrogenation reactions (viz. the slow hydrogenation of cinnamaldehyde and the fast hydrogenation of 3-methyl-1-pentyn-3-ol) and two flow conditions (viz. cocurrent upflow and cocurrent downflow of gas and liquid). For the studied conditions the simulation results show that for fast reactions the Hairy Foam is an attractive alternative for the conventional packed bed of (porous) spherical particles, due to the high conversion and selectivity and the low total pressure drop. In the case of the slow reaction, the pressure drop for the Hairy and solid foam packings is, however, significantly larger than for the bed of particles. This is due to the lower solids holdup, and thus lower catalyst concentration, of the Hairy Foam and solid foam packings. Therefore, larger reactors are required for the Hairy Foam and solid foam packings, resulting in a higher total pressure drop.