Search

Increasing the stream of recycled plastic necessitates an approach beyond the traditional recycling via melting and re‐extrusion. Various chemical recycling processes have great potential to enhance recycling rates. In this Review, a summary of the various chemical recycling routes and assessment via life‐cycle analysis is complemented by an extensive list of processes developed by companies active in chemical recycling. We show that each of the currently available processes is applicable for specific plastic waste streams. Thus, only a combination of different technologies can address the plastic waste problem. Research should focus on more realistic, more contaminated and mixed waste streams, while collection and sorting infrastructure will need to be improved, that is, by stricter regulation. This Review aims to inspire both science and innovation for the production of higher value and quality products from plastic recycling suitable for reuse or valorization to create the necessary economic and environmental push for a circular economy., Plastic fantastic: Plastic can rise again and again as a new product. Researchers now know methods with which new plastics can be produced from 100 % recycled material, products that can even be used for food applications. This development is possible thanks to chemical recycling, through which polymer chains are first broken to then be reformed into new molecules, such as plastics but also other chemicals.

Oscillating gas bubbles can be created in a liquid by exposing it to ultrasound. These gas bubbles implode if the sound pressure is high enough. This process is called cavitation. Interesting phenomena take place during the collapse. The gas and vapour inside the bubble are compressed and reach temperatures of several thousand Kelvin and pressures of several hundred bars, a dense plasma is formed inside the bubble, light is emitted, reactive radical molecules are formed, and there is a liquid flow around the bubble which can be utilised for mixing or even scission of polymers by strain rates. An eroding jet is formed if the bubble collapses near a wall. The pressure and temperature are at ambient conditions around the bubble during this process. There is a plethora of applications which can be operated or intensified by cavitation, e.g. micro mixing, catalyst surface renewal and material synthesis, treatment of kidney stones, waste water treatment and other radical-induced chemistry such as polymerisation, and polymer weight distribution control. In this thesis, several forms of cavitation have been investigated. Special attention is paid to the influence of gas and vapour content in the cavitation bubble. The gas and vapour content of the bubble play a crucial role in the extent of the effects of cavitation. First and foremost the thermal properties adiabatic index and heat conductivity determine the maximum temperature during the collapse. Gases with a high adiabatic index, such as noble gases, and liquids with a low vapour pressure and a high adiabatic index, such as water or sulfuric acid, yield the highest hot spot temperatures and therefore the most intense effects. Some gases such as oxygen also participate in chemical reactions. Another difference between cavitation effects of several gases may be caused by changing gas solubility in the liquid. This changes the concentration gradient around the bubble, and the mass transport to and from the bubble. The effects of the gas on the cavitation process depend on the ultrasound frequency. At frequencies above 20 kHz air has a higher efficiency of radical formation than argon as a saturation gas. This has been measured by following the oxidation of potassium iodide to iodine spectroscopically. The carbon dioxide in the air contributes to this increase at low ultrasound input power. This is surprising since carbon dioxide addition in the cavitation bubble gas phase likely suppresses the hot spot temperature. The enhancement of radical production by carbon dioxide only occurs when it is present in low quantities (

Binding of multivalent ions is of major interest for many fields of water purification technology, ranging from household applications to drinking water processes. Current technology is based on the use of ion exchange resins, which have the major drawback that a chemical regeneration step is necessary during the process, leading to environmentally undesired waste streams. In this thesis, two novel polymer micellar systems are described that have the special ability to reversibly bind multivalent ions, thereby avoiding the need of a chemical regeneration step. Both systems contain thermo-responsive polymer blocks and anionic polymer end-groups, which give these systems the ability to bind and release multivalent ions as a function of temperature. The first system consists of carboxylic acid end-standing triblock Pluronic molecules. At temperatures below the critical micellization temperature (CMT) the Pluronic molecule and multivalent cations are present in solution as unimers or "free molecules". The affinity of isolated anionic groups for multivalent cations is very low. Increasing the temperature above the CMT brings the anionic groups present on different surfactant chains together, thus allowing for the sequestering of the multivalent cations. This binding has been described and analyzed using isothermal titration calorimetry and a conductometric analysis method. The second surfactant system is constituted of a hydrophobic polystyrene (PS) block and a hydrophilic block. Two different hydrophilic monomers were used to generate the hydrophilic block: poly(N-isopropylacrylamide) (PNIPAM) and poly(di-ethyleneglycolacrylate) (PDEGA). The synthesis of this surfactant system involves the Atom Transfer Radical Polymerization of the different monomers. In order to bind multivalent ions, the synthesized surfactants are provided with a negatively charged functional end group. Results show that the synthesis of the surfactant of PS and PDEGA was most successful. The aggregation behavior of the anionic PS/PDEGA block copolymers in water is evaluated by construction of a Zimm plot, by light scattering and cryoTEM measurements. With respect to size and aggregation behavior, the results of these methods endorse each other. The thermo-sensitive properties of the block copolymers in water are investigated using UV-Vis spectroscopy. The block copolymer has a Lower Critical Solution Temperature (LCST), which value depends on the block length and the concentration in water. Micelles consisting of the PS/PDEGA surfactants form negatively charged surfaces as a result of the carboxylate end-groups of the surfactant molecules. Therefore, they have the ability to bind positively charged ions. Because of the LCST of PDEGA, there is a sudden change in the solubility of the micelles when the temperature is increased. As a result, the dielectric permittivity of the water around the micelle changes, which leads to an apparent pKa-shift of the carboxylate groups. Results of dialysis experiments show that the surfactant binds and releases calcium ions reversibly by altering the temperature of the water to values below and above the LCST. The hydrophobic-induced pKa-shift of the PS/PDEGA micelles is confirmed by a Self Consistent Field analysis. In this analysis, the properties of the above described polymer micelles is investigated as a function of the solvent quality for the corona chains. The use of PS/PDEGA micelles is probably most feasible in water softening applications, since the micelles can bind hardness ions at room temperature. The available water streams at an elevated temperature can be used to regenerate the water softener. Possible modifications to make this novel technique commercially more attractive involves an increase in charge density of the anionic active groups and immobilization of the thermo-reversible system on an inert, insoluble carrier. Considering the worldwide trend towards environmental friendly applications and the growing interest of thermo-reversible polymers, the long-term perspective of anionic thermo-reversible polymers with a hydrophobic pKa-shift seems promising.

The increased awareness for environmental issues as well as for public health, demands "greener" processes as well as products. In the polymer industry, one of the major concerns relate to residual monomer in polymer (latex) products. Low molecular species are toxic. When monomer units diffuse out of their polymeric matrix, they contaminate the surrounding area. Decreasing the residual monomer level in polymer (latex) products is especially of great importance in food packaging, toys for children, etc. Conventional techniques to reduce residual monomer in polymer (latex) products are energy intensive, time consuming and not able to fulfill the increasingly stringent requirements for further reduction of residual monomer in polymer products, i.e.

The objective of the work reported in this thesis is to evaluate the potential of microstructured systems for the production of µm-sized polymer particles. For this purpose, we investigated a method of emulsifying monomer followed by a controlled suspension polymerization process in which the droplet size distribution can be maintained. Additionally, the option of emulsifying dissolved polymer to exclude the complexity of a suspension polymerization process was also taken into consideration. For emulsification experiments, we selected straight-through microchannel (MC) arrays as they show a promising performance in terms of both monodispersity of the emulsions produced and potential for scale-up. The operating window for emulsification of monomer with straight-through MC arrays is reported in Chapter 2. Coefficients of variation (CVs) below 5% have been obtained with hydrophobic monomers such as styrene or n-butyl methacrylate or mixtures of hydrophobic solvents such as toluene and hydrophobic monomers. Mixtures of hydrophobic and more hydrophilic monomers such as methyl methacrylate result in broader droplet size distributions. The emulsification process is quite stable and variations in the cross-flow velocity and applied pressure up to a certain blow-up point do not significantly affect the final droplet size distribution. Chapter 3 describes the suspension polymerization for styrene-in-water emulsions prepared with straight-through MC arrays. The main focus is on maintaining the size distribution of the original monomer droplets during polymerization. It has been shown that significant broadening of the droplet size distribution occurs already on time scales of the emulsification process. In addition, secondary nucleation of submicron particles occurs. The molecular weight distribution of these secondary nucleated particles points to emulsion polymerization as the dominating mechanism. The initiator concentration as well as the monomer volume fraction in the reaction mixture have a strong influence on secondary particle nucleation. Even the addition of a water soluble inhibitor such as NaNO2 can only partly suppress secondary nucleation. The potential for scale-up of styrene emulsification with straight-through MC arrays with subsequent suspension polymerization is evaluated in Chapter 4. At this point, feasibility is only expected for a small scale process, such as an annual production of 103 kg of monodisperse polystyrene microparticles. A preliminary economic evaluation points to a required selling price between 847 and 540 €/kg polystyrene microparticles. The required selling price is mainly influenced by the pore activity and the applied pressure. The influence of fouling on emulsification performance has also been investigated. Fouling occurs in a time frame of a few days and causes a strong deterioration in the droplet size distribution of the final emulsion. The rate of deterioration is stronger for sodium dodecyl sulfate (SDS) than for sodium dodecyl benzene sulfonate (SDBS), indicating limitations in the choice of the surfactant system. Direct emulsification of a polymer solution or melt can avoid complex issues such as secondary particle formation during suspension polymerization. The viscosity reduction required for emulsification of polymer is usually achieved by dissolving the polymer in an organic solvent. In Chapter 5, the plasticizing effect of CO2 is investigated as an alternative for organic solvents. In a setup with a system pressure of up to 100 bar and a constant pressure difference over the MC module in the order of 0.1 bar, emulsification of CO2-plasticized polystyrene is only possible for oligomers. The combined knowledge on emulsification with straight-through MC arrays is applied in Chapter 6 to produce monodisperse perfluorohexane emulsions for a completely different area of science, i.e. quantitative targeted ultrasound contrast imaging. For this purpose, strictly monodisperse echogenic perfluorohexane emulsions with a preset droplet size are required, since too small droplets will not generate sufficient ultrasound reflection and droplets which are too large will not provide sufficient coverage of specific ligands foreffective adhesion under physiological shear stress conditions. Standard perfluorocarbon emulsions do not satisfy both requirements, i.e. adhesion and echogenicity. Ultrasound experiments demonstrate that the perfluorohexane emulsions produced with straightthrough MC arrays clearly enhance echogenicity. Surface functionalization of the perfluorohexane droplets with a biotinylated fluoro-surfactant enables them to bind avidine-coated SiO2 particles. So we demonstrated that these emulsions can provide novel insights into the development and early detection of important vascular diseases, such as atherosclerosis or, more general, in the field of selective targeting. In retrospect, the author of this thesis is convinced that emulsification of monomer with straight-through MC arrays is a promising technique for the production of small-scale specialty microparticles. This potential is not limited to monodispersity, it also comprises particles with advanced functionality and morphology. Understanding and controlling all of the mechanisms during polymerization will be a key concern for future developments.

Externally triggered on-demand drug release from an implant can significantly improve the efficiency of the drug therapy since it enables the patient or physician to control the dosing to the patient’s needs and releases the drug only at the required location in the human body. Therefore, patient compliance and efficacy will increase and toxic side effects decrease as untargeted locations are not exposed to significant drug levels as is often the case in systemic drug administration. In this work, the externally triggered drug delivery system is a thermoresponsive polymeric implant triggered using an alternating magnetic field. The thermal switch is based on a significant change in diffusivity of a solute around the glass transition temperature (Tg) of a polymer. At a temperature below the glass transition temperature of the polymer (TTg), the polymer becomes rubbery. This significantly increases the flexibility and free volume of the polymer resulting in release of the active. Since the glass transition is a reversible transition, subsequent lowering of the temperature significantly decreases the drug release rate from the implant, enabling pulsatile drug administration. The temperature of the implant is increased using an externally applied alternating magnetic field. In order to increase the temperature of the implant using a magnetic field, the use of superparamagnetic iron oxide nanoparticles (SPION) is explored. These nanoparticles are used as MRI contrast agents and to locally increase the temperature in hyperthermia treatment, the destruction of tumors by elevated temperature. The particles have no remanent magnetization, are biocompatible and are able to generate thermal energy using an alternating magnetic field because of N´eel and Brown relaxation. N´eel relaxation is the reorientation of the magnetic moment within the particles, generating thermal energy by crossing an anisotropy barrier, and Brown relaxation the reorientation of the magnetic particles itself, generating thermal energy by viscous friction with the carrier fluid. Since the nanoparticles are used for heating a polymer implant, different preparation methods for an iron oxide - polymer nanocomposite have been investigated. Freeze drying a mixture of a ferrofluid with a poly(methyl methacrylate) (p(MMA)) latex and subsequent compounding, results in an optimal distribution of the particles. It is expected that the particles do not agglomerate because of the combination of stabilization of both the iron oxide particles and polymer latex by surfactants, and the lack of mobility during freeze drying. Other methods used, e.g. solvent casting and direct injection of the ferrofluid into the compounder, result in significant agglomeration of the particles. Subsequently, the particle distribution has been shown to have a significant effect on the heating of the particle. An optimal distribution of the particles results in the highest specific absorption rate (SAR), the amount of thermal energy generated per gram of iron oxide, because of a minimum in interparticle interactions. Since the nanoparticles incorporated in a polymer are immobilized, the particles are not able to generate thermal energy by Brown relaxation. By a direct comparison of the specific absorption rate of particles suspended in liquid and incorporated in p(MMA) using the optimal freeze drying method, the contribution of both N´eel and Brown relaxation to the heating of SPION has been investigated. Since the observed SAR is identical in both situations, it is concluded that at the frequency used (745 kHz), N´eel relaxation is the only relaxation process that contributes to the heating of the particles in ferrofluid, because of the significantly shorter relaxation time for N´eel relaxation. Using a cylindrical core of iron oxide - p(MMA) nanocomposite, coated with a thermoresponsive poly(butyl methacrylate-stat -methyl methacrylate) (p(BMAMMA)) layer, externally triggered on-demand drug release has been investigated. A model drug, ibuprofen, has been incorporated in the thermoresponsive p(BMAMMA) coating. Upon exposure of the sample to an alternating magnetic field (on situation), the drug release rate is significantly increased compared to the release rate without the magnetic field (off situation). After the magnetic field is removed, the release rate decreases back to the rate prior to the exposure, demonstrating the reversibility of the system. Multiple consecutive exposures to the external trigger result in similar increases of the release rate. Increasing the iron oxide concentration in the core of the device increases the release rate upon exposure, whereas the release rate without exposure is not influenced, therefore increasing the on/off ratio, because of a higher temperature increase upon exposure. Even though externally triggered pulsatile drug release has been shown, the maximum on/off ratio obtained is only 16.5. This relatively low ratio is primarily due to the suboptimal nature of the used commercially available iron oxide and the relative high off release rate of ibuprofen from p(BMA-MMA). In order to increase the on/off ratio, a cylindrical iron core has been used, coated with an ibuprofen incorporated poly(styrene-stat -butyl methacrylate) (p(S-BMA)) layer. The heat generated in the iron core upon exposure to the magnetic field is due to induction heating. Externally triggered pulsatile drug release has been shown using this concept with on/off ratios exceeding 2000, where both the on/off ratio and the release rate are affected by the concentration of ibuprofen. Generally, decreasing the base temperature of the release experiments from 37 to 25 ¿C significantly increases the on/off ratio. The effect of the orientation of the cylindrical iron rod with respect to the direction of the magnetic field on the heating of the device has been investigated using a Comsol model. Even though the effect of orientation is limited at small angles, a significantly lower surface temperature has been shown for larger angles, up to ~20 ¿C. This can result in a several orders of magnitude difference for the diffusion coefficient of ibuprofen in the polymer. Subsequently, the requirement of alignment between the sample and the magnetic field has been circumvented by the use of a macroscopic spherical iron core, coated with ibuprofen incorporated p(S-BMA). The absence of an alignment effect has been shown using 1 sample and 2 samples in line with the magnetic field, as this does not influence the release rate and on/off ratio, normalized to the surface area available for release in on-demand release experiments. Therefore, it is possible to use multiple samples to increase the attainable drug dose. Increasing the size of the spherical iron core and, therefore, decreasing the polymer thickness, only increases the release rate upon exposure, resulting in higher on/off ratios. In the case of a thinner polymer layer, the distance between the heating core and the outer surface of the polymer is smaller, resulting in a higher temperature of the outer layer. The solubility of a solute in a polymer is predominantly important for the release characteristics of that solute from the polymer. Therefore, the solubility of ibuprofen in p(S-BMA) has been investigated. Even though samples of p(S-BMA) with an ibuprofen concentration above 31 wt% show a clear phase separation, indicating maximum solubility at 31 wt% ibuprofen, measurement of the glass transition temperature of composites show that the system of p(S-BMA) with ibuprofen concentrations below 31 wt% is in a meta-stable state. In conclusion, repetitive on-demand drug release from a polymeric implant can be externally triggered using an alternating magnetic field. Due to their biocompatibility and the absence of an alignment effect, superparamagnetic iron oxide nanoparticles are preferable for the required heat generation. However, more optimal nanoparticles are required for high on/off ratios, as has been shown using another material for heat generation.

By non-invasive external triggering of drug release from an implant, pulsewise administration can be realized according to the patient’s needs and at specific locations in the human body. In comparison to more traditional delivery forms (e.g. oral or by injection), externally triggered drug release potentially offers the advantage of increased patient compliance due to easy and painless dosing, higher bioavailability, and lower doses of potent drugs, thereby reducing the toxic side effects commonly associated with systemic administration. The main focus in this thesis is on externally triggered on-demand release of relatively small drug molecules using the large change in diffusivity occurring near the glass transition temperature (Tg) of a polymer matrix. When the polymer is in the glassy state (T Tg) the high amount of free volume and flexibility of the polymer chains allows diffusion of the drug out of the implant. The rate of diffusion greatly depends on the size, shape and flexibility of the drug, as well as its relation to the flexibility of the polymer chain. To a certain extent, these phenomena can be described by the free volume theory (FVT). However, the FVT has mainly been applied to describe diffusion of relatively small solutes in rubbery polymer-solute systems, whereas only a limited amount of research has dealt with its application to large solutes, such as drug molecules. Chapter 5 demonstrates that to use the FVT for diffusion of larger molecules approaching the size of polymer jumping units, a better understanding of polymer/drug dynamics is required.Non-invasive thermal triggering of drug release is achieved using near-infrared radiation (NIR), bordering on visible red light in the electromagnetic spectrum. The low absorption of water and tissue chromophores in this region allows for a relatively high penetration of NIR into the human body of up to several centimeters, making it attractive for biomedical applications. Since most polymers do not intrinsically absorb NIR, an NIR-absorbing compound is required. Here, a quaterrylenebis(dicarboximide) derivative, exhibiting a maximum in absorption around 780 nm, is incorporated into polymer matrices by conventional polymer processing techniques without affecting the Tg. UV-Visible spectroscopy reveals the formation of face-to-face stacked molecular aggregates on incorporation of this dye by solvent casting, which can be suppressed by incorporation or post-processing at elevated temperature. Although dye aggregation affects the absorption spectrum, the absorption at the wavelength of the NIR-source can effectively be controlled by increasing the dye concentration. Using a 785 nm NIR-laser, initial heating rates of over 3ºC s-1 and steady state temperature increases as high as 80ºC are reached. To investigate the effect of NIR-dye distribution on heat generation inside the polymer matrix in aqueous environment, a finite element (FE) model for NIR-induced heating is developed. Since the main resistance to mass transfer is located at the polymer/water interface, heat generation should mainly be located at the surface of the implant. Coating the implant with a dye containing layer therefore yields the best combination of high surface temperatures without exposing the inside of the implant to excessive temperatures. The FE-model is subsequently used to estimate the temperatures that can be achieved in vivo using NIR, taking into account scattering and absorption by skin and subcutaneous tissue. Compared to aqueous environment (in vitro), higher temperatures are induced in vivo due to a lower amount of heat dissipation. By using an insulating layer, the tissue can be protected from high temperatures, thus preventing tissue damage from occurring and enabling NIR-induced drug release. Experimentally, both in vitro and in vivo, repeated and reproducible on-off release of ibuprofen is demonstrated by non-invasive selective heating of a polylactide implant using near-infrared radiation, attaining on-off ratios in excess of 1,000. Supersaturation of the implant ensures replenishment of released ibuprofen by partial dissolution of the drug crystals providing a constant concentration of ibuprofen dissolved in the polymer and, consequently, a constant driving force for release. Although the large plasticizing effect of ibuprofen lowers the Tg of the polymer to about 0ºC (i.e. the bulk polymer is in a rubbery state), release of ibuprofen at body temperature is extremely low, attributed to the spontaneous formation of a glassy surface layer. Drug release can thus be controlled by changing the temperature of this switching layer around Tg. In addition, this self-sealing mechanism provides an intrinsic safety precaution, as rupture of the implant will not result in dose dumping. The simple design of an implant based on the glass transition switch thus offers an appealing drug delivery system. It should be realized, however, that strict requirements are imposed on the polymer, and that new solubility and diffusivity data needs to be generated for each new drug/polymer combination. Although the Tg can be effectively used to switch on and off the release of ‘simple’ drug molecules, controlling release of large molecules, such as peptide drugs, is impossible. For this reason, an alternative mechanism is explored based on the reversible swelling capability of a hydrogel around the so-called lower critical solution temperature (LCST). A drug reservoir is covered by a thermoresponsive composite membrane, consisting of a microporous polymeric support and a hydrogel graft layer. Above the LCST of the hydrogel (on-state), the shrunken state of the graft layer appears to only partially cover the membrane, allowing protein permeation through the uncovered pores. Provided the grafting degree is high enough, the swollen hydrogel covers the membrane completely below the LCST (off-state), thus preventing protein permeation. In this way, reversible on/off-switching of protein permeation through a thermoresponsive composite membrane with negligible permeation in the off-state is demonstrated. The on-demand release mechanism proposed in this thesis is based on switching the membrane surface coverage rather than previously reported switches based on effective pore size or hydrogel mesh size, thus allowing for higher fluxes in the on-state, since permeation is not limited by pore-narrowing. Both the Tg- and the LCST-concept permit long-term and patient-friendly administration of a wide variety of medicines, ranging from systemic dosing of drugs with low oral bioavailability to drugs required locally. This thesis presents a proof of principle for on-demand drug delivery using a laser as NIR-source. From a commercial point of view, it is expected that the use of cheaper light emitting diodes (LEDs) in the NIR range can lead to the design of hand-held devices suitable for near-infrared induced drug delivery in both a clinical and home environment.

The increased awareness for environmental issues and concomitant environmentally conscious governmental policies has prompted the chemical process industry to implement "greener" production and synthesis methods. In particular, the reduction of the emission of harmful, often organic, substances, reduction of the production of waste, and increasing the energy efficiency are three important aspects in the development of environmentally benign chemical production processes. For the chemical and chemical engineering academic community this has given rise to a new direction, where the concept of "green chemistry" is being explored. Supercritical fluids have been established as promising substitutes to organic solvents. Carbon dioxide is of particular interest as an alternative solvent as it has a low toxicity, is non-flammable and has an accessible critical temperature and pressure. In addition, catalysis is an important tool for the optimization of atom efficiency of a chemical conversion, and therefore for the reduction of waste production. Additionally, catalysis allows for reactions to take place under milder conditions, which can also contribute to an increase in energy efficiency. In particular, soluble molecular organometallic catalysts allow chemical conversions with a higher rate and a better selectivity than their heterogeneous counterparts. The difficult separation of a homogeneous catalyst from reaction products, without deactivating the catalyst, is one of the main obstacles for their application on an industrial scale. Nanofiltration using a microporous ceramic membrane has the potential to be a solution to this problem. A large enough catalyst molecule will be retained while reaction products and solvent can permeate across the membrane. In the field of separation technology membranes have emerged as an energy efficient alternative to conventional separation methods, like distillation and extraction. Ceramic membranes are seen as one of the most promising candidates to purify process streams under demanding conditions. The main objectives of this thesis are the evaluation of the possible advantages of using supercritical carbon dioxide as a solvent as an alternative for organic solvents, and the investigation into the potential of membrane technology for the retention of homogeneous catalysts. The hydroformylation of 1-octene, which is an example of a homogeneously catalyzed reaction on an industrial scale, is considered as a model reaction. To perform the hydroformylation in supercritical carbon dioxide an experimental procedure has been developed, which allows for catalyst preparation under hydroformylation conditions and for carrying out the hydroformylation reaction with a well-defined starting point. It has been demonstrated that with this experimental procedure it is possible to obtain highly reproducible results. Furthermore, a relationship between the change in pressure and the change in reaction mixture composition as a function of time has been established. Using this experimental procedure the effect of total pressure, temperature, concentration of reactants, and concentration of catalyst precursors on the reaction rate, chemoselectivity, and the regioselectivity of the hydroformylation of 1-octene has been studied. The concentration of carbon dioxide had an effect on the regioselectivity of the reaction. Therefore, the same density of solvent has been used for each experiment rather than the more common approach of applying the same total pressure for each experiment. Based on the results obtained by the variation of the reaction parameters a kinetic model has been developed. An optimization method has been applied to find the model parameter values that best describe the experimental data. The observed kinetics for the catalyst based on rhodium(I) dicarbonyl acetylacetonate and tris(3,5-bis(trifluoromethyl)phenyl)phosphine shows resemblance to that observed for the hydroformylation where bulky phosphites have been used as the ligand. For this catalyst a high activity in the order of 5×103 to 12×103 mol1-octene molRh-1 h-1 has been observed at 70 °C. Organometallic complexes based on rhodium with phosphine ligands with a varying number of trifluoromethyl groups have been screened for their activity and selectivity for the hydroformylation of 1-octene. Furthermore, the effect of the type of solvent: carbon dioxide, hexane, and toluene has been included in this study. An increase in the number of trifluoromethyl substituents on the triphenylphosphine ligand results in an increase in 1-octene conversion rate and a decrease in the overall selectivity towards aldehydes. This behaviour is observed in all three solvents. For supercritical carbon dioxide or hexane, as the solvent, the outcome of the hydroformylation reaction in terms of activity and selectivity shows great similarity. By following the hydroformylation of 1-octene in time, it was observed that during batch operation rhodium catalysts with trifluoromethyl-substituted triarylphosphines showed a higher differential regioselectivity than based on the overall regioselectivity at the end of the reaction. For the hydroformylation in carbon dioxide this effect was most pronounced. Both the mode of operation, batch or semi batch, and the type of solvent had a significant influence on this phenomenon. The transport of a supercritical fluid across a microporous alumina supported titania membrane has been investigated. The dependence of the permeation of carbon dioxide across the titania membrane on the feed pressure is similar to what has been previously observed for microporous alumina supported silica membranes. At high feed pressure viscous flow appears to be the main mechanism of mass transport across the membrane. Furthermore, the titania membrane shows a reasonable stability over a period of operation of about at least six months in varying conditions. Finally, the first continuously operated experiment has been performed, in which hydroformylation of 1-octene and separation of the catalyst have been integrated using a membrane reactor. During a 27.5 h of operation of the membrane reactor, spread over four consecutive days, a maximum conversion of 17 % and a maximum regioselectivity of 5 in terms of n:iso ratio has been observed. The conversion and the n:iso ratio, which is the ratio between the linear and branched aldehyde product, decrease as a function of the number of permeated reactor volumes indicating a loss of catalyst. Permeation of free ligand and the catalytic species through the membrane appear to be the main reasons for the decrease in activity and selectivity. A good match between membrane retention characteristics and the size of the catalyst and its precursors is not found yet. However, a number of feasible improvements can be made to improve the retention of the catalyst. Using a membrane for retention of a homogeneous catalyst in combination with the application of carbon dioxide as a solvent for the continuous hydroformylation of 1-octene has great potential. Successful application of the envisioned membrane reactor process can have implications for other homogeneously catalyzed reactions of which asymmetric hydrogenation is a commercially relevant example. As a result of the experimental methods used in this thesis the potential benefits of using carbon dioxide as an environmentally benign alternative to organic solvents could be further extended.

The industrial application of solvent resistant nanofiltration (SRNF) membranes for solvent purification has grown considerably in the last decade. The number of solvent/membrane systems is increasing and consequently, the need for characterization of chemical and physical phenomena, i.e. sorption, swelling, compaction and mobility. This thesis provides a tool box with characterization methods that are aimed to be independent of the solvent and membrane material and to provide results focused on the transport determining layer of the membrane. In Chapter 2 the determining transport mechanism of three similar solvents (methanol, ethanol and 1-butanol) in a silicon-type SRNF membrane has been studied. The results suggest that diffusion alone can not explain the transport of alcohols and an additional transport mechanism takes place. Assuming the additional mechanism to be viscous in nature allows a fit of a flux expression to the experimental data using one single fitting parameter. For 1-butanol, the contribution of the additional mechanism is only small. In contrast, for the smaller methanol and ethanol the contribution of viscous flow is very high. This is in contrast with observations in literature for similar systems, signifying a limited applicability of state-of-the-art models for predicting solvent transport through SRNF membranes. Chapter 3 presents a generic method that allows the determination of retention behavior in non-aqueous solvent filtration. The method has limited restrictions with respect to solute solubility and is independent of the solvent studied. Using a low concentration of polyethylene glycol 1000 with a broad molecular weight distribution, combined with a powerful analytical technique such as mass spectroscopy, a single retention experiment produces a full retention curve. In this way, the molecular weight cut-off (MWCO) of a membrane can be directly determined and the cost and time of the experimental procedure is reduced. Retention has been shown to be mostly dependent on the rate and mechanism of solvent transport. For transport that is solely or partially viscous, dragging effects cause the retention of small PEG oligomers to be low. For transport determined by diffusion, dragging effects are insignificant and retention is high (98%) irrespective of the size and shape of the solute molecule. Chapters 4 and 6 show the versatility of attenuated total reflectance spectroscopy infrared (ATR-IR) to determine in-situ sorption and compaction in the active layer of a cellulose acetate (CA) membrane and a silicon-type SRNF membrane, respectively. Preferential sorption of water has been measured in a CA membrane for water/methanol and water/ethanol mixtures. The results are in accordance with literature. The extent of compaction of the transport determining layer of the commercial SRNF membrane has been determined under different pressure conditions. Results show that reduction in thickness of the entire membrane can not be directly related to the reduction in density of the active layer. Comparison with permeation experiments under the same conditions show the importance of measuring both swelling and compaction simultaneously. Chapter 5 shows the development of two straightforward techniques to measure swelling: mechanically using a micrometer and optically using an interferometer. Both methods allow the study of the dynamic swelling behavior and yield similar results. Different polymer layers present in the membrane show different contributions to swelling. Comparison of the dynamics of swelling and sorption reveals that these two phenomena, although interdependent, do not start simultaneously and are not comparable in magnitude. For this reason, if the extent of sorption is interpreted as a direct measure for the extent of swelling, incorrect conclusions can be drawn. Chapter 7 concludes the thesis by taking a look at the impact of the work developed. Furthermore, the future possibilities for developments regarding the characterization, prediction and process monitoring are discussed.

Particulate fouling is defined as the accumulation of particles on a heat exchanger-surface forming an insulating powdery layer. Particulate fouling in biomass gasifiers is a major problem that may lead to inefficient operation. As observed in a large-scale biomass gasifier the character of the fouling layer is related to the local gas temperature and velocity. At high gas-side temperatures, the fouling layer structure changes from a fragile powdery form to a coherent sintered structure. Particulate fouling of heat exchangers shows an asymptotic growth rate with a levelling off increase of the thickness. The fouling growth rate is determined by the difference between the deposition rate and the removal rate of particles on and from the fouling layer respectively. Till now, most attention has been given to the deposition process. The objective of this research project is to study and model removal of particles from particulate fouling layers as a function of the fouling layer structure. By integrating the removal rate and the depositing rate a complete perspective can be given about the fouling process. Two mechanisms contribute to the removal of particles from fouling layers: the transfer of kinetic energy from an incident particle to bed particles or the forces exerted on the bed of particles by a shear flow. To study the removal of particles from fouling layers by shear flow, fouling experiments in a heat exchanger set-up have been done with particles of different sizes and different materials running under different gas speeds. It is found that the smallest particles in the flow deposit first on the tubes of the heat exchanger at areas of minimum flow velocities. Then the larger particles deposit and the fouling layer starts to build up. The fouling layer thickness and growth over the heat exchanger tube is influenced by the flow speed. As the flow speed in the heat exchanger increases, the thickness and the surface area of the fouling layer deposited over the heat exchanger tube are reduced. There is a limiting flow speed above which fouling is avoided. This limiting speed appears to be related to the critical flow velocity required to roll a particle resting on a flat surface. To prevent fouling, the gas speed in a heat exchanger should be larger than the critical flow velocity that corresponds to the particle size most likely to stick on the heat exchanger tube. It appeared that the removal of particles from fouling layers by shear flow becomes important at high gas speeds. The second mechanism, which is the removal of particles by an incident particle impact, is the dominant removal mechanism at low gas speeds. The removal of particles from powdery fouling layers due to an incident particle impact is investigated numerically and experimentally. A numerical model is developed to simulate the interaction between a particle hitting a bed of particles. The numerical model is based on the discrete element method. The forces during collision are based on the concepts of contact mechanics. The numerical model can predict the particle velocity at which an incident particle starts either to stick, rebound or remove other particles from the impacted bed of particles. The critical sticking and removal velocities of a particle hitting a powdery layer become independent of the layer thickness, if the thickness is larger than a certain limit. This limit is of a thickness of 2 particles in case of a monodisperse particulate layer and when the particles in the layer are arranged in an orthorhombic structure. It is found that the ejection time of particles from a bed of particles due to an incident particle impact is proportional to its diameter and to the square root of the number of bed layers. To validate the numerical model, experiments are carried out in a vacuumed column. In the experiment, incident particles drop on a bed of particles and the sticking, bouncing and removal behaviour is measured as a function of the incident particle impact speed. The numerical model predictions regarding the critical sticking and removal velocities are in agreement with the experimental results. During operation of heat exchangers, particulate fouling layers may sinter due to the high gas-side temperature. The influence of the gas-side temperature on the fouling layer structure and consequently on the removal and deposition of particles are investigated. The fouling layer structure is dependent on the gas-side temperature in relation to the minimum sintering temperature. Impaction experiments are carried out to determine the sticking and removal velocities for an incident particle hitting a bare tube surface, a powdery fouling layer and a sintered layer. The change in the heat exchanger surface from a bare tube to a powdery one increases the critical sticking velocity with at least one order of magnitude, which consequently speeds up the fouling process. The further change in the heat exchanger surface from a powdery form to a sintered one lowers significantly the ability of an incident particle to stick on the fouling layer or to remove particles out of the fouling layer. Particles that are still able to deposit on the sintered fouling layer will not sinter immediately and can therefore be removed, due to an incident particle impact. Sintering reduces the fouling rate of heat exchangers by lowering the deposition rate of new particles and increasing the removal rate of newly deposited particles, such that the fouling process becomes as slow as the formation of a single layer on a bare tube during the initiation period. When the initiation period is longer than the characteristic sintering time, the newly deposited particles become sintered and we revert again to the sintered case, which leads to a very slow fouling process known as the asymptotic behaviour of particulate fouling layers. The numerical model is extended to account for an incident particle hitting a sintered layer. Summarizing, the removal and deposition of particles from particulate fouling layers as a function of the fouling layer structure have been numerically modelled and the model is experimentally validated. By combining the numerical model together with a CFD-particle transport model, the growth rate of the particulate fouling layer around a heat exchanger tube can be quantitatively predicted.