Controlling Crystal Size by Drop-Based Microfluidic Crystallization

Crystallization is used to produce a number of solid products, ranging from common materials like sugar to high value products like pharmaceuticals. Control of crystal properties such as the size distribution, shape and polymorph is important in order to maintain product quality. Emulsion based methods have been demonstrated as effective means for producing polymeric and inorganic particles of controlled size. Previous studies of the control of organic or protein crystal size by solution crystallization in emulsions are complicated by polydisperse drop size distributions or the addition of a surfactant to stabilize the drops. This thesis investigated a drop-based microfluidic crystallizer as a new method for producing crystals of controlled size. In the microfluidic crystallizer, the mother liquor is dispersed as monodisperse drops and crystallization is confined to the drop domain. The maximum crystal size is limited by the amount of material in the drop. If only one crystal is grown in each drop, then the crystals will have a narrow size distribution near this maximum size. A new microfluidic system for producing monodisperse drops, the HPLC T-junction, was developed for use in the study of drop-based crystallization. The HPLC T-junction was capable of producing drops with diameters from 50 to over 300 µm that were stable against coalescence without any added surfactant. The coefficient of variation (CV) of drop size was less than 3%, indicative of very narrow size distributions. The HPLC Tjunction had superior solvent resistance, more favorable wetting properties for producing aqueous drops and higher pressure tolerance compared with conventional PDMS microfluidic devices. The HPLC T-junction did not require the expensive and timeconsuming fabrication steps or the specialized fabrication equipment associated with the production of conventional PDMS microfluidic devices. Mathematical models of crystallization in monodisperse drops were developed using population balance and Monte Carlo methods to predict the crystal size distribution and number of crystals per drop and to identify the important factors that determine the product crystal size distribution. Model predictions indicated that the drop-based microfluidic crystallizer would produce the narrowest crystal size distribution when each drop contained one crystal and that crystal had grown to the mass balance limited size. Modeling showed that the crystallization behavior in drops was determined by the relationship between the nucleation and growth rates of the solute. Two dimensionless time scales, the dimensionless growth and nucleation times were proposed to quantify the relationship between the two rate processes. Rapid crystal growth led to a high fraction of drops containing only one crystal and a very narrow size distribution. If crystal growth was slow relative to nucleation, the crystallization in drops yielded many drops with more than one crystal and a broader size distribution. Experimental case studies of crystallization in monodisperse drops were conducted using lysozyme and lactose. The HPLC T-junction and a Teflon tubular crystallizer were used as the drop-based microfluidic crystallizer in the studies. Drop-based crystallization experiments were conducted by filling the tubular crystallizer with mother liquor drops and measuring the number and size of the crystals in each drop by image analysis after certain residence times. Lysozyme was crystallized by the addition of NaCl to induce supersaturation. Cooling crystallization was used to generate supersaturation in the lactose experiments. Lysozyme crystals grew rapidly and quickly depleted the supersaturation in drops, thereby preventing formation of a second crystal. It was possible to produce over 90% of drops containing single lysozyme crystals in the microfluidic crystallizer. Lysozyme crystals with a CV of crystal size as low as 14% were produced. A substantial portion of the spread in the measured size distribution was likely due to orientation effects of the plate-shaped crystals during observation. Unlike lysozyme, lactose crystals grew slowly in drops, leading to a maximum of 40% of drops containing one crystal and a substantial fraction of drops containing multiple crystals. The CV of crystal size was as low as 7% in those drops that contained one crystal, indicative of a very narrow size distribution. In a non-optimized, isothermal process, the CV the total crystal size distribution produced by the drop-based microfluidic crystallizer was as low as 16%, compared with 40% produced from bulk crystallization with common history seed. Nucleation rates for lysozyme and lactose were estimated from measurements of the fraction of empty drops over time using the drop-based microfluidic crystallizer and were correlated to an expression based on classical nucleation theory. Lysozyme nucleation rates measured in the microfluidic crystallizer showed good agreement with literature data. Measured nucleation rates were used to estimate the interfacial tension for primary nucleation in classical nucleation theory for each material. The interfacial tension of lysozyme was estimated as 0.65 mN/m, which agrees with literature values. The interfacial tension for primary nucleation of lactose was estimated as 5.9 mN/m, the first such estimate for lactose. The study of lactose crystallization in the microfluidic crystallizer showed that isothermal crystallization will not always lead to a high fraction of drops with one crystal. Two nonisothermal control methods, fines dissolution and nucleation-growth thermal cycling, were proposed. Fines dissolution increased the fraction of drops with one crystal, reduced the fraction of drops with multiple crystals and produced a narrower crystal size distribution. Nucleation-growth thermal cycling minimized the formation of more than one crystal in a drop, yielding a narrower crystal size distribution. The results of the lysozyme and lactose case studies indicate that the drop-based microfluidic crystallizer is an effective means for controlling the size distribution of organic and protein crystals. The mathematical models are useful for optimizing the microfluidic crystallizer and extending drop-based crystallization to control the crystal size of materials other than lactose and lysozyme.