Crystallization solvent selection

Salting-out crystalli tion operates through the addition of a nonsolvent to the magma ia a crystallizer. The selection of the nonsolvent is based on the effect of the solvent on solubiHty, cost, properties that affect handling, iateraction with product requirements, and ease of recovery. The effect of a dding a nonsolvent can be quite complex as it iacreases the volume required for a given residence time and may produce a highly nonideal mixture of solvent, nonsolvent, and solute from which the solvent is difficult to separate. 

Deviations from Raonlt s law in solution behavior have been attributed to many charac teristics such as molecular size and shape, but the strongest deviations appear to be due to hydrogen bonding and electron donor-acceptor interac tions. Robbins [Chem. Eng. Prog., 76(10), 58 (1980)] presented a table of these interactions. Table 15-4, that provides a qualitative guide to solvent selection for hqnid-hqnid extraction, extractive distillation, azeotropic distillation, or even solvent crystallization. The ac tivity coefficient in the liquid phase is common to all these separation processes. 

The thienothienoimidazolium salts 29 were prepared by the reaction of thiophanes 362 with HX (X = halogen) and crystallization from solvents selected from ketones, aromatic hydrocarbons, and halohydrocarbons. l-(—)-3,4-(l, 3 -dibenzyl-2 -ketoimidazolido)-2-(u -ethoxypropyl)tetrahydrothiophene 362 was reacted with HBr at 99-103 °G for 2h and crystallized from methyl-Tro-butyl ketone to give l-(—)-3,4-(T,3 -dibenzyl-2 -ketoimidazolido)-l,2-trimethyle-nethiophanium bromide 29 (95%, 98.7% purity) (Scheme 75) <2001JAK100477>. 

Separation can also be accomplished by solvent extraction, adsorption, and crystallization. Solvent extraction is accomplished by selectively dissolving certain hydrocarbon components. Adsorption is similar to solvent extraction but uses a solid to separate out various components selectively based on their tendency to adhere to the surface of the solid adsorbent. Crystallization uses the differing melting points of the components during cooling, which causes some of its compounds to solidify or crystallize, and separate out of the liquid. 

It is important to understand this behaviour when selecting a crystallization solvent and the operating temperature range for a new process. The temperature range for the crystallization should be selected so that the desired polymorph is always the thermodynamically stable form. If this is not feasible then kinetic studies will be required to confirm that the polymorphic form that crystallizes first is subsequently converted to the desired form during processing. It is also possible that more than two polymorphic forms could be involved in this type of system, increasing its complexity. 

Whenever the solubility curve is crossed for the less stable Form II there is a risk that it will nucleate and contaminate the product. This situation is very probable when the solubility curves of the two polymorphs lie close together, as shown in Figure 21 of the Cimetidine case study. The addition of seed crystals of Form I, close to its solubility curve, and minimization of the supersaturation during the growth process is a good method of control in this instance. Solvent selection to extend the width of the Form II metastable zone would also be desired, as discussed in section 2.4.4. 

The selection of a potential list of crystallization solvents is made against the following criteria ... 

In systems where the liquid phase interaction between the solute and solvent is close to ideal, then Eq. 2 can be used successfully on it s own to fit and extrapolate solubility data with respect to temperature. The technique is valuable in an industrial setting, where time pressures are always present. Solubility data points are often available without any additional effort, from initial work on the process chemistiy. The relative volume of solvent that is required to dissolve a solute at the highest process temperature in the ciystallization is often known, together with the low temperature solubility by analysis of the filtrates. If these data points fit reasonably well to the ideal solubility equation then it can be used to extrapolate the data and predict the available crystallization yield and productivity. This quickly identifies if the process will be acceptable for long term manufacture, and if further solvent selection is necessary. 

The non-random two-liquid segment activity coefficient model is a recent development of Chen and Song at Aspen Technology, Inc., [1], It is derived from the polymer NRTL model of Chen [26], which in turn is developed from the original NRTL model of Renon and Prausznitz [27]. The NRTL-SAC model is proposed in support of pharmaceutical and fine chemicals process and product design, for the qualitative tasks of solvent selection and the first approximation of phase equilibrium behavior in vapour liquid and liquid systems, where dissolved or solid phase pharmaceutical solutes are present. The application of NRTL-SAC is demonstrated here with a case study on the active pharmaceutical intermediate Cimetidine, and the design of a suitable crystallization process. 

For the purpose of this case study we will select Isopropyl alcohol as the crystallization solvent and assume that the NRTL-SAC solubility curve for Form A has been confirmed as reasonably accurate in the laboratory. If experimental solubility data is measured in IPA then it can be fitted to a more accurate (but non predictive) thermodynamic model such as NRTL or UNIQUAC at this point, taking care with analysis of the solid phase in equilibrium. As the activity coefficient model only relates to species in the liquid phase we can use the same model with each different set of AHm and Tm data to calculate the solubility of the other polymorphs of Cimetidine, as shown in Figure 21. True polymorphs only differ from each other in the solid phase and are otherwise chemically identical. 

NRTL-SAC has been demonstrated through the case study on Cimetidine as a valuable aid to solubility data assessment and targeted solvent selection for crystallization process design. The average model error is typically 0.5 Ln (x) [1] and is sufficient as a solvent screening tool. Methods that can deliver greater accuracy would increase the value and utility of these techniques. It is impressive in the case of Cimetidine that the NRTL-SAC correlation is capable of reasonable accuracy and predictive capability on the basis of just 2 fitted parameters. Further work to extend the solvent database and optimize the descriptive parameters will be beneficial, and are planned by the developers. 

Other aspects related to the packing of these grid structures, the presence of a multitude of interactions between the tetranuclear cations and the counter-anions not hosted in the cavities, and with the crystallization solvent molecules, are beyond the scope of this discussion. In general all of these solids can be considered as microporous and the anions and solvent molecules located within these pores are normally very disordered. Figure 3.7 shows two selected examples, where channels are formed along the c axis in complexes 49 and 40. In the former, disordered triflate anions located in the channels are drawn in grey. In the second example, the regular... 

One of the characteristic properties of rod-like polymers is that their concentrated solutions form lyotropic liquid crystals51. Such examples among synthetic polymers are polyamides52,53 and polyisocyanates54 which form cholesteric or nematic liquid crystals in selected solvents. 

Active ester formation by the mixed anhydride method is accompanied by the side reaction of esterification at the carbonate moiety of mixed anhydride 51 which generates mixed carbonate 52 (Scheme 12).This decreases the yields, but is more of a nuisance than an obstacle as the side products do not interfere with crystallization of the esters as the former are soluble in the crystallizing solvent. More mixed carbonate is formed from derivatives of the hindered amino acids and proline none is formed from a-unsubstituted acids. A-Hy-droxysuccinimide gives rise to much less byproduct than 4-nitrophenol other phenols generate intermediate amounts. Less byproduct is generated when the reagent is isopropyl chloroformate. The impurity can be readily removed from a solution of the ester by adsorption of the compounds on reverse-phase chromatography beads followed by separation by selective displacement. ... 

If crystals do not form on cooling, you rfiay have formed a supersaturated-solution and you should -- scratch with a Pyrex glass rod, as described for -single solvent selection (Box 13.1). 

The first series of compounds assayed directly by CD detection were the morphine alkaloids. They were supported in aqueous solutions, in a chiral cholesteric liquid crystal solvent, and mixed in pellet form with solid KBr. ° Contrary to expectations, the homogeneous aqueous solution medium gave the best selectivity among 10 related opiates and the most quantitative results. The pH-dependence of phenol substituted analogs, which in some instances caused the sign of the CD signal to invert, enhanced the selectivity. Heroin was assayed both directly and as the morphine hydroly-sate.f Direct multicomponent analyses were made for prepared mixtures of morphine, codeine, thebaine, noscapine, and opium extracts. ... 

Solvent selection is usually very important in preparing both high-quality product and the desired crystal form. Details may be found in Chapter 11. 

Equilibrium reactions may be optimized by solvent selection if one component of the equilibrium is removed from the reaction, as occurs with crystallization. The raloxifene intermediate 18 was prepared in 80% yield by the sequence in Figure 8.9 [9], Studies showed that at thermal equilibrium the ratio of 17 18 was 10 90, but by adding heptane near the end of the isomerization, the desired product 18 crystallized out of solution affording a final ratio of 2 98. 

Crystallization in or from solution brings another variable, solvent selectivity, into play. If the solvent is selective for the crystalline block, it can swell the crystalline lamellae (Tm is obviously also reduced). In contrast, if the solvent is selective for the non-crystalline block, it can precipitate out of solution in a non-equilibrium structure. 

---