# Control of Preferential Crystallisation Processes

## Contents

### Abstract

This project addresses control problems for preferential crystallistation processes which are used for the seperation of enantiomers. We investigate two innovatove configurations — a cyclic batch scheme, where production of the enantiomers is performed in an alternating manner, and a simultaneous scheme, where both enantiomers are produced at the same time in two vessels.

### Description

Preferential crystallisation is used for the separation of enantiomers — substances with identical physical and chemical properties but different metabolic effects. The basic idea in preferential crystallisation is quite simple. Both enantiomers are initially dissolved, and the solution is kept in a temperature range where primary nucleation is of much less importance than secondary nucleation. Hence, after seeding one of the two enantiomers, say E1, it will be almost exclusively the seeded enantiomer that will crystallise — existing crystals will grow and sencondary nucleation will generate further E1 crystals. This will of course consume E1 in the liquid phase, reduce supersaturation and therefore "slow down" the desired crystallisation process. To ensure required product purity, the process has to be stopped before — via primary nucleation — too many crystals of the counterenantiomer E2 are being generated. This basic single-batch procedure can be extended to a cyclic batch scheme for the production of both enantiomers. A possible configuration, consisting of two batch crystallisers, is shown in Fig.1.

Figure 1: Cyclic batch process for enantiomer separation.

In the beginning, racemic mixture is filled into one of the two vessels (represented by an Failed to parse (PNG conversion failed; check for correct installation of latex, dvips, gs, and convert): A

in the ternary phase diagram in Fig.1. Seeding E1 crystals initiates the single batch process described above (Failed to parse (PNG conversion failed;


check for correct installation of latex, dvips, gs, and convert): A \rightarrow B ). After stopping the process, crystalline E1 is harvested by transfering the liquid into a second crystalliser vessel. Racemic mixture is added (Failed to parse (PNG conversion failed; check for correct installation of latex, dvips, gs, and convert): B \rightarrow C ), and E2 crystals are seeded. This initiates a second batch process where crystalline E2 is produced (Failed to parse (PNG conversion failed; check for correct installation of latex, dvips, gs, and convert): C \rightarrow D ). Stopping the process, transferring the liquid contents into the first vessel and again adding racemic mixture will finish the first cycle of the process (Failed to parse (PNG conversion failed; check for correct installation of latex, dvips, gs, and convert): D \rightarrow A ) [1].

Apart from a minimum required purity, control specifications include quality requirements related to the shapes of the product CSDs for E1 and E2. The resulting control problem is intrinsically hybrid: single batch process dynamics is continuous, but discrete events are obviously of paramount importance for the overall cyclic batch scheme. Control inputs include the crystalliser (or heat jacket) temperature signal, the seed CSD [2] and the amount of added racemic mixture for each single batch process, and the inter-batch switching pattern.

Crystal growth is driven by supersaturation, which, in turn, can be influenced via the crystalliser temperature. However, in the cyclic batch scheme, this dependence cannot be exploited properly: each attempt to keep supersaturation for the currently desired enantiomer at an appropriate level for a long period of time would lead to an unacceptable increase of supersaturation for the (undesired) counterenantiomer. This is the motivation for investigating the alternative configuration shown in Fig.2. There, E1 and E2 are crystallised simultaneously in two separate vessels, and crystal free solution is continuously exchanged between the two crystallisers. This implies that concentrations of E1 and E2 are simultaneously reduced in the liquid phase, and it will therefore be possible to keep supersaturation for both enantiomers at higher levels. The resulting increase in productivity does not come for free, however, as additional hardware is needed to guarantee that the exchanged liquid does not contain any crystals. Degrees of freedom for this control problem include temperature signals and seed CSDs in both crystallisers and the flow rate between the vessels.

Figure 2: Simultaneous batch-crystallisation.

### Publications

1. M.-P. Elsner, E. Alonso Muslera, I. Angelov, D. Fernandez Menendez, H. Lorenz, D. Polenske, U. Vollmer, J. Raisch, A. Seidel-Morgenstern. Analysis of different crystalliser configurations to perform preferential crystallisation. In Proc. 16th Int. Symp. Industrial Crystallization (ISIC-16), pages 829–835, Dresden, Germany, 2005.
2. Ivan Angelov, Jörg Raisch, Martin-Peter Elsner, Andreas Seidel-Morgenstern. Optimization of the initial conditions for preferential crystallization. Industrial & Engineering Chemistry Research, 45 (2):759–766,