1. Introduction
Recently, continuous crystallization has again become the focus of many research activities. Efforts have been made to develop design rules based on comparing it with the corresponding batch process [
1,
2] or to elucidate basic mechanisms, for example, the impurity incorporation and carryover [
3]. Additionally, novel concepts, applying to example slug-flow [
4], oscillatory baffled [
5,
6], Couette-Taylor [
7,
8] or fluidized bed crystallizers [
9,
10,
11,
12], as well as columns with static mixers [
1] or coiled flow inverters [
13], have been developed lately. These concepts exploiting tubular crystallizers in general are not superior nor applicable for every substance system or separation problem. They have specific benefits and drawbacks [
1] compared to the well-known mixed-suspension, mixed-product removal (MSMPR) concept but have given new input to the topic of continuous crystallization and have widened the field of application for the food, agriculture, and pharmaceutical industries.
Fluidized bed crystallizers (FBCs) have recently been successfully exploited in wastewater treatment to recover phosphate [
9,
10], sulfates and magnesia [
11], or boron [
12] from diluted aqueous solutions. All of these studies aimed at a full conversion of the pollutant ions in the liquid phase to form a salt and utilized afterward the surface of fluidized, inert particles to collect the fine precipitate. FBCs can be also applied, however, for separations where the substance of interest has a similar or the same concentration as the impurity, as is the case for enantioseparations [
14].
This process is then based on Preferential Crystallization, which is often applied as an efficient and inexpensive option for the production of pure enantiomers from the racemic, i.e., 50:50, mixture [
2,
15]. The classical variant is, nevertheless, an unstable process carried out in the metastable zone of the respective ternary substance system, consisting in this case of two enantiomers, D and L, dissolved in a solvent. Seeds of the desired enantiomer are utilized to selectively remove this species from the liquid phase. Hence, the dissolved mass of the seeded component decreases while the mass fraction of the counter-enantiomer increases. Consequently, the supersaturation of the counter-enantiomer also increases in most cases, which forces this species to nucleate after a certain induction time. A complete separation with the maximum yield is, hence, only achievable if the process is interrupted at the right time shortly before the nucleation.
To increase robustness, productivity, and yield of the classical variant, two opposite Preferential Crystallizations can be carried out in two vessels only connected via a continuous liquid phase exchange [
16]. In this way, the selective removal of the enantiomers in the respective crystallizer is counterbalanced and a racemic liquid phase composition maintained, which reduces the risk of nucleation.
The application of fluidized bed crystallizers together with the coupled Preferential Crystallization principle, as shown in
Figure 1, provides two additional benefits. Firstly, the flow rate of the feed entering the columns from the bottom drags small crystals out of the process at the top. Hence, contamination of the product can be avoided to a certain extend even if the counter enantiomer nucleates since the nuclei follow the liquid flow. Secondly, a product classification can be achieved by a conical shape of the columns. Then the liquid velocity profile varies over the height of the crystallizers and only a specific crystal size is present at the product outlet. This process allows, hence, for a continuous production of pure enantiomers with a high productivity and a specific, adjustable product crystal size. In previous studies, the feasibility of steady-state operation of this process was shown without considering the process dynamics [
17] and only based on the liquid phase composition for a specific operation point [
14,
17]. However, no systematic investigation also based on the periodically harvested product crystals has been conducted.
In this work we, first, prove the achievability of steady-state operation on the example of D-/L-asparagine monohydrate based on the liquid phase composition as well as the periodically harvested product crystal size distribution. Furthermore, the required startup time to reach the steady-state will be estimated for different operation conditions and the high reproducibility of the results will be demonstrated. On this basis, consecutively the dependence of productivity, purity, and yield on the process conditions, feed flow rate, supersaturation, and crystallization temperature, will be systematically studied.
2. Materials and Methods
2.1. Substances
For the fluidized bed crystallization experiments, racemic asparagine monohydrate (rac Asn∙H2O) and the respective enantiopure D- and L-Asn∙H2O were purchased from Sigma Aldrich (Purity >99%, Steinheim, (Baden-Württemberg), Germany). Both were used without further treatment. Deionized water (Millipore, Milli-Q Advantage A10) was used as solvent and for washing of the crystalline product together with ethanol, which was purchased from VWR Chemicals (Purity >99.7%, Fontenay-sous-Bois, France).
Solubility data of the D-/L-Asn∙H
2O/ water system was mainly determined in [
18]. The accuracy was confirmed by additional measurements in [
19] where also the parameters of polynomial (Equation (1)) were estimated to describe the saturation concentration, x
sat, as a function of temperature and composition. As shown in
Figure 2, asparagine monohydrate forms a conglomerate, as a requirement for the application of Preferential Crystallization directly from a supersaturated racemic liquid phase.
where
for
The selective removal of the seeded enantiomer during Preferential Crystallization leads to an altering of the liquid phase composition during the course of the separation process. Hence, the driving forces of both enantiomers change and a different supersaturation calculation also need to be applied. Thus, Equation (2) is utilized in the present study to describe the supersaturation, S, as a function of temperature and composition.
for
2.2. Experimental Setup
All separation experiments were performed in the setup shown in
Figure 3. The plant consisted of two tubular crystallizers (C1 and C2,
Figure 3) with an individual volume of approximately 0.5 L and a total height of 1130 mm. Each double-jacketed tubular crystallizer was composed of a conical section in the lower part and a cylindrical section in the upper part of the columns. Thus, the fluid velocity of a liquid phase passing through the crystallizers changes with the height of the columns due to the increasing diameter in the conical section. At the transition between the upper and lower section, where the diameter is the largest, the fluid velocity becomes constant again over the height of the cylindrical part.
The temperature inside the columns was measured at the inlet, in the middle, and at the outlet via resistance thermometers (Pt100s). Temperature control was ensured by two thermostats connected to the respective middle Pt100s. In this study, the middle temperature is referred to as the crystallization temperature for the sake of simplicity, even though a temperature gradient will be present over the height of the columns due to the fluid flow.
Both crystallizers were coupled via heated steel pipes and flexible heated hoses with a double-jacketed feed tank on the scale of 7.5 L. A propeller-type stirrer was utilized in this tank during the separation experiments to suspend a certain excess mass of racemic solid phase in a solution saturated at a defined initial temperature. This excess of solid phase served to re-saturate the depleted mother liquors from the crystallizers during the process to ensure constant initial conditions at the inlet of the columns. The temperature within the feed tank was again measured by a Pt100 and controlled by a thermostat.
Clear saturated racemic solution was withdrawn during the process from the feed tank via glass filters and was pumped continuously from the bottom through the tubular crystallizers using gear pumps. The rotation speed of these gear pumps was controlled by Coriolis mass flowmeters to ensure a constant volumetric flowrate.
To harvest the product, each column had a product outlet at a height of 365 mm, close to the end of the conical section. The product removal was realized periodically due to the scale of the plant, using solenoid valves, which were controlled by photoelectric barriers as described in the next section.
Due to the mild mixing conditions of the particulate phase, secondary nucleation will not be sufficient to counterbalance the loss of crystals of the periodically continuous product removal. Hence, a continuous seeding strategy was necessary, which also supports Preferential Crystallization by providing enantiopure crystals. Thus, the largest particles present at the bottom of the crystallizer columns were withdrawn via peristaltic pumps into a bypass. They were pumped, subsequently, through high-speed dispersers, which were utilized as mills working on the rotor-stator principle. Afterward, the ground crystals were fed back as seed material to the process.
2.3. Experimental Procedures and Operation Parameters
In the following the experiments performed are described with their operation parameters and objectives, and performance parameters for process evaluation are introduced (
Section 2.3.1). Since starting and operation procedure is crucial for achieving a successful continuous racemate resolution, the procedure along each experiment, especially the crystallization progress and the periodical product removal, are explained and the used analytics are depicted (
Section 2.3.2)
2.3.1. Study of Operation Parameters and Process Evaluation
Altogether, seven experiments were planned and carried out for 8 h in this first attempt to systematically investigate the influence of the operation conditions on the process performance. In continuation of the previous study [
14], a saturation temperature of 35 °C was chosen for most of the experiments (T
sat,
Table 1). The first three processes (Exps. 1–3,
Table 1) were used to determine the operation window of the process with respect to the crystallization temperature (T
crys,
Table 1).
The data of Exp. 2 were additionally utilized to determine the required time to reach steady-state operation. It was found that after two product withdrawals (process time approximately 2 h) all process characteristics became constant. Hence, some of the later experiments were split into two periods, where two sets of process parameters (compared, for example, to the volume flow of Exp. 4) were tested for 4 h each, to enhance the time efficiency of the investigation. Exps. 4 and 5 were carried out to evaluate stepwise the influence of the volume flow rate (F,
Table 1) between 10 and 14 L/h. In Exp. 6, the supersaturation was kept constant but the saturation and crystallization temperature were reduced by approximately 10 K to investigate the influence of reduced crystallization kinetics. In Exp. 7 the central test point was run again, T
sat = 35 °C, T
crys = 30 °C, and F = 12 L/h, to evaluate the standard deviation of all process characteristics for a longer time period.
Every reached operation point was evaluated based on the normalized volume related product crystal size distributions, q
3 (Equation (3)), their mean sizes,
(Equation (4)), and their respective standard deviation, s
L3 (Equation (5)), as well as the achieved yields, Y, and productivities, Pr (Equations (6) and (7), respectively) calculated from the product masses and the time window between two withdrawals.
where
µk = mass fraction of sieve class k [-] |
zk = characteristic length of sieve class k [µm] |
Δzk = width of sieve class k [µm] |
where
mprod = mass of withdrawn product crystals [g] |
mtheo = theoretical maximum of product mass [g] |
Δt = time window between two withdrawals [h] |
Vtot = total volume of tubular crystallizer [L] |
(C1 = 0.478 L, C2 = 0.511 L) |
2.3.2. Start and Operation Procedure, Crystallization Progress, and Analytics Used
The starting point of all experiments was the preparation of the feed suspension. Therefore, the solvent was filled into the feed tank and circulated via the gear pumps through the columns and via the peristaltic pumps through the comminution bypasses. The required racemic solute mass for saturating the aqueous solution at the desired temperature of each experiment (
Table 1, T
sat) plus 200 g racemic excess solid was afterward filled in the feed tank, and the suspension heated to T
sat, which was kept constant throughout the whole experiments. To avoid any risk of nucleation within the columns during the preparation step, the coolant temperature of the double jackets was set to one Kelvin above T
sat. After the mother liquor within the feed tank reached T
sat, the coolant temperature of the double jackets was set to one Kelvin below T
sat. In parallel, the heating hoses were set to 40 °C to prevent any nucleation or clocking within the tubes: all other steel pipes (
Figure 3) were appropriately heat traced to prevent clocking. The liquid flow was stopped for seeding each column from the top (
Figure 3, port) with a certain mass (
Table 1, m
Seed) of one of the pure, crystalline enantiomer once the respective temperatures were reached. In all experiments, crystallizer C1 was seeded with D-Asn∙H
2O and crystallizer C2 with L-Asn∙H
2O. Immediately after seed addition, the pumps were started again to fluidize the present crystals and to prevent nucleation in the stagnant fluid phase. After this initial seeding, both crystallizers were subsequently cooled to the planned operation temperature (
Table 1, T
crys). The fluidized seed crystals in both columns started to grow according to the present supersaturation. Hence, they settled constantly due to their increasing mass towards the crystallizers’ bottoms. All crystals, which reach the fluid phase inlet, were withdrawn into the inline seeding bypass, where they were ground. The generated crystal fragments were fed back to the process merged with the liquid inlet stream.
If the process conditions are chosen appropriately, the crystal fragments will grow (
Figure 4a) large enough to settle within the cylindrical part and thus stay in the process. Then the overall crystal mass increases and the process progresses. The growing seeds start to settle, subsequently, on the top of the initial solid mass and the crystal bed increases in height.
The product withdrawal was initiated when the crystal bed first blocks the upper photoelectric barrier (L1,
Figure 4a). The solenoid valve (V1,
Figure 4a) was opened at this point and the crystal product is harvested until the lower photoelectric barrier (L2,
Figure 4a) was not blocked by the fluidized crystal phase anymore. Afterward, the valves were closed and the crystallization progressed until the crystal bed height was large enough for the next product withdrawal. This periodic procedure continued and the process progressed towards its dynamic steady-state. Immediately after the product withdrawal, the product crystals were filtered and washed with an ethanol-water mixture (40 wt% EtOH, 60 wt% H
2O). The dried product crystals (
Figure 4b) were afterwards analyzed via high performance liquid chromatography (HPLC), X-ray powder diffraction (XRPD), and sieve analysis to evaluate each operation point with respect to process performance parameters (purity, yield, productivity, and size distribution, Equations (3)–(7)). The specifications of the applied analytics are given in
Table 2.
The recycled, concentration-depleted mother liquor causes partial dissolution of the solid racemic excess within the feed tank. To minimize additional distortions of the feed concentration, and thus of the process, a piecewise addition of new solid excess material is required. Since the average product crystal mass of one withdrawal is approximately 20 g per column, subsequently, 50 g of racemic solid excess material was added to the feed tank after every product withdrawal.
4. Conclusions
In the present experimental parameter study, seven experiments were performed to investigate the continuous fluidized bed crystallization at twice 0.5 L scale for racemate resolution at its steady-state. Each experiment was conducted for 8 hours to ensure constant conditions. As verified, after a relatively short operation time (approximately 2 hours), the utilized pilot plant reaches a cyclic steady-state, where product crystals with constant crystal size distribution and productivity can be periodically withdrawn. The reproducibility of the steady-state results was proven and sensitivities of the utilized pilot plant on the steady-state results were identified. In particular, it was observed that changes of the high-speed disperser slightly effect the steady-state results in terms of productivity and product crystal size distribution. Nevertheless, the steady-state results were shown in the present study to have exceptionally good reproducibility.
It was proven that the steady-state product crystal size mainly depends on the volumetric flowrate, and thus can be easily adjusted (in this work between 260 and 330 µm). All products, withdrawn at steady-state, have a narrow crystal size distribution (standard deviation <60 µm) and a low fines content. Thus, the size classifying effect of the conically shaped tubular crystallizers and its selectivity is verified. Productivity and yield increase with supersaturation, whereby it was shown that the continuous racemate resolution is limited by a certain maximal and minimal supersaturation. At too low supersaturation, the seed crystals grow insufficiently and are mainly discharged at the top of the crystallizer. At too high supersaturation, the respective counter-enantiomer nucleates and, thus, contaminates the resolution product. Furthermore, the limitations regarding the supersaturation are not fixed values and also depend on the volumetric flowrate. In particular, the nucleation probability increases with higher supersaturation and lower volumetric flowrates. A decrease of productivity was observed at lower supersaturation and higher volumetric flowrates, and thus, higher fluid velocities. These observed correlations enhance the expectation that the location and width of the operation window and, thus, the process performance, are tunable via geometrical aspects of the conically shaped tubular crystallizer [
20].
The utilized pilot plant enables continuous racemate resolution with enantiomer purities above 97% and productivities up to 40 g/L/h for each enantiomer, which is far above productivities documented by other studies [
14,
21]. Since the process was not optimized at all, productivities higher than 40 g/L/h are to be expected. Thus, the coupled fluidized bed crystallization was proven to be an excellent technology for continuous enantioseparation, which facilitates high purities and the robust production of both enantiomers simultaneously.