*3.2. Reproducibility of Steady-State Results*

The central point (DoE, design of experiments) of the investigated operation window (operation parameters, Tsat = 35 ◦C, Tcrys = 30 ◦C and F = 12 L/h) was evaluated during Exps. 4, 5, and 7. The corresponding operation parameters were used for the first period of experiments 4 and 7 (4 h), though different seed crystal sizes (Exp. 4: 212–300 μm, Exp. 7: 250–355 μm) were added for initialization. In Exp. 5, another operation point was tested before going to the central point. Figure 7 depicts the steady-state product crystal size distribution, q3 (Equation (3)), and the steady-state productivity, Pr, for both crystallizers, C2 and C1 (Figure 7a,b) for the three experiments. The average mean values, L3, and standard deviations, sL3, of the steady-state product crystal size distribution are also given. The comparison of Exps. 4 and 5 shows that differences between the product crystal sizes are below 4% and thus negligible. In contrast, the scattering range of the productivity values for each evaluated operation point is approximately 8–10 g/L/h (≈ 30%). Responsible for this broad scattering are the dynamics of the periodic product withdrawal, since during each withdrawal an unknown and varying crystal mass remains within the outlet ports of the crystallizer. This crystal mass is removed from the tube after each withdrawal via a cleaning step with pure solvent and is, thus, not quantifiable. Several experimentations to capture and quantify this crystal mass led to the assumption that it varies between 1 and 3 g, which consequently reduces the evaluated productivity by approximately 2–10 g/L/h. However, the productivity values of Exps. 4 and 5 scatter within a common range, considering the previously mentioned deviation.

**Figure 7.** Steady-state crystal size distributions, q3, with their respective mean values and standard deviations, as well as the corresponding productivities, Pr, of crystallizer C2 (**a**) and C1 (**b**) for Exps. 4, 5, 7.

As indicated in Figure 7, the high-speed dispersers were maintained before Exp. 7. During the process, the maintained disperser tools were placed in cylindrical steel vessels, through which the suspended crystals of the bypass flow. Since the disperser tools do not cover the whole cross section area of these vessels, crystals carried by the surrounded flow might not be comminuted. Thus, the maintenance could cause changes in the amount and size distribution of the generated seed crystals and affect the cyclic steady-state and its reproducibility. As seen in Figure 7, the product crystal size distributions and the productivities of Exp. 7 slightly differ from the respective results of Exps. 4 and 5, which shows exceptionally good reproducibility.

Comparing the results of crystallizer C2 with the corresponding data of crystallizer C1 (Figure 7a,b) shows that C2 produces somewhat smaller product crystals with higher productivities. This trend was persistent throughout all experiments. It indicates that both crystallizers differ in their hydrodynamics due to small differences resulting from their hand manufacturing. Furthermore, the total volume of C2 is 511 ml and, thus, significantly larger than C1 with a total volume of 478 ml. Due to the classifying effect of the conically shaped crystallizers, a larger diameter of C2 would lead to smaller product crystals assuming similar flow rates. These smaller crystals have a higher specific crystal surface, which enhance the total crystallization rate and, thus, the productivity.

The comparison of Exps. 4, 5, and 7 prove that the utilized pilot plant generates reproducible results. Furthermore, the three experiments clearly prove that, when starting from different initial conditions, the same operation points can be independently reproduced. The reproducibility of a further operation point (operation parameters, Tsat = 35 ◦C, Tcrys = 30 ◦C, and F = 10 L/h) was also verified (see Appendix A, Figure A1).

#### *3.3. Influence of Volumetric Flowrate*

To study the influence of the volumetric flowrate and the resulting residence time of the liquid phase, three operation points with different volumetric flowrates, F = 10, 12, and 14 L/h, were investigated during Exps. 2, 4, 5, and 7. A constant saturation temperature (35 ◦C) and a constant crystallization temperature (30 ◦C), were ensured during all experiments. It was observed that at the lowest flowrate, F = 10 L/h, incrustations at the top of both crystallizers occurred and spread significantly faster than at the other operation points. The incrustation layer detached after a certain time, and

settled towards the crystallizer bottom and the milling bypasses, where they were comminuted. The subsequently detected decrease of the product purity indicates that the incrustations are induced by nucleation of the counter-enantiomer. To achieve a successful racemate resolution at lower flowrates and, thus, higher residence time of the liquid phase, the risk of nucleation could be reduced by setting lower supersaturation.

Figure 8a depicts the mean product crystal size distributions with their respective mean values and standard deviations for the three volumetric flowrates, from Exps. 4 and 5. As seen, the product crystal size distributions differ regarding the obtained product crystal size; with increasing flowrate the mean crystal size increases significantly from 267 to 300 and 330 μm. As expected, higher volumetric flowrates and, thus, higher fluid velocities, cause larger product crystals. Thus, our results verify the results of former studies [17].

**Figure 8.** Mean steady-state crystal size distributions, q3, with their respective mean values and standard deviations (**a**), productivities, Pr (**b**), and yields, Y (**c**), for three different volumetric flowrates, F = 10, 12, and 14 L/h. Given results are from crystallizer C2 for Exps. 4 and 5.

Figure 7b,c shows the influence of the volumetric flowrate on productivity and yield. As shown, a higher volumetric flowrate leads to lower productivity and yield. This can be correlated to the larger crystals and the resulting lower total crystal surface of the fluidized bed. A second reason is that the higher volumetric flowrate reduces the residence time of the seed crystals and, thus, reduces their growing period. Consequently, a higher amount of seed crystals is discharged at the top of the crystallizers. Comparing productivity and yield with respect to the total decrease shows that productivity decreases by approximately 33% whereby yield decreases by approximately 50%. This disproportion is attributed to the increasing throughput rate of the mother liquor, and consequently its decreased depletion.
