Two-Stage Crystallizers Deliver Bigger, Cleaner Drug Crystals Without Extra Equipment

…

Crystallization occupies a distinctive position among pharmaceutical manufacturing operations because it is the step in which the active ingredient first acquires the solid-state properties that will follow it through the rest of the process. Unlike most unit operations, crystallization simultaneously determines chemical purity, the identity of the crystal polymorph, the distribution of particle sizes, and the shape of the crystals themselves. Each of these attributes propagates downstream: particle size and shape govern how quickly a slurry can be filtered and dried, polymorphic form controls dissolution behavior and stability, and variability in any of these properties can compromise the consistency of the final drug product. The transition from batch to continuous crystallization is therefore not simply a matter of equipment preference but a question of whether steady-state operation can hold these interdependent attributes within tighter tolerances than cyclic batch processing allows.

The mixed-suspension mixed-product-removal crystallizer has long served as the conceptual and practical workhorse of continuous crystallization. In an MSMPR vessel, the slurry is assumed to be well mixed, so that crystals of every size experience the same supersaturation environment, and the product stream withdrawn from the vessel has the same composition as the well-mixed contents. This idealization yields a particularly elegant theoretical consequence: the steady-state crystal population density distribution, when plotted on a logarithmic scale against particle size, becomes a straight line whose slope is directly related to the ratio of growth rate to residence time. This relationship has allowed generations of crystallization researchers to extract nucleation and growth kinetics directly from steady-state particle-size measurements, and it underpins the population-balance modeling framework used in the present study. At the same time, the well-mixed assumption introduces the very limitation that staging seeks to address, because a broad residence-time distribution means some crystals spend far longer in the vessel than others, and back-mixing can carry fine particles into environments where they continue to compete for supersaturation.

The tension at the heart of single-stage operation deserves elaboration. Supersaturation, the thermodynamic driving force for crystallization, must be generated somewhere in the process, whether by cooling, antisolvent addition, or evaporation. In a single vessel, that driving force is spent on competing processes: nucleation of new crystals, growth of existing ones, and, in polymorphic systems, the selection and maintenance of the desired solid form. High supersaturation favors nucleation, which multiplies the crystal population and divides the available solute among more particles, yielding smaller crystals with wider distributions. Low supersaturation favors growth on existing crystal surfaces, which produces larger and more uniform particles but leaves more solute unconverted and reduces the productivity of the equipment. These are not merely engineering inconveniences; they reflect the fundamental kinetics of how crystals are born and how they grow, and no single set of operating conditions can optimize both simultaneously in one vessel.

Staging resolves this conflict by temporal separation rather than by compromise. In a two-stage configuration, the first vessel can operate at conditions that favor controlled nucleation and early growth, while the second vessel operates at lower supersaturation, where the crystal population that survived the first stage can grow to its target size without the disruptive generation of new nuclei. The comparison reported for paracetamol and carbamazepine is methodologically significant precisely because it holds total active volume and overall residence time constant between the single-stage and two-stage configurations. Without this constraint, any improvement observed in a staged system could simply be attributed to the additional equipment volume or longer processing time. By matching these variables, the study isolates the effect of how supersaturation is distributed in time and space, and the results make a clear argument: the benefit of staging lies in the controlled distribution of the driving force, not in added capacity.

The choice of model compounds strengthens the generality of these findings. Paracetamol crystallizing as Form I represents a relatively well-behaved cooling crystallization system, one that has become a standard benchmark in crystallization research because its solubility behavior, metastable-zone width, and growth kinetics have been extensively characterized. Carbamazepine crystallizing as Form III presents a more demanding challenge, because carbamazepine is notorious among pharmaceutical polymorphs for the complexity of its solid-form landscape, with multiple anhydrous and solvated forms that can interconvert depending on temperature, solvent composition, and mechanical history. Antisolvent-assisted cooling crystallization of carbamazepine must therefore manage not only particle size but also the risk of nucleating or transforming into an undesired polymorph. That the two-stage configuration improved polymorphic selectivity in both systems, despite their contrasting crystallization behavior, suggests that the underlying mechanism, namely the suppression of secondary nucleation through staged desupersaturation, operates across chemically distinct APIs.

The experimental design employed here reflects the maturing methodology of quality-by-design in pharmaceutical development. The Box-Behnken design is a response-surface approach that efficiently samples a four-factor experimental space with fewer runs than a full factorial design while avoiding the extreme corner combinations that might push the process into inoperable regimes. By applying independent designs to each compound, the study acknowledges that paracetamol and carbamazepine respond differently to the same process levers, and that a compound-specific operating space is more informative than a generic one. The four factors chosen, residence time, relative supersaturation, seed loading, and power input per unit volume, span the principal degrees of freedom available to a crystallization engineer. Seed loading deserves particular attention, because continuous crystallization differs fundamentally from batch operation in how seeding is implemented: rather than charging a single dose of seed crystals at the start, continuous operation requires a continuous stream of seed material, which raises practical questions about seed preparation, seed quality, and the long-term consistency of the seed supply.

The population-balance modeling component illustrates both the power and the requirements of mechanistic crystallization modeling. A population-balance equation describes how the number of crystals in each size class evolves over time under the combined influences of nucleation, growth, breakage, agglomeration, and removal. When calibrated against experimental data and, critically, validated against data not used in the fitting process, such models can interpolate within the experimental domain and support scenario analysis for operating conditions that were not directly tested. The reported coefficients of determination between 0.894 and 0.943, with normalized root-mean-square errors between 1.8 and 8.8 percent, indicate that the models captured the dominant behavior of both systems with useful quantitative fidelity. The emphasis on external validation is important, because crystallization models calibrated on one dataset can appear deceptively accurate while failing to generalize, particularly when kinetic parameters are correlated or when the model structure omits relevant phenomena such as breakage or size-dependent growth.

The multimodal process analytical technology deployed in the study reflects a broader trend toward real-time monitoring in continuous pharmaceutical manufacturing. Attenuated total reflection Fourier-transform infrared spectroscopy and Raman spectroscopy probe the liquid and solid phases respectively, allowing dissolved concentration and polymorphic content to be estimated without withdrawing samples. Focused-beam reflectance measurement provides chord-length distributions, which are related to but not identical with true particle-size distributions, since the measured chords depend on particle shape, orientation, and the concentration of the suspension. The careful validation of these online measurements against independent offline references, such as laser diffraction and microscopy, is what transforms them from trend indicators into quantitative process monitors. This validation work, though less glamorous than the optimization results themselves, is essential for regulatory acceptance of real-time release testing and for the closed-loop control strategies that continuous crystallization ultimately promises.

The attention paid to fouling and downstream processability distinguishes this work from much of the continuous-crystallization literature, which has historically prioritized yield and particle size. Fouling, the accumulation of crystal deposits on vessel walls, impellers, and heat-transfer surfaces, is a persistent operational challenge in continuous crystallizers. It degrades heat transfer, alters the effective working volume, disturbs mixing and residence-time distribution, can blind optical PAT probes, and eventually forces shutdown for cleaning, undermining the very continuity that justifies continuous operation. The finding that two-stage operation reduced fouling by roughly 42 to 46 percent for the two compounds is therefore operationally meaningful, and the mechanism is plausible: by lowering the supersaturation experienced in each individual vessel, staging reduces the thermodynamic driving force for heterogeneous nucleation on surfaces, which is a principal pathway for fouling formation.

The downstream consequences of particle attributes are equally concrete. Filtration time declined by 26.3 percent for paracetamol and 34.4 percent for carbamazepine when the two-stage configuration was used, a direct consequence of the larger median particle sizes and narrower distributions. The relationship between particle size and filtration follows from classical cake-filtration theory, in which the specific cake resistance varies inversely with particle size, so that modest increases in median size can produce disproportionate reductions in filtration time. Narrower distributions further improve washing efficiency and reduce mother-liquor retention, which affects both drying time and final product purity. Elongated or needle-like crystals, which often result from high-supersaturation conditions, pack poorly in filter cakes and create high resistance; the reported improvements in morphology under staged operation therefore compound the benefits of size alone.

The trade-off between product quality and space-time yield deserves honest acknowledgment. Longer residence times improved particle size, distribution width, and polymorphic content, but reduced space-time yield, meaning that a given equipment volume produced less material per unit time. This is not a flaw in the study but a genuine feature of crystallization physics: crystal growth takes time, and time spent growing is time during which the vessel is not processing new feed. The practical resolution of this trade-off depends on which constraints dominate a given manufacturing scenario, whether the bottleneck is downstream filtration capacity, equipment volume, or product quality specifications. The value of the response-surface models developed here is precisely that they allow such trade-offs to be quantified and navigated rather than discovered by trial and error.

The preservation of chemical purity and mass-balance closure across all configurations provides an important baseline assurance. Mass-balance closure, the accounting identity that solute entering the system must equal solute leaving in product and mother liquor, is a stringent test of experimental rigor in continuous operation, because it requires accurate measurement of flows, concentrations, and solids contents over extended steady-state periods. That closure was maintained while fouling was present in the single-stage configuration suggests careful experimental technique, and it lends credibility to the comparative conclusions. Similarly, the extended polymorphic stability testing addresses a concern specific to continuous operation: because continuous crystallizers operate at steady state for long periods, any slow polymorphic transformation that would be invisible in a batch process could accumulate over time and compromise the solid form of the product.

Looking beyond the two compounds studied, the findings suggest a design principle for continuous pharmaceutical crystallization that is likely to generalize: the distribution of supersaturation across stages is a primary design variable, comparable in importance to residence time or temperature profile. The authors are appropriately measured in their claims, presenting the work as a reproducible demonstration for two APIs with contrasting behavior rather than a universal framework. Extending the approach to compounds with different solubility temperature-dependence, different growth kinetics, or different polymorphic landscapes would test the boundaries of the principle, as would configurations with more than two stages or with interstage seeding or fines dissolution. The combination of structured experimentation, externally validated mechanistic modeling, multimodal PAT, and operational metrics such as fouling and filtration time offers a template for how continuous crystallization research can connect molecular-level phenomena to manufacturing-relevant outcomes, and it is this end-to-end perspective that gives the study its practical significance for the pharmaceutical industry’s ongoing transition toward continuous manufacturing.

Subject of Research: …

Article Title: Continuous crystallization of pharmaceutical compounds in mixed suspension mixed product removal systems

Article References: Continuous crystallization of pharmaceutical compounds in mixed suspension mixed product removal systems. (n.d.). https://doi.org/10.3389/fchem.2026.1947756

Image Credits: AI Generated

DOI: 10.3389/fchem.2026.1947756

Keywords: continuous crystallization, MSMPR, pharmaceutical manufacturing, paracetamol, carbamazepine, polymorphism, population-balance modeling, process analytical technology, particle-size distribution, fouling, filtration, quality by design