Crystal clear batch
Fiber optic probe helps control batch crystallization in pharmaceutical applications
By Richard S. Harner and Nick D. Gipson
The pharmaceutical industry uses batch crystallization routinely for separation of intermediates and purification of the active pharmaceutical ingredient (API). The process is ubiquitous within FDA approved protocols, but it is critical to maintain fine control of the crystallizer operation. Small changes in the raw materials, poorly controlled process variables, and certain impurities can affect the final crystal habit and all downstream unit processes. One solution, a fiber optic probe technology, enables nucleation and dissolution detection as well as an indirect measurement of super-saturation (the driving force for crystal growth), which are valuable capabilities in designing crystal-lization/purification systems, scaling-up a defined process, and monitoring batch-to-batch variability in a large-scale, high-volume production environment. (See details on fiber optic probe in the related sidebar, “New fiber optic probe use revealed.”)
Other changes in raw materials could include slurry exhibiting a poor particle size distribution (PSD), which could double or triple the time required to separate the phases through filtering. Even worse, it might create the wrong polymorphic form, triggering a rework of the batch or possible scrapping of the product. Undetected, it could lead to poor bio-availability of the API or complete loss of efficacy. Fiber optic technology enables users to identify, quickly scale up, and reproducibly execute an optimal crystallization process.
Clearing up crystallization
Crystallization is an energy-efficient process for creating a very pure solid from impure mother liquor. The process typically creates slurry (a two-phase mixture), where the API solids can efficiently separate from the API depleted liquid phase using common techniques, including filtering and centrifugation. Crystallizations can have a variety of purposes—to isolate a solid, to purify a product, or to create a consistent particle size, as in a final step for API. Regardless of the key goal, it is important to understand the fundamentals driving crystallization (solubility curve, super-saturation limit, and kinetic affects) to fully optimize the process. Once you determine the limits, you can control a crystallization system using a fiber optic probe response to consistently create the desired result.
Planting a seed
In one case study, a traditional seeded crystallization profile produced a bulk pharmaceutical API already in clinical trials. We added seeds after generating a certain level of super-saturation and holding temperature constant to allow full development of the initial crop. We accomplished the main growth phase by controlled cooling. A fiber optic probe signal indicates the degree of completion of each of these three steps. This process yielded large and pure crystals that met users’ existing product specifications. However, formulations development work indicated users needed a smaller particle size to create the required drug release profile. A subsequent milling study revealed futile efforts to mechanically grind the material to meet the desired PSD. We needed a smaller bulk API directly from the crystallizer to feed to the mill.
We now had a new process to modify the desired particle size in the critical path to drug approval. We quickly initiated studies to evaluate the solubility, cooling rates, and nucleation characteristics using a new fiber optic technology at the lab scale.
We initially trialed a much simplified cooling profile (without the need of seeding) and found it to give material somewhat smaller than the desired PSD. The faster cooling curve produces a high level of super-saturation and a large crop of small crystals due to spontaneous nucleation. While these nuclei do grow throughout the balance of the cooling curve, the final PSD is much smaller than the original recipe that produced large crystals, which filtered easily but failed formulation requirements. We modified the cooling profile to include a partial digestion after nucleation.
Digestion is a critical step, using Ostwald Ripening, a phenomenon that relieves the level of super-saturation temporarily to allow the smallest nuclei to dissolve and grow the larger nuclei, which receive thermodynamic favor due to their greater volume-to-surface ratio. This also alters the PSD of the nuclei. To further shift the PSD, it is possible to use multiple digestion steps, although they can be time consuming. A fiber optic probe makes it possible to control digestion. The probe responds primarily to total surface area of solids. Although touted strictly as a control device, it reliably generates an in-situ piece of information that is otherwise unknown during crystallizer operation.
It detects nucleation via a patented fiber optic response, then the slurry heats under jacket control to reduce super-saturation. The fiber optic response continues to increase, reaches a maximum, and starts to decrease as the Ostwald effect shifts the instantaneous PSD of the nuclei. After reaching a set point percentage (calculated percentage of the recorded maximum), it initiates a new cooling profile or resumes the original profile. In the latter case, a 50% digestion step yielded a tightly controlled PSD with a mean size [D(v,0.5)] increased by a factor of two. We could now achieve the desired product PSD through a non-aggressive milling step.
Once a lab scale vessel produces acceptable PSD based on a reproducible temperature profile (both jacket and internal), it is important to scale up the crystallization process to equipment suitable for making clinical quantities of the bulk API. One stumbling block for scale-up is temperature profiles cannot transfer directly from a 1-liter lab reactor to a 500-gallon pilot reactor. Large vessels have slower heat transfer rates from jacket to wall and wall to contents. Agitation parameters also change significantly and affect the heat transfer capability. Some level of impurity and carryover (potential nucleation sites) is always suspect in large-scale process vessels with many ports and installed instruments.
The fiber optic probe in use at lab scale is functionally identical to the one suspended in a process vessel. Like a temperature probe, it responds only to the material that passes through its field of detection (a few cubic millimeters adjacent to the window). The jacket temperature profile (the heat transfer driving force) of the large vessel can adjust to generate a fiber optic response profile similar to the lab scale. This procedure generates a similar super-saturation level in the bulk (assuming efficient mixing) and produces the desired product PSD. Following this methodology, the first batch in the pilot reactor achieved the desired PSD. We fine-tuned subsequent batches and realized success in less than six weeks from initial lab development to completion of a pilot scale campaign ready for formulation. Saving months of development and scale-up work enabled an earlier launch of the drug, which translated to $15 million of additional revenue for users.
After a successful pilot run and subsequent clinical trial, we scaled the process again to a commercial production plant (~5x pilot scale). Again, the fiber optic technology was a key in identifying the temperature control settings needed to obtain the same product attributes.
Even if you implement a well-defined and controlled crystallization at the commercial scale, the fiber optic measurement is valuable to track subtle changes in the process. The on-line fiber optic profile of each batch, if compared with previous batches, can provide an early indication of an impending process upset. You can monitor profile parameters and set thresholds for automatic response or alarm. The ability to detect nucleation temperatures or to apply multivariate techniques for real-time profile comparison allows a plant to detect when something has changed.
In one case, a final crystallization yielded product that failed an infrared spectroscopic identification test. The vessel contained a fiber optic probe, but the results did not see use for control and close monitoring did not occur. An isolation transfer error in the previous batch left an unnoticed product heel in the crystallizer. The additional product resulted in an abnormally high initial concentration of API for the next crystallization, which caused favor toward an undesired polymorph. A review of the stored process data indicated the fiber optic profile was noticeably different (possibly indicating a different polymorph). Also, the fiber optic profile during the previous batch indicated a failure to remove all solids. Had the data undergone close monitoring or full automation, users could have avoided both events. This $200,000 incident indicates just how valuable it is to have a fiber optic eyes-in-the-reactor probe for detecting early signs of a process upset in a high-throughput, high-volume, and high-value production plant.
ABOUT THE AUTHORS
Richard S. Harner (RSHarner@dow.com) is an online spectroscopy specialist in process analytical chemistry at The Dow Chemical Company. He is a member of ISA and the ISA12.21 Committee on Fiber Optics in Hazardous Areas. Nick D. Gipson (DGipson@dow.com) is a production coordinator in Dowpharma Designed Polymers at The Dow Chemical Company.
New fiber optic probe use revealed
By Richard S. Harner and Nick D. Gipson
During the past 20 years, a custom fiber optic backscattering probe has seen use to monitor and control batch crystallizations at The Dow Chemical Company. U.S. Patents issued to Dow cover the basic probe design, methodology for detecting the onset of crystallization (nucleation), and an energized gasket sealed probe, which survives crystallizer operation. Over the years, we have implemented other practical improvements, but the basic optical design remains unchanged. However, we have never published specific information about how best to use the probe to facilitate successful crystallization batch after batch, until now.
The Dow fiber optic backscatter probe is a simple but rugged design with no moving parts.
It is a typical ½” diameter, 1½” long body containing multiple optical fibers terminated behind a stepped window. The fibers mount precisely in a two-piece insert at an angle and separation to create an intersection zone in the process at the window surface and beyond. The fibers are optically polished on the flat insert face and geometrically located adjacent to the inner surface of a sapphire window. We spring-loaded an assembly comprising the stepped window, optical fibers, and flat gaskets inside the tubular probe body to create an improved primary process seal capable of withstanding differential component expansion over a wide temperature range of crystallizer operation. While there are no moving parts, it accommodates temperature-induced expansion and contraction of all components without breaching the primary process seal. You can choose internal components for ideal optical, structural, or binding properties because they are protected from the process. Yet the precision optical alignment of fibers and window maintains over the specified process conditions, because reproducibility of the probe response is critical for batch-to-batch control of process crystallizers.
A version of the Dow probe is available commercially, but the advertised application is measurement of turbidity in harsh environments. While you can calibrate the probe in standard turbidity units, and a batch cooling crystallizer is indeed a harsh environment, the probe does not see use to determine conventional turbidity, even at Dow.
We designed this probe specifically for process vessels in which the range and repetition of the temperature change leads to premature failure of other poorly sealed sensor devices. Fiber optic probes have survived for over 10 years in active use when the crystallizer is dedicated to a single API/solvent system. Given the simple optical design (fibers retained behind a stepped window), gasket seals, energizing mechanism, and a body fabricated from nearly any chemically resistant material, many options and configurations are possible.
The process interface is also critical to reliable long-term operation in a chemical production environment. A fiber optic dip pipe sees use routinely to support the small probe inside a commercial scale crystallizer. It must withstand not only the process temperature and pressure, but the mechanical forces generated when agitating dense slurry. Another useful process interface is the fiber optic thermowell. Here, a standard thermowell with ½” diameter tip modifies to incorporate the optical elements and energized seal. You could implement a fiber optic installation point at an existing thermowell port by incorporating a thin film RTD in the insert to preserve the temperature monitoring capability of the well. While Dow fiber optic probes and process interfaces have seen fabrication for diverse applications, including organic acid recycle streams, bio-fermentation reactors, and wastewater sludge clarifiers, only a handful of options see routine use in process crystallizers.