Chapter 13 Special Applications
13.1 INTRODUCTION
While solution crystallization is the main focus of this book, there are applications which are special cases of the crystallization methods described in earlier chapters. For example, a process requiring sterile crystallization is a special example of an antisolvent addition and is presented in Example 13.1 . In this particular example, it is necessary to produce fine particles in order to meet drug dissolution time specifications in the final single‐dose vials. Furthermore, the material is thermally unstable. Therefore, special precautions have to be taken to minimize any degradation during processing.
Integration of crystallization with other process operations, specifically reaction, is another aspect of crystallization that has significant potential to improve overall process economics. In Examples 13.2 and 13.3 , two cases are presented in which products are selectively crystallized during reaction. As a result, reaction selectivity, yield, and the cost of raw materials are significantly improved. Furthermore, solvent recovery is simplified since the same solvent is used in both reaction and crystallization. This type of operation can provide significant improvement in process economics and should be part of the development of crystallization processes when applicable.
Other crystallization techniques that are less frequently applied in the pharmaceutical industry, such as melt and freeze crystallization, may be applicable for some processes. In Example 13.4 , purification of dimethyl sulfoxide (DMSO) is presented. In this case, low‐level impurities, primarily dimethyl sulfide, are removed by controlled fractional crystallization from the melt (DMSO is a liquid above 18.45°C), in combination with adsorption of impurities from the unfrozen liquid. In the feed DMSO prior to the crystallization step, the impurities, while unacceptable, are at too low a level to be removable by adsorption alone.
In Example 13.5 , freeze crystallization of imipenem, which has lower stability in solution at room temperature, is presented. In this process, the product is rapidly frozen at an amorphous solid state to conserve its chemical purity. The temperature is then raised (still below the freezing temperature at this stage), and the amorphous solid converts to a crystalline solid over time. After the completion of the solid‐state transition phase, the lyophilization drying cycle is initiated.
Resolution of optical isomers via preferential crystallization is outlined in Chapter 7 , Example 7.6 , as an example of the use of tightly controlled supersaturation in a cooling crystallization. This process is discussed in greater detail in Example 13.6 . The process for resolution of optical isomers utilizes crystallization kinetics, instead of equilibrium solubility, to accomplish the desired isomer separation. It is a proven technique and has been in long‐term operation at the manufacturing scale. Although this application has a specific focus, the crystallization principles to achieve the required all‐growth conditions are broadly applicable.
Another unique technique is solid dispersion, where active pharmaceutical ingredient (API) is dispersed in pharmaceutical excipients, and the resulting solid dispersion material can exhibit advantageous physical and chemical properties over the original APIs. It should be mentioned that there is another recent approach named API‐excipient co‐processing. The focus is on modifying solid powder properties of API for better formulability via addition of non‐active excipients (Schenck et al. 2020 ). Since solid dispersion involves co‐processing of API and excipients, solid dispersion can be viewed as part of API‐excipient coprocessing in a broader sense. In the book, authors will focus on solid dispersion (Section 13.5 ) with a specific example presented in Example 13.7 .
Other unique crystallization techniques, such as supercritical fluid (SCF ) crystallization and sonocrystallization (ultrasound in crystallization), are also mentioned. Potential applications for these emerging technologies are present in the pharmaceutical industry.
13.2 CRYSTALLIZATION WITH SUPERCRITICAL FLUIDS
SCFs are gases and liquids above their critical point. In this state, they are single‐phase fluids with some advantageous properties of both liquids and gases. These properties enable them to be used in a unique manner to engineer, or design particles with proper process manipulation.
SCF, while possessing high enough density to solubilize many substances, are highly compressible, especially near the critical point. In many processes using these fluids, solubility of the desired compound can be manipulated by relatively minor changes in temperature and pressure. Additionally, materials dissolved in them benefit by possessing much lower viscosities and higher diffusivities than they do in conventional liquids, allowing very rapid solid phase formation. In fact, this may be so rapid that pains must be taken to avoid the creation of an amorphous, rather than a crystalline, solid (assuming crystallinity is desired).
A comprehensive review of particle design using SCF is provided by Gupta (2006 ), York et al. (2004 ), and Jung and Perrut (2001 ). York (1999 ) and Subramaniam et al. (1997 ) give excellent reviews on use of this technology with pharmaceutical compounds.
SCF processing on a large scale had its earliest success with extraction processes. The decaffeination of coffee and tea is carried out in tonnage quantities, and methylene chloride residues are thus eliminated. Extraction of essential oils and flavors has been commercially successful. Because SCF often have only limited solvation properties for some desired materials, nonsupercritical solvents can be added into the process. Because of its low toxicity and relatively low temperature and pressure critical point, carbon dioxide is the dominant SCF used in the pharmaceutical industry. At its best, an SCF process can produce, in a single step, a pure, dry crystalline solid with high productivity.
The common types of SCF processing and their common acronyms, are:
Rapid expansion of supercritical solutions (RESS): The dissolved product is nucleated by rapid expansion through a nozzle.
Gas (or SCF) antisolvent (GAS or SAS): SCF is used as antisolvent.
Aerosol spray extraction system (ASES): Very small droplets are sprayed into SCF antisolvent to produce, micro‐ or nanoparticles.
Particles from gas‐saturated solutions (or suspensions) (PGSS): SCF is dissolved in liquid product or solution in solvent, followed by rapid depressurization.
One variant of the GAS or SAS process (SCF as antisolvent) is solution enhanced dispersion by SCFs (SEDS). Coaxial nozzles are used to introduce drug solution and carbon dioxide at the desired temperature and pressure. In this case, the SCF carries out both droplet breakup and antisolvent functions. SEDS has been tested for a number of pharmaceutical compounds. As noted above, this is a continuing effort.
13.3 RESOLUTION OF STEREO‐ISOMERS
In recent years, advances in selective stereochemistry have reduced the need for isomer separation. However, many synthetic processes continue to have racemic output (desired and undesired stereoisomers). Separation of these products reduces the possibility of side effects from the inactive isomer. In addition to the reduction of side effects, separation of these isomers often provides the possibility of greatly improved yield if the inactive entity can be racemized and recycled.
13.3.1 Option 1: Use of a Chiral Additive to Create a Diastereoisomeric Set of Compounds
Diastereoisomers have two chiral centers, and those created with the compound being separated will not have the same solubilities and other properties as the original isomers.
This is the most common option exercised in practice. However, it can be costly, because the chiral compound being added to carry out the separation can be expensive. Also, for resolution of racemic mixture which contains 50/50 stereoisomer ratio, the maximum yield would be 50% which is low and can be economically unacceptable.
13.3.2 Option 2: Chiral Chemistry to Improve Reaction Chiral Selectivity of the Desired Isomer
This is a very desirable solution and has attracted very extensive research efforts over decades (for example Akiyama and Ojima 2022 ; Nag 2018 ; Gawley and Aube 2012 ). This approach can greatly improve the chiral purity in chemical synthesis for example from 50 to 90% or higher. Under this scenario, traditional crystallization technique such as cooling, antisolvent, etc., with or without the chiral resolving agents can be applied in a straightforward manner for chiral purification. As an example, the desired isomer can be crystallized out and undesired isomer together with an equal amount or slight excess of desired isomer will remain in the mother liquor.
13.3.3 Option 3: Kinetic and Dynamic Resolution
For the case of kinetic resolution, the desired stereoisomer, for example D, is preferentially crystallized out from the solution in D crystallizer (Figures 7.26 and 7.27 ). The undesired stereoisomer, for example S, will also be preferentially crystallized out from the solution in L crystallizer (Figures 7.26 and 7.27 ).
For the case of dynamic resolution, the desired stereoisomer, for example D, is preferentially crystallized out from the solution in D crystallizer. The undesired stereoisomer, for example S, is epimerized in the solution and in equilibrium with R in the dissolver. Since R is preferentially crystallized out, all R/S in the system will convert R over time in D crystallizer (Pellissier 2008 , 2022 ). The process scheme will be very similar to Figure 7.25 (or Figure 7.24 ).
On the crystallization side, because enantiomers R or S can have the same thermodynamic and kinetic properties, including solubility, crystal growth and nucleation characteristics in nonchiral solvents, special care in maintaining preferential crystal growth of R, while maintaining metastability of supersaturated S in the crystallizer is a must. More details will be provided in Example 13.6 to address these issues.
13.3.4 Option 4: Use of Chromatography, Membrane, Enzyme, or Other Separation Technology
This is also desirable but often costly.
13.4 WET MILLS IN CRYSTALLIZATION
The use of wet mill has been extensively addressed throughout this book (Chapters 4 , 5 , 6 , 9 , and 10 ). Wet mills, for example sonicator or rotor–stator homogenizer, can be used to break up crystals and alter the crystal morphology. By breaking up crystals, wet mill can effectively shorten the aspect ratio of crystals and improve many solid flow properties. In addition, by using wet mill to initiate nucleation in a controlled manner, for example in situ seed generation, it can facilitate the massive seeding and subsequent crystal growth. The improvement in product size distribution and reduced agglomeration can potentially result in fewer inclusions of impurities and a number of advantages in mechanical handling.
As highlighted below, wet mills can be used beneficially in several key areas of crystallization:
Initiation of primary nucleation, narrowing the metastable zone width, especially for massive seeding.
Secondary nucleation
Crystal habit and perfection
Crystal size distribution
Reduced agglomeration
Improved product handling
There are various types of wet mills, for examples bead mill, sonicator, rotor–stator homogenizer, high pressure cavitation mill, etc. The mechanism for nucleation and particle breakage may not be the same. Sonication (Suslick 1988 ; Young 1989 ; Suslick et al. 1990 ) or high pressure cavitation mill create local high intensity cavitation. Bead mill breaks the particles by mechanical grinding, and high speed rotor–stator create high local shear and frequent collision among particles and rotor/stator to initiate nucleation and particle breakage.
Among these mills, sonicator and rotor–stator homogenizer are probably two most commonly reported tools. Example 7.3 , 7.6 , 9.7 , 10.1 , 10.2 , and 10.4 highlight the usage of sonicator and rotor–‐stator homogenizer wet mills. Both tools can be easily set‐up in the laboratory for quick investigation. However, the design guidelines for large‐scale operation is more straightforward for rotor–stator homogenizer, and less defined for sonicator, specifically the sonication power intensity, intensity zone, residence time in the recycle loop, overall time cycle, etc. Another concern of sonicator is probe shedding. The sonication probe tip, normally made of titanium, can suffer from pitting and erosion during use, which requires special attention especially for long term usage.
Example 13.6 provides additional details on the application of sonicator at the manufacturing scale to address various aspect of scale‐up, including the issue of probe shedding. It is possible to avoid the complication of shedding of horn metal into the product. This is accomplished by (i) controlling power density at the horn surface (W/ cm2 ) and (ii) providing proper maintenance of the horns, including periodic inspection of the probe(s) and machining away tip erosion. Erosion on an ultrasonic horn occurs more rapidly on an already damaged surface.
Further references on ultrasound in crystallization have been reported [by, among others, Ruecroft et al. (2006 ), McCausland et al. (2001 ), Thompson and Doraiswamy (2000 ), Price (1997a ), Anderson et al. (1995 ), Price (1997b ), Martin et al. (1993 ), and Hem (1967 )].
13.5 COMPUTATIONAL FLUID DYNAMICS IN CRYSTALLIZATION
Computational fluid dynamics (CFD) is concerned with obtaining numerical solutions to fluid flow problems by using computers. The equations governing the fluid flow problem are continuity (conservation of mass), Navier–Stokes (conservation of momentum), and energy equations. These form a system of coupled nonlinear partial differential equations (PDEs). Solving a particular problem generally involves first discretizing the physical domain that the flow occurs in, such as the interior of a stirred tank. Commercial CFD software, such as FLUENT, is currently available. It has been used to assess the effect of mixing variables and system geometry on mixing performance. Figures 5.9 and 5.10 shows examples of applying CFD to illustrate the mixing in stirred tank, Figure 7.13 shows one example of applying CFD to calculate the flow patterns of two streams in a mixing elbow.
CFD is becoming a more useful tool for describing solid–liquid mixing in crystallization. It can predict the flow patterns, local solids concentration, and local kinetic energy values, taking into account the effects of vessel and agitator shape. Since additional particle population balance equations including crystal growth and nucleation kinetics are required, simplifications such as ignorance of the impact of solid particles on the fluid flow are inevitably applied (Myerson 2001 , chapter 8 ; Wang and Fox 2004 ; Woo et al. 2006 ). Therefore, CFD results need to be examined against the actual experimental data before accepting fully their validity. Despite its limitation, CFD is an excellent educational tool for quick learning and a rapid screening/ diagnostic tool for process development.
13.6 SOLID DISPERSION—CRYSTALLINE AND/OR AMORPHOUS DRUGS
Solid dispersion is a unique material science and particle engineering technique which API is co‐processed and dispersed with pharmaceutical excipients in solid state. It can be prepared by melting (fusion), solvent, or melting‐solvent methods, as well as other nontraditional technique, such as co‐grinding (Chiou and Riegelman 1971 ; Kaneniwa and Ikekawa 1975 ). Solid dispersion generates materials with properties significantly different from the original APIs. In particular for water‐insoluble APIs, the resulting solid dispersion material can greatly improve the water solubility (Jermain et al. 2018 ). It also improves drug–drug and drug–excipient compatibility (Nie et al. 2017 ). Blending/mixing of a drug or drugs in a solid diluent or diluents by traditional mechanical mixing is not considered as a solid dispersion.
Solid dispersions can be broadly divided into two categories: crystalline‐based and amorphous‐based solid dispersions (Leuner and Dressman 2000 ; Craig 2002 ; Singh et al. 2011 ; Tung 2018 ). The crystalline‐based dispersion includes eutectic‐mixture, solid‐solution, and (nano)crystalline‐suspension, where API exists in crystalline state. Chapter 2 contains more discussion on solid solution and eutectic mixture. For simplicity, we will focus more on eutectic mixture in the following discussion. The amorphous‐based dispersion include glass‐solution and glass‐suspension, where API exists in amorphous state. Glass‐solution is typically called as amorphous solid dispersion, i.e. ASD.
Crystalline‐based solid dispersions generally possess better stability but inferior dissolution performance in comparison to amorphous‐based solid dispersion (Brough and Williams 2013 ). For eutectic‐mixture, the drug loading is limited by eutectic composition which is typically <50 wt% and there is no specific control of API particle size (Law et al. 2003 ; Cherukuvada and Nangia 2014 ). If the drug loading is above the eutectic composition, it can lead to nonuniformly distributed crystalline API and additives, which will reduce bioavailability. For (nano)crystalline‐suspensions, drug or API loading can be greater than 50 wt% (Kawabata et al. 2011 ). However, achieving uniformly dispersed nano/micro size API particles with minimum particle agglomeration under high drug loading can be challenging (Bhakay et al. 2018 ).
Amorphous‐based solid dispersions generally possess better dissolution performance due to the amorphous nature of API. However, amorphous‐based solid dispersions have higher degree of instability. If drug loading is higher than API’s solubility/miscibility in the additive, it maybe kinetically stable over a certain period, but is thermodynamically unstable, and can result in phase separation or uncontrolled crystallization of API during storage, and/or in vitro/in vivo dissolution. Consequently, drug loading is generally <50 wt% in order to reduce the risk of instability (Shah et al. 2014 ; Newman 2015 ).
Hybrid solid dispersion, which can consist of both crystalline‐based and amorphous‐based solid dispersion is also explored recently (Rahman et al. 2019 ; Tung 2021 ). Hybrid solid dispersion offers an opportunity to overcome the constraints of solid dispersion, i.e. drug loading and instability, and in the meanwhile keep the benefits of solid dispersion, i.e. better dissolution/bioavailability and drug–drug compatibility. Example 13.7 presents one case study using ibuprofen as one model compound.
13.7 PROCESS DESIGN AND EXAMPLES
EXAMPLE 13.1 Sterile Crystallization of Imipenem
Goal : Achieve desired physical properties and chemical purity
Issues : Sterile operation, narrow particle size distribution (PSD) small mean particle size (20 microns)
The final step in a multistep synthesis of imipenem, an injectable antibiotic, is a sterile recrystallization.
COMPLICATING FACTORS
In this example, control of growth and narrow PSD are essential because the dried product is filled in single‐dose vials and then redissolved for injection at the time of use. The redissolution step in the vial must be rapid (<20 seconds in an aqueous solution) to meet specifications. Therefore, the growth strategy must be designed to achieve a maximum particle size of <20 μm and a narrow PSD to satisfy the rapid dissolution requirement. Milling to reduce particle size is not a desirable option because of temperature instability of the molecule, the need to avoid particulate contamination from the milling operation, and the complexities of sterile milling (delumping the dried product is acceptable because of the lower energy input).
In addition to these specifications, the downstream operations of filtration, washing, and drying require a filterable crystal with low solvent retention in order to facilitate drying at low temperature.
PROCESS DEVELOPMENT
These requirements impose conflicting restraints on the development of a manufacturing process. The common goal of achieving maximum growth for downstream operation efficiency is not consistent with the goal of controlling mean particle size at 20–30 microns. The most difficult limit to control could be the upper limit of particle size since dissolution of particles or agglomerates 40 microns would be prohibitively slow.
The temperature stability profile of imipenem effectively rules out cooling or evaporative crystallization. Antisolvent addition is the obvious choice. The preferred solvent is water because of the need for sterile filtration of the solution to be fed to the crystallizer. A good antisolvent is known to be acetone. In this case, the antisolvent has the important function of reducing the metastable zone width as well as the solubility.
PREPARATION OF AN AQUEOUS SOLUTION
Issue : Low solubility in water at operating temperatures. The solubility curve is shown in Figure 13.1 , including an estimate of the width of the metastable zone in aqueous solution.
Figure 13.1 Solubility curve in aqueous solution showing the width of the metastable zone.
Feed Preparation, Option 1
A solution of 10 g/ l can be made and concentrated to 30 g/ l by reverse osmosis (evaporation of an aqueous solution is not viable because of temperature instability).
This solution is supersaturated, as can be seen, but is well within the metastable zone, as shown in Figure 13.1 . This stability under supersaturated conditions is critical for concentration by reverse osmosis and for the following sterile filtration (0.22 micron sterilizing filters). Following sterile filtration, the antibiotic can be crystallized by addition of acetone.
Option 1 has the serious disadvantage of low yield because of the limit on time and the limit of temperature on concentration by reverse osmosis at the manufacturing scale to prevent unacceptable decomposition. After addition of acetone to the concentrate to crystallize the product, the solubility is unacceptably high for yield purposes.
Feed Preparation, Option 2
A second option is possible only because of the high supersaturation that can be achieved with imipenem in aqueous solution.
In this option, a higher concentration, 60 g/ l at 60°C, is achieved by a heat/ cool system, as shown in Figure 13.2 . The short contact time for heating to achieve dissolution and cooling to a stable temperature, ~one minute total, is consistent with chemical stability.
The high supersaturation at filtration temperature, S = 5 at 15°C, is acceptable because of the increased width of the metastable zone in aqueous solution combined with a longer induction period—also in aqueous solution. During this time, sterile filtration to the crystallizer without nucleation can be completed—a gift of nature to the development team. Note, however, that premature nucleation from aqueous solution in the feed line to the sterile filter will cause slow sterile filtration and jeopardize successful completion of the run.
Figure 13.2 Flowsheet for Example 13.1 ; feed preparation and crystallization of an antibiotic under sterile conditions.
Figure 13.3 (a) The very large cubic crystals have grown in aqueous solution where the smaller ones were formed after acetone addition. (b) SEMs of the final crystals at 100×. (c) SEMs of the final crystals at 2000×.
In addition, if nucleation from an aqueous solution occurs in the crystallizer—either by excess time before addition of acetone or by subcooling below 15°C in the crystallizer—the resulting crystals will have different crystal morphology. The crystals from an aqueous solution grow rapidly, once nucleated, and produce crystals that dissolve slowly and thereby fail to meet the dissolution time and PSD specifications. An example of the crystals that can nucleate and grow in aqueous solution—before acetone addition—is shown in Figure 13.3 a, where the very large cubic crystals have grown in an aqueous solution, whereas the smaller ones were formed after acetone addition.
Note, however, that these agglomerates show evidence of growing out from a core of an aqueous derived crystal or from drops of oil that are known to form readily before the onset of crystallization.
Note : The initiative of the Merck Plant in Elkton, Virginia, in obtaining Figure 13.3 and the ones to follow, is gratefully acknowledged. These scanning electron micrographs (SEMs) are from recent (2004) production batches of imipenem produced in the sterile area of the plant.
The large crystals that appear in the crystallizer cannot be removed and would cause failure of the dissolution time and PSD specifications. Milling and screening cannot be utilized to break or separate the large crystals because of the realistic restrictions of sterile operation.
Crystallization
The imipenem can now be crystallized by the addition of acetone.
Crystallization Option 1: Slow Addition without Seed
All attempts to nucleate without seed failed because of the initial formation of oil followed by uncontrolled formation of amorphous product. This option was pursued because of the complications of sterile powder seeding but without success.
Crystallization Option 2: Addition of Acetone with Seeding and Controlled Growth
Seeding with product from a batch prepared for this purpose was initially utilized, with satisfactory results. However, the PSD was somewhat broader than optimal.
Crystallization Option 3: Use of Milled Seed
As indicated above, milling of this sterile product presents many regulatory and operating problems. An intramuscular dosage form requiring 5–10 micron particles was subsequently introduced, however, requiring milling. Milled, sterile material with a reproducible PSD then became available for seed.
Final Process
Acetone, 7% by volume, is added to the sterile‐filtered aqueous solution. This amount of acetone dramatically narrows the width of the metastable zone.
Milled seed (0.4%) is added. The ideal amount of seed is calculated to provide the approximate number of 5–10 micron particles that will double in size during essentially all growth (see Chapter 5 ).
The remaining acetone is added over a one hour period to achieve a final acetone/ water ratio of 65/ 35.
The crystals are filtered, washed, dried, and delumped, and the bulk product is dry‐filled into single‐dose vials.
The uniform shape and size achieved by this process are required for these downstream steps to be successfully completed. Key issues are filtration rate, low retention of solvent by the wet filter cake (<10%), drying at <25°C, and flowability of the delumped crystals. SEMs of the final crystals at 100× and 2000× are shown in Figure 13.3 b, c.
This process is also successful in achieving growth of crystals that exhibit solid‐state stability. Figure 13.4 shows photomicrographs of crystals of this compound that have different stabilities, as indicated. A difference in the sharpness of the end faces is apparent, which is believed to be the cause of this difference. These crystals were grown in isopropanol/ water.
Figure 13.4 Photomicrographs of antibiotic crystals; the sharpness of the crystal surfaces is an indication of solid‐state stability.
EXAMPLE 13.2 Enhanced Selectivity of a Consecutive–Competitive Reaction by Crystallization of the Desired Product During the Reaction
BACKGROUND
As introduced in Chapter 10 , Section 10.3.8 , the selectivity of a chemical reaction may be increased above that predicted from its rate constant ratio if the reaction is run under conditions that cause the product to crystallize while the reaction is in progress. A related example is presented in Example 13.3 . A common system that can be modified by crystallization is the classic consecutive–competitive reaction as follows:
where R is the desired product and S is the over‐reaction, undesired product.
For a consecutive reaction that is pseudo‐first order in R , the effect of crystallization of R is to reduce the rate of reaction to the undesired over‐reaction product S by reducing the concentration of R in the reaction mixture as the reaction proceeds. The topic of increasing reaction selectivity in the presence of, or by the creation of, a second phase, e.g. by crystallization, has been discussed by Sharma (1988 ) and Paul et al. (2003 , chapter 13 ).
The qualitative effect of crystallizing R during the reaction is shown in Figure 13.5 , where the solution concentrations of the reaction components are shown as a function of mole ratio as B is added to A. In this solvent system, the solubility of R is sufficient to prevent crystallization. Under these conditions, the system is following a homogeneous reaction pathway and the maximum yield of R is determined by the mole ratio and the rate constant ratio, as indicated in Eq. (13.1) and as shown in Figure 13.5 a.
where κ = k 2 /k 1 and Y exp is the expected yield under conditions of perfect mixing.
Figure 13.5 Concentration profiles for a consecutive–competitive reaction showing the effect without (a) and with (b) the crystallization of the desired product, R , and the resulting change in overreaction to S .
Figure 13.6 Effect of the solubility of the desired product, R , in the reacting system on the yield/ selectivity of R .
Note : Equation 13.1 applies to perfectly mixed conditions. A decrease in the yield of R (increased S ) can result from fast reactions with imperfect mixing, as discussed by Baldyga and Bourne (1999 ) and Paul et al. (2003 , chapter 13 ). Reactions that are fast enough to be affected by mixing can also be improved by crystallization, but the results may be a function of the mixing issues discussed in these references.
If the solvent system can be changed, or the reaction run at higher concentrations exceeding the solubility of R , R can be crystallized as the reaction proceeds. The selectivity can thus increase, as shown by a decrease in S in Figure 13.5 b. Note that although the amount of R as shown in Figure 13.5 b appears to decrease under these conditions, the R curves are for solution concentration and do not include the amount of R that has crystallized. This effect on yield/ selectivity is also shown in Figure 13.6 .
This application of reactive crystallization to achieve increased selectivity is illustrated below.
PROCESS
Goal : Maximize selectivity (yield) of a competitive–consecutive chemical reaction
Issues : The reaction produces a major product, R , which overreacts to an unacceptable amount of by‐product, S .
A common method to increase selectivity by limiting the mole ratio of B to A cannot be used because the reaction must be forced to >98% conversion of A because A cannot be removed in subsequent steps (Note : This restriction in the B/A ratio is not an issue in Example 13.3 .)
Original Process
One of the reaction steps in a multistep synthesis produces the desired mono‐addition product, R, which overreacts to a bis‐addition by‐product, S in a phase transfer reaction system. The solvent system is water/ isopropyl acetate. The reaction is run by feeding reagent B over a several‐hour period to a two‐phase mixture of the solvents containing the phase transfer catalyst.
As originally run, the amount of solvent was sufficient to solubilize all of the reactants and products throughout the reaction. The rate constants, k 1 and k 2 , were determined and the ratio correctly predicted the expected yield using Equation (13.1) . This yield was achieved experimentally, and as predicted, the amount of bis (S) formed was unacceptably high (~18%).
Alternatives
Change the reaction system to reduce k 2 , thereby increasing the yield of R and decreasing S —Reacting systems were found that achieved reduced k 2 at a favorable rate constant ratio. However, the primary reaction rate was also reduced, as is common for competitive–consecutive reaction systems, and was too slow for manufacturing purposes.
Reduce the conversion of A, thereby reducing S formation, at reduced mole ratios—This method is commonly used if downstream processing can remove and possibly recycle unreacted A. However, in this case, downstream separation was not feasible since A could not be separated from R in the subsequent purification by extraction.
Change the apparent rate constant ratio by crystallizing R to reduce its concentration in solution, thereby reducing S formation—This method was investigated, as it offered the best opportunity for success at the minimum change in operations.
Final Process
The solubilities of R and S in the reaction mixture were determined and an appropriate volume of water/ isopropyl acetate was found that could solubilize S but allow R to crystallize soon after the start of addition of B to A . The reaction is run in the fed batch (semibatch) mode with B added to A. Two critical crystallization parameters are controlled to ensure the minimum concentration of R (maximum crystallization) during the addition:
Seed addition as early in the addition of B as possible (as soon as the conversion of A creates an R concentration in the metastable zone).
Slow addition of B to minimize the concentration of R in solution by allowing time for nucleation and/ or growth (particle size is not a key issue in this case because the R is resolubilized after completion of the reaction for the subsequent purification by extraction. R is crystallized only for minimizing S and maximizing yield). The effect of addition rate on yield/ selectivity is shown qualitatively in Figure 13.7 , where it can be seen that a critical rate can be determined at which the crystallization rate can maintain the minimum concentration of R. At higher addition rates, selectivity decreases as the R concentration increases.
These process changes resulted in a decrease in S from 18 to 3% and a corresponding increase in the yield of R. These results were realized on scale‐up to manufacturing.
Figure 13.7 Effect of the addition time of reagent B on the yield/ selectivity of the desired product, R, resulting from its effect on the crystallization of R.
Message
Crystallization of a reaction product as the reaction proceeds can be effective in increasing reaction selectivity in a complex reaction. It may therefore be advantageous to cause crystallization by changing the concentration or the solvent system, even though there is no other reason to crystallize the product at this point in the synthesis. If crystallization is not possible, addition of an immiscible solvent can also have this “protecting” effect, as discussed in the references above. This process advantage is unique to heterogeneous systems.
EXAMPLE 13.3 Applying Solubility to Improve Reaction Selectivity
Goals : To demonstrate the application of solubility and crystallization to improve reaction selectivity and yield
Issues : Solubility, reaction selectivity, yield, crystallization, filtration, recycle
In the synthesis of a drug candidate, one intermediate monoaldehyde (Mono) was formed by reacting two reagents, 7‐chloroquinaldine (7‐chloro) and isophthalaldehdye (Iso). As shown in Figure 13.8 , reagent Iso has two equally reactive carbonyl groups. One of these can react with the single methyl group of 7‐chloro to form the condensed product mono. Since the product Mono has one remaining reactive carbonyl group, it will further react with the starting material Iso to form an inert byproduct bis‐adduct. This reaction scheme can be modeled by the following consecutive competitive reaction system:
where
A = starting material, isophthalaldehyde (Iso), which contains two reactive carbonyl groups
B = starting material, 7‐chloroquinaldine (7‐chloro), which contains one methyl group
R = product, monoaldehyde (Mono), which contains one reactive carbonyl group
S = by‐product, bis‐adduct (Bis), which contains no reacting group and is chemically inert
k 1 , k 2 = rate constants. In this reaction system, the ratio of k 1 /k 2 is ~2 (Tung et al. 1992 ; Hobbs et al. 1997 ).
Figure 13.8 Chemical reaction of Example 13.3 .
To ensure a high reaction selectivity for this type of reaction, a high A/B ratio is required (Levenspiel 1972 ).
For convenience and consistency of the presentation, the abbreviated names Iso, 7‐chloro, Mono, and Bis will be used in place of the generic A, B, R, and S throughout the rest of this discussion.
Option 1: One‐Pass Process
As shown in Figure 13.9 , in the original process, solvent, Iso, and 7‐chloro were charged to a reactor first. The batch was heated and aged at the reaction temperature above 100°C. After the reaction was complete, the batch was cooled to approximately 90°C and the majority of by‐product Bis was precipitated. The by‐product slurry was filtered to remove the precipitated by‐product Bis. The filtered clear solution was transferred to a crystallizer and slowly cooled to crystallize the product Mono. The product slurry was filtered, and wet cake was washed and vacuum‐dried.
A typical reaction yield of this approach was approximately 73% based upon the charge of 7‐chloro. The charge ratio of Iso/ 7‐chloro was 1.5 to 1. Excess Iso was charged to increase the reaction selectivity. However, unreacted Iso was lost in the mother liquor. The final yield after the crystallization was approximate 67%. The isolated cake contained the desired product Mono with approximately 4–5 wt% Bis‐adduct. A single solvent (toluene) was used in the process.
Figure 13.9 Flowsheet of the original process of Example 13.3 .
Option 2: Two‐Filtration Process with Recycle of Second Filtration Mother Liquor
As mentioned earlier, it is necessary to have a high A/ R ratio to ensure high reaction selectivity. This can be accomplished by increasing the charge of Iso over 7‐chloro over the current procedure and crystallizing Mono during the reaction. Meanwhile, it is also critical to recover the excess charge of Iso as much as possible in order to minimize the loss of unreacted Iso in the mother liquor. To access the feasibility of alternative options, the first step is to obtain the solubility information on Iso and Mono.
Solubility
Solubility data of the starting material isophthalaldehyde and the product monoaldehyde in one solvent system are presented in Figure 13.10 . It is evident that Iso has a much higher solubility than Mono. The solubility difference between the product and the starting material plays a key role in the design of the two‐filtration recycle system. A high ratio of Iso over Mono can be maintained by initially charging a large molar excess of Iso and selectively crystallizing the product Mono during the reaction, while the excess Iso remains solubilized in the reaction mixture. Following the reaction, the recovery of excess Iso is further simplified as a result of the solubility difference. After isolating the cake which contains both the product Mono and the unreacted Iso, the excess Iso can be easily recovered in the filtrate after slurry‐washing the cake.
It should be mentioned that the by‐product Bis has a lower solubility than both Iso and Mono. Because Bis is chemically inert, its presence does not affect the reaction performance and its solubility data are not included in this communication.
Process Description
Figure 13.11 shows the flowsheet of the two‐filtration process with recycle of second filtration mother liquor. For the recycle process, at the beginning of the batch, the starting materials Iso and 7‐chloro were charged to a reactor. The Iso charged included the recycled Iso from the previous batch and the fresh Iso. The batch was heated and aged at the desired reaction temperature. After the reaction was complete, the batch was transferred to a crystallizer and cooled over a period of time. Both Mono and the unreacted Iso, along with the by‐product Bis, crystallized at this stage. The hot filtration of by‐product Bis was omitted due to its reduced formation in the recycle process compared to the original process. The slurry was filtered and the wet cake was washed. The isolated cake contained both Mono and Iso, as well as the by‐product Bis.
Figure 13.10 Solubility of Mono and Iso in the toluene mother liquor of Example 13.3 .
Figure 13.11 Flowsheet of the two‐filtration process with recycle of second filtration mother liquor.
The isolated wet cake after the first filtration was reslurried in a solvent system which dissolved completely the excess Iso and a small amount of Mono. The slurry was filtered and the wet cake washed. This isolated cake after the second filtration contained Mono and Bis, but not Iso. The mother liquors and washes of the second filtration contained all the unreacted Iso. The second filtration mother liquors and washes were concentrated and recycled to the next batch.
As indicated earlier, the by‐product Bis was not removed in the recycle process, primarily due to a much lower level of formation in comparison to the original process. This did not create any additional complication since Bis was chemically inert and could be easily rejected in the next step of the synthesis.
Several key changes were made in the recycle process. First, the level of starting material Iso charged at the beginning of the reaction was increased. Up to three molar equivalents of Iso versus 7‐chloro were charged to the batch versus 1.5 molar equivalents charge in the original process. This change significantly increased the reaction selectivity and yield. Second, a slurry wash/ filtration operation was included in the recycle process to effectively recover the unreacted Iso. Due to the solubility difference between Iso and Mono, a simple slurry‐washing operation recovered all the excess unreacted Iso and reused it in the next cycle. Third, a binary solvent mixture of n ‐heptane and toluene was used which prompted the crystallization of Mono during the reaction and reduced the loss of Mono and Iso in the mother liquor.
Reaction Yield
The experimental results of both the original and recycle processes are listed in Table 13.1 . Both the reaction yield and the isolated yield are reported. The reaction yield is equivalent to the isolated yield plus the mother liquor loss in the first filtration.
As shown in Table 13.1 , the average reaction yield is only 72.2% for the original process. It is 84.4% for the recycle process.
In the recycle process, the increase in the reaction yield is a direct result of increasing the Iso/ 7‐chloro ratio from 1.5 : 1 to 3 : 1, which effectively increases the ratio of Iso/ Mono and crystallization of Mono during the reaction. It should be pointed out that due to the difficulty of sampling and filtration of slurry at the reaction temperature, the solid phase cannot be easily assayed to quantify the percentage of Mono present.
Isolated Yield
As shown in Table 13.1 , the isolated yield of the recycle process is 82.2% versus 66.1% for the original process. As discussed above, the improvement is primarily due to three factors, all of which contribute to a higher reaction yield: (i) a higher ratio of Iso/ Mono, (ii) Mono crystallization, and (iii) a small amount of Mono in the recycle second filtration mother liquor.
It should be pointed out that reduction of mother liquor loss should be carefully balanced against effective rejection of other minor impurities that are also present.
Table 13.1 Experimental results of the original process versus the two‐filtration process.
Process Batch no. Initial mole ratio Iso/7‐Chloro/ Mono Isolated yield, % Reaction yield, % Mono purity, wt% First filtration loss, % Mono Iso Original 1 1.5/ 1/ 0 64.4 69.6 94.7 5.2 35.3 2 66.7 73.2 97.6 6.5 44.0 3 67.4 73.9 95.6 6.5 48.4 Average 1.5/ 1/ 0 66.1 72.2 96.0 6.1 42.6 Recycle 169 3/ 1/ 0.15 80.6 82.7 89.7 2.1 21.2 176 3/ 1/ 0.09 83.5 85.0 90.0 1.5 20.8 178 3/ 1/ 0.14 81.6 83.9 90.2 2.3 24.3 180 3/ 1/ 0.13 82.9 84.9 88.7 2.0 19.0 184 3/ 1/ 0.13 82.4 84.7 89.7 2.3 20.7 190 3/ 1/ 0.13 82.1 84.4 91.0 2.3 20.2 Average 3/ 1/ 0.13 82.2 84.4 89.7 2.1 21.2
Raw Material Iso Requirement
As shown in Table 13.1 , the fresh charge of raw material Iso for each process is 1.5 and 1.22 equivalents versus 7‐chloro, respectively. The excess charge of raw material Iso is reduced from 50% in the original process to 22% in the new process. Since Iso is a key component of the overall process cost, the reduction of Iso requirement provides a substantial saving.
Message
By applying solubility information to the reaction, it is possible to improve the reaction selectivity and overall yield. Process integration of crystallization and reaction can be a very powerful tool to improve the process economics.
EXAMPLE 13.4 Melt Crystallization of Dimethyl Sulfoxide
Goal : Remove the objectionable impurities
Issues : DMSO absorption of water from the atmosphere, DMSO decomposition when stressed, intense mixing required at the liquid–solid interface
DMSO is a well‐known chemical and has been used as a solvent for many materials in industry. Totally dry DMSO is a liquid above 18.45°C. The excellent solvency of this material led to its use in a variety of pharmaceutical applications. In particular, DMSO was found, in the 1960s, to facilitate the absorption of drugs transcutaneously at a very high rate, and a number of pharmaceutical companies investigated the use of DMSO in this manner.
Unfortunately, commercially available DMSO has a characteristically objectionable odor which is caused by low‐level impurities, considerably lower than 1%, largely dimethyl sulfide (DMS) but also a small number of other sulfur‐containing compounds. These compounds have an effect on taste as well as smell. They are particularly measurable by UV absorption at 275 mm. Technical Grade DMSO (pharmaceutically objectionable odor) has a UV275 > 0.25. Lowering UV275 to <0.1 makes DMSO essentially odorless.
PROCESS OPTIONS 1: REMOVE THE OFFENDING IMPURITIES BY ADSORPTION ON A SUBSTRATE SUCH AS ACTIVATED CARBON OR RESIN
This option was tested and did not work as a standalone process because the DMSO was too dilute in these impurities. The measured adsorption isotherms indicated that unrealistically large amounts of adsorbent would be required, even for marginal results.
PROCESS OPTION 2: REMOVE THE IMPURITIES BY A DISTILLATION PROCESS
This is the commercial means used by manufacturers of Technical Grade DMSO. The product of these very carefully controlled distillation processes still contains unacceptable levels of DMS and the other odor‐causing materials. If continual reworking of the pure DMSO stream is carried out, the impurity concentration stabilizes at a low‐level steady‐state value, which is still unacceptable by pharmaceutical standards.
The reason distillation is limited in its ability to reach the very low impurities level required is that the highly reactive DMSO decomposes at a finite rate as boilup heat is provided.
Solution
Freeze out (crystallize) pure solid DMSO, leaving impurities behind in the mother liquors (unfrozen melt).
Remove the impurities from their higher concentration in the mother liquors by an adsorption process with activated carbon.
Return the treated mother liquors to the crystallizer.
The above process (Allegretti and Midler 1967 ) was successful in producing purified DMSO with essentially none of its characteristic odor and UV275 < 0.10 at a yield of ~100%.
Carbon Isotherms
Figure 13.12 a presents room temperature adsorption isotherms for some representative activated carbon samples on Technical Grade DMSO, showing the difficulty of reducing UV275 to the desired level of <0.10 without additional purification. Figure 13.12 a is a schematic of the melt crystallization—carbon column recycle system which was employed to get around this problem. The higher concentration of impurities in the unfrozen melt altered the equilibrium concentration on the activated carbon. In the steady state (Figure 13.12 b) a reasonably sized carbon column could produce effluent suitable for further freeze crystallization, and the yield of the total process was close to 100%.
Freezing Point and Water Removal
DMSO has extreme affinity for water, which was present in feed Technical Grade DMSO at about 0.5% and which was also absorbed from the air despite precautions. For the recycle purification process, water is rejected by the crystalline front of DMSO and accumulates in unfrozen liquid returned to the process. For material balance removal of water, and because the freezing point of DMSO is sensitive to moisture content (Figure 13.12 c), a dehydration step had to be inserted into the recycle loop.
Vapor pressures of DMSO and water are shown in Figure 13.12 d. Based on appearances, they could be separated by a single‐stage batch distillation (under vacuum and in a short time to avoid decomposition). Because of the affinity between these two liquids, however, a single‐stage water stripper removed too much DMSO in the overhead stream. The solution to this problem was to insert a reflux condenser in the vessel with warm water in the jacket to allow return of the entrained DMSO in the vapor stream back to the boilup. In accordance with Figure 13.12 d, the dehydration system pressure was maintained at about 50 mm Hg and the condenser jacket temperature was set at about 50–60°C, which successfully removed the water and maintained the freezing point near 18°C in the crystallizer.
Mixing in the Freeze Crystallizer
Previous discussion in this book, primarily in Chapter 6 , points out the necessity for controlling conditions in the localized region where nucleation and crystal growth processes are taking place. The region of concern in the growing DMSO crystallizer was at the advancing front of the freezing product. In order to ensure effective heat and mass transfer at the growing front, a crystallizer was fabricated (Figure 13.12 e) with an outer jacket and a center tube containing flowing coolant. The temperature driving force for crystallization was held at about 5°C (13°C cooling surface, 18°C freezing point). A double vertical impeller moved the moderate‐viscosity fluid rapidly for the entire length of the center tube and jacket to ensure effective rejection of impurities.
Approximately 50% of the frozen DMSO was crystallized in each cycle. After removal of the unfrozen impurities concentrate, the frozen (odorless, low‐UV275 ) product was thawed and packaged.
Figure 13.12 (a) DMSO and active carbon absorption isotherms at room temperature. (b) Schematic of the melt crystallization with a carbon column recycle system. (c) Effect of water on the freezing point of DMSO. (d) Vapor pressure of DMSO and water. (e) Crystallizer for melt crystallization of DMSO.
Early in development, a roller drum freeze crystallizer was constructed in which the melt container was fitted with an ultrasonic transducer to sonicate the growing DMSO ice front. Although the unit worked mechanically, the ultrasonic energy caused excessive DMSO decomposition, and the more conventional stirred crystallizer described above was fabricated.
Message
Crystallization from the melt, often considered in recent decades for avoidance of solvents, can be effective in manufacturing very high purity product even when solvents are not an issue. The need to pay attention to mass and heat transfer principles, and to provide suitable agitation to meet system requirements, bears many similarities to more conventional crystallization from solution.
EXAMPLE 13.5 Freeze Crystallization of Imipenem
Goals : To investigate the formation of crystalline imipenem from amorphous imipenem during freeze crystallization.
Issues : freeze drying, freeze crystallization, solid–liquid (S–L) equilibrium, amorphism, and crystallinity.
Freeze drying, or lyophilization, is useful in the pharmaceutical industry because compounds which are heat‐sensitive and exhibit poor stability in solution can, in many cases, be freeze‐dried in order to prepare rapidly soluble and sterile injectable pharmaceuticals with minimal degradation. Basic studies of freeze drying include the work of Deluca and Lachman (1965 ) on the physical–chemical parameters of eutectic temperature and solubility, MacKenzie (1965 ) on sublimation, MacKenzie (1977 ) on the physical‐chemical basis for freeze drying, MacKenzie (1977 ) on nonequilibrium freezing behavior of aqueous systems, MacKenzie (1985 ) on the fundamentals and applications of freeze drying, and Franks (1990 ) on predicting successful freeze drying.
Freeze‐dried samples, however, are often found to be amorphous, exhibiting a low degree of crystallinity as determined by X‐ray powder diffraction analysis. For some pharmaceuticals, a low degree of crystallinity decreases stability, as in the case with imipenem, which, although readily soluble in this amorphous state, is unstable and readily decomposes. Thus, it is of great interest to study and modify freeze‐drying procedures to induce crystallization during processing of pharmaceutical entities (Cise and Roy 1979 ).
Option 1: Direct Lyophilization
In the process for the preparation of imipenem (Figure 13.13 ), the drug is first dissolved in a solvent mixture of acetone/ water, sterile‐filtered into single‐dose vials, and frozen at a temperature of −40°C. Sodium bicarbonate is included in the solution make‐up to improve the chemical stability of imipenem. Freezing is observed to take place within 15 minutes of the solution’s being charged to the freeze dryer, resulting in the formation of a solid matrix of frozen solvent and amorphous drug. The lyophilization cycle is initiated by placing the system under vacuum, and the temperature is slowly raised to remove all the frozen solvent by sublimation. Imipenem solid derived from this process is 100% amorphous by X‐ray powder diffraction analysis (Crocker and McCauley 1995 ).
Option 2: Freeze Crystallization
This process is described by Connolly et al. (1996 ). Similar to the process described in Option 1 above, the antibiotic is first dissolved in a solvent mixture of acetone/ water and frozen at a temperature of −40°C. The freeze crystallization cycle is then started by raising the temperature to between −10 and −30°C and holding the system for several hours. The system is then frozen again to below −40°C. The lyophilization cycle is initiated by placing the system under vacuum, and the temperature is slowly raised to remove all the frozen solvent by sublimation.
Figure 13.13 Chemical structure of imipenem.
Table 13.2 shows the freeze crystallization cycles for the study. Three different freeze crystallization temperatures, −10, −20, and −30°C, and three different levels of acetone, 10, 20, and 30 vol%, were studied. The output variables were the degree of crystallization of imipenem products and the percentage of the liquid phase at the freeze crystallization temperature. For the aging time of stage 3, a value of 99.95 hours was shown in order to hold the freeze dryer shelf temperature at −40°C overnight.
DSC Thermograph
Figures 13.14 –13.17 show the differential scanning calorimetry (DSC) thermograph of the acetone/ water solvent mixture with and without the presence of imipenem. The DSC curves indicate endothermic peaks which correspond to the phase transitions. The peaks around 0 and −95°C represent the melting of the solid water and acetone phases, respectively. The peaks around −19°C represent the melting of the solid clathrate phase (Rosso et al. 1975 ). The clathrate is a solid phase complex of 17 water molecules surrounding a single molecule of acetone, 17H2 O* (CH3 )2 CO. In addition, the DSC thermograms indicate only slight temperature fluctuations for the endotherms with the addition of imipenem and the sodium bicarbonate. Therefore, the acetone/ water binary phase behavior is not affected significantly by the addition of imipenem and sodium bicarbonate.
Table 13.2 Freeze crystallization cycles (all acetone levels).
Stage Starting temperature, °C Ramping time, hr End temperature, °C Aging time, hr 1 −40°C 1 −10, −20, −30 10 2 5 −30 0 3 1 −40 99.95
Figure 13.14 DSC thermograms for 10, 20, and 30% acetone/ water solutions.
Figure 13.15 DSC thermograms for 10% acetone/ water and imipenem/ NaHCO3 in a 10% acetone/ water solution.
Figure 13.16 DSC thermograms for 20% acetone/ water and imipenem/ NaHCO3 in a 20% acetone/ water solution.
Figure 13.17 DSC thermograms for 30% acetone/ water and imipenem/ NaHCO3 in a 30% acetone/ water solution.
Crystallinity of Imipenem and Phase Behavior of Acetone/Water
The results of the crystallization trials are summarized in Table 13.3 . The acetone/ water binary phase diagram was used to approximate equilibrium conditions for the solvent systems at the freeze crystallization conditions in the absence of a multicomponent phase diagram for the acetone/ water/ imipenem/ NaHCO3 system.
These results indicate some obvious patterns of crystallization. High degrees of crystallization occur independently of the initial acetone composition of the solvent at −10°C, and the observed degree of crystallinity decreases significantly as the freeze crystallization temperature is lowered. In addition, at the lower crystallization temperature, the observed degree of crystallinity increases with increasing initial acetone composition of the solvent. This trend is particularly significant at −30°C. The degree of crystallinity increases from 21 to 80% as the initial acetone composition of the solvent is increased from 20 to 30 vol%.
Using the acetone/ water binary phase diagram (Rosso et al. 1975 ), the relative weight percentages of liquid and solid phases observed at equilibrium were calculated as shown in the last column of Table 13.3 . According to the DSC curves and the phase diagram, the formation of a solid clathrate occurs at −19°C. Since two freeze crystallization temperatures at −30 and −40°C are well below the clathrate formation temperature, the weight percentages of the liquid phase observed at these two temperatures were reported, assuming that the clathrate forms. However, for the crystallization temperature at −20°C, clathrate formation is assumed not to occur since the temperature of the system fluctuates under the experimental conditions.
The calculated weight percentages of the liquid phases observed at equilibrium provide further insight into the observed crystallization patterns. As shown in Figure 13.18 , the data suggest a direct correlation between the presence of a liquid phase during the freeze crystallization stage and the degree of crystallinity of the resulting product. For imipenem crystallized under conditions in which significant levels of liquid are present at equilibrium, high degrees of crystallinity are observed, and vice versa.
Table 13.3 Imipenem crystallinity and phase behavior of acetone/ water.
Acetone, vol% Freeze crystallization temperature, °C Degree of crystallinity, % Liquid level wt% in acetone/ water system 10 −10 99 32.2 −20 78 12.6 −30 8 0 −40 6 0 20 −10 99 64.4 −20 77 26.3 −30 21 0.1 −40 14 0.1 30 −10 100 96.6 −20 93 39.4 −30 80 14.4 −40 25 12.8
Figure 13.18 Relationship between the degree of crystallinity of imipenem and the equilibrium wt% of liquid in the system.
The solubility values for imipenem in the 30% acetone/ water solution are 0.72, 0.40, and 0.22 g/ l at −10, −20, and −30°C, respectively. Values for solubilities in the equilibrium solutions at −20 and −30°C could not be measured. These solubility results indicate that imipenem is slightly soluble in the liquid phases present at equilibrium under the freeze crystallization conditions. Therefore, it may be assumed that the transition from amorphous form to crystalline form is facilitated by the presence of a liquid phase.
Message
Transition of amorphous imipenem to crystalline material can be greatly facilitated in the presence of liquid even at the frozen stage. The degree of crystallinity shows a clear correlation with the degree of the liquid level in the frozen solution.
EXAMPLE 13.6 Continuous Separation of Stereoisomers
Goals : Kinetic separation of isomers by crystal growth of the desired isomer on the seed bed of that isomer (desired: all‐growth process).
Issues : Minimize nucleation of the undesired isomer—strict control of supersaturation Maintain particle balance—create new growth centers in the absence of nucleation.
Kinetic resolution is the least expensive alternative, compared to chiral chemistry, creation of diastereoisomers, or separation devices such as membranes. Additionally, it produces the highest optical purity product of all these methods. In general, kinetic resolution requires the presence of a racemic mixture (conglomerate) and the absence of a (generally lower‐solubility) racemic compound (both enantiomers in the crystal lattice). The key features include
Heavily seeded, all‐growth, limited residence time crystallizers.
A continuous (steady‐state) process for tight control of supersaturation, unchanging with time.
Additionally, fluidized bed crystallizers to avoid the need for heavy‐magma, high‐flux filtration equipment.
Continuous kinetic resolution of stereoisomers was put into production, using stirred tank crystallizers, at Merck & Co. in 1962 magazine‐[Chemical Engineering staff (1965 ); Krieger et al. (1968 ); Lago et al. (1966 )]. New chemistry was accompanied by a switch to production‐scale fluidized bed crystallizers in 1967, using ultrasonic cleavage of seed particles to maintain population balance [Allegretti and Midler (1970 ), Midler (1970 , 1975 , 1976 )]. High‐speed film studies on crystal cleavage in the ultrasonic field were described by Klink et al. (1971 ). A patent by the Ajinomoto Company (Ito et al. 1966 ) using column crystallizers to resolve amino acids indicated that visible nucleation in the crystallizers was acceptable, which is contrary to the experience at Merck.
Stirred tank and fluidized processes for stereoisomer resolution share the following characteristics:
All locations in the process stream (including the walls of heat exchangers) are maintained at low supersaturation, well within the metastable region, to preserve isomeric purity within each seed bed by avoiding nucleation.
All configurations contain a provision for generation of new seed particles to replace those removed in the product harvest streams. This can be carried out by controlled attrition, as, for example, in the high shear regions in the in‐line mixer of the recirculation loop, or with superimposed attrition by devices inserted for this purpose.
Fluidization
Design of fluidized bed crystallizers requires estimates of the required seed bed volume and the quality of fluidization.
Fluidization behavior of seed was measured in the laboratory for monosized cuts. For a given volumetric flow rate through any given fluidized bed crystallizer, there is a minimum and maximum particle size which will result. The minimum is that below which the particles will elutriate out the top of the column. The maximum is determined by the size at which removal or controlled attrition (see below) takes place at the bottom.
Figure 13.19 shows three flow systems capable of carrying out stereoisomer (or other) resolution by crystallization.
The single‐crystallizer system has the benefit of simplicity but requires periodic harvesting of the dissolver.
Series operation always has the desired product in the downstream crystallizer. It decreases the chance for isomer contamination in that unit by reducing supersaturation of the unwanted isomer.
Parallel operation is easier to run because both crystallizers are operated in the same way. Harvesting and other operations carried out in one operation do not affect the other, and any washout of crystals from either of the crystallizers returns to the dissolver rather than possibly contaminating the other isomer seed bed. For these reasons, parallel operation is generally the method of choice.
It is desirable to measure the fluidization behavior of the particular solids in the column, ideally in actual mother liquors. A typical set of curves is shown in Figure 13.20 a. Each particle size is a relatively monosized sieve fraction. The x ‐intercept of each curve is the terminal settling velocity (slurry concentration 0 g/ l), but it is more accurate to measure the terminal settling velocity separately with a stopwatch and a graduated cylinder. Such a data set is shown in Figure 13.20 b.
The solid in liquid‐fluidized bed systems, as distinct from most fluidized with gas, is generally suspended uniformly. Liquid flow, with reasonable column geometry, can be uniformly mixed radially with minimal backmixing. Even in a column of constant diameter, the particles tend to classify according to size (largest at the bottom). Plug flow can be further enhanced by having a substantial tapered section in the bottom.
Figure 13.19 Flowsheet of stereoisomer resolution systems.
Figure 13.20 (a) Fluidization behavior of particular solids in a (liquid) fluidized bed. (b) Terminal settling velocity of particles in liquid.
The crystallizers described in this section were designed with a 3 : 1 diameter ratio (9 : 1 ratio in linear velocity). Plug flow approximation for these fluidized bed crystallizers was confirmed by point insertion studies with a radioactive tracer.
Crystal Growth
The fluidization studies noted above for monosized crystal populations showed that the minimum particle size in the fluidized bed crystallizers should be 150 microns and the maximum 600 microns. In an all‐growth process (absence of nucleation), the smallest should grow to become the largest with a distribution like that of the “Idealized” curve in Figure 13.21 a. The histogram plot of actual column data (Actual) in the same figure shows that the fluidized bed crystallizer run under the conditions of that experiment, corresponds to an all‐growth situation.
Figure 13.21 (a) The histogram plots the actual column data versus the idealized curve. (b) An exception (2.5‐order) of relative growth rate as a function of supersaturation for different acid concentrations.
Growth rate kinetics for most of the stereoisomer resolution processes run at Merck have been essentially first‐order with respect to supersaturation. Figure 13.21 b shows an exception (2.5 order). The parallel plots in Figure 13.21 b are for different acid concentrations used to solubilize the compound, while supersaturation was generated by a temperature difference between the dissolver and crystallizer.
As presented earlier in Chapter 7 , Table 7.5 , the fluidized bed system showed considerably faster growth rate kinetics than the stirred tank, presumably by convective enhancement in the boundary layer on the surface of the crystals. Additionally, since nucleation suppressants nearly always also inhibit the growth rate, the slower growth kinetics in this case are not unexpected.
Sonicator, Seed Generation, and Particle Population Control
Since separation of stereoisomers requires that this type of crystallization be all‐growth, some means for creation of new seed particles must be provided for any crystallizer type. These particles can be added from an external source or can be created internally by attrition.
Stirred tanks with a recycle loop have natural attrition from both the pump and the agitator. The extent of this breakdown is clearly dependent on the types of pump and impeller employed. Fluidized beds have a special requirement: any attrition process must minimize fines, because they wash out of the seed bed by elutriation and thus reduce the surface area for growth. A study of possible attrition and wet‐milling options showed that sonication created the fewest fines because of the nature of crystal disruption
In this case, sonication is the means used in the fluidized bed crystallizers for maintaining the number of seed particles in the magma to replace those removed in the product and concurrently preventing the formation of overly large crystals. Excessive particle size starves the seed bed of crystal surface area for growth and, in the case of fluidized bed crystallizers, causes sluggish solids movement, which can cause the particles to grow together.
As presented earlier in Chapter 7 , Table 7.6 estimates the amount of ultrasonic energy required to run in the scale‐up crystallizer. It is shown that breakage was essentially first‐order with sonication power and slurry concentration.
As presented earlier in Chapter 7 , Figure 7.28 shows the unsonicated and sonicated particles. The unsonicated crystals (left) are disrupted by sonication along cleavage planes (right), a mechanism which produces few fines to elutriate out of the seed bed. For this product, the sonicators were located in the column bottom (Figure 13.22 a). Some crystals are more friable, particularly when they grow in a relatively needle‐shaped morphology. More fines may be produced under sonication. In addition to creating some washout from the bed, the fines can adhere to the other particles, as shown in Figure 13.22 b. The fines can completely coat the particles, and the resulting seed bed becomes exceedingly dilute and unproductive.
Figure 13.22 (a) Sonicators located at the bottom of the column. (b) Fines generated by sonication can adhere to other particles.
Figure 13.23 The sonicator is an external circulation loop.
By installing an external flow sonication system, the fine coating issue was successfully dealt with the slurry return above the top of the fluidized seed bed, as shown in Figure 13.23 . In this system, fines generated by sonicator are washed out at the top rather than passing through the bed, and the coating of particles is minimized. Depending on the actual amount of fines produced, they can either be filtered as product (they are pure isomer) or carried back to the dissolver for resolubilization.
As presented earlier in Chapter 7 , Figure 7.29 shows unsonicated crystals entering (left) and sonicated crystals leaving (right) the flow sonicator. Although not necessary, slurry return fines disengagement can be enhanced by a tangential flow device like that shown in Figure 13.23 .
Whether the sonicators are located internally or in a slurry flow system, the crystal population fed to the sonicators is always drawn from the bottom of the seed bed, where the largest particles are located.
Fluidized Bed Crystallizer Scale‐Up
When scaling up column geometry, it is desirable to know the localized slurry concentrations within the crystallizer. Excessively high concentrations can result in sluggish fluidization and possible agglomeration, and excessively low concentrations decrease productivity. These local concentrations can be calculated, assuming a classified bed (largest particles on the bottom), and either a measured or estimated size distribution.
The results of these calculations for the particles are shown in Figure 13.24 . For the laboratory (3 inch diameter) column shown at the left, the maximum slurry concentration (most sluggish fluidization) was calculated to be in the region just above the top of the tapered section. This corresponded to visual observation of operation in the (glass) columns. Pilot plant crystallizers were designed with a similar geometry.
Figure 13.24 The calculations for the particles in a fluidized bed.
Comparable calculation of a factory column with a similar ratio of taper volume to that in the straight section indicated that fluidization would be excessively sluggish in parts of the crystallizer, and the relative taper volume was increased. Calculation of the ultimate plant design localized slurry concentrations is presented at the right of Figure 13.24 , showing a relatively uniform slurry concentration. Samples taken from various taps in the factory operation confirmed these results.
As noted above in the section on fluidization, it is appropriate to minimize the angle of the conical section from the vertical to minimize backmixing. Different relative scales of some actual fluidized bed crystallization operations are shown in Figure 13.25 a.
In a growth‐dominated fluidized bed crystallizer, the interface between the fluidized bed and the mother liquors should be sharp, as the particles either grow larger or wash out at the top. Figure 13.25 b shows a sight glass view of such an interface in a factory‐scale crystallizer.
As presented earlier in Chapter 7 , Figure 7.30 , it shows a pair of factory fluidized bed crystallizers, one for each stereoisomer, in construction. These correspond to the calculation at the bottom of Figure 13.24 . Internal sonicators (Figure 13.22 a) were installed in the column bottoms. The blowers shown at the bottom of each column were used to cool the sonication units.
Figure 13.25 (a) Different relative scales of some actual fluidized bed crystallization operations. (b) A sight glass view of a sharp interface in a factory‐scale crystallizer.
Message
When close control of an operating parameter (e.g. supersaturation) is required for a process, a continuous operation should be considered. The success of sonication in this case proves, once again, that the manufacture of high‐value products (such as pharmaceuticals) can often tolerate technologies which are unusual in the production of less expensive materials.
Goals : To investigate hybrid dispersion as an enabling technology for drug delivery and manufacturing.
Issues : Drug loading, bioavailability, stability, drug–drug compatibility, and manufacturing excellence.
As discussed in earlier Section 13.6 , the current crystalline‐based and amorphous‐based solid dispersion have its advantages and constraints. The crystalline‐based approach have higher drug loading and stability, but also less bioavailability than amorphous‐based approach. On the other hand, amorphous‐based approach have higher bioavailability, but also less drug loading and stability than crystalline‐based approach. Therefore, it is desirable to develop a hybrid solid dispersion to synergistically integrate these technologies to maximize its effectiveness for drug delivery.
From the chemical manufacturing control (CMC) perspective, the hybrid solid dispersion co‐processes API and excipients in one step. It simplifies manufacturing steps of drug substance and drug product, thus can also reduce the overall production cost.
Process Design and Rationale
The process design comprises two key steps as shown below:
Generating a slurry mixture of API, additive(s) and solvent(s) via crystallization, wet milling and/or combination. Specifically, in the slurry mixture, a portion of API exists as micro/nano‐sized crystalline particles which are uniformly dispersed and suspended in the solvents and a portion of API is dissolved in the solvents. In the slurry mixture, all additives are full dissolved in the solvents.
Evaporating solvent(s) off from the above mixture to generate hybrid solid dispersion composite. Specifically, after solvents are evaporated, the suspended API crystalline particle form crystalline‐suspension. The dissolved API and additives can form ASD, eutectic‐mixture (including solid solution), or a mixture. All are uniformly dispersed in the hybrid solid dispersion.
In comparison to conventional ASD under high drug loading, where drug loading exceeds the limit of miscibility/solubility, the hybrid design resolves the uncontrolled phase separation or crystallization of API issues fundamentally. By design, any excessive amount of API beyond the limit will be crystallized out in the first step in a properly controlled manner in order to generate API crystals of desired polymorphs, salts, and co‐crystals with target PSD.
In comparison to conventional (nano)crystalline‐suspension, the hybrid design also intelligently solves the issues of particle agglomeration. In the second step drying, the dissolved API and additives will form ASD (and/or eutectic mixture, etc.). Due to the strong molecular interaction, the ASD or eutectic mixture can act as binder to bind the nano/micro size crystalline particles into easily flowable solid granules. They also form a thin layer coating on these granule surface for improvement of drug–drug compatibility. Upon dissolution, these binder can dissolve easily and granules will disintegrate back into uniformly dispersed nano/micro‐size particles.
Needless to say, the above two‐step approach also simplifies the manufacturing of drug substance and drug product by coprocessing of API and excipients together, and thus reduces the cost.
Model Compound and Experiments
Racemic ibuprofen is chosen as the model compound. It has a low aqueous solubility of ~27 μg/ml and a high dose from hundred mgs to grams. Also, in vitro–in vivo dissolution correlation is well established (Newa et al. 2007 ).
Ibuprofen is well‐studied in the literature with various solid dispersion approaches. In this example, HPMC E3 is chosen to form the ASD with ibuprofen. Nicotinamide is chosen for evaluating the drug–drug compatibility because it forms Ibuprofen‐nicotinamide co‐crystal.
For generation of hybrid solid dispersion, either crystallization or slurry approaches were employed. For generation of ASD, spray drying method was employed. For both scenarios, various ratios of ibuprofen/additives were explored.
Drug Loading, Amorphism, and Crystallinity
Analysis of API crystallinity in these dispersion samples is shown in Table 13.4 . For ASD samples at drug loadings of ibu/HMPC ¼ and ½, essentially all ibuprofen exists in amorphous state. For ASD samples at higher loadings of ibu/HPMC 2/1 and 4/1, ~70 and ~60% of ibuprofen remain in amorphous state. Or equivalently, 30 and 40% of ibuprofen now exists in crystalline state. Since ASD samples are prepared by spray drying of clear solution containing fully dissolved ibuprofen and HPMC in solvent, the crystalline ibuprofen in the ASD sample reflects the physical instability risk of ASD. The uncontrolled crystallization of API may occur during the spray drying operation or after the drying in storage.
Table 13.4 API crystallinity in ASD and hybrid solid dispersion.
Source: Tung (2020 ).
Ibuprofen—HPMC system Amorphous API (in glass solution), % Crystalline API (in crystalline suspension), % ASD 1/4 ~100 Non‐detected ASD ½ >95 <5 ASD 2/1 ~70 ~30 ASD 4/1 ~60 ~40 Hybrid, 2/1 ~20 ~80 Hybrid, 4/1 ~10 ~90
On the other hand, for hybrid solid dispersion, at drug loading of ibu/HPMC 2/1 and 4/1, the level of amorphous ibuprofen is ~20 and ~10%. Or equivalently crystalline ibuprofen is up to 80–90%. This high level of crystalline API is determined by the design of the hybrid solid dispersion, specially the level of undissolved API in the first step of the hybrid process. As to the level of amorphous API, it is determined by the dissolved API which is at or below the API solubility/miscibility in the additive.
Dissolution Behaviors
Figure 13.26 compares the dissolution profiles of hybrid solid dispersion and ASD under various drug loadings. For ASD, at a low drug loading of ibu/HPMC ¼ w/w (or equivalently 20 wt% API), it shows an excellent rapid dissolution profile. As drug loading increases from ¼ to ½, 2/1 and 4/1 (or equivalently 20, 33, 66 to 80 wt%), the dissolution slows down progressively. This type of dissolution trend is fully expected per existing knowledge. For hybrid dispersion at drug loading of 2/1 and 4/1 (or equivalently 66 and 80 wt%), the dissolution profiles are very comparable to that of ASD profile of ¼ (or equivalently 20 wt%). The results are very encouraging to confirm the power and benefit of hybrid dispersion.
Figure 13.26 Comparison of dissolution profile of hybrid solid dispersion and ASD.
Dissolution behaviors are further examined under the microscope (Figures 13.27 and 13.28 ). Upon dispersing these samples in water, the ASD sample convert to irregular amorphous (gummy) and/or crystalline clusters. When the HMPC level in high, i.e. ¼ or 80%, these clusters can remain amorphous which is still highly soluble. When the HPMC level decreases, i.e. ½ or 66%, there are increasing risks of formation of crystalline cluster which will greatly decrease the solubility and dissolution rate. These types of behaviors are fully expected.
On the other hand, the hybrid solid dispersion samples are large solid granules. They possess nice solid properties for formulation. Upon dispersing these granules in water, very encouragingly, they disintegrate rapidly and are converted back to original uniformly dispersed nano/micro‐size particles. The well‐dispersed non‐agglomerated particles should facilitate the dissolution.
Figure 13.27 Microscope images of ASD samples before and after dispersed in water.
Figure 13.28 Microscopic Images of hybrid samples before and after dispersed in water.
Stability and Drug–Drug Compatibility
To compare the drug–drug or drug–excipient compatibility, physical mixture of ibuprofen and HMPC, ASD, and hybrid solid dispersion samples are physically blended with an equal weight of nicotinamide to ibuprofen. The compatibility study is conducted at 40°C/75% RH for one month under open dish conditions. Figure 13.29 shows the DSC results for cases with a ratio of ibuprofen/HPMC 2/1. The appearance of ibu‐nic cocrystal endotherm at ~90°C serves as an index for the drug–drug interaction of ibuprofen and nicotinamide. The solid line is before the stressing and dotted line is after the stressing. It is clear that physical mixture has the highest level of drug–drug interaction, ASD has some degree of drug–drug interaction, and hybrid solid dispersion has the minimum degree of drug–drug interaction. These data support the benefit of improved stability and drug–drug compatibility.
Message
In comparison to conventional ASD or eutectic‐mixture, the hybrid dispersion design resolves the uncontrolled phase separation or API crystallization issues under high drug loading. In comparison to conventional (nano)crystalline‐suspension, this design resolves the challenging particle agglomeration issue during drying. Hybrid solid dispersion further improves drug–drug or drug–excipient compatibility, and facilitates downstream formulability of drug product.
The coprocessing of API and excipients together can effectively simplify the manufacturing of drug substance and drug product, and thus reduces the manufacturing cost.
Figure 13.29 DSC profiles of physical mixture, ASD and hybrid solid dispersion with nicotinamide.