6.3.1.1 Seed Crystal Form and Quantity

It is a typical requirement that crystal form of seed should be the same as desired crystal form of final product after crystallization. Seed with the desired crystal form will facilitate the formation of desired crystal form and minimize the formation of undesired crystal form during crystallization. Conversion of crystal form during crystallization should be avoided as it essentially falls back into the non‐seed scenario. For anhydrates or desolvates which are generated after drying, one approach is to slurry some of these crystals in the crystallization solvent first to convert them back to hydrates or solvates for use as seed.

Four levels of seeding can be described, depending on the purpose of seed addition as follows:

• Pinch, to hopefully avoid oiling out and/or uncontrolled nucleation, crashing, or snowing out. It may be satisfactory in the laboratory but is rarely effective or reliable on scale‐up.
• Small (<1%), to hopefully aid in more controlled nucleation but not adequate to achieve primary growth on scale‐up. It is subject to additional nucleation and bimodal distribution of the product. However, this approach can be adopted in the in situ seed generation approach, which generates massive seeding for the subsequent crystallization operation, thus maintaining the benefit of massive seeding but minimizing the amount of external seed used.
• Large (5–10%), to improve the probability of growth with the possibility of preventing further nucleation and bimodal distribution.
• Massive (>10%), to provide maximum opportunity for all growth. (See Examples 7.6 and 13.6 on resolution of optical isomers.)

Seed can be critical in the control of enantiomer separation and selection between polymorphs and hydrates/solvates. In these applications, the seed will maximize the growth of the desired product where nucleation of undesired isomer/polymorph/hydrate/solvate are minimized/prevented.

6.3.1.2 Estimation of Seed Size and Quantity

The impact of seed size and quantities on final product particle size increase if an all‐growth process were calculated (Table 4.3) by a simple relationship. Just a note that validity of this calculation depends on an all‐growth process. Nucleation—which is virtually always present to some degree — will result in a reduction in the actual particle size of the product as well as an increase in total particles per unit volume. This increase can come in the form of a bimodal distribution made up of growth on seed and nucleated smaller crystals.

6.3.1.3 Seeding Procedures

There are several sources of seed crystals that may be used with varying degrees of effectiveness. Each seeding application requires analysis to determine which of the following sources will be most satisfactory for the operation in question. These sources and methods of preparation include the following:

• Seed from the previous batch—added separately or as heel recycle; effective if the physical and chemical attributes and stability remain satisfactorily constant or can be normalized by intermediate treatment.
• Seed prepared in a batch for that purpose and used in many batches; effective provided that the seed so prepared has the necessary physical and chemical attributes and stability.
• Seed prepared as above and then (dry) milled to achieve a mean particle size and PSD as required; effective for achieving increased control of physical attributes.
• Seed generated in situ from the current batch or heel recycling, both with wet milling, to control the mean particle size and PSD; one of the most effective seeding procedures when applicable (see Example 10.1).

The issues involved in the necessary record keeping, good manufacturing practice (GMP), and regulatory control that are associated with a seeded process in the pharmaceutical industry are sometimes cited as reasons not to seed. In response to an article endorsing this premise (Pessler 1997), the following rebuttal by C.B. Rosas (1997, personal communication) summarizes his and the authors’ views on this topic.

Not only is timely seeding an excellent way to do away with batch‐to‐batch vagaries in the width and sensitivity of the metastable, supersaturated region, but the avoidance of nucleation is often crucial for achieving crystallization elegance—a requisite for enhanced rejection of impurities. To trade away such a powerful tool for the sake of trivially lessened inconveniences in record keeping in a GMP plant is most unsound as an operating principle. For those frequent and difficult purification tasks that crystallization from solution does so well, seed early, seed often and, above all, seed always.

6.3.2.1 Seeding Point

The obvious problem of adding the seed before reaching saturation is that of dissolution of some or all of the seed, and the problem of adding the seed after reaching saturation is that of already experiencing nucleation. These difficulties are greatly exaggerated in cooling crystallization of solute with steep solubility dependence on temperature, as described in Example 7.1. In this example, suitable timing of addition of seed was considered to be unachievable in the manufacturing operation and an alternate strategy utilizing seed heel recycling, as described in the example, was utilized.

This problem can also be severe in evaporative methods particularly at reduced pressure, as is common. For these and other reasons described in Chapter 8, evaporative methods are considered to be the most difficult for predictable control of mean particle size and PSD on scale‐up.

For anti‐solvent addition methods, the seed can be added to a small part of the anti‐solvent, which is then added as slurry as the saturation point is approached. Although some of the seed may dissolve, this technique is considered to be more reliable than adding the seed all at once.

For reactive crystallization, the solubility of the product is usually low and seed can be added before the operation is initiated without concern for dissolution. However, since this method generally produces the smallest particles and is subject to agglomeration, the effectiveness of seeding is critically dependent on providing sufficient surface area for growth at the outset to prevent nucleation at the high local supersaturation conditions at the point of addition and throughout the resulting slurry. The reader is referred to Examples 10.1 and 10.2 for descriptions of operations in which nucleation was minimized in favor of growth.

6.3.2.2 Instrumentation for Seeding Timing

Online instrumentation can be a critical aid in detecting the correct seeding point (Zhou et al. 2006). Measurement of solution concentration by, e.g. Fourier transform infrared (FTIR), near infrared (NIR), UV/visible or Raman spectroscopy, can be very effective in determining the seeding point. After seeding, in order to determine whether or not seeding was effective, particle count and size distribution measurement can be made by an in situ particle size and counting instrument. Sampling and laboratory measurement can be used, but changes that can occur in the sample and the time delay can often introduce unreliability.

Depending on the outcome of the particle number and size determination, corrective measures may be initiated in the event that the seed is dissolved or there was excessive nucleation. Reseeding can correct the former and reheating to dissolve excess nuclei may be applicable in the latter.

6.3.2.3 Seed Aging and Annealing

Many crystallization operations can benefit from a seed age in which the temperature is held constant or the anti‐solvent addition is temporarily halted while the seed crystals grow and deplete the supersaturation before continuing. This procedure can help in normalizing the PSD resulting from the initial steps of seeding and supersaturation generation. The use of online instrumentation can again be very effective in determining when the seed age is complete and the operation can continue.

In some cases, a crystallization procedure is unsatisfactory for producing satisfactory seed crystals for subsequent use. This is particularly true for reactive crystallization because of the fine particles normally produced by this method. Since no amount of fine particle seed is satisfactory for significant growth in succeeding batches, the next possibility is to increase the particle size of the seed. To accomplish this, it is necessary to grow seeds in a separate operation from the reactive crystallization. Other applications may also require separate preparation of grown seed. The following procedure is one possible method.

In a separate operation, starting with fine needles, these crystals are subjected to many heat/cool treatments with sonication/wet milling between cycles. A suitable solvent is required in which the solubility approximately doubles on heating (e.g. a temperature increase of 20–50°C). During heating, the finer crystals will dissolve. During slow cooling, growth on the remaining crystals may be achieved. Sonication/wet milling after cooling can break the needles lengthwise, creating more particles. In subsequent cooling cycles, the diameter of the needles, the slowest growth dimension, can slowly increase eventually producing three‐dimensional crystals.

This procedure is applicable to other shapes. Crystal breakage by sonication/wet milling may result in other types of crystal cleavage but three‐dimensional growth may also be achieved in these cases.

The needles shown in Figure 10.7a were subjected to this heat–cool treatment resulting in the large three‐dimensional crystals shown in Figure 10.7a. Success in growing seed crystals both prepares seed for larger batches and establishes that this compound will actually grow. It is then necessary to determine if it will grow at a practical rate in the actual reactive crystallization system.

Several authors indicate the importance of the condition of the surface of seed crystals, and some qualitative suggestions can be made. These include (i) making a seed slurry to help condition the surface by dissolution of shards and providing some surface activation (which is also a good method of seed addition) and (ii) avoiding the use of dried seed. Milled seed may be necessary for particle size control, but milled seed should either be reslurried or the milling should be done by wet‐milling methods.

6.3.2.4 Continuous Operation

The most effective seeding is achieved in semicontinuous and continuous crystallizations by the nature of the operations themselves, in which the seed is always present and in large quantity. Although common in large industrial operations, these techniques have found more limited application in the pharmaceutical industry. As observed, there are more pharmaceutical applications using the in situ wet seed generation approach (Chapter 4) which has the advantages of mass seeding as in the continuous crystallization, but without the burden of large industrial operations.

There are excellent examples as detailed in Examples 7.6 and 13.6, on the continuous resolution of optical isomers in fluid bed crystallizers, and in Example 10.1, which presents a semicontinuous method of utilizing seed heel recycle in reactive crystallization to achieve primary growth. Also, Example 9.7 represents a semicontinuous method utilizing in situ generated seed in anti‐solvent crystallization. Because they are primary growth processes, these applications require reduction in seed crystal size to prevent oversized crystals from limiting the surface area. The methods documented in these examples include sonication and wet milling.

6.3.2.5 Method of Seed Addition

Seeding with dry powder seed through a vessel head nozzle, while widely practiced in the past, is now limited because of safety and exposure considerations. Slurry seed additions by pump or from seed tanks are preferred and are superior to powder addition for the reasons discusse 9.7d above. The advantages of retaining the seed in the system as utilized in continuous operation or in heel recycling are also indicated above. Needless to say, for in situ generated seed it is no longer an issue.

6.4.2.1 Oiling Out

Oiling out, beyond as an equilibrium phenomenon as discussed in Section 6.2.1, can be a kinetic phenomenon in crystallization by any of the methods of creating supersaturation and becomes increasingly possible under several conditions, including:

• high supersaturation
• rapid generation of supersaturation
• high levels of impurities
• presence of crystallization inhibitors even at low levels
• absence of seed
• inadequate mixing (high local supersaturation)

Kinetically, when supersaturation is achieved rapidly such that the concentration is beyond the upper metastable limit—as can often be the case in a nucleation‐based process—the substrate is forced to separate into a second phase by the creation of the resulting high solution concentration. However, crystallization is delayed by a slow crystallization rate. This combination may result in the creation of a nonstructured oil or possibly an amorphous solid. The rates of phase separation and nucleation are relative to each other such that “slow nucleation” implies only that nucleation was not fast enough to create discrete particles before oil separation.

Transition of an oil or an amorphous solid or a crystal can then occur. However, this type of operation can be difficult to control, and scale‐up is treacherous because the oil droplets may coalesce into masses and/or form gum balls and increase in size to intolerable levels.

Oiling out may be minimized or eliminated by control of supersaturation and seeding (Deneau and Steele 2005), as discussed below. Seeding has been proven to be essential to prevent oiling out in some systems because, although the oil may not be the thermodynamically stable phase, the transformation to crystals may be sufficiently slow and uncontrolled to cause severe processing problems, as discussed above.

The initial formation of an oil, gel, gum, or amorphous, solid conforms to the Ostwald step rule discussed in Myerson (2001, p. 39). This rule states that in any process, the state that is initially obtained is not the most stable state but rather the least stable state that is closest in terms of free energy change to the original one. It has been postulated that the initial state of crystallization processes is amorphous clusters and that the difference in time constants for the transformation to more stable crystalline states (nuclei) is a key determining factor in the course of the crystallization. It is difficult to distinguish between cluster formation, nucleation, agglomeration, and growth in the early stages (Mersmann 2001, p. 235). The following possibilities can be recognized qualitatively as determined by the time constants and the physical chemistry of the specific compound and system.

As is well known, some compounds have never been crystallized, and phase separation results in a stable oil or an amorphous solid. The search for solvents and conditions, or the introduction of foreign particle seeds (e.g. by scratching a glass test tube) to induce crystal formation for a new compound, becomes a matter of trial and error. Combinatorial techniques continue to be developed that can aid in this evaluation. A critical factor for success may be removal of impurities to achieve a very high level of purity, because the effect of even very low levels of impurities on homogeneous nucleation will not be known at this stage.

High supersaturation can lead to very small (nano‐sized) particles and result in agglomeration and gel formation. This phenomenon has been discussed by Mersmann (2001, p. 295) and Mullin (2001, p. 317). Although difficult to define or predict, one mechanism of phase separation leading to agglomeration may be a combination of gelling, to produce a colloidal system, followed by oiling out and nucleation in which the oil serves as a bridge between developing nuclei. Operation at high global or local supersaturation ratios is expected to promote this effect. Mixing can also be a key factor. Although increased mixing may result in breakup of agglomerates in some cases, it may also cause an increased rate of formation by impact between “particles.” The effect of mixing on agglomeration of particles smaller than the Kolmogoroff length scale has been termed by Smoluchowski (1918) as “orthokinetic.” This would predominate in a stirred vessel. The other term used is “perikinetic,” pertaining to Brownian motion in a static fluid and when particles are in the submicron size range.

6.4.2.2 Agglomeration and Aggregation

The distinction between agglomeration and aggregation has been described differently by various authors. Both have received further study by several investigators, including Myerson (2001, pp. 110–111, 146), Mullin (2001, pp. 316ff.), and Mersmann (2001, pp. 235ff, 527). In agreement with these authors, the differences between them are not significant and agglomeration will be the term used in this discussion.

One mechanism for agglomerate growth is the collision of growing nuclei followed by “cementing” together from continuing growth between two or more crystals. Although simultaneous collision of more than two particles is not statistically important, the addition of a large number of nuclei to an original two‐crystal agglomerate can readily occur by ongoing collisions, leading to very large agglomerates. Aggregation is weak bonding of colloidal particles. Aggregates are relatively easily separated

Several investigators have developed models for the effectiveness of collisions that lead to agglomeration including Nyvlt et al. (1985) and Sohnel and Garside (1992). This complex interaction of hydrodynamics and crystallization physical chemistry is difficult to predict or describe but can be critical to the successful operation and scale‐up of a crystallization process. In particular, for reactive crystallization in which high supersaturation levels are inherently present, agglomeration is very likely to occur as the precipitate forms. Careful control may be necessary to avoid extensive agglomeration, as outlined below and in Examples 10.1 and 10.2 for reactive crystallization.

The difficulties that can result from agglomeration include:

• entrapment of solvent and/or impurities in the crystal mass
• reduced effective surface area for true growth
• subsequent breakup of agglomerates into small crystals that were captured during nucleation without an opportunity for growth
• difficulties in downstream processing because of these small crystals

For these reasons, agglomeration is generally to be avoided. The use of additives (Myerson 2001, p. 146) may be considered for minimizing agglomeration. However, the use of additives in the pharmaceutical industry—particularly for final products—is generally not done for regulatory reasons barring extreme need.

There are operations, however, which may intentionally generate agglomerates for a particular purpose. These operations are described as flocculation and/or coagulation. However, a discussion of the purposeful generation of these clusters is beyond the scope of this section.

6.4.2.3 Minimization of Agglomeration

As indicated above, the primary process variables that can be manipulated to minimize agglomeration are

• operation within the metastable region
• controlled rate of supersaturation generation
• removal of crystallization inhibitors from the feed stream
• appropriately high seed levels
• mixing conditions to achieve particle suspension

These factors are all discussed in the applicable sections of this book and are the same as those recommended for achieving growth.

Both the formation and disruption of agglomerates are functions of mixing conditions and local shear. The reader is referred to the detailed treatment by Mersmann and Braun (Mersmann 2001, pp. 235ff.) in their chapter on “Agglomeration” for a comprehensive analysis of the forces involved in these phenomena. This discussion includes the distinction between attrition and disruption, which are both functions of mixing. Disruption refers to breakup of agglomerates that formed under conditions of low supersaturation and can be broken because the binding forces are small. Agglomerates that are formed under conditions of high supersaturation are much stronger and not subject to disruption by mixing. Attrition refers to breakup of large primary crystals that were formed by growth at low supersaturation and is a function of local shear and crystal lattice energy.

6.4.3.1 Nucleation Issues

In some cases, a process that is dominated by nucleation can result in an acceptable process outcome. The process may appear satisfactory in a laboratory‐scale operation. However, there are several potential problems with this type of process for laboratory and pilot plant operation or when scaling up for manufacturing, including:

• fine crystals and/or wide PSD
• high surface area
• low bulk density
• the risk of large batch‐to‐batch variation
• occlusion of solvent and impurities
• agglomeration/aggregation
• lack of control of hydrates, solvates, and polymorphs

In addition, scale‐up of a nucleation‐dominated process is difficult to predict, unless the generation of supersaturation is well controlled.

A principal reason for the difficulty in controlling local supersaturation and nucleation is the differences in the local energy dissipation rates within a stirred vessel. But under high mixing intensity with in‐line mixer, the issues of variability of metastability of a supersaturated solution for nucleation can be minimized.

With an extremely high nucleation rate, for example, one with a characteristic time in milliseconds, the degree of control required to prevent formation of the large number of nuclei, may not be achievable in a stirred vessel, in part, because of the inherent properties of the compound and/or because of mixing scale‐up issues (the most common being the geographical distribution and magnitude variation of the energy dissipation rate). Extreme examples of this are found in ionic reactions to precipitate inorganic salts. Organic acid–base reactions can also generate nuclei at high rates. An in‐line type of crystallization approach may be required for anti‐solvent and/or reactive crystallization, both of which may be dominated by nucleation in the absence of strategies to promote growth. The processes referred to above for creating fine particles (Examples 9.5–9.7) are carried out using in‐line mixer/crystallizers.

Nucleation must also be minimized by tight control of supersaturation in processes involving resolution of optical isomers. In some cases, nucleation must be avoided to prevent the formation of undesired polymorphs.

6.4.3.2 Nucleation Rate

As discussed in Chapter 4, the nucleation rate is both species‐specific and a function of the supersaturation ratio. The relation between nucleation rate, growth rate, and particle size as a function of the supersaturation ratio is illustrated qualitatively in Figure 6.3. The actual rate and supersaturation characteristics, such as metastable zone width, are system specific and can vary over wide ranges. In typical stirred vessels, it has been observed that the nucleation rate may vary from milliseconds to hours, and the metastable zone width may vary from less than 1 mg/ml to tens of mg/ml.

In addition, for a specific compound, the nucleation rate is also dependent on the solvent(s) system, impurities, and mixing. These factors combine to cause the difficulties that are often encountered in controlling a nucleation‐based crystallization process, especially on scale‐up.

The variation in supersaturation range, as indicated by the variability in width of the metastable region, can play a critical role in the operability of a process. An example of this type of variability is described in Example 13.1. In this process for the crystallization of an antibiotic, a high degree of supersaturation can be achieved and maintained for several minutes (S ~ 15) in an all‐aqueous system allowing completion of sterile filtration at this high degree of supersaturation (high concentration). The subsequent addition of 3% acetone, however, effectively increases the nucleation rate and narrows the width of the metastable region (as defined in Section 2.2). Crystallization is then completed by the addition of sufficient acetone to achieve the desired yield.

As noted above, nucleation rate differences can be extreme within the same crystallization process. With small changes, a nucleation‐based process can be made to run predictably when the nucleation rate is low (characteristic nucleation time is tens of minutes or hours), and can be essentially an all‐growth process. At the other extreme, nucleation control can be virtually impossible when high rates are encountered (characteristic time of seconds or less).

By increasing the mixing intensity or energy dissipate rate per unit volume by orders of magnitudes (see Table 5.2) using in‐line mixer such as impinging jet, rotor–stator homogenizer, the variability of metastable zone can be significantly reduced and results become more consistent (Examples 9.5–9.7). As a result, the in‐line mixer is considered to be leader for controlling the nucleation, especially. for generation of seed in situ.

6.4.4.1 Growth Issues

A growth‐dominated process may have several advantages, including:

• larger particle size and lower surface area
• higher bulk density
• improved solvent/impurity rejection
• improved control of hydration, solvation, and polymorphs
• decreased sensitivity to mixing
• improved reproducibility on scale‐up
• adaptability to continuous operation

All scales of operation are dominated by the effects of supersaturation, although the outcome is usually more critical on scale‐up to pilot plant and plant operations. The local and global supersaturation ratios that are experienced over the course of a crystallization operation are critical because they determine the balance between nucleation and growth, not only at the onset of crystallization, but throughout the course of a batch or semi‐batch operation. This balance, in turn, determines the resulting physical properties and, in many cases, the distribution of chemical impurities between the crystals and the liquors.

6.4.4.2 Growth Rate and Growth Characteristics

Inherent crystal growth rates are system specific and can vary over several orders of magnitude for different compounds. Measurement techniques for growth rate are presented by Myerson (2001, pp. 58ff.) and Mersmann (2001, pp. 220ff.). A laboratory procedure for measuring the growth rate in a fluidized bed—chosen to minimize nucleation—is presented in Chapter 4. Readers can also read Example 7.2, which discusses crystal growth impact on particle size.

The factors influencing the crystal growth rate of a specific compound are discussed by Myerson (2001, pp. 93ff.). In addition to molecular structure and the solvent system, the growth rate can be greatly modified by the presence of dissolved impurities that may either compete for growth sites or block these sites. As with the nucleation rate, these differences can be so extreme as to make growth impracticably slow, or at the opposite extreme, the rate may be sufficient to achieve an essentially all‐growth process with careful control of supersaturation and growth area. In practice, it has been observed that the release rate of supersaturation by crystal growth can vary from less than a second to several days.

Supersaturation has an important effect on the growth rate, as shown in Figure 6.4. As can be seen, the increased growth rate that can be achieved at higher supersaturation may come at the expense of increased nucleation, leading to broader PSD and possibly a bimodal distribution.

The effect of supersaturation also can depend on the solvent system, as shown in Figure 6.4 for hexamethylene tetramine (Davey et al. 1982).

In the design of a crystallization process, therefore, the balance that is achieved between nucleation and growth rates is critical to particle size under the operational constraints of equipment and facilities. The supersaturation ratio can be controlled to limit nucleation in order for growth to predominate. This becomes increasingly difficult at lower inherent growth rates, since it will extend the batch time cycle substantially and because the nucleation rate becomes more critical at lower growth rates.

Solvent and impurity effects must also be considered. Solvent effects are important and may play a key role in inclusions and in affecting the width of the metastable zone, as discussed in Example 13.1. However, variations in impurity composition can have more influence and can dominate the course of crystallization in many ways.

Impurities can:

• • slow the nucleation rate, leading to high supersaturation and create oiling (see below)
• • retard or stop growth
• • co‐crystallize and/or form solid solutions (see Chapter 2)

Experimentation is required to evaluate these effects. The most useful experiments utilize spiking with known impurities when they can be isolated for this use. However, as is often the case, the number and possibly the low concentration of impurities often make this impractical. An experimentally simpler method is to recrystallize the compound with and without spiking of the mother liquors obtained from the process isolation. Differences in nucleation and growth may readily be observed by comparing photomicrographs of the resulting crystals. Both size and shape can be expected to be affected. If no significant differences are observed, the impurities from the process may not cause any nucleation or growth changes, and the inherent properties of the compound may be assumed to prevail.

Ideal steps in determining growth potential include the following:

• purification to the highest possible extent (using chromatography if necessary),
• selection of a solvent with solubility <50 g/l and some dependence on temperature,
• preparation of a clear solution with low supersaturation,
• aging of this solution with minimal or no mixing in the presence of some seeds, and/or
• subjecting the solution to heat/cool cycles (some fines dissolve during each heating cycle, and some growth may occur on slow cooling).

This procedure may show that growth is possible. A growth rate can then be determined by various methods, including the fluid bed method described in Chapter 4 using the crystals from the heat/cool experiments as seed.