2

Crystallization of Fats and Oils

Serpil Metin1 and Richard W. Hartel2
1 Cargill Inc. Minneapolis, Minnesota 2 University of Wisconsin Madison, Wisconsin

1. INTRODUCTION 1.1. Control of Lipid Crystallization In many food products and even some processing operations, it is important to be able to control lipid crystallization to obtain the desired number, size distribution, polymorph, and dispersion of the crystalline phase. In most foods, it is crystallization of triacylglycerols (TAG) that is most important, although, at times, crystallization of other lipids (i.e., monoacylglycerols, diacylglycerols, phospholipids, etc.) may also be important to product quality. Proper control of the crystalline microstructure leads to products with the desired textural properties and physical characteristics. For example, tempering of chocolate prior to molding or enrobing is designed to control crystallization of the cocoa butter into a large number of very small crystals that are all in the desired polymorphic form. When controlled properly, the cocoa butter crystals in chocolate contribute to the desired appearance (shine or gloss), snap, avor release, meltdown rate upon consumption, and stability during shelf life (fat bloom). Similar
Baileys Industrial Oil and Fat Products, Sixth Edition, Six Volume Set. Edited by Fereidoon Shahidi. Copyright # 2005 John Wiley & Sons, Inc.

45

46

CRYSTALLIZATION OF FATS AND OILS

arguments can be made for other products such as butter, margarine, whipped cream, ice cream, shortening, peanut butter, and a host of others. During processing of fats, crystallization is often used to modify the properties of the fat. For example, winterization of vegetable oils is needed to ensure that the oil remains a clear liquid even when stored at low temperatures for extended time periods. The process of fractionation of fats to produce components of natural fats with different melting properties also requires control of crystallization to optimize the separation process. Many fats, including palm oil, palm-kernel oil, milk fat, and tallow, are fractionated by crystallization to produce different functional fats. 1.2. Crystallization of Natural Fats There are several aspects of lipid crystallization that make it unique from crystallization of other components in foods (like water, sugars, salts, etc.). These are related to the complex molecular composition of natural fats and the orientation of the triacylglycerol molecules. Fats are made up primarily of TAGs, approximately 98%, with the remainder of the fat being more polar lipids like diacylglycerols (DAGs), monoacylglycerols (MAGs), free fatty acids (FFAs), phospholipids, glycolipids, sterols, and other minor components. In rened fats, these minor lipids are much lower in concentration than in unrened fats. Although the TAGs form the main crystalline phase, the minor components, or impurities, can often play a large role in how crystallization occurs and crystallization may be substantially different in a rened oil than in the unrened starting material. Natural fats also contain a wide range of TAG species with fatty acids of different chain length and degree of unsaturation. Milkfat, for example, contains hundreds of different TAG species with no single species present at greater than about 5%. TAGs are composed of three fatty acids arranged on a glycerol molecule, and with variations in chain length and degree of saturation of the fatty acids, a wide range of components is possible. This range of composition leads to interesting complexities in crystallization. The nature of the TAG molecule is such that it can often take multiple forms in a crystal lattice. That is, the same molecule can crystallize into different crystalline forms dependent on processing conditions. The phenomenon is called polymorphism. Although there are numerous molecules that exhibit polymorphism in nature (many in the pharmaceutical eld), polymorphism is somewhat unique to lipids in the food industry (although some sugar alcohols also form polymorphs). In this chapter, the complex nature of lipid crystallization, primarily related to TAG, will be discussed. 2. LIPID PHASE BEHAVIOR 2.1. Nature of the Liquid Phase It is important to understand the nature of the liquid phase prior to crystallization to understand how crystals form. It is widely recognized that lipids retain some degree

LIPID PHASE BEHAVIOR

47

of ordering in the liquid phase, with temperatures well above the melting point needed to fully dissociate this ordering. When melting fats, this liquid ordering is termed a crystalline memory effect, where subsequent recooling leads to formation of a different (usually more stable) phase than would occur if the fat was heated to higher temperatures to destroy the liquid memory (13). In nucleation, or the formation of the crystalline phase from the liquid, some organization of molecules is expected. In lipids, the natural ordering of the liquid phase leads to crystal formation. In fact, rapid cooling of liquid lipids results in the formation of a diffuse crystalline phase (low-energy polymorph) because of the ordering structure in the liquid phase. Such rapid cooling of other systems, most

Figure 1. Proposed mechanism (highly schematic) for nucleation of triacylglycerols (TAGs). Straight chains indicate crystallized TAGs, whereas bent chains indicate uid TAGs (4).

48

CRYSTALLIZATION OF FATS AND OILS

notably sugars and starches, often results in the formation of a glassy state consisting of molecules that are randomly organized together with no long-term ordering. Upon slower cooling from the liquid, the lipid molecules have time to organize into lamellae (1) and eventually can form coherent, three-dimensional crystals (shown schematically in Figure 1). The arrangement of the molecules into the crystalline state depends on such factors as the cooling rate, the temperature at which crystallization occurs, the agitation rate, and the composition of the lipid phase. 2.2. Polymorphism Polymorphism is the ability of a molecule to take more than one crystalline form depending on its arrangement within the crystal lattice. In lipids, differences in hydrocarbon chain packing and variations in the angle of tilt of the hydrocarbon chain packing differentiate polymorphic forms. The crystallization behavior of TAG, including crystallization rate, crystal size, morphology, and total crystallinity, are affected by polymorphism. The molecular structure of the TAG and several external factors like temperature, pressure, rate of crystallization, impurities, and shear rate inuence polymorphism (5). TAGs are oriented in a chair or tuning fork conguration in the crystalline lattice. The TAG can take either a double or triple chain-length structure as seen in Figure 2. The fatty acids of TAG pairs overlap in a double chain-length structure whereas in triple chain packing, the fatty acids do not overlap. The height of these chair structures and the distance between the molecules in the chair structures are found by using the X-ray spectra as the long and short spacings, respectively.

Figure 2. Packing arrangements of triacylglycerol molecules in the crystal lattice (4).

LIPID PHASE BEHAVIOR

49

TABLE 1. Identication of Polymorphic Forms of Fats Based on X-ray Analysis of Short Spacings (6). Polymorphic Form a b0 b Unit Cell Hexagonal Orthorhombic Triclinic Lines and Short Spacings (A ) A single strong and very broad @ 4.15 Two strong lines @ 4.2 and 3.8 A strong line @ 4.6

The polymorphic forms of fats are often simply classied into three categories, a, b0 , and b, in increasing order of stability. The a form is the least stable polymorph with the lowest melting point and latent heat of fusion. The b form is the most stable, with the highest melting point and latent heat. Each polymorphic form has distinct short spacings (the distances between parallel acyl groups on the TAG) that are used to distinguish the polymorphic forms based on their X-ray diffraction patterns, as summarized in Table 1. Based on the unique conguration of the molecules within the crystal lattice, each polymorph has a different crystallographic unit cell, also shown in Table 1. In general, TAGs with three saturated fatty acids crystallize in double chainlength packing, whereas triple chain-length packing is obtained if the TAG contains fatty acids with different structures (chain length and unsaturation). Lutton (7) stated that if the fatty acids of a TAG differ in length by more than four carbons, it forms a triple chain-length structure. Triple chain-length packing is also observed in TAG containing a cis-unsaturated fatty acid because this causes a kink in the structure, as seen in Figure 2. Cis-unsaturated fatty acids do not mix in one layer with saturated fatty acids, and triple chain-length crystals are formed (8). It should be noted that trans-unsaturated fatty acids incorporate into a crystal structure in the same way as the saturated fatty acids (8). The chain-length structure inuences the mixing-phase behavior of different types of TAGs in solid phases (5). The triple chain-length structure has greater long spacings than does the double chain-length structure. Lipids exhibit monotropic polymorphism, where unstable forms are the rst to crystallize in a subcooled fat because of their lower energy state, according to the Gibbs free energy (5). Subsequent transformation of unstable polymorphs into more stable forms occurs over time until, eventually, the most stable polymorph for a given lipid is reached. Transformation of unstable to stable polymorphs can be achieved by a slight increase in temperature above the melting point of the lessstable forms. This increase in temperature rst causes the melting of the unstable forms and then solidication in a more stable form. Transformation to a more stable form can also take place without melting as seen in Figure 3. The difference in Gibbs, free energy between polymorphs is the driving force for this transformation, as the molecules become more tightly arranged in the crystal lattice. It is assumed that the chair structure is maintained during polymorphic transformations (9). The layer arrangement of the a polymorph does not change when it is transformed to the b0 polymorph, although its lateral chain packing and angle of tilt changes during polymorphic transformation.

50

CRYSTALLIZATION OF FATS AND OILS

Figure 3. Monotropic polymorphism of lipids where (Tm)a, (Tm)b0 , and (Tm)b are the melting temperatures of the a, b0 , and b polymorphs, respectively.

The hydrocarbon chain packing of the b polymorph is denser than that of the a polymorph. The denser chain packing in the b polymorph gives increased stability compared with the a polymorph. In addition, stable polymorphs have higher melting point and higher heat of fusion than the less-stable forms. The different polymorphic forms typically crystallize at rates in order of their stability (a < b0 < b). Thus, the least-stable polymorphic form typically crystallizes rst in a strongly subcooled molten fat because of the lower surface energy (10). The rate of polymorphic transformation depends on the length of the fatty acid chain and is the greatest for TAGs with short-chain fatty acids (10). Natural fats usually contain a large number of TAGs; thus, the transformation of unstable to stable forms is often very slow. As mentioned previously, the a form is generally formed rst in a rapidly cooled liquid fat, but it is usually very unstable and rapidly transforms to the b0 form. The b0 form may remain for an extended time (hours to days), although in many fats, it eventually transforms into the b polymorph, which is usually the most stable form. However, in many natural fats, the b0 polymorph can exist for long periods of time because of compound or solid solution formation (11). That is, in some mixed-acid TAGs, no b polymorph may form and b0 is the most stable. In other cases, two b forms may be present (5). For example, SOS, a mixed-acid TAG, has ve polymorphic forms in which two b forms are present. The molecular structures of the ve polymorphic forms have been identied using XRD, differential scanning colorimetry (DSC), and Fourier-transformed infrared spectroscopy (FT-IR) techniques (5). In addition, two liquid crystalline phases called LC1 and LC2 were found for SOS using time-resolved synchrotron radiation X-ray

LIPID PHASE BEHAVIOR

51

TABLE 2. Polymorphic Forms of Cocoa Butter. Melting Temperature ( C) Form I II III IV V VI g a b0 2 b0 1 b2 b Wille and Lutton (13) 17.3 23.3 25.5 27.5 33.8 36.3 Davis and Dimick (13) 13.1 17.7 22.4 26.4 30.7 33.8

diffraction (SR-XRD) analysis (12). The researchers stated that the crystallization properties of SOS polymorphic forms were somehow inuenced by the presence of the two liquid crystal phases. Additionally, more than one subtype within the main polymorphic grouping has been identied in some fats. For example, six different polymorphic forms have been identied in cocoa butter, although there is still some debate whether they are all truly unique polymorphs (Table 2). Two b0 and two b forms have been identied for cocoa butter. These polymorphs have slightly different melting points, but they have X-ray spectra that t within the denition of that polymorph. Different nomenclatures have been used for denoting polymorphic forms, as seen in Table 2 for cocoa butter. In the Greek nomenclature, where polymorphs are given a Greek letter, the most stable form within a polymorph type is given the subscript 1, and other polymorphs within that form are ordered in decreasing stability or melting temperature. For example, cocoa butter has two b0 forms, with the b0 1 form having the highest melting point (most stable). It is also common to see a hyphenated number following the Greek letter, usually 2 or 3, stating the chain-packing arrangement (double or triple chain packing, respectively). Wille and Lutton (13) denoted the different polymorphs of cocoa butter with Roman numerals, ordered in increasing melting point. The time-temperature relationships governing the polymorphic behavior of cocoa butter (in the temperature range of 20 C to 40 C and a time range of 10 days) were investigated by using real-time XRD (15). The g, a, and b0 polymorphs crystallized directly from the melt, and formation of b0 is much quicker when it transforms from a compared with its formation from the melt. The leaststable polymorph g stayed unchanged at solidication temperatures (Tp) below 10 C for 10 days. At higher Tp, the g polymporph transformed to a within a short time. The g always transformed to a and a transformed to b0 . The a phase transformed into b0 phase within 1 hour or less at temperatures above 6 C. They noted that two b phases (polymorphs V and VI) were obtained via direct transformation from the b0 phase only, not from the melt. Direct b phase formation from melt is only viable if the melt has a memory effect. Their observation of two different b phases from the b0 phase is contradictory to the results of the work of SchlichterAronhime and Garti (16) who stated that b-V can be directly formed from the melt and that b-VI can be formed only from the transformation of b-V.

52

CRYSTALLIZATION OF FATS AND OILS

2.3. Phase Behavior In order to understand and control lipid crystallization, one should know the thermodynamic driving force for crystallization. In a pure system, like a single TAG, the melting point, Tm, denes the driving force and a temperature below Tm is required to induce crystallization. That is, the subcooling or the melting temperature minus the actual temperature (Tm T) denes the driving force for crystallization. When two TAGs are mixed together, each species can inuence the melting properties of the other and a phase diagram is needed to dene the crystallization driving force at any condition. Rossell (17) summarized the phase behavior of binary mixtures of various TAGs. Depending on molecular differences (chain length and degree of unsaturation), most binary TAG mixtures had either monotectic, eutectic, or peritectic behavior (Figure 4), where the melting temperatures (liquidus

Figure 4. Phase behavior in binary systems: (a) monotectic, continuous solid solution; (b) eutectic; (c) monotectic, partial solution; and (d) peritectic (18).

LIPID PHASE BEHAVIOR

53

lines) of the species with the higher melting point decreased with increasing addition of the species with the lower melting point. Wesdorp (19) used a thermodynamic approach to predict phase behavior of each of the polymorphs for different binary mixtures of TAGs. At the liquidus line on the phase diagram, the chemical potential of the crystallizing species in the liquid state is equal to the chemical potential of that species in the crystalline state (the denition of equilibrium). If one of the species in a binary mixture is a liquid (oil or solvent), the other species (higher melting point) will dissolve to some extent into the solvent (the liquid oil can be considered a solvent in this case too). For example, a certain amount of trisaturated TAGs (SSS) dissolves in solvent (either organic solvent like acetone or hexane or a liquid oil), with the solubility concentration increasing with temperature in the normal fashion (as shown schematically in Figure 5). In this case, a binary mixture of SSS and solvent can be supersaturated with SSS once its concentration exceeds the saturation concentration at any temperature, as indicated by line AC in Figure 5. Thermodynamically, the driving force for crystallization is the difference in chemical potential of SSS at point A and the chemical potential at saturation (point C). Often, this crystallization driving force is approximated as the difference in concentrations between points A and C. When more than two TAG species are mixed together, the phase behavior is signicantly more complicated. For mixtures of three TAGs, a ternary phase diagram (sometimes called a triangle diagram) can be used to denote phase behavior at any temperature. The effects of temperature on phase behavior, however, must be taken into account in yet another dimension, and thus, characterizing phase behavior in ternary systems gets very difcult very quickly. The situation is even more complex when there are greater than three TAG components, as occurs when a natural fat is crystallized. Natural fats are mixtures of numerous TAGs, containing perhaps 10 to 12 different TAGs (as in cocoa butter) to well over 100 (as in milkfat). In natural

SSS Content (g SSS/g solvent)

Temperature (C)

Figure 5. Schematic of a solubility diagram for a high-melting fat (SSS) in a liquid oil or solvent. Line AC represents supersaturation for system at point A.

54

CRYSTALLIZATION OF FATS AND OILS

fats, the complex interactions among mixtures of various TAGs with different fatty acids (chain length and degree of unsaturation) and having different melting points result in melting over a range of temperatures. This range of temperatures may be fairly narrow (as for cocoa butter) or may be broad (as for milkfat). At a temperature above the melting point of the highest melting component, the entire lipid is melted and the natural fat is in a liquid state. This highest melting point, often characterized as the clear point (the temperature at which the last crystal melts under carefully controlled heating conditions), is actually the melting temperature of the TAG with highest melting point in the specic mixture of the other TAG. Some researchers use this highest melting point, or some measure of melting point like the Mettler dropping point, to dene the driving force for crystallization when the fat is cooled (20, 21). However, when the natural fat contains a wide range of TAGs with different melting points, cocrystallization of different TAGs into compound crystals is dependent on the temperature of crystallization. Thus, the highest melting point does not necessarily represent the true driving force for crystallization of the TAG species that are cocrystallizing. If the fat is cooled to some point below the melting point of the highest melting component and allowed to fully equilibrate (crystallize to the maximum extent in the most stable polymorph), there will be some ratio of solid to liquid fat dependent on the nature of the TAG mixture in the natural fat. This solid fat content (SFC) is often measured by a pulsed nuclear magnetic resonance (NMR) technique. A plot of the maximum amount of fat crystallized (SFC) at sequentially higher temperatures
100 90 80 Solid Fat Content (%) 70 60 50 40 30 20 10 0 0 10 20 30 40 50 Temperature (C)
Figure 6. Solid fat curves for milkfat, cocoa butter, and their mixtures (4).

100% cocoa butter 20% milk fat 50% milk fat 75% milk fat 100% milk fat

LIPID PHASE BEHAVIOR

55

gives a melting prole that represents a type of phase equilibrium for a natural fat. Some fats, like cocoa butter, have a very high SFC at low temperatures (about 90% at 0 C) and then melt very sharply over a narrow temperature range (2535 C). Other natural fats, like milkfat, have lower SFC at low temperatures (about 50% at 0 C) and melt gradually with increased temperature. These SFC melting curves are dependent on the specic molecular composition of the natural fat, as seen in Figure 6 for cocoa butter and milkfat. Although SFC melting curves denote a certain aspect of phase behavior, they are not true phase diagrams because the composition of the crystalline phase changes as temperature increases. Nevertheless, melting proles are useful tools for understanding the crystallization behavior of natural fats. In mixtures of two or more natural fats, as often occurs in processed foods (e.g., milkfat and cocoa butter in chocolate), it is even more difcult to characterize the true phase behavior for crystallization of fats. One approach that has been used to characterize compatibility of fat mixtures is the isosolids diagram (22). SFC melting curves are obtained (by NMR) for various mixtures of the two fats, as seen for cocoa butter and milkfat in Figure 6. Lines of constant SFC for different temperature and composition are calculated and plotted on an isosolids diagram (Figure 7).
40 % SOLID 5 30

Temperature (C)

10 20 60 70 10

20 30 40 50

80 0 100 % ICB 0% AMF

50:50 Composition

100 % AMF 0% ICB

Figure 7. Isosolids diagram for mixtures of anhydrous milkfat (AMF) and cocoa butter (ICB) (4).

56

CRYSTALLIZATION OF FATS AND OILS

Eutectic behavior is seen where the SFC of a mixture falls below the SFC for either of the two individual components, as seen between 30% and 70% milkfat in Figure 7. Isosolids diagrams allow phase compatibility to be studied (4), but they do not provide a thermodynamic measure of driving force for crystallization. Again, because the crystal phase composition may be different at different temperatures (and mixture ratios), isosolids diagrams do not represent true phase diagrams. Recently, attempts have been made to characterize the driving force for crystallization of natural fats by considering classes of TAG (high-melting, low-melting, etc.). For example, milkfat contains three primary fractions that crystallize nearly independently. The effective solubility of the high-melting fraction (HMF) in the low-melting fraction (LMF) was found by using a turbidimetry technique (23). Through chemical analysis of the major TAG constituents of HMF, an effective solubility curve in terms of chemical composition of HMF in LMF was developed and used to characterize the driving force for crystallization, as shown in Figure 8. Such an effective solubility takes into account the intersolubility of different TAGs as well as the melt behavior of individual TAGs. Although this approach is still somewhat empirical, it provides a reasonable approximation of the crystallization driving force in complex lipids. Further work is needed in this area to truly dene the driving force for crystallization in natural fats.

0.80

Critical Concentration metastability

solubility

0.75 Long-chain TAG Content (g(C46-C52)/g(<=C40))

L+S Labile

0.70 AMF 0.65

L + S(conditional)
supersaturation at 30C

0.60

0.55 Critical temperature for AMF 0.50 20 25 30 35 40 45 Temperature (C)

Figure 8. Operational phase diagram for high-melting components of milk fat dissolved in low-melting components of milk fat based on triacylglycerol composition (acyl carbon number) (4).

Liquid

CRYSTALLIZATION BEHAVIOR

57

3. CRYSTALLIZATION BEHAVIOR 3.1. Nucleation Nucleation, or the formation of a crystalline phase from the liquid state, is probably the most important factor in controlling crystallization. The nucleation rate is the major determining factor in the number and size of crystals formed, their polymorphic form, and the ultimate distribution of crystalline solids. Crystallization cannot occur until the phase is supersaturated or subcooled. However, attaining the supersaturated or subcooled state is not necessarily sufcient to promote crystallization because a certain energy barrier exists to formation of nuclei. A nucleus is the smallest crystal that can exist in a solution at a certain temperature. The formation of a nucleus from the liquid phase, or the nucleation process, requires the molecules to organize into a crystal lattice. There is a free-energy barrier opposing this transition, but when nucleation does occur, there is a release of energy (latent heat of fusion) as the molecules assume the lower energy state in the crystal lattice. Based on these energy considerations, a free-energy maximum exists that must be overcome for nucleation to occur (24). At this maximum free energy, there is a critical size for a stable nucleus. Above this critical size, a stable nucleus is formed that continues to grow, whereas clusters smaller than the critical size can potentially disperse into the liquid state (4, 24, 25).
3.1.1. Nucleation Theories Nucleation is generally classied according to pri-

mary nucleation, which may occur either homogeneously or heterogeneously, and secondary nucleation mechanisms. The presence of foreign nucleating sites catalyzes the formation of heterogeneous nucleation, whereas homogeneous nucleation occurs without the assistance of outside surfaces. Secondary nucleation occurs when crystals in a subcooled system spawn new nuclei, generally because of contacts between two crystals, or between a crystal and a surface such as a stirrer or a solid wall (4).
3.1.1.1. Homogeneous Nucleation

Homogeneous nucleation is based on accretion of molecules in the liquid phase. Single species (molecules or ions) come together and form dimers. Dimers become trimers by addition of a molecule, and this accumulation process continues until eventually a stable nucleus forms depending on temperature and supersaturation. According to the classic nucleation theory, a free-energy barrier must be overcome to form a stable nucleus. The energy needed to form a crystal is proportional to the interfacial tension, g, and the surface area. However, once a nucleus is formed, there is a release of energy (latent heat) associated with the phase change. The free-energy change for the formation of the crystal surface is positive and proportional to the surface area (r2) and interfacial tension (g) between the crystal and the surrounding uid. The free-energy change for formation of the bulk of the crystal is negative because energy is released because of latent heat of fusion and proportional to volume (r3). The total free-energy change during nucleation is the

58

CRYSTALLIZATION OF FATS AND OILS

sum of these free-energy terms for the formation of the crystal surface and the crystal volume. Thus, a maximum in free energy occurs during nucleation at some critical nucleus size, rc. The critical nucleus size is the minimum size for a stable nucleus. Above this critical size, a stable nucleus is formed, whereas clusters of molecules smaller than this critical size can potentially redisperse into the liquid phase (4, 2426). Homogeneous nucleation, however, rarely occurs under commercially important conditions. In practice, nucleation is usually dominated by a heterogeneous mechanism, where a foreign surface serves to reduce the energy barrier to nucleation.
3.1.1.2. Heterogeneous Nucleation

Typically, nucleation of fats (as well as most other substances) occurs by a heterogeneous process catalyzed by foreign nucleating sites. The presence of these foreign nucleating sites, like dust particles, vessel walls, and other foreign particles in the system, reduces the free energy required for nucleation. Even though the exact mechanisms of heterogeneous nucleation are not clearly understood, it most likely results from the interactions at the interface between the solid particle and the supersaturated uid. These interactions result in a local ordering of molecules of the crystallizing species; thus, the free energy of formation of a critical size for a stable nucleus is decreased. For example, nucleation on a surface irregularity at a wall results in a decrease in the surface energy required to form a stable nucleus. In general, the capability of a foreign surface to catalyze nucleation is thought to depend on the degree of lattice matching between the solid surface and the crystals of the nucleating species (26), although this trend is not always observed (24). In general, a closer lattice match indicates a greater likelihood that a surface will catalyze heterogeneous nucleation. Because the foreign surface provides some of the energy needed to overcome the formation of the crystal surface, heterogeneous nucleation occurs at lower crystallization driving force (supersaturation or subcooling) than homogeneous nucleation (24, 27). Interestingly, there is an aging effect on the ability of a heterogeneous nucleation site to catalyze nuclei formation (26). That is, the same material nucleated multiple times under identical conditions results in a spread of nucleation capabilities. This variability in heterogeneous nucleation leads to difculties in controlling lipid crystallization. The formation of new nuclei in the presence of existing crystals is called secondary nucleation. Secondary nucleation may occur whenever microscopic crystalline elements are separated from an existing crystal surface (24), although contact secondary nucleation is probably the main mechanism in commercial fat crystallization processes. As a crystal slurry is agitated in a vessel, crystals contacting with other crystals, vessel walls, or stirrer may lead to attrition or fracture of the existing crystal structure, and consequently, secondary nuclei are formed (Figure 9). Contact secondary nucleation has been explained by two possible mechanisms, namely, the adsorption layer theory and microattrition (4). The adsorption layer theory involves displacement of a surface layer of organized molecules (precrystalline)

3.1.1.3. Secondary Nucleation

CRYSTALLIZATION BEHAVIOR

59

Crystal - Wall

Crystal - Crystal

Crystal - Stirrer

Figure 9. Potential sources of contact nuclei in a stirred crystallizer (4).

as a result of crystal interactions or collisions. Thus, precrystalline embryos are dispersed into the crystallizing medium, where under conditions of secondary nucleation they survive and develop into stable nuclei (24). Microattrition theory involves the dispersion of broken pieces of a crystal into the uid, which remain as stable nuclei (28). Production of secondary nuclei may also result from growing crystals containing dislocations, inclusions, or defects (24, 28). As expected, secondary nucleation is also dependent on the crystallization driving force (supersaturation or subcooling), with more stable nuclei being formed at higher supersaturation (25, 27). Secondary nucleation may also occur in static conditions under certain circumstances (11). In lipid systems, needle-like or dendritic crystals that form under certain conditions may lead to the formation of secondary nuclei. Heat dissipation and/ or concentration of noncrystallizing species in certain regions may lead to melting/ dissolution at the base of the branches of dendritic crystals and result in the formation of numerous nuclei centers. Although the exact mechanisms for this type of secondary nucleation are not fully understood, it is undoubtedly important for

60

CRYSTALLIZATION OF FATS AND OILS

nucleation in emulsions and in certain cases during seeding of bulk solutions (as in tempering of chocolates). During fractionation of fats, secondary nucleation is undesired because the small crystals, formed in the presence of larger ones means that subsequent separation is not efcient. Thus, stirring or agitation during fractionation is usually kept to the minimum needed to facilitate heat transfer. Secondary nucleation is inuenced by numerous parameters, including the driving force for crystallization, temperature, additives, impurities, agitator, agitation rate, the number and size of existing crystals, and roughness of the crystallizer surface. The parameters affecting nucleation and nucleation rate will be reviewed in a subsequent section.
3.1.2. Nucleation Kinetics Nucleation rate is generally measured as the rate of

formation of nuclei (numbers formed per unit volume per unit time). Sometimes the induction time, or the time necessary for the onset of nucleation once the subcooled state has been attained, is used for calculation of nucleation rate because the actual rate is often very difcult to measure. Induction time for nucleation will be reviewed later in this section. In some cases, as in crystallization of viscous materials from the melt, the FisherTurnbull equation (29) is often used to describe nucleation of lipids (20, 30) ( )     16pg3 Tf2 NkT Gd J exp : exp h kT 3kTHf2 Tf T2 1

Here, N is the number of molecules (monomers) per mole, k is the Boltzman constant, T is absolute temperature, h is Plancks constant, Gd is a term denoting the mobility of the lipid molecules, g is interfacial tension, Tf is melting temperature, and Hf is latent heat of fusion. The rst exponential term in Equation 1 has been related to the ability of a lipid molecule to attain the necessary conformation to become attached to the crystal lattice, and it is often given as (20) Gd aS ; kT R 2

where a is the fraction of molecules with the correct conguration to be incorporated into the crystal lattice, S is the decrease of entropy associated with incorporation of one mole of lipid, and R is the ideal gas constant. Kloek (31) determined that 80% of TAG molecules were in the correct conformation for incorporation into a nucleus. According to the classic theory, nucleation is a very strong function of crystallization driving force. At low driving forces (low supersaturation or high temperatures), nucleation rate is essentially zero. After some critical driving force is attained, nucleation becomes spontaneous and occurs almost instantaneously once the critical driving force has been attained. In natural fats, cooling below a

CRYSTALLIZATION BEHAVIOR

61

certain temperature results in massive nucleation with numerous nuclei being formed. For fats, the nucleation rate also depends on the type of polymorph formed, because each of the polymorphs has a different melting point and interfacial tension. The a polymorph, the least stable of the common polymorphic forms of fats, has the lowest interfacial tension, heat of crystallization, and melting point temperature. The b0 and b polymorphs have increasing interfacial tensions, heats of crystallization, and melting point temperatures. Thus, as a liquid fat is cooled, the polymorph that forms rst depends on the properties of the different polymorphs. For example, Hernqvist (2) showed that the rst polymorph to appear as trisaturated triacylglycerols (with fatty acids from lauric to stearic acid) were cooled was either the a or b0 polymorph, depending on the chain length, even though the nucleation temperature was well below the melting point of the b polymorph (Figure 10). In the case of tristearin, formation of the a polymorph occurred even though both the b0 and b polymorphs were subcooled to a greater extent (higher driving force). The formation of a less-stable polymorph under conditions where a more stable polymorph is subcooled to a great extent has been explained by the difference in interfacial tensions of the different polymorphs (28). A small difference in interfacial tension can result in a large difference in nucleation rate (25), and this effect

80

mp mp
60 Temperature (C)

mp
liq - liq - 1 liq - 1 liq - 1

40

20

1 -

0 12 14 16 18

nc

Figure 10. Onset temperature of nucleation and polymorphic form of monoacid triacylglycerols with different chain lengths (nc) at slow cooling rate (0.4 C/min). amp, b0 mp and bmp represent the melting temperatures of the different polymorphs (2).

62

CRYSTALLIZATION OF FATS AND OILS

Nucleation Rate

Driving Force
Figure 11. Nucleation rate (highly schematic) of lipid polymorphs (4).

generally is greater than the effect of temperature driving force. Thus, nucleation rate of lipid polymorphs is often considered to follow the general trend shown in Figure 11. Kellens et al. (32) studied the nucleation rate of the b0 polymorph of tripalmitin (PPP) by using a microscope counting technique. An increase in temperature from 45 C to 52 C led to a decrease in nucleation rate, as expected. A semilogarithmic plot of nucleation rate versus the inverse of the square of the subcooling, according to the general form of Equation 1, gave a straight line over the range from 45 C to 50 C. Above 50 C, a different straight line was obtained indicative of formation of a different polymorph (conrmed from the change in crystal habit observed microscopically). Another important kinetic aspect of nucleation is the induction time, dened as the time required for a system to nucleate once a certain subcooling has been attained. That is, induction time for the onset of nucleation is the time required for detection of the rst nuclei in a supersaturated or subcooled system. In reality, induction time includes the true time required for nucleation plus the time required for detection of crystallization by the experimental technique. Techniques that have been used for studying lipid nucleation include microscopy, refractive index, light scattering, calorimetry, viscosity, turbidimetry, laser polarized-light turbidimetry, and NMR (4). Each method has its advantages and limitations for studying lipid nucleation (33). Herrera et al. (34) showed that light microscopy could detect a crystal with a minimum size of 0.2 mm, whereas laser polarized-light turbidimetry detected a smaller size of nuclei. Thus, the laser polarized-light turbidimetry technique was more accurate and suitable when size of nuclei is very small. Any method of studying induction time for nucleation must be used with caution (35).

CRYSTALLIZATION BEHAVIOR

63

120 melt 100

80 Induction Time (s) melt 60 melt

40 20 0 35 40 45 50 Temperature (C) 55 60 65

Figure 12. Induction time kinetics for onset of nucleation of different polymorphis forms of tripalmitin. Melting temperatures of each polymorph indicated by straight line (4).

The induction time, t, is a function of subcooling and reects the time necessary for a critical size of nucleus to be developed in the liquid. The induction time is also dependent on the size at which nuclei are detectable and the growth rate at this early stage. Despite this limitation in measurement methods, induction times are often considered to be inversely proportional to nucleation rate (4) ta J 1 : 3

Induction times for nucleation of a tripalmitin melt at different temperatures are shown in Figure 12 (36). The tripalmitin melt was cooled quickly from 80 C to the different crystallization temperatures indicated on the gure and induction time measured as the rst point of detection of crystals on a polarized light microscope. The relative time scales for the onset of nucleation are clearly shown, with the less-stable a form taking signicantly less time to nucleate than the b0 polymorph. The induction time for the most stable b polymorph was substantially longer than for either of the less-stable polymorphs.
3.1.3. Nucleation in Lipid Emulsions

In many foods, the lipid phase appears in emulsion form, or small droplets of fat dispersed in a continuous aqueous phase,

64

CRYSTALLIZATION OF FATS AND OILS

as for example found in cream (37). The nature of the fat crystals in cream plays an important role in determining the physical properties and quality characteristics of butter. Thus, nucleation of fats in emulsion form is an important commercial phenomenon. When a fat is emulsied, nucleation is substantially altered compared with the same fat in bulk liquid form. This is primarily because of the distribution of heterogeneous nucleation sites among the emulsion droplets. If there are more droplets than heterogeneous nucleation sites, then some of the droplets will nucleate by a homogeneous nucleation mechanism. That is, as a nely dispersed emulsied system is cooled, one population of droplets nucleates at relatively higher temperatures because of heterogeneous nucleation, whereas another population nucleates at substantially lower temperature because of homogeneous nucleation. It is widely recognized that the size of the emulsion droplets is an important factor in the extent of subcooling (11). Smaller droplet size leads to nucleation at a lower temperature (greater degree of subcooling). Thus, the probability of nucleation within an emulsion droplet is lower than in the bulk fat (38). The dispersity of droplet sizes, however, did not change the critical subcooling required for onset of nucleation (39). Crystallization from the emulsied state may lead to different nucleation processes than observed for the same fat in bulk liquid form. It has been suggested that nucleation often occurs at the interface of the droplet where surface-active agents are located. The general similarity of the lipophilic components of surfactants oriented at the surface may provide some ordering and structure for the lipid molecules within the droplet and enhance nucleation, as found for example by Kaneko et al. (40) for a hydrocarbon emulsion. Walstra (11) also suggests that formation of compound crystals from emulsions of natural fats may be different than the same fat crystallized from bulk liquid. The initial polymorph formed may also be different, with more stable polymorphs more likely to form in the emulsion (38). 3.2. Crystal Growth Once nuclei have formed, they grow by the incorporation of other TAG molecules from the liquid phase. The incorporation of a new TAG molecule into an existing crystal lattice depends on the probability of it having the correct conguration at the correct site on a crystal surface. When a molecule diffusing from the liquid phase reaches the crystal surface, it may bind into the crystal lattice or return to the supersaturated system, depending on its conguration. Growth continues as long as there is a driving force for crystallization. Eventually crystal growth ceases when the system attains phase equilibrium or the entire system is crystallized (4). For growth to occur, molecules from the liquid phase must migrate to the surface of the crystal, where rearrangement and orientation takes place. A growth unit (either an individual molecule or a cluster of molecules) then migrates across the crystal surface until it nds an appropriate site for incorporation into the lattice. Once a growth unit has become incorporated, there is a release of latent heat and this energy must be diffused away from the growing surface or else the temperature

CRYSTALLIZATION BEHAVIOR

65

will increase to the point where no further growth can occur. General theories of crystal growth have been developed for crystallization of pure substances (4, 24). These theories are based on one or more of the steps in crystal growth being the rate-limiting step. Further details of these theories can be found in the references by Mullin (24) and Hartel (4). In natural fats, the different TAG species come together to form mixed or compound crystals. The likelihood of two TAG crystallizing together depends on the similarities or differences in molecular conguration (chain length, degree of unsaturation, nature of any double bonds, and arrangement of the fatty acids on the glycerol backbone). TAG species that are similar tend to cocrystallize, but under certain conditions (e.g., very rapid growth), even different TAG species can cocrystallize in a loosely organized crystal lattice (g or a polymorphs). In fact, it is this molecular diversity that results in some natural fats remaining in the metastable b0 polymorph for extended periods of time. Growth of TAG crystals is typically very slow (41). There may be several reasons for slow growth rate of TAG crystals:  The incorporation of a TAG molecule into a crystal lattice requires a very large loss in conformational entropy, and thus, a long time is needed for the TAG molecule to t into the crystalline lattice. In addition, the TAG molecule may be detached before the crystalline lattice before it is fully incorporated into the crystalline lattice. For example, for growth of tristearin (SSS) in triolein (OOO), linear growth rates of the order of 108 to 107 m/s have been observed (41).  In a multicomponent fat, there is a vigorous competition between similar molecules for a vacant site in a crystal lattice. Multicomponent fats crystallize more slowly than pure TAG at the same crystallization driving force. However, crystal growth in multicomponent fats may be enhanced by the formation of compound crystals. Compound crystals usually occur in the a or b0 forms and rarely in the b form. According to Timms (25), more stable polymorphs grow faster than unstable ones at any given temperature. This is because of the higher melting point of the more stable polymorphs, which means that the more stable polymorph has a higher degree of subcooling at any given temperature.

3.3. Modeling of Crystallization Kinetics of Fats Crystallization data have typically been treated theoretically using either the FisherTurnbull model or the Avrami equation. These analyses not only allow lipid crystallization to be modeled but may also shed some light on the mechanisms of nucleation and growth. However, there is some recent debate about the validity of such models, especially the application of the Avrami equation (42) to accurately depict crystallization of lipids.

66

CRYSTALLIZATION OF FATS AND OILS

Recently, Foubert et al. (43) developed a new, empirical model (Foubert model) to predict the kinetics of fat crystallization. Other authors have used a reparameterized Gompertz equation (Gompertz model) to empirically describe crystallization kinetics of fats (44, 45).
3.3.1. Avrami Analysis

The Avrami equation, a general approach for description of isothermal phase transformation kinetics originally developed for polymers (46), is often used for describing nucleation and crystal growth in fats. The Avrami equation is given as 1 X expfktn g; 4

where X is fraction of crystal transformed at time t during crystallization, k is crystallization rate constant that depends primarily on crystallization temperature, and n, the Avrami exponent, is a constant relating to the dimensionality of the transformation. The values of n and k are calculated from the linear form of the Avrami equation (Equation 5) as the slope and intercept at ln t 0, respectively ln ln1 X lnk nlnt: 5

The Avrami exponent (n) is a function of the number of dimensions in which growth takes place, and it reects the details of nucleation and growth mechanisms. For most transformations, the n is found to be constant over a substantial temperature range (47). Christian (48) tabulated some values of n expected for various crystallization mechanisms. For example, an n of 4 indicates heterogeneous nucleation and spherulitic growth from sporadic nuclei, whereas an n of 2 indicates high nucleation rate and plate-like growth (i.e., two-dimensional growth). Metin and Hartel (49) applied the Avrami equation to the isothermal crystallization of binary mixtures of cocoa butter with milk fat or milk fat fractions at 15 C. Avrami analysis indicated an n value of 4 for cocoa butter crystallization, so the suggested mechanism was heterogeneous nucleation with spherulitic growth from sporadic nuclei. For milk fat, the value of n was 3, suggesting that the crystallization mechanism was instantaneous heterogeneous nucleation with spherulitic growth. For milk fat fractions, the n value was 2, which suggested that the mechanism was high nucleation rate at the beginning of crystallization decreasing with time, and plate-like growth. The crystallization rate constant (k) is a combination of nucleation and growth rate constants, and is a strong function of temperature (47). The numerical value of k is directly related to the half time of crystallization, t1/2, and therefore, the overall rate of crystallization (50). For example, Herrera et al. (21) analyzed crystallization of milkfat, pure TAG fraction of milkfat, and blends of high- and low-melting milkfat fractions at temperatures from 10 C to 30 C using the Avrami equation. The n values were found to fall between 2.8 and 3. 0 regardless of the temperature and type of fat used. For temperatures above 25 C, a nite induction time for crystallization was observed, whereas for temperatures below 25 C, no induction time was

CRYSTALLIZATION BEHAVIOR

67

found (crystallization was instantaneous)). Calculation of crystallization rate constant, k, and half time for crystallization based on the Avrami analysis were in line with the two different behaviors observed in SFC values of the fats. Even though the Avrami model has been the most frequently used model to describe the isothermal kinetics of fat crystallization, there are some concerns about the use of the model in fat crystallization. Theoretically, integer values should be obtained for the Avrami exponent, n. However, generally fractional values of n were obtained in crystallization of fats and oils. Additionally, the linear format of the Avrami equation should give a single slope associated with the value of the Avrami exponent. However, in some studies, two regions of different slopes were obtained. Moreover, secondary nucleation during crystal growth is not considered in the Avrami model, which may in part explain the noninteger values of the Avrami exponent.
3.3.2. Fisher-Turnbull Analysis The activation free energy for nucleation, Gc,

may be found from the FisherTurnbull equation given in Equation 1. The term in the second exponential of Equation 1 is often given as Gc =kT. Combination of Equations (1) and (3) allows development of the following equation:  tT ( )    16pg3 Tf2 h aS exp exp : Nk k 3kTHf2 Tf T2

Based on Equation 11, a plot of tT versus {1/T(T)2} leads to a straight line for nucleation of a given polymorph. The critical free energy for nucleation, Gc , is then found from the slope of that straight line, s, as Gc sk Tf T2 : 7

For a given fat system, although the slope is constant, Gc varies with crystallization temperature. The FisherTurnbull approach has been used to compare nucleation of various lipid systems. Ng (51) and Herrera et al. (34), for example, have used this approach to characterize crystallization of palm oil and hydrogenated sunower oil, respectively. The use of the FisherTurnbull approach to characterize nucleation leads to a better understanding of the energy changes needed for onset of nucleation and can be used to compare nucleation in different systems. However, this approach is based on a crystallization driving force dened by a single melting point, which may only occur in cases where a single TAG component (or a TAG grouping with narrow range of melting temperature) crystallizes from a liquid oil. It also applies only when the subcooling is low (typically less than 10 C). In cases where massive cocrystallization and compound crystal formation occurs, this approach does not work.

68

CRYSTALLIZATION OF FATS AND OILS

3.4. Crystalline Microstructure The dispersion of the crystalline fat phase in a material determines the physical and textural properties of a lipid-based product. For example, the hardness, snap, and glossy appearance of chocolate is caused by crystallization of cocoa butter in the form of numerous, very small (1 mm or less) crystals of the most stable polymorph (b form). The size distribution (mean size and range of sizes), polymorphic form, and shape of the fat crystals, as well as the network formed among the crystals, all play important roles in determining physical attributes of lipid-based products. In the case of lipid fractionation, however, a different crystal size distribution is desired. As the fat crystals are to be separated from the liquid phase, uniform crystals of distinct size and shape are needed for the most efcient separation. For the most efcient separation by ltration, reasonably large (200 to 300 mm) crystals of fairly uniform size (narrow distribution of sizes) are needed. Fractionation technologies carefully control nucleation and growth to produce this uniform distribution of crystals to enhance ltration and separation of the high-melting stearin phase from the low-melting olein phase. In crystallization of most natural fats, the rst crystals formed are often observed as thin and fairly long platelets (41). For example, cooling of melted milkfat leads to initial formation of small b0 crystals in needle or platelet shape. As these initial crystals grow, they aggregate into spherulites (52) consisting of the needles arranged radially and ranging in size from a few microns up to about 300 mm. If crystallization is very slow (slow cooling), very large spherulitic crystals form. In contrast, rapid cooling to a low temperature results in the formation of numerous small crystals, often found in a random orientation (53). Thus, cooling rate is one of the most important factors inuencing crystalline microstructure. Further details on lipid crystalline microstructure are given in Chapter 4.

4. CONTROLLING CRYSTALLIZATION 4.1. General Principles of Controlling Crystallization To truly control crystallization to give the desired crystalline microstructure requires an advanced knowledge of both the equilibrium phase behavior and the kinetics of nucleation and growth. The phase behavior of the particular mixture of TAG in a lipid system controls both the driving force for crystallization and the ultimate phase volume (solid fat content) of the solidied fat. The crystallization kinetics determines the number, size, polymorph, and shape of crystals that are formed as well as the network interactions among the various crystalline elements. There are numerous factors that inuence both the phase behavior and the crystallization kinetics, and the effects of these parameters must be understood to control lipid crystallization.

CONTROLLING CRYSTALLIZATION

69

4.2. Parameters Affecting Crystallization Parameters that affect crystallization may inuence either the thermodynamic behavior or the crystallization kinetics (or both). Parameters that inuence lipid crystallization include chemical composition, subcooling, cooling rate, agitation, minor components of fats (mono- and diacylglycerols, polar lipids, etc.), and scale of operation. The effects of these parameters on lipid crystallization will be reviewed briey in this section. More detailed information about the effects of these parameters on lipid nucleation and crystal growth may be found elsewhere (4, 24, 28, 54).
4.2.1. Compositional Parameters

Natural fats are composed of a wide range of TAG that contain fatty acids of differing chain length, degree of unsaturation, and positional arrangement on the glycerol backbone. The fatty acid composition of fats may be broad, as in milkfat, or may be limited, as in cocoa butter. It might be expected that a faster nucleation rate occurs in molecularly similar fats compared with the ones with complex structure (wide range of fatty acid species), but this is not necessarily true. Metin and Hartel (55) observed that the induction times for nucleation of milkfat were signicantly faster than that for cocoa butter at the same isothermal temperatures (and approximately the same melting point). The faster induction time for milkfat may be a result of a higher driving force (even though the difference between crystallization temperature and nal melting point is about the same), or it may be because the TAGs in milkfat more readily come together into mixed crystals. As both are likely to form in a mixture of a and b0 polymorphs, the differences in nucleation rate cannot be attributed to the formation of different polymorphs. Furthermore, when two fats added together are crystallized from the liquid state, the nucleation rate of the mixture often decreases. For example, the addition of milkfat or milkfat fractions to cocoa butter is widely known to retard crystallization of cocoa butter, with higher addition levels having a greater effect. This effect is commercially important because milk chocolate must be processed at lower temperatures to generate the same level of crystallization as dark chocolate. Metin and Hartel (55) documented the inhibitory effects of milkfat and milkfat fractions on induction time for nucleation of cocoa butter. Martini et al. (56) measured the induction time for nucleation for addition of sunower oil to a high-melting milkfat fraction. As the level of sunower oil increased to 40%, the melting point decreased only by a few degrees, but induction time increased by more than a factor of two. This suggests that the effect of sunower oil on inhibiting nucleation of the milkfat was primarily caused by a true inhibition rather than to a decrease in the driving force for crystallization.
4.2.1.1. TAG Composition 4.2.1.2. Minor Constituents

Minor constituents in fats that can inuence crystallization of TAG include the more polar lipids like DAG, MAG, free fatty acids,

70

CRYSTALLIZATION OF FATS AND OILS

phospholipids, and sterols, although there may be trace amounts of other components that can inuence crystallization as well. These constituents have long been considered as active agents for affecting crystallization. In some cases, the presence of these components may enhance crystallization, whereas in other systems, an inhibition is observed. Nucleation of fats may either be enhanced or inhibited by the presence of these minor components. Dimick (57) has argued that the phospholipids in cocoa butter, with higher melting point than the cocoa butter TAG, crystallize rst and subsequently catalyze formation of cocoa butter TAG. The appearance and chemical composition of cocoa butter crystals formed from rened cocoa butter (phospholipids removed) was different from that of the initial crystals formed in nonrened cocoa butter. Recent studies where these minor components have been separated and then added back to the puried TAG have shown that they invariably inhibit nucleation (21). There are three potential mechanisms by which addition of minor lipids might affect crystallization. They may limit mass transfer rates of crystallizing TAG to the appropriate site for incorporation into the lattice, they may adsorb on the surface of the growing crystal or cluster and inhibit further incorporation of the crystallizing TAG, or they may actually be incorporated into the crystal lattice as a crystal forms and grows (4). Through any of these mechanisms, the minor constituents in a fat may affect the polymorphic form that is crystallized and often affects the crystal microstructure through preferential inhibition on certain crystal faces (28). However, in some cases, increased crystallization rate may be observed in the presence of minor constituents. If a macrocrystallizing substance and an additive have a similar structure or form similarly structured complexes to the lattice of the crystallizing substance, then new growth sites on the crystal lattice can be formed by the adsorbed addition. These active sites may be energetically more favorable for incorporating further substances, resulting in an increased crystallization rate (58). For example, Smith et al. (59) found that addition of monolaurin and lauric acid enhanced the crystal growth rate of trilaurin by decreasing facet and crystal size. However, addition of dilaurin decreased the crystal growth rate and altered crystal morphology. They postulated that the varying effects were observed because of the varying sizes and shapes of the additives.
4.2.1.3. Seeding At times, crystallization of natural fats may be promoted by the

addition of a solid seed material, either of the desired crystallizing species or a foreign particle with nucleating properties. If seeds of the desired crystallizing species are added, they can promote further nucleation and/or provide a surface area for additional crystal growth. Smith (60) reported that addition of b0 or b seed crystals to cooled palm oil initiated crystallization at lower degrees of subcooling (higher temperatures) than in the absence of these seeds. In a sense, tempering of chocolate is done to create a small (<3%) population of seed crystals in the melted chocolate, which catalyze further crystallization of the cocoa butter when the chocolate is subsequently cooled. Through the tempering process, seed crystals in the b polymorph are formed. These stable crystals then

CONTROLLING CRYSTALLIZATION

71

promote formation of numerous small cocoa butter crystals, also in the stable b polymorphic form, as the chocolate is cooled. In this case, the existing seed crystals are thought to spawn additional nuclei through secondary nucleation, although the exact mechanism for this process is not clearly understood. A similar effect is observed upon addition of the high-melting TAG, behenic-oleic-behenic (BOB), to chocolate (61). In this case, the BOB molecules, with very high melting point (53 C), catalyze formation of the b polymorph of cocoa butter crystals, eliminating the need for tempering of chocolate.
4.2.2 Operating Parameters 4.2.2.1. Subcooling or Crystallization Temperature Arguably, the most impor-

tant parameter that inuences lipid crystallization is subcooling, or the temperature to which the lipid is cooled below the equilibrium point. As subcooling increases, nucleation rate increases and induction time for crystallization decreases. In many natural fats in bulk liquid form (as opposed to emulsied form), only a few degrees of subcooling are necessary to induce crystallization because of the presence of nucleation sites. These sites catalyze nucleation by lowering the energy required for the formation of nuclei. If subcooling is small, molecules only with the correct conguration (spatial orientation, fatty acid composition, positional arrangement of fatty acids, etc.) are incorporated into a crystal because molecules have sufcient time to orient themselves perfectly. At low subcoolings (crystallization at temperatures within a few degrees of the melting temperature), crystallization rate is slow and only the more stable polymorphs form. When the subcooling is large, incorporation of molecules to the crystal surface is faster, resulting in imperfect attachment of TAG molecules to the surface. Different TAGs can cocrystallize if their chain length and melting points are reasonably close to each other. Consequently, TAGs of different conguration are more easily incorporated into the crystal. The result is more rapid crystallization, but at the cost of formation of compound crystals and lower stability polymorphs.
4.2.2.2. Cooling Rate Fat crystallization is greatly inuenced by the cooling rate

(62). Rapid cooling generally leads to nucleation occurring at a lower temperature than for slow cooling. That is, during slow cooling, the temperature is higher for a longer time and the TAGs have more opportunity to rearrange into a crystal lattice. Cooling rate also affects nucleation rate, which governs crystal size. Rapid cooling to a low temperature promotes a higher nucleation rate, which leads to formation of numerous small crystals (62). When a fat is cooled very slowly, large crystals form. Cooling rate also inuences crystalline microstructure. Marangoni and Hartel (53) used confocal microscopy to show that slowly cooled milkfat formed spherulitic crystals, whereas rapidly cooled milkfat formed random crystalline strands.
4.2.2.3. Agitation (Shear) The speed of mixing is generally thought to promote

both nucleation and crystal growth (4). However, the effects of agitation rate may

72

CRYSTALLIZATION OF FATS AND OILS

be complex because it is sometimes difcult to separate the effects of mixing and cooling rate on crystallization (higher agitation often results in faster cooling rate). Thus, higher agitation rate may inuence crystallization time and crystal size without necessarily inuencing nucleation and growth (41). Agitation may promote nucleation because of the mechanical disturbance that supplies energy to overcome the energy barrier for nucleation (24). Agitation aids cooling, crystallization, and formation of small crystals. Slow cooling rate and slow agitation of fats may result in increased number of mixed crystals; thus, melting range is increased. Higher agitation rate results in a higher crystallization rate and formation of small crystals. Agitation also promotes secondary nucleation, primarily by detachment of small particles from crystal structures. Thus, Herrera and Hartel (62) found that higher agitation rates led to the formation of smaller fat crystals in a milkfat model system. The structure of the crystal network in fats and oils is strongly inuenced by cooling and shear rates, the degree of subcooling, and annealing time. For example, crystalline orientation and acceleration of phase transitions induced by shear in different fats (cocoa butter, milkfat, stripped milkfat, and palm oil) were demonstrated using synchrotron XRD (63). The fats were crystallized under static conditions and under shear (90 s1 and 1400 s1) from the melt (50 C) to 18 C at a rate of 3 C/min. During static crystallization (20 C after 1 day), the initial nucleation was characterized by the appearance of platelet-like nuclei far apart from each other. As they grew, the system became a dispersed suspension of rapidly growing crystals. Eventually, clusters of crystals were formed. The introduction of a moderate shear eld to the fat system seemed to prevent the formation of these clusters. The presence of shear eld resulted in the formation of small asymmetric crystals. Weak or no orientation of the crystals was observed at low shear rates either because of a random distribution of anisotropic crystals or the formation of spherical particles upon aggregation. They also stated that the shear forces accelerated solid-state phase transformations. The effects of agitation rate on crystallization kinetics of butter fat were studied by Grall and Hartel (64). In a 2 L batch crystallizer, increased agitation rate caused an increase in nucleation rate (more crystals generated per unit time) and an increase in total crystallization (mass deposition) rate. However, the effects of agitation on growth rates of individual crystals were dependent on temperature of operation. At 30 C, increased agitation led to a decrease in growth rate, whereas for crystallization at either 15 C or 20 C, increased agitation caused an increase in growth rate. These results may be related to the different composition effects at the different temperatures (different TAGs cocrystallize). Garbolino et al. (65) studied the effects of shear rate on crystallization of a confectionery coating fat (hydrogenated and fractionated mixture of soybean and cottonseed oils) using ultrasonic sensors. They hypothesized that primary nucleation is less likely to be affected by shear and suggested that crystal nuclei probably form from heterogeneous nucleation sites (dust particles or other suspended insoluble materials and imperfections in the container walls). They also suggested that

REFERENCES

73

growth of crystals and their interactions are more likely to be affected by stirring because of the occurrence of frequent interparticle collisions. Thus, from the contradictory results available in the literature, it is clear that our understanding of the effects of heat and mass transfer on crystallization processes is still not complete.
4.2.2.4. Scale of Operation The size of the batch being crystallized may inu-

ence rate of crystallization. For example, crystallization from an emulsion generally occurs at a lower temperature than for the bulk fat based on the separation of catalyzing nucleation sites. In an emulsion, the catalyzing nucleation sites are more dispersed (spread through the number of droplets) and this leads to nucleation at a lower temperature than the same fat in bulk phase. Grall and Hartel (64) studied crystallization of milkfat at different scales of operation (2 L and 20 L) and found induction times for nucleation were lower but individual crystal growth rates were higher in the larger scale crystallizer. Other crystallization parameters (total crystal number, mean size, yield, and nucleation rate) were not signicantly inuenced by this difference in crystallizer size. As scale of operation changes, mixing rates and heat transfer rates change as well, which can inuence crystallization processes. Scale up of fat crystallization processes is still somewhat of a trial and error process because of the lack of fundamental understanding of the effects of heat and mass transfer on lipid crystallization.

5. SUMMARY Controlling lipid crystallization in foods has proven to be a technical challenge over the years. Despite a considerable amount of study, controlling the complex interactions between the various lipid components during crystallization remains essentially an empirical process of studying the effects of various operating parameters on crystal formation. Further work on the fundamental principles of lipid nucleation, growth, and polymorphic transformation is needed to truly control crystallization of lipids in foods.

REFERENCES
1. K. Larsson, Fette Seifen Anstrichmittel., 74, 136142 (1972). 2. L. Hernqvist, Fette Seifen Anstrichm. 86, 297300 (1984). 3. L. Hernqvist, in N. Garti and K. Sato, eds. Crystallization and Polymorphism of Fats and Fatty Acids, Marcel Dekker, New York, 1988, pp. 97137. 4. R. W. Hartel, Crystallization in Foods. Aspen Publishers, Inc., Gaithersburg, MD, 2001. 5. K. Sato, Chem. Eng. Sci., 22552265 (2001).

74

CRYSTALLIZATION OF FATS AND OILS

6. K. Larsson, in S. Friberg and K. Larsson, eds. Food Emulsions, 3rd ed. Marcel Dekker, New York, 1997, pp. 111140. 7. E. S. Lutton, J. Am. Oil Chem. Soc., 27, 276281 (1950). 8. K. Larsson and P. Dejmek, in S. Friberg and K. Larsson, eds. Food Emulsions. 2nd ed. Marcel Dekker, Inc., New York, 1990, pp. 97125. 9. K. Larsson, Acta Chem. Scand., 20, 22252260 (1966). 10. A. E. Bailey, in Melting and Solidication of Fats, Interscience Publishers, New York, 1950, pp. 2129. 11. P. Walstra, in J. M. Blanshard and P. Lillford, eds. Food Structure and Behaviour, Academic Press, Orlando, Florida, 1987, pp. 6785. 12. S. Ueno, A. Minato, H. Seto, Y. Amemiya, and K. Sato, J. Phys. Chem. B, 101, 68476854 (1997). 13. R. L. Wille and E. S. Lutton, J. Am. Oil Chem. Soc., 43, 491496 (1966). 14. T. R. Davis and P. S. Dimick, in Proceedings of the Pennsylvania Manufacturing Confectioners Association (PMCA) Production Conference, Center Valley, PA, 1986, pp. 104108. 15. K. Van Malssen, A. van Langevelde, R. Peschar, and H. Schenk, J. Am. Oil Chem. Soc., 76, 669676 (1999). 16. J. Schlichter-Aronhime and N. Garti, in N. Garti and K. Sato, eds. Crystallization and Polymorphism of Fats and Fatty Acids, Marcel Dekker, New York, 1988, pp. 363393. 17. B. Rossell, Prog. Lipid Res., 23, 138 (1967). 18. R. E. Timms, Progress Lipid Res., 23, 138 (1984). 19. L. Wesdorp, Ph.D. Dissertation, Technical University of Delft, The Netherlands, (1990). 20. W. Kloek, P. Walstra, and T. Van Vliet, J. Am. Oil Chem. Soc., 77, 643652 (2000). 21. M. L. Herrera, M. de Leon Gatti, and R. W. Hartel, Food Res. Int., 32, 289298 (1999). 22. G. Bigalli, Manufacturing Confectioner, 68, 6670, 7980 (1988). 23. R. W. Hartel and K. E. Kaylegian. in N. Garti and K. Sato, eds. Crystallization Processes in Fats and Lipid Systems, Marcel Dekker, New York, 2001, pp. 329355. 24. J. W. Mullin, Crystallization. 3rd ed., Butterworth-Heineman Ltd., Oxford, U.K., 1993, pp. 172260. 25. R. E. Timms, in R. J. Hamilton, ed. Developments in Oils and Fats, Chapman & Hall, Glasgow, U.K., 1995, pp. 204223. 26. A. G. Walton, in A. C. Zettlemoyer, ed. Nucleation, Marcel Dekker, New York, 1969, pp. 225308. 27. J. Garside, in J. M. Blanshard and P. Lillford, eds. Food Structure and Behaviours. Academic Press, Inc., Orlando, FL, 1987, pp. 3545. 28. R. Boistelle, in N. Garti and K. Sato eds., Crystallization and Polymorphism of Fats and Fatty Acids, Marcel Dekker, New York, 1988, pp. 189226. 29. D. Turnbull and J. C. Fisher, J. Chem. Phys. 17, 7173 (1949). 30. R. F. Strickland-Constable, in R. F. Strickland-Constable, ed. Kinetics and Mechanisms of Crystallization, Academic Press, London, U.K., 1968, pp. 74129. 31. W. Kloek, Ph.D. Dissertation. University of Wageningen, The Netherlands, 1998. 32. M. Kellens, W. Meeussen, and H. Reynaers, J. Am. Oil Chem. Soc., 69, 906911 (1992).

REFERENCES

75

33. A. J. Wright, S. S. Narine, and A. G. Marangoni, J. Am. Oil Chem. Soc., 77, 12391242 (2000). 34. M. L. Herrera, C. Falabella, M. Melgarejo, and M. C. Anon, J. Am. Oil Chem. Soc., 75, 12731280 (1998). 35. S. Martini and M. L. Herrera, J. Am. Oil Chem. Soc., 79, 411412 (2002). 36. K. Sato and T. Kuroda, J. Amer. Oil Chem. Soc., 64, 124130 (1987). 37. D. Precht, in N. Garti and K. Sato, eds. Crystallization and Polymorphism of Fats and Fatty Acids, Marcel Dekker, New York, 1988, pp. 305362. 38. P. Walstra and E. C. H. Beresteyn, Neth. Milk Dairy J., 29, 3565 (1975). 39. D. N. Kaschiev, Kaneko, and K. Sato, J. Colloid Interface Sci., 208, 167177 (1998). 40. N. Kaneko, T. Horie, S. Ueno, and K. Sato, J. Crystal Growth, 197, 263270 (1999). 41. P. Walstra, W. Kloek, and T. van Vliet, in N. Garti and K. Sato, eds. Crystallization Processes in Fats and Lipid Systems, Marcel Dekker, New York, 2001, pp. 289328. 42. A. G. Marangoni, J. Am. Oil Chem. Soc., 75, 14651467 (1998). 43. I. Foubert, P. A. Vanrolleghem, B. Vanhoutte, and K. Dewettinck, Food Res. Int., 35, 945 956 (2002). 44. W. Kloek, P. Walstra, and T. Van Vliet, J. Am. Oil Chem. Soc., 77, 389398 (2000). 45. B. Vanhoutte, K. Dewettinck, I. Foubert, B. Vanlerberghe, and A. Huyghebaert, Eur. J. Lipid Sci. Tech., 104, 490495 (2002). 46. M. Avrami, J. Chem. Phys., 8, 212224 (1940). 47. J. W. Graydon, S. J. Thorpe, and D. W. Kirk, J. Non-Crystalline Solids, 175, 3143 (1994). 48. J. W. Christian, The Theory of Transformation in Metals and Alloys, 2nd ed., Pergamon Press, London, U.K., 1975. 49. S. Metin and R. W. Hartel, J. Am. Oil Chem. Soc., 75, 16171624 (1998). 50. I. Arvanitoyannis and J. M. V. Blanshard, J. Food Sci., 59, 197205 (1994). 51. W. L. Ng, J. Am. Oil Chem. Soc., 67, 879882 (1990). 52. H. Mulder and P. Walstra, Products and Comparable Foods, Commonwealth Agric. Bureaux, Bucks, England, 1974, pp. 3352. 53. A. G. Marangoni and R. W. Hartel, Food Technol., 52, 4651 (2000). 54. D. Aquilano and G. Sgualdino, in N. Garti and K. Sato, eds., Crystallization Processes in Fats and Lipid Systems, Marcel Dekker, New York, 2001, pp. 152. 55. S. Metin and R. W. Hartel, J. Thermal Anal., 47, 15271544 (1996). 56. S. Martini, M. L. Herrera, and R. W. Hartel, J. Agric Food Chem., 49, 32233229 (2001). 57. P. S. Dimick, in N. Widlak, ed. Proceedings of International Conference on Physical Properties of Fats, Oils and Emulsiers with Applications to Foods, Am. Oil Chem. Soc. Press, Champaign, IL, 2000, pp. 138161. 58. J. Nyvlt, O. Sohnel, M. Matuchova, and M. Broul, The Kinetics of Industrial Crystallization, Elsevier Science Publishers, New York, 1985, pp. 189191. 59. P. R. Smith, D. J. Cebula, and M. J. W. Povey, J. Am. Oil Chem. Soc., 71, 13671372 (1994). 60. K. W. Smith, in N. Garti and K. Sato, eds. Crystallization Processes in Fats and Lipid Systems, Marcel Dekker, New York, 2001, pp. 357380. 61. T. Koyano, I. Hachiya, and K. Sato, Food Structure, 9, 231240 (1990).

76

CRYSTALLIZATION OF FATS AND OILS

62. M. L. Herrera and R. W. Hartel, J. Am. Oil Chem. Soc., 77, 11771187 (2000). 63. G. Mazzanti, S. E. Guthrie, E. B. Sirota, A. G. Marangoni, and S. H. J. Idziak, Crys. Growth Design, 3, 721725 (2003). 64. D. S. Grall, and R. W. Hartel, J. Am. Oil Chem. Soc., 69(8), 741747 (1992). 65. C. Garbolino, G. R. Ziegler, and J. N. Coupland, J. Am. Oil Chem. Soc., 77, 157162 (2000).