Phase Transitions and Polymorphism of Cocoa Butter

C. Loisela,b, G. Kellera, G. Lecqb, C. Bourgauxc, and M. Ollivona,*

a
Laboratoire de Physico-Chimie des Systèmes Polyphasés, Université Paris-Sud, 92296 Châtenay-Malabry, France,
b
Centre Jean Thèves, groupe DANONE, Branche Biscuits, 91207 Athis-Mons Cedex, France, and cLaboratoire
pour l’Utilisation du Rayonnement Electromagnétique, Université Paris-Sud, 91405 Orsay, France

ABSTRACT: The polymorphism and phase transitions of Natural fats are mainly composed of triacylglycerols (TAG)
cocoa butter (CB) have been reexamined separately by differen- in which some minor compounds are solubilized. Their TAG
tial scanning calorimetry (DSC) and X-ray diffraction as a func- composition determines their physical, in particular their ther-
tion of temperature (XRDT) at scanning rates between 0.1 to mal, properties. While most, but especially animal, fats are
5°C/min and 0.1 to 2°C/min, respectively. A new instrument, composed of complex mixtures of numerous TAG, some veg-
which allowed simultaneous DSC and XRDT recordings from
etable fats, such as cocoa butter (CB), have a simpler compo-
the same sample by taking advantage of the high-energy flux of
a synchrotron, was employed for characterization of the inter-
sition. This fat is composed of a mixture of three classes of
mediate phase transitions. These techniques allowed us to con- TAG (saturated, monounsaturated, and polyunsaturated
firm the existence of the six polymorphic forms of CB (called I TAG), the monounsaturated being by far the major compo-
to VI) by in situ characterization of their formation in the DSC + nent because they represent more than 80% of the total. More-
XRDT sample holder. A detailed study of Form I structure led us over, only three TAG [1,3-palmitoyl-2-oleoylglycerol (POP),
to propose a liquid-crystal organization in which some of the 1-palmitoyl-2-oleoyl-3-stearoylglycerol (POSt), and 1,3-
chains displayed sharp long-spacing lines (d001 = 52.6 ± 0.5 Å) stearoyl-2-oleoylglycerol (StOSt)] account for more than
and a β′ organization (4.19 and 3.77 Å), while the others re- 95% of this fraction. Polyunsaturated and trisaturated TAG
mained unordered with broad scattering (maxima at about 112 correspond to about 13 and 3%, respectively, of the TAG con-
and 36.5 Å). The organization of this liquid crystalline phase is tent. This specificity gives CB a thermal and structural behav-
compared to that of fat and oil liquids. This liquid crystalline
ior similar to that of a pure compound. This behavior is nev-
phase progressively transformed on heating into a more stable
phase (Form II, α type, d001 = 48.5 ± 0.5 Å and short-spacing at
ertheless complex because all major TAG in its composition,
4.22 Å). Form III was only observed in a sharp temperature do- like most lipids, themselves show complex polymorphism
main through its specific short-spacings. The existence of the (e.g., six different polymorphic varieties have been observed
six species has been essentially related to the crystallization of for POP) (1).
monounsaturated triacylglycerols (TAG), while trisaturated Fat polymorphism and TAG crystallization have been
species were found partly solid-soluble in these six polymor- studied for more than a century. Polymorphism results from
phic forms. An insoluble fraction crystallized independently of the different possibilities of lateral packing of the fatty acid
the polymorphism of the monounsaturated TAG in a separate chains and of the longitudinal stacking of molecules in lamel-
phase with long-spacings that were either of the α (49.6 ± 0.5 lae (1–6). These two levels of organization are easily identifi-
Å) or β (44.2 ± 0.5 Å) form. In mixture with Form V, this frac- able from the short- and long-spacings observed by X-ray dif-
tion melts and solubilizes in the liquid phase at 37.5°C. Isola-
fraction (XRD) at wide and small angles, respectively. The
tion of these high-melting crystals shows a melting point of
about 50°C. High-performance liquid chromatography analysis
three main organizations frequently observed for the lateral
of this fraction confirmed an increase from 3.0 to 11.3% of sat- packing of TAG, called α, β′ and β, in the order of their in-
urated TAG and their association with part of the 1,3-stearoyl- creasing stability, have been related to different subcells that
2-oleoylglycerol preferentially to 1-palmitoyl-2-oleoyl-3- have been described in detail (2–6).
stearoylglycerol and (1,3-palmitoyl-2-oleoylglycerol). The lipid long-spacings correspond to the repeat distance
JAOCS 75, 425–439 (1998). in the direction perpendicular to the lamellae. For TAG, long-
spacings are commonly double or triple chainlengths (2L or
KEY WORDS: Chocolate crystallization, cocoa butter, DSC, 3L). A fourth crystalline structure, often called sub-α, al-
liquid crystals, palm oil, phase transitions, polymorphism, tria- though it contains a β′ subcell and would be less stable than
cylglycerols, X-ray diffraction. the α form, is also reported in the literature (7). Moreover, the
existence of a liquid state of TAG, which is also organized as
a liquid crystalline phase, is still under debate (3,8,9).
The polymorphism of CB, which is a vegetable fat mainly
*To whom correspondence should be addressed at Laboratoire de Physico-
Chimie des Systèmes Polyphasés, CNRS URA 1218, Université Paris-Sud, used by chocolate manufacturers, has often been discussed in
92296 Châtenay-Malabry, France. E-mail: ollivon@psisun.u-psud.fr. the literature because it is related to the organoleptic and

Copyright © 1998 by AOCS Press 425 JAOCS, Vol. 75, no. 4 (1998)
426 C. LOISEL ET AL.

TABLE 1
Long- and Short-Spacings (Å) of the Different Polymorphic Forms of Cocoa Butter (literature values)a
I (2L)b II (2L)b III (2L)b IV (2L)b V (3L)b VI (3L)b
Reference β′-sub(α) α β′2 β′1 β2 β1
10 4.19 (vs) 49 (vs)c 49 (vs)c 45 (vs)c 63.1 (m)c 63.1 (vs)c
3.70 (s) 16.3 (s)c 4.25 (vs) 14.87 (s)c 32.2 (vs)c 32 (?)c
4.24 (vs) 3.86 (s) 4.35 (vs) 12.8 (s)c 12.76 (s)c
4.15 (vs) 4.58 (vs) 4.59 (vs)
3.98 (vs) 3.70 (s)
3.67 (w)
15 54 (s)c 51 (vs)c 51 (vs)c 49 (vs)c 66 (s)c 63 (s)c
4.17 (s) 4.20 (vs) 4.20 (vs) 14.8 (w)c 33 (s)c 31 (s)c
3.87 (m) 3.87 (w) 4.32 (s) 12.8 (m)c 12.7 (mw)c
4.13 (s) 4.58 (vs) 4.53 (vs)
3.98 (ms) 3.84 (m)
3.65 (ms) 3.67 (s)
16 45c 64c 63.8c
4.35 4.58 4.59
4.15 3.98 3.7
a
vs = very strong, s = strong, ms = medium strong, mw = medium weak, w = weak.
b
2L, 3L = number of hydrocarbon chains in the crystalline structure.
c
Long-spacings.

physical characteristics of the final products (snap, molding TAG (67.7%), promote the crystallization of CB. Moreover,
contraction, gloss and blooming during storage). It often has it has been found that CB crystallization is related to its com-
been compared to that of POP, POSt and StOSt (about 17, 37, position (27,28). Cebula et al. (29) observed that trisaturated
and 27% of the fat, respectively) and that of their mixtures TAG did not hinder the formation of the appropriate amounts
(1,6,10–13). CB polymorphism is commonly described in the of symmetrical TAG seeds during chocolate tempering.
literature in terms of six different polymorphic forms, noted Recently, access to high-flux sources (synchrotron) has
from I to VI in increasing order of melting points, according permitted study of the thermal behavior of pure TAG by XRD
to the nomenclature of Wille and Lutton (10,14–20; Tables 1 as a function of temperature (XRDT) (30–33). This technique
and 2). Although these six forms have been confirmed by allows quantitative monitoring of the evolution of both long-
other authors (15,17–20), the existence of some of them is de- and short-spacings as a function of temperature, for instance
bated. According to some authors (10,19,21,22), Form III during phase transitions that occur on heating of unstable
corresponds to a mixture of Forms II and IV and is not a sep- species of pure compounds. More recently, the polymorphism
arate crystalline variety. In the same way, Form I could be a and phase transitions displayed by complex TAG mixtures
phase mixture (10), and Form VI may result from phase sep- also have been characterized by XRDT and compared to dif-
aration within a solid solution (19,21,23). More recently, van ferential scanning calorimetry (DSC) recordings (34). Lavi-
Malssen (24) criticized the designation of solid states of com- gne (35) used these techniques to study the thermal behavior
plex mixtures, crystallized as polycrystalline aggregates, as of anhydrous milk fat and its fractions. Chemical TAG analy-
different polymorphs when the chemical compositions of the sis of each fraction allowed the complex thermal profiles
different crystallites are not identical. According to him, sev- recorded by DSC to be decomposed into elementary phase
eral polymorphs of CB and at least the phases sub-α, α, and transitions and to identify the TAG involved in each of them.
β′ probably correspond to nonhomogeneous crystalline states. A model of the molecular organization of the β′ phase of
Davis and Dimick (25,26) postulated and verified by the complex fat mixtures was proposed (35). This demonstrates
analysis of seed crystals that some minor components, such that the combination of these two techniques of thermal
as glycolipids (11.1%), phospholipids (6.6%) and saturated analysis is especially suitable for the characterization of co-

TABLE 2
Melting Point (°C) of the Different Polymorphic Forms of Cocoa Butter (literature values)
References
14 10 15 17 18 19 20
18 (γ) 17.3 (I) 14.9–16.1 (I) 13 (VI) 16–18 (γ) 13.1 (I)
23.5 (α) 23.3 (II) 17–23.2 (II) 20 (V) 20.7–24.2 (α) 17.7 (II)
25.5 (III) 20.7 (III) 22.8–27.1 (III) 23 (IV) 22.4 (III)
28 (β′′) 27.5 (IV) 25.6 (IV) 25.1–27.4 (IV) 25 (III) 26–28 (β′) 26.4 (IV)
33 (β′) 33.87 (V) 30.8 (V) 31.3–33.2 (V) 30 (II) 33.7–34.9 (β) 30.7 (V)
34 (β) 36.3 (VI) 32.3 (VI) 33.8–36 (VI) 33.5 (I) 33.8 (VI)

JAOCS, Vol. 75, no. 4 (1998)

 COCOA BUTTER POLYMORPHISM 427

existing organizations that result from the phase separations TABLE 3

frequently occurring in complex fat mixtures. Concentration of TAG in Initial and Fractionated Cocoa Butter (wt%)a
The aim of this paper is to extend this approach to a fat that Low-melting Initial cocoa High-melting
TAG fraction butter fraction
is considered to behave as a pure compound, in studying, by
combined and simultaneous DSC and XRDT, the complex OLO 0.4 0.4 0.4
PLO 0.4 0.5 0.3
phase transitions that occur during CB crystallization and PLP 1.7 1.6 1.1
melting. The high energy of the synchrotron source allows OOO 0.2 0.4 0.2
possible phase separations to be detected and the competition POO 2.8 2.8 2.2
between the different polymorphic species to be followed PLSt 1.5 1.4 1.1
quantitatively, even at fast scanning rates, in the short-spac- StOO 3.3 3.4 3.1
StLSt 3.2 3.0 2.6
ing domain at large angles as well as in the long-spacing do- POP 17.3 17.0 14.8
main at very small angles. POSt 37.3 36.6 34.1
StOSt 27.3 27.3 27.4
StOAr 1.8 1.9 1.4
MATERIALS AND METHODS PPSt 1.3 1.4 3.4
PStSt 0.6 1.1 4.8
Samples. The CB used was a standard factory product that orig- StStSt 0.4 0.5 3.1
inated from the Ivory Coast. Its composition, determined by Saturated TAG 2.3 3.0 11.3
gas chromatography (norm ISO 5508/5509) and high-perfor- Monounsaturated TAG 83.7 82.8 77.7
mance liquid chromatography (HPLC) (IUPAC 2.324) for Polyunsaturated TAG 13.5 13.5 11.0
TAG analysis and by atomic absorption (IUPAC 2-423; see Total 99.5 99.3 100.0
a
Ref. 6, Chapter 14, for these methods) for phospholipid con- Abbreviations: Ar: arachidic acid, (C 20:0); L: linoleic acid, (C 18:2); O:
oleic acid, (C 18:1); P: palmitic acid, (C 16:0); St: stearic acid, (C 18:0).
tent, was as follows: TAG 97%; diacylglycerols 1.1%; monoa-
TAG, triacylglycerols.
cylglycerols 0.2%; free fatty acids 1.3%; phosphatides 0.15%;
others 0.25%. The TAG composition is reported in Table 3.
Pure palm oil and its fractions (low- and high-melting frac-
tions) were donated by Sanofi-Soprorga (Saint Denis, France). associated with a cryostat (Lauda, RCS6; Königshofen, Ger-
Tristearin (StStSt) and POP were synthesized in the laboratory many), these sample holders allow heating or cooling of the
(36), POP was further purified by HPLC (35) (the purity of sample at rates ranging from 0.1 to 10°C/min. Gas-filled lin-
both was about 99.9%). Triolein (99%) was from Sigma (l’Isle ear detectors (1024 channels, either filled with Ar or a
d’Abeau, France). Xe–CO2 mixture) were used for data collection. Standardiza-
CB fractionation. CB fractions were obtained after 3 wk of tion was carried out with the form β of pure StStSt, which is
storage at 30.0 ± 0.5°C, in tubes capped under argon gas, after characterized by a long-spacing of 44.98 Å at room tempera-
initial melting at 80°C over about 10 min. At the end of this in- ture (6).
cubation time, the crystals [high-melting fraction (HMF)] and All XRD patterns were recorded by transmission through
the liquid [low-melting fraction (LMF)] were isolated by cen- glass capillaries (GLAS, Berlin, Germany). Samples were
trifugation at 30 ± 1°C and 64,000 × g (Sigma; model 3K30) in prepared by filling these glass capillaries with about 20–30
a preheated rotor. The separated HMF was isolated by aspirat- µL of melted fats at temperatures about 20°C above their final
ing most of the liquid fraction with a pipette and then eliminat- melting points and centrifuging them immediately before fat
ing the residual liquid with blotting paper. This separation took solidification could take place. Quenching of these capillaries
less than 2 min at room temperature (21 ± 1°C). Samples were with the melted samples was first accomplished either by
stored at 4°C before thermal and XRD analysis. TAG composi- placing them rapidly in contact with a metal block cooled to
tions of these fractions were determined as above. −30°C and then transferring them to the precooled sample
XRD. XRD was performed with either a classical installation holder (−10°C), or by directly introducing them into the lat-
(Enraf-Nonius, Cu, 40 kV, 20 mA; Gagny, France) and a go- ter. A more convenient technique was developed to prevent
niometer (Seifert; Alexandre Labo, Toulouse, France) or with breakage of the capillary and uncontrolled solid–solid heat
the high-energy synchrotron beam at L.U.R.E. (Laboratoire transfer. Very fast quenching was attained by plunging capil-
pour l’Utilisation du Rayonnement Electromagnétique, Orsay, laries into a vessel with ethanol maintained at −30 or −70°C.
France) by using the D24 station (about 5·109 photons/s/mm2) Form V of CB was prepared by rapid crystallization at 21
operated at λ = 1.488 Å. Both installations were equipped with ± 1°C in a glass capillary and storage for 1 wk at this temper-
Peltier temperature-controlled sample holders, which recently ature. Depending on heating or cooling rates, the time of ex-
have been described (34). posure corresponding to one recording of a diffraction pattern
Briefly, single-stage Peltier modules (Melcor, AMS Elec- was between 60 s (2°C/min) and 300 s (0.1°C/min). Two ex-
tronic, Les Ulis, France) were used to cool these devices perimental setups were employed, and most of the XRDT ex-
(below −20°C in the absence of heating), while heating was periments were recorded twice, the first at short sample-de-
monitored by a temperature controller (Eurotherm 902P, tector distance (d = 290 mm) to determine events at both
Dardilly, France) with a type E thermocouple as sensor. When short- and long-spacings simultaneously, and then at long dis-

JAOCS, Vol. 75, no. 4 (1998)

428 C. LOISEL ET AL.

tance (d = 1450 mm) to obtain a better resolution of long-

spacing lines at small angles (0–2°).
Because the intensities of the diffraction peaks varied as a
function of temperature during XRDT recordings, they were
normalized as follows. At temperature T, the intensity of a line
IT was determined either by an X-ray analysis computer pro-
gram (37) or by quadratic regression analysis of the data that
corresponded to the peak maximum and its surroundings. The
normalized intensities were calculated from that of maximum
Imax and minimum Imin : IT% = 100 · (IT − Imin)/ (Imax − Imin).
DSC. Thermal analyses were conducted on a Perkin-Elmer
(St. Quentin en Yvelines, France) differential scanning
calorimeter (DSC-7) in aluminum pans of 40 µL (pan, part
#B014-3024 and cover, part #B014-3004). Calibration was
made with lauric acid (purity >99.9%), so that temperature and
enthalpies could be corrected for each heating rate, as previ-
ously described (38). Unstable species were studied by quench-
ing melted CB (80°C) at −40°C (in less than 1 s) and subse-
quent heating at 0.1, 1, and 2°C/min. Intermediate species of FIG. 1. Small-angle X-ray diffraction (XRD) recordings of Form V of
fractionated and pure CB were studied by cooling the melted cocoa butter (CB) (solid line) and of its high-(HMF) and low-melting
CB at −2°C/min and heating it at 5°C/min. The peak tempera- (LMF) fractions. Insert shows Form V of CB on an enlarged y scale.
tures were taken at the maxima of the endotherms instead of
the onsets because of peak overlapping. Therefore, the melting 34°C and at 0.3°C/min from 34 to 41°C. The long- and short-
points observed for the different polymorphic varieties at these spacings observed at 25°C with the short sample-detector dis-
peak maxima were systematically higher than the values re- tance (LS = 66 and 33 Å; SS = 4.58, 3.98, 3.89, 3.77, and 3.67
ported in the literature or measured at onset (39). Å) confirmed, as expected, the crystallization of CB into a
β-3L structure (Form V, Table 1). A weak line at about q = 0.14
Å−1, corresponding to a long-spacing of 44.4 Å, was also ob-
RESULTS
served at 25°C and was confirmed on recordings obtained at
More stable forms. XRD. XRDT of CB (Form V) was long sample-detector distance (Fig. 1). The evolution of the
recorded every minute during heating at 1°C/min from 25 to intensity of this line as a function of temperature (between 25

FIG. 2. Evolution of the intensities of the small-angle XRD spacings of Form V of CB between 25 and 40°C. For
each spacing, a series of lines is observed, the intensities of which have been expressed as a percentage of the peak
maximum. Absolute intensities observed for Form V of CB are shown in Figure 1. Relative intensities of the peaks
are given on the graph within parentheses: 100% corresponds to the strongest peak. Insert: enlarged scale. See Fig-
ure 1 for abbreviations.

JAOCS, Vol. 75, no. 4 (1998)

 COCOA BUTTER POLYMORPHISM 429

and 41°C) was compared to that of the β-3L (V) structure (Fig. spectively) and its LMF (2.3 and 13.5%, respectively). TAG
2) by using data recorded with the short sample-detector dis- analysis shows that StStSt, PStSt, and PPSt concentrations
tance. Above 34.5°C, all lines (long- and short-spacings) that were increased about 8, 7, and 2.5 times, respectively, in the
characterize Form V disappeared (they were progressively re- HMF compared with LMF. On the other hand, a decrease of
placed by the bumps that correspond to scattering from the liq- monounsaturated TAG was observed in HMF (77.7%) com-
uid organization), except the line at 44.4 Å, which finally van- pared with LMF (83.7%), which was specific for POP
ished at 37.5°C. The observation at 25°C of a line at 44.4 Å (14.8/17.3) and for POSt (34.1/37.3), while StOSt was found
was first interpreted as the presence, in β-3L (Form V), of a to be constant in the two fractions (27.4/27.3). Fatty acid
trace amount of the β′ form (Form IV) because the latter is analysis confirmed these results and showed an increase of
characterized by a similar long-spacing at about 45 Å (Table stearic acid from 36.4 to 41.6%, from LMF to HMF, while
1). However, the melting point of this form, within the temper- palmitic, oleic, and linoleic acids decreased from 25.5 to
ature interval 25–28°C (Table 2), is clearly incompatible with 24.6%, 33.1 to 29.3%, and 2.6 to 2.1%, respectively. So, XRD
the persistence of the line observed by XRDT above 37°C and lines at 66 and 33 Å, observed in the HMF (Fig. 1), can be at-
rules out this hypothesis, even taking into account the heating tributed to the crystallization of a β-3L form of monounsatu-
rate of 1°C/min, because melting of Form V ends at 34.5°C. rated TAG, while the line at 44.4 Å is related to that of the
These results show that a lipid segregation occurs in CB StStSt.
on cooling and/or during storage. If the segregation occurs Thermal analysis. Thermal analysis was successively con-
during cooling, it would be a rapid process because sample ducted at −2°C/min on cooling between 60 and −40°C and at
crystallization was observed in the capillary after a few sec- +5°C/min in the same temperature range, with pure CB,
onds (see the Materials and Methods section). The fraction HMF, and LMF samples. In the DSC recordings reported in
that crystallizes in addition to the usual β-3L form (V) ex- Figure 3A, three peaks appeared during pure CB crystalliza-
hibits a higher melting point than Form V and a long-spacing tion, at 19.2, 14.2, and 1.7°C. During the crystallization of
(44.4 Å) that corresponds to a 2L packing close to that of the
β form of StStSt (about 45 Å, see above).
This behavior may be considered in the light of recent
DSC observations made by Davis and Dimick (25,26) on CB
crystallization. They found that crystallization at 26.5°C
starts by the formation of seed crystals that are characterized
by a high melting point (up to 72.4°C). Compared to the orig-
inal CB, these crystals are enriched in glycolipids, phospho-
lipids, and the saturated TAG [1,2-palmitoyl-3-stearoyl glyc-
erol (PPSt), 1-palmitoyl-2,3-stearoyl glycerol (PStSt), and
tristearoylglycerol (StStSt)]. We also observed the formation
of a cloudy fraction in the liquid CB, initially isotropic, after
a few hours’ storage at 26.5°C. At the same time, it was nec-
essary to increase the sample temperature to above 45°C to
melt this fraction. Taking into account the similar conditions
of crystal formation in all samples, we decided to further
characterize this fraction by XRD, DSC, and chemical analy-
sis.
Chemical and XRD analysis of the fractions. Partial crys-
tallization of a CB sample was carried out by conditioning its
melt at 30.0 ± 0.5°C for 3 wk. As above, a cloudy and dense
fraction was separated from the liquid. Two fractions, HMF
and LMF, were isolated at 30.0 ± 1.0°C and analyzed as de-
scribed in the Materials and Methods section. Small-angle
XRD analysis of these two fractions, reported in Figure 1,
showed that the HMF long-spacings were characterized by a
stronger intensity of the line at 44.4 Å than that of the initial
CB. This diffraction line was not detected in the LMF. No sig-
nificant difference was observed in the short-spacings be-
tween HMF, LMF, and initial CB.
The TAG compositions of these two fractions are reported
in Table 3. The HMF showed a higher concentration of trisat- FIG. 3. Differential scanning calorimetry (DSC) recordings of crystal-
urated TAG (11.3%) and a lower content in polyunsaturated lization (A, cooling rate = 2°C/min) and melting (B, heating rate =
TAG (11%) compared with the initial CB (3.0 and 13.5%, re- 5°C/min) of CB. See Figure 1 for other abbreviations.

JAOCS, Vol. 75, no. 4 (1998)

430 C. LOISEL ET AL.

HMF (Fig. 3A), the first peak at 19.2°C increased to 34.7°C,

while it was absent from LMF (Fig. 3A). Furthermore, the
peak at 1.7°C increased to 4°C for LMF and was absent from
HMF. The single peak at 14.2°C for the pure CB gave a dou-
ble peak at 15 and 17°C for HMF and at 15.2 and 16.3°C for
LMF. Indeed, the large increase of the initial crystallization
temperature on DSC recordings confirmed that phase separa-
tion occurred between a minor fraction with a high melting
point and a major fraction composed of the other TAG. Simi-
larly, the weak exothermic events observed at about 4°C, the
enthalpy of which increased for LMF, are probably related to
another phase separation that occurs between polyunsaturated
and monounsaturated TAG.
The corresponding melting recordings (Fig. 3B) show that
the increase of trisaturated TAG content in HMF led to the
formation of an endotherm at about 50°C (Fig. 3B), while its
decrease in LMF (Fig. 3B) led to the disappearance of the
peak observed with pure CB at 24.9°C (Fig. 3B). The origin
of this endotherm will be discussed below (see the Intermedi-
ate forms section). The exotherm at 37°C for HMF (Fig. 3B)
shows that recrystallization occurred at this temperature, and
melting of the species formed at this point spread up to about
50°C. The incorporation of 2% StStSt in pure CB confirmed
that this polymorphic transition was related to trisaturated
TAG crystallization (see below). On this occasion, a different
analysis carried out with CB enriched with StStSt (data not
presented here) showed that a decrease of the cooling rate
from 5 to 0.5°C/min induced increases of the peak tempera-
tures shown in Figure 3A (HMF curve ) from 24.0 to 28.9°C
for the first exotherm and from 13 to 15.4°C for the second.
These temperature shifts were comparable to those observed
between HMF and CB curves (Fig. 3A). Whatever the cool-
ing rate, the last endotherm observed on heating at 5°C/min
(Fig. 3B, HMF curve for comparison) was found at a constant
temperature (49.8 ± 0.1°C). However, a decrease of the cool-
ing rate induced an increase of the exotherm temperature (ob-
served on Fig. 3B, HMF) from 35.8 to 38.2°C. This indicates
that trisaturated TAG probably crystallize into an unstable
form (likely to be α) before transformation to their more sta-
ble form (β).
The formation of Form VI of CB, its thermal and structural
properties, and its coexistence with variety V already have
been addressed previously and therefore will not be reexam-
ined in this study (16,34,40).
Unstable forms. XRD. Metastable forms were studied by
quenching melted CB from 80 to −10°C at about 100°C/s.
Small-angle XRD, recorded at −10°C, shows that the long-
spacings of CB decomposed into two sharp lines that corre-
spond to spacings of about 52.6 Å (q = 0.12 Å−1) and 26.3 Å
(q = 0.24 Å−1) and two broad scattering peaks that are cen-
tered, respectively, at about q = 0.17 Å−1 (intense, 36.4 Å) and
q = 0.06 Å−1 (weak, 112 Å) (Fig. 4A). Short-spacings ob- FIG. 4. Three-dimensional plot for the evolution of the long-spacings of
CB as determined by small-angle XRD during heating of a sample that
served together with long-spacings during a second experi-
was obtained by liquid quenching at 100°C/s from −10 to 40°C at
ment, performed at a shorter sample-detector distance but 1°C/min. Part A shows the long-spacings observed at −10°C, before, and
under the same conditions, are given in Figure 4B and show at 20°C, after transition to the α form. Part B shows the corresponding
the existence of two spacings at 4.19 and 3.77 Å above a broad short-spacings. ——, −10°C; ------, 20°C. See Figure 1 for abbreviations.

JAOCS, Vol. 75, no. 4 (1998)

 COCOA BUTTER POLYMORPHISM 431

scattering peak at about q = 1.4–1.5 Å−1. Figure 4 also shows

the evolution of the long-spacings recorded at 1°C/min be-
tween −10 and 40°C. We see that the sharp and broad peaks,
the initial repeat distances of which are 52.6 and 36.4 Å, tend
progressively to merge into a single large peak at 48.5 Å (the
third order of which was also observed at 16.1 Å during the
second experiment).
To determine the correspondence of the different lines,
variations of their intensities between −10 and 40°C were
plotted on the same graph (Fig. 5A and 5B). It appears that
the line at 52.6 Å may be related to the line at 3.77 Å because
they decrease simultaneously. The broad peak at 36.4 Å does
not seem to correspond to the same polymorphic form, be-
cause its evolution is different from the evolution of the for-
mer lines, but this evolution is probably related to the increase
of the line at 48.5 Å. The lower intensity of the line at 112 Å
prevents precisely plotting its evolution, which apparently
follows that of the three other lines. The disappearance of the
long-spacings at 52.6 Å was accompanied by the appearance
of a new line at 48.5 Å and by the evolution of the short-spac-
ings at 4.19 and 3.77 Å toward a single line at 4.22 Å. How-
ever, this evolution was preceded by a shift of the 3.77 Å
spacing toward about 3.85 Å. This observation may explain
the 3.87 Å value reported by Chapman et al. (15) for Form I
of CB (Table 1). Further heating above 20°C slowly shifts the
spacing at 48.5 Å to give a weak line at 44.9 Å, which in turn
completely disappears below 30°C (arrow in Fig. 5A).
Thermal analysis, recorded simultaneously with XRD
from the same sample during heating, achieved through FIG. 5. Simultaneous XRD as a function of temperature (XRDT) and DSC
XRDT–DSC coupling, is reported in Figure 5C. The two recordings. Evolution of the main (A) long- and (B) short-spacings,
recordings presented correspond to the two different sample- recorded in Figure 4 during heating at 1°C/min of two samples
detector distances. Exotherms at 0°C result from melting of quenched at −10°C; the corresponding DSC recordings (C) were ob-
some ice crystals formed on the capillary tube during quench- tained from the two different samples used for A and B recordings but
with a different sensitivity of the recorder; for explanation of dotted line
ing of CB. Figures 5A and 5B show that the decreases of the see text. See Figure 3 for other abbreviation.
lines corresponding to spacings at 52.6 and 3.77 Å are corre-
lated with the increase of the line at 48.5 Å (also seen for third
order, data not shown). However, no clear endotherm is asso- the peak that might have been recorded in absence of the
ciated with this transition, in contrast to what was observed exotherm (41). The respective enthalpies of the peaks and
beyond 15°C for further transitions. This would mean that their XRDT intensities indicate that the monotropic α(II)
this polymorphic transition is energetically low and/or ex- →β′(IV) transition was not complete, so that only part of the
tends over a large domain of temperature. Its irreversible more stable form (β′) was formed under the experimental
character can be deduced from the fact that cooling of the α conditions.
form, directly obtained by the transformation of this phase on Intermediate forms. Low heating rate (0.1°C/min). A sim-
heating, did not restore it, even after 1 d at low temperature ilar experiment was carried out by quenching the CB at
(data not shown). Identification of the transitions that corre- −30°C but heating it at 0.1°C/min between −10 and 38°C, in-
spond to each endotherm is reported in Figure 5C. The en- stead of 1°C/min as above. Contrary to the previous experi-
dotherm between 16 and 19°C was related to the melting of α ments and because of the duration of the run at 0.1°C/min
(form II, according to Tables 1 and 2) and thus associated (about 8 h), only one XRDT recording was performed with
with the decrease of the lines at 48.5 and 4.22 Å. The last en- the synchrotron beam at short sample-detector distance. The
dotherm, recorded above 25°C, corresponds to melting of the initial XRD pattern shows the same intense sharp lines at
variety that exhibits a spacing at 44.9 Å, presumably some 52.4, 26.2, 4.19, and 3.74 Å and two broad peaks at 111.2 and
form IV (according to Tables 1 and 2). In between, overlap- 36.5 Å. Figures 6A and 6B report intensity variations of the
ping of the exotherm that corresponds to the formation of this main spacings as a function of temperature. At 0.1°C/min, the
last form (also indicated by α → β′) and of the endotherm lines that corresponded to the spacing at 52.4 Å completely
maximum of α melting results in this complex recording in vanished in the range of 0–15°C, while the line at 48 Å
which the peaks are not separated. The dashed line indicates reached its maximal intensity at 15°C. The broad peak, the

JAOCS, Vol. 75, no. 4 (1998)

432 C. LOISEL ET AL.

yond 38°C at low heating rates confirms that high-melting

crystals may coexist within the liquid and be in equilibrium
with the latter.
According to Tables 1 and 2, the lines at 48 and 4.22 Å, 45
and 4.33 Å, 64.8 and 4.58 Å can be identified as the Forms II,
IV, and V of CB, respectively. Attribution of the other lines is
more difficult and will be discussed later.
The corresponding thermal analysis, either recorded si-
multaneously with XRDT as above or later with the DSC-7
are reported in Figure 6C. The two recordings were identical
and can be interpreted with the help of the XRDT recordings
as above. The large endotherm observed between 27 and
34.5°C corresponds to melting of Form V. The complex
exotherm preceding this peak is mainly attributable to crys-
tallization of this Form V starting at about 25°C. The ex-
pected enthalpy of this exotherm should have been, more or
less, the same as that of the melting endotherm. This indicates
that melting of Form IV and crystallization of Form V over-
lap, which is also indicated by XRDT (Figs. 6A and 6B). As
above, only the resulting thermal event is recorded (41). The
beginning and the end of these Z-shaped curves correspond
to melting of the less-stable species (endothermic) and to
crystallization of the more-stable one (exothermic). The in-
termediate part of the curve results from the sum of the two
thermal events. The same analysis can be tentatively applied
to the two preceding Z-shaped resulting curves. The first ob-
servable transition (between 15 and 18°C) is an endotherm
that corresponds to the melting of Form II (α), which pre-
cedes an exotherm (seen as a first decrease of the α en-
dotherm around 16–18°C) and presumably corresponds to
FIG. 6. Simultaneous XRDT and DSC recordings. Evolution of the long- partial recrystallization of the liquid into Form III. Crystal-
and short-spacings of CB as determined by (A) small- and (B) wide- lization of Form IV starts at 18°C before the completion of
angle XRD during heating at 0.1°C/min of two samples quenched from crystallization of Form III at 19°C, as shown by both Figures
80 to −10°C. The arrow indicates the inflection point in the intensity in-
6B and 6C. The two exotherm transitions, corresponding to
crease of the line at 44.8 Å. Relative intensities of the peaks are given
on the graph within parentheses (100% corresponds to the strongest the formation of Forms III and IV, are close and overlap with
peak) or noted as vs (very strong), s (strong), m (medium), w (weak), and melting of II and III, respectively. Some of the variations of
vw (very weak). (C) Corresponding DSC recordings (upper curve, simul- intensity reported in Figures 6A and 6B cannot be used reli-
taneous XRDT and DSC recordings; lower curve recorded 1 d later with ably for quantitative evaluation of the existing varieties be-
a Perkin-Elmer (DSC-7; St. Quentin en Yvelines, France). See Figures 3
cause there is also overlapping of peaks in XRDT (for in-
and 5 for other abbreviations.
stance, decreases of the lines characterizing the α form, at
maximum of which was observed at 36.5 Å, seemed to de- both 4.19 and 48 Å, observed above 15°C cannot be used to
crease later than the lines at 52.4 Å and 3.74 Å (the tempera- follow its melting because of their closeness to other lines).
ture-dependent evolutions of which are again associated), Nevertheless, the inflection point, observed in the increase of
while the intensity of the line at 4.19 Å stayed constant with a the intensity peak recorded at 44.8 Å (Fig. 6A), delimits two
small shift to 4.20 Å (not shown). Starting from 18°C, the domains of crystal growth. The first one can be associated
spacing at 48 Å was progressively transformed to yield that with either the presence of a line at 3.86 Å, and thus con-
at 44.8 Å, which reached its maximum at 25°C. During this nected to the formation of Form III, or to the formation of
phase transition, a small line at 4.33 Å appeared between 18 Form IV from the liquid phase. The domain of temperature in
and 28°C with an evolution similar to that of the spacing at which the spacing at 3.86 Å was recorded corresponds exactly
44.8 Å. At 23°C, a spacing at 64.8 Å began to increase to to the first intensity jump (the amplitude of which is less than
reach its maximum between 27 and 29°C, just after the spac- that of the second and explains the respective enthalpies/peak
ing at 44.8 Å vanished, and it disappeared at 33–34°C. This surfaces observed in Fig. 6C); it vanishes at exactly the in-
spacing can be correlated to the line at 4.58 Å, which was ob- flection point of the spacing at 44.8 Å (Figs. 6A and 6B, re-
served between 26 and 33.5°C. The increase of the line at spectively). The absence of an inflection point during the de-
43.8 Å above 28°C corresponds to crystallization of trisatu- crease of the line at 48 Å indicates that Form III develops at
rated TAG discussed above. The persistence of this line be- the expense of Form II but with the same long-spacing. The

JAOCS, Vol. 75, no. 4 (1998)

 COCOA BUTTER POLYMORPHISM 433

second part of the curve is associated with the presence of the

line at 4.33 Å and corresponds to variety IV of CB (Tables 1
and 2).
Also, a small endotherm, observed between 42.5 and 47°C
on the DSC-7 recordings, might be related to melting of the
trisaturated TAG (43.8 ± 1 Å). At higher heating rates, such
as 0.5 or 1°C/min, a broad peak, centered at about −3°C, was
clearly recorded between −10 and 5°C (not shown). This
hump, which was also observed around −10°C at 0.1°C/min
in the DSC-7, may be interpreted as an endotherm associated
with the decrease of lines at 52.4, 3.74, and 36.5 Å and their 20
)
evolution to give the line at 48 Å. e(
°C
tur
Intermediate cooling and heating rates (2°C/min). The pe
ra
m
formation of intermediate forms also has been studied at in- Te

termediate cooling rates, close to those used in chocolate

manufacturing, to get a better understanding of industrial
crystallization processes. A liquid CB sample was first cooled
from 40 to 0°C at 2°C/min and reheated to 40°C at the same
rate. Figures 7A and 7B show the three-dimensional plots ob-
tained by XRDT for crystallization and fusion, respectively.
As previously, the intensity of each short- and long-spacing
was plotted as a function of temperature (Figs. 7C and 8).
Crystallization begins by the appearance of a small line at
49.6 Å, observed from about 20–22°C, which shifts slowly to
about 48.5 Å as it develops. This line can be related to the
exotherm that appears at about 19°C in the DSC recording
(Fig. 3A), which was interpreted as the crystallization of
trisaturated TAG. The spacing of 49.6–48.5 Å is close to the
value given by Lavigne et al. (31) for the α form of StStSt
(50.6 Å) and confirms that trisaturated TAG first crystallize
in their α form before transforming into the β form.
Figures 7A and 7C also show that the major part of CB
crystallizes at about 15°C into a highly unstable form that is
characterized by the lines at 52.9 and 52.9/2 Å after partial
crystallization in the α form (48.5 Å). The growth of the lines
at 52.9 and 52.9/2 Å was accompanied by formation of a
bump between the two sharp lines. This unstable form corre-
sponds to the liquid-crystalline organization of the TAG dis-
cussed above (Fig. 4A). During cooling (Figs. 7A and 7C),
the unstable form progressively transformed into α (49.5 Å)
between 10 and 0°C. Evolution of the short-spacings recorded
as a function of temperature shows that the lines at about 53
and 48.5 Å were related to those at 3.84 and 4.18 Å, respec-
tively. During heating (Figs. 7B and 7C), these unstable forms
successively vanished at about 9 and 20°C, while a spacing at
45 Å, which corresponds to the β′ form (IV) of CB or to the β
form of trisaturated TAG, persisted until about 28°C. The
constant value observed for the line at 3.84 Å between 10 and
22°C is due to closeness of the neighboring peak at 4.18 Å.
Another experiment was carried out under the same condi-
tions with a CB sample that contained 2% of StStSt to ob- FIG. 7. Wide-angle XRD as a function of temperature. The three-dimen-
serve the influence of trisaturated TAG on its crystallization sional plots were obtained from the XRD patterns recorded during crys-
(Fig. 8). The trisaturated TAG started to crystallize in their α tallization and subsequent fusion of CB at cooling (A) and heating (B)
rates of 2.0°C/min. (C) Evolution of the long- and short-spacings during
form at a higher temperature than in the absence of StStSt be- crystallization (left) and subsequent fusion (right) of CB during experi-
tween 30 and 25°C, but crystallization of monounsaturated ments (A) and (B). (Intensities of the lines at 49.6 and 48.5 Å are plotted
TAG of CB in Forms I (sub-α, 53 Å) and II (α, 48.5 Å) was together). See Figure 1 for abbreviations.

JAOCS, Vol. 75, no. 4 (1998)

434 C. LOISEL ET AL.

FIG. 8. Evolution of the long- and short-spacings during crystallization FIG. 9. DSC recordings on cooling and heating at 2°C/min to show dif-
(left) and subsequent fusion (right) of CB with 2% tristearin (StStSt) as ferent crystallization and fusion observed with a CB sample that con-
determined by wide-angle XRD as a function of temperature during tained 2% StStSt. See Figures 1, 3, and 8 for abbreviations.
cooling and heating at 2.0°C/min. (Intensities of the lines at 49.6 and
48.5 Å are plotted together.) See Figure 1 for abbreviations.

not accelerated. Furthermore, during heating at 2°C/min, the TAG, which remain liquid at low temperature and only crys-
melting temperatures of these unstable forms were not shifted tallize at about 4°C on cooling, should also be considered be-
by more than 1°C. The main consequence of StStSt addition cause it might play an important role in the evolution of un-
was evolution of the line at 44 Å, which increased above stable CB polymorphs, as well as in chocolate blooming (40).
30°C, with a maximum at 33°C, and only vanished at 48°C. Although the main TAG segregation in the solid state has only
This intensity increase results from the polymorphic transi- been observed when CB crystallizes into the Forms II and V
tion of the trisaturated TAG from α to β form as observed pre- (the form in which chocolate is usually commercialized), it
viously (Fig. 3B, HMF). The slow kinetics of the transition, probably also occurs with some other forms, such as VI
as well as the temperature at which it was observed, suggests (19,21,22). This behavior may be explained by the low solu-
that it is favored when the monounsaturated TAG are com- bility of the trisaturated TAG within the monounsaturated
pletely melted, or in others words, when the liquid content is TAG, which represent more than 80% of CB (42). These re-
high enough to partly solubilize the trisaturated TAG. Figure sults agree with the “memory effect” of CB as reported by van
9 shows the DSC curves of a CB that was enriched with 2% Malssen (24). This author observed that, depending on its ori-
of StStSt as recorded during cooling followed by heating at gin, CB needed to be heated beyond a minimal temperature
the same rates as above (2°C/min) (Fig. 8). With the help of between 34 and 38°C to restore its original properties. Below
XRDT recordings, all thermal phenomena were identified (as this temperature, high-melting crystals may persist in liquid
reported in Fig. 9), although two of them are still tentative. CB and modify its crystallization behavior during storage at
The endotherm between 28.5 and 30°C was related to melt- 25°C. However, the above observations show that these high-
ing of the β′1 variety of CB (Form IV, Table 1), although the melting crystals are mainly composed of trisaturated TAG
line at about q = 0.14 Å−1 (45 Å) was not clearly observed (it and not StOSt (24). Thus, the lower and higher memory melt-
was probably hidden by the large peak at 48.5 Å and/or fused ing points reported by this author correspond to trisaturated
with the line at 44.2 Å). The small endotherm, observed be- TAG contents of 1.0 and 2.8%, respectively, which correlates
tween 21 and 22°C on heating, probably results from the perfectly with our results.
overlapping of two thermal events, corresponding to melting The separate crystallization of trisaturated TAG also
of the α form (48.5 Å) and recrystallization of liquid into the agrees with observations of Dimick and Manning (43) and
β′1 form (about 45 Å, form IV, Table 1). later of Davis and Dimick (25,26) who reported that CB crys-
tallization starts from that of HMF which are composed
mainly of complex lipids and these TAG. However, the phase
DISCUSSION
separations that they observed were obtained during a static
Thermal XRD and DSC analysis indicated that phase separa- crystallization at 26.5°C of several hours. Under these condi-
tions systematically occur during CB crystallization. The tions, the high-melting crystals, observed at 72.4°C, appar-
main TAG segregation involves a trisaturated fraction, which ently correspond to the crystallization of the β form of trisat-
partially phase-separates from the mono- and polyunsaturated urated TAG. Our results show that, during the tempering
species by crystallizing, as shown by chemical analysis of the process of chocolate when CB crystallizes within a few min-
fractions. However, the phase separation of polyunsaturated utes, the crystallization of the trisaturated TAG probably oc-

JAOCS, Vol. 75, no. 4 (1998)

 COCOA BUTTER POLYMORPHISM 435

curs via the α form. This crystallization was only revealed at liquid crystalline organization that combines both types of
the long-spacing level; no short-spacing (expected at about characteristics.
4.20 Å) was observed. This was also true above 35°C for The coexistence of two distinct structures would have
melting of the β form of trisaturated TAG, for which a strong shown two different patterns of thermal behavior, which were
short-spacing was expected at 4.58 Å. This might be due to a not observed. Moreover, no sharp transition was observed
lack of sensitivity at wide angles, while the concentration of when this variety transformed into α. On the contrary, accord-
crystals to detect is very small despite the high flux of the X- ing to XRDT and DSC recordings (Figs. 4 and 5), the transi-
ray source, rather than to the fact that long-spacings are orga- tion was gradual and, whatever the heating or cooling rate
nized first during crystallization (K. Sato, personal communi- used, it always occurred between −10 and 15°C. Although the
cation). sample-detector distance used (almost 1.5 m) allowed excel-
The observation by XRDT, and also by DSC, of the for- lent line separation, it was not enough to achieve full separa-
mation of the different forms during heating, as previously tion of the 48.5 and 52.5 Å lines and to determine the transi-
reported by Wille and Lutton (10), including Form III, the ex- tion temperature more precisely by XRDT. In fact, in contrast
istence of which has been discussed several times to the usual solid-liquid-solid transition, which is observed
(7,10,19,21,22), unambiguously confirms the pertinence of for monotropic substances at the melting point of an unstable
their description of CB polymorphism. form (31), the metastable form is here progressively replaced
Fast cooling (about 100°C/s) of melted CB results in the by α. Moreover, some intermediate recordings show the two
formation of a phase that is less organized than the α form, X-ray patterns superimposed (in the range of 5–10°C, Fig. 4),
but it transforms irreversibly into the latter on heating. At making line separation difficult without a specific mathemati-
−10°C, its diffraction/scattering pattern (Fig. 4) decomposes cal model. This gradual transition agrees with the observa-
into a set of two strong and sharp diffraction lines (∆1/2 = tions made by Riiner (7) that it is time- and temperature-de-
half-height widths of lines, expressed in Å−1) at q = 0.12 pendent.
Å−1 (52.6 Å, ∆1/2 = 0.004 Å−1) and 0.24 Å−1 (26.3 Å, ∆1/2 = This thermal behavior favors the hypothesis of the exis-
0.006 Å−1), and a series of broad peaks at q = 0.06 Å−1 (112 tence of a single form rather than that of a mixture of two.
Å, ∆1/2 = 0.040 Å−1) and 0.17 Å−1 (36.4 Å, ∆1/2 = 0.057 Å−1). However, how can the coexistence of the two organizations
The short-spacings, which indicate a packing more compact discussed above in the same phase be explained? The fact that
than the hexagonal arrangement of the α form, correspond to the whole organization is less stable than α and nevertheless
a β′ type (4.19 Å, q = 1.5 Å−1 and 3.77 Å, q = 1.67 Å−1, ∆1/2 = that part of it corresponds to a phase much more ordered than
0.05 Å−1). Such a pattern, showing the coexistence of sharp α indicates, as a counterpart, that only a part of the structure
and broad lines, cannot be explained by a single organization is ordered, while the rest is not. This is coherent with the ex-
of the chains. Therefore, assuming that the two sharp peaks istence in the X-ray pattern of several broad lines that corre-
correspond to the d001 and d002 diffraction lines of a lamellar spond to scattering from the unorganized component of the
structure (with a 52.6 Å spacing), the values of half-height structure. The disordered moiety exhibits broad peaks that re-
widths found for these peaks, both for long- and short-spac- semble those scattered by micellar structures with aliphatic
ings (∆1/2 = 0.005 ± 0.001 and 0.05 Å−1, respectively), are chains in the liquid state (44). The observation of a broad
less than that found for α (48.5 Å, ∆1/2 = 0.016 Å−1 and 4.22 bump at the bottom of the short-spacing lines, probably due
Å, ∆1/2 = 0.06 Å−1). This indicates that the packing of this to some smectic liquid-crystalline organization of part of the
moiety of the sub-α organization is more compact than that chains, confirmed this interpretation (Fig. 4A). The organiza-
of α. The broad peaks, the main one of which is centered at tion of the ordered moiety is probably lamellar, as indicated
about 36.5 Å, probably result from scattering by poorly ori- by the presence of sharp first- and second-orders for the long-
ented chains. This organization, which displays maxima at q spacings, with an O⊥ subcell. Thus, this coorganization might
= 0.06 Å−1 (112 Å) and q = 0.17 Å−1 (36.4 Å), is less orga- correspond to a partly crystallized, partly liquid crystalline
nized than the α form but more than the liquid, as shown by structure, the chains in the latter being organized by their co-
both lines’ positions, for instance at smaller angles than the valent links with the crystalline lattice (as in a brush). In such
hexagonal short-spacing of α, and larger half-height widths an organization, the mobility of the liquid chains is only al-
(∆1/2 = 0.09 for the broad peaks of the liquid phase of CB and lowed at methyl-end group extremity.
∆1/2 = 0.057 Å−1 for the less-organized moiety of the liquid These peaks also can be compared to those found for the
crystalline phase). TAG in the liquid state, which are located at mean q = 0.27
Some of the lines of this pattern have been observed pre- Å−1 (23.75 ± 0.10 Å) ∆1/2 = 0.09 Å−1 and q = 1.36 Å−1
viously. They were assigned to the sub-α (Table 1) and/or β′ (4.57 ± 0.01 Å−1) ∆1/2 = 0.31 Å−1 at 30°C for CB. These last
(sub-α) (7) form of CB by these investigators, to focus either peaks are broader and weaker in intensity than those found
on the relative metastability of this form compared to α for the less-ordered part of this phase, as expected according
and/or on its typical short-spacing pattern. In fact, such a to the above hypothesis, for liquid chains that are connected
complex pattern may be attributed either to the coexistence to a strongly organized (crystalline in this case) lattice. The
of two different structures, each of them displaying part of the main difference between the two liquids originates in the pla-
scattering/diffraction peaks discussed above, or to a single nar nature of the lattice limit compared to the probably fluc-

JAOCS, Vol. 75, no. 4 (1998)

436 C. LOISEL ET AL.

tuating surface made by the glycerol groups of the TAG in the compared with, for instance, its homolog 1-stearoyl-2,3-palmi-
liquid state (3,8). Such an organization also can be compared toylglycerol (StPP), which preferentially exhibits the β form, is
to that shown by phospholipids, such as phosphatidylcholine, a good illustration of the close relationship between the posi-
in the liquid crystalline state (Lα). Above the temperature of tioning on the glycerol of fatty acids of different chainlengths
their main transition, the phospholipid chains are liquid and and the orientation of their crystallization toward a particular
maintained partly organized by the electrostatic links of the subcell (4,46). Indeed, the fast formation of the β′ organization
polar head groups (45). However, the X-ray pattern shown by is in favor of some liquid surrounding the molecule extremi-
egg phosphatidylcholine in Lα in excess water, although sim- ties.
ilar at the short-spacing level, is much more ordered in the On the other hand, the presence of a liquid crystalline moi-
chain direction, as shown by its peak width. ety in the structure also would explain its progressive transi-
Although calculation made from model organization is tion into the α form because it is well known that the pres-
needed to solve the structure of this phase, a scheme of a hypo- ence of a liquid favors the evolution of unstable species to-
thetical arrangement of the TAG molecules is given in Figure ward more stable forms (47). On cooling at 2°C/min, the α
10. The possibility of an arrangement of the TAG within ag- form starts to crystallize before the liquid-crystalline phase
gregates has not been addressed because the sharp decrease of (Figs. 7A, 7C, 9). Therefore, the latter probably corresponds
the intensity observed at low q (Fig. 4A) indicates that they are to aggregation of the molecules not frozen in α form. The
rather large, if they exist. Also, this scheme does not explain sub-α → α transition, which is favored at slow cooling rates
the exact origin of the sharp lines observed at a spacing of by the presence of some α in mixture with sub-α, occurs more
about 52.6 Å, which are incompatible with the usual model of or less rapidly in the range of −10 to 15°C and thus raises the
TAG crystallization, this period being too long for a 2L organi- question of chain mobility and organization at lower temper-
zation and too short for 3L. It associates two levels of organi- atures in relation with the stability of this phase.
zation, a crystalline and a liquid-like moiety. Such a structure The second consequence of the existence of such a compact
may explain why, on rapid cooling starting from the liquid fat, and well-structured (sharp long-spacings) organization is that
only part of the chains has time to crystallize while the other presumably the liquid from which it forms could be already
part does not. Moreover, the degree of freedom in the molecu- preorganized. So, taking into account the respective tempera-
lar mobilities left by the liquid part of the organization allows tures of crystallization of chains, the saturated ones would crys-
the other moiety to crystallize rapidly into a compact subcell. tallize first, leaving the unsaturated chains in the liquid state.
For comparison, formation of the β′ form of StStSt takes a min- This preorganization assumes that: (i) the liquid state of CB, as
imum of about 10 min and has complex time–temperature con- other liquid fats, is liquid crystalline and rather of the smectic
ditioning requirements; the situation is similar for the TAG-like type (3,48) than nematic (9), (ii) its organization corresponds
1,3-palmitoyl-2-stearoylglycerol (PStP), which displays only to layers (or flat aggregates) made from saturated chains, while
the β′ form (and α). Moreover, the behavior of this last TAG, other parts contain the unsaturated chains. The liquidus shapes
observed in some binary phase diagrams of TAG, almost hori-
zontal on the side of the thermally more stable component with
a marked change of slope when approaching the concentrations
rich in the second component, already have been proposed as
reflecting some kind of phase separation in the liquid medium
(36). Such a liquidus shape is frequently observed for satu-
rated-unsaturated TAG (or saturated-partly unsaturated) binary
phase diagrams on the saturated side (for instance with StStSt-
triolein or StStSt-StOSt) (42).
Comparison with palm oil. The existence of a liquid crys-
talline phase, obtained by very fast cooling, was not found in
pure TAG, such as triolein, StStSt and even POP, which is nev-
ertheless one of the major constituent of CB, because they all
readily crystallize in the α phase (1,13). However, we found
that palm oil and lard, which are, like CB, mainly composed of
a mixture of monounsaturated and saturated TAG, also show
partial crystallization. The propensity of fats to crystallize in
sub-α(β′) carefully has been investigated by Riiner (7) by
means of temperature-programmed XRD camera recordings.
Among 30 oils and fats of food technological interest, he found
FIG. 10. Scheme of a possible arrangement of the triacylglycerol (TAG)
that five display this property, namely, beef tallow, CB, lard,
molecules in the sub-α form. A crystalline moiety with a planar organi-
zation involves the glycerol backbone and saturated fatty acid chains, palm oil, and sheanut butter. Except this last, all these fats have
while the remaining acyl groups (mono- and polyunsaturated) adopt a intermediate iodine values; fats and oils with lower and higher
liquid-like structure. iodine values do not display such behavior.

JAOCS, Vol. 75, no. 4 (1998)

 COCOA BUTTER POLYMORPHISM 437

The relationship between the formation of this liquid crys-

talline phase and the fat composition is still unclear. Obtaining
this phase seems related to: (i) the presence of a mixture of
TAG and (ii) a specific unsaturated/saturated fatty acid ratio.
Thermodynamic considerations. Further evidence for an or-
ganization of the liquid that was not mentioned before is pro-
vided by calculations from thermodynamic data. First, the plot
of melting enthalpies of TAG of the same family, such as trisat-
urateds with a single fatty acid (trilaurin to tribehenin), as a
function of carbon number revealed that a constant number of
methylene groups (about two per aliphatic chain) and the glyc-
eryl carbons do not participate in the melting enthalpy. (The
negative term found in the linear relationship between melting
enthalpy of the β form of trisaturated TAG cannot be attributed
to some disorganization of this solid phase because it has been
shown by XRD that all these TAG display the same basic struc-
ture, derived from that of LLL.) Then, taking into account the
importance of this nonmelting fraction, it must be related to
some liquid organization (41,49).
Second, evidence arises from the plots of the molar vol-
ume of the trisaturated TAG in the solid (β form) and liquid
states (at 80°C) as a function of their molecular weights.
These molar volumes were deduced from their liquid and
solid densities taken from the literature (45, from Table 10-
3). Both series of molar volumes vary linearly with the mole-
cular weights of the corresponding TAG.
The two equations deduced from the linear regression are
shown with their corresponding plots in Figure 12. The nega-
tive value (−53.4 cm3) observed for the plot of the liquid state
already shows that a rather high density is expected for the
polar headgroup packing, while the positive value found for
the β form tends to indicate the reverse for this solid phase
(neglecting, for this comparison, the dilation of β solid phase
between the temperatures at which molar volumes were mea-
FIG. 11. Small-angle X-ray patterns obtained for pure fats and their olein sured and 80°C). Taking into account the fact that, except for
and stearin fractions, CB (A) and palm oil (B). See Figure 1 for abbrevia-
tions.

The small-angle XRD patterns of palm oil and its fractions

(olein and stearin) were recorded, at −30°C after quenching as
above, for comparison with that of CB obtained by fractiona-
tion (see above) (Fig. 11A,B). Weak differences were observed
between HMF, LMF, and CB itself. HMF showed a broad peak
at about 36 Å (q = 0.18 Å−1), reduced in intensity, but an in-
crease of the spacings at 52.1 Å (q = 0.12 Å−1), associated with
important broadening compared to the other fats. Palm oil and
its fractions exhibited similar behavior, except that, for the
stearin fraction, the liquid-crystalline phase (55, 55/2 Å, and
broad peak at 36 Å) was mixed with the α form (48.5 Å). This
pattern also revealed a weak but sharp peak at 111 Å (Fig. 11B),
which seems to be related to the sharp peaks at spacing 55 Å
rather than to the α form. For all samples, whatever the fat and
the fractions considered, the intensities of the sharp and scat-
tering peaks are approximately in the same proportions. How- FIG. 12. Plots of the molar volumes of TAG in the liquid (solid line) and
ever, pure CB exhibits the highest intensity for the scattering solid (dashed line) states deduced from literature data. See Figure 10 for
peak, even compared to its fractions. abbreviations.

JAOCS, Vol. 75, no. 4 (1998)

438 C. LOISEL ET AL.

the liquid phase, the data are not taken at the same tempera- Danone for Ph.D. preparation (50). We thank G. Barratt for her crit-
ture (but thermal expansion is rather low in the solid phase), ical reading of the manuscript.
the intercept of the two straight lines is found at a high mole-
cular weight (about 400). That the volumes below this inter- REFERENCES
cept are occupied by polar groups in both states and are
1. Sato, K., T. Arishima, Z.H. Wang, K. Ojima, N. Sagi, and H.
equivalent (about 400 cm3/mole) indicates that the first car-
Mori, Polymorphism of POP and SOS. I. Occurrence and Poly-
bons of the chains are not disorganized by chain melting. morphic Transformation. J. Am. Oil Chem. Soc. 66:664–674
This confirms that the polar groups and parts of the chains of (1989).
trisaturated TAG in the liquid state are as dense as the equiv- 2. Chapman, D., The Structure of Lipids by Spectroscopic and X-
alent moiety of the crystalline form in its most compact pack- ray Techniques, Methuen, London, 1965, pp. 221–315.
3. Larsson, K., Molecular Arrangement in Glycerides, Fette Seifen
ing. These two independent lines of evidence for liquid orga-
Anstrichm. 74:136–142 (1972).
nization favor the hypothesis of a smectic organization of the 4. Lutton, E.S., Lipid Structures, J. Am. Oil Chem. Soc. 49:1–9.
liquid, as reported by Larsson (8), in particular because the (1972).
high density found for the polar head- groups is unlikely to 5. Garti, N., and K. Sato, Crystallization and Polymorphism of
correspond to their dilution in a paraffinic environment. Fats and Fatty Acids—Surfactant Science Series, Vol. 31, Mar-
cel Dekker Inc., New York, 1988.
In conclusion, fast cooling of fats, mainly composed of
6. Ollivon, M., Triglycérides, in Manuel des Corps Gras,
monounsaturated TAG, such as CB and palm oil, leads first Lavoisier, Paris, 1992, pp. 469–503.
to the formation of a liquid crystalline phase, which is proba- 7. Riiner, U., Investigation of the Polymorphism of Fats and Oils
bly partly organized as a crystal with a compact subcell of β′ by Temperature Programmed X-Ray Diffraction, Lebensm.
type and partly as a liquid. This organization, which likely Wiss. Technol. 3:101–106 (1970).
8. Larsson, K., On the Structure of the Liquid State of Triglyc-
also forms from other monounsaturated-rich TAG fats, is also
erides, J. Am. Oil Chem. Soc. 69:835–836 (1992).
obtained for CB at moderate cooling rates below 15°C. In 9. Cebula, D.J., D.J. McClements, M.J.W. Powey, and P. Smith,
spite of the presence of a more organized moiety, the exis- Neutron Diffraction Studies of Liquid and Crystalline Trilaurin,
tence of a liquid-crystalline part in the structure makes it on Ibid. 69:130–136 (1992).
the whole less stable than α, into which it progressively trans- 10. Wille, R.L., and E.S. Lutton, Polymorphism of CB, Ibid. 43:
491–496 (1966).
forms when maintained in the range of 0–15°C.
11. Kunutsor, S.K., and M. Ollivon, Ternary Phase Diagram of β Sta-
The partial fractionation of the StStSt TAG of CB observed ble Forms of Major Triglycerides of CB (POP, POS, SOS), 16th
in this study develops independent of the presence of this phase ISF Research World Congress, edited by J. Hollo, Budapest, 1983,
and others. The persistence of crystals, mainly composed of p. 11.
trisaturated TAG when the other moiety of CB is liquid, ex- 12. Sato, K., Physical and Molecular Properties of Lipid Poly-
morphs—A Review, Food Microstructure 6:151–159 (1987).
plains the specific thermal behavior previously observed
13. Arishima, T., N. Sagi, H. Mori, and K. Sato, Polymorphism of
(25,26). This phase separation in the solid state does not affect POS. I. Occurrence and Polymorphic Transformation, J. Am. Oil
the validity of the nomenclature proposed by Wille and Lutton Chem. Soc. 68:710–715 (1991).
(10), especially because the six forms that they described have 14. Duck, W., The Measurement of Unstable Fat in Finished Choco-
been reidentified. However, it should serve as a reminder that late, Manufacturing Confectioner 35 (6):67–72. (1964).
15. Chapman, G.M., E.E. Akehurst, and W.B. Wright, Cocoa But-
even for CB, which is a specific fat mainly composed of three
ter and Confectionery Fats. Studies Using Programmed Tem-
similar monounsaturated TAG, it only describes the behavior of perature X-Ray Diffraction and Differential Scanning Calorime-
its main fraction and not the whole fat. Therefore, for natural try, J. Am. Oil Chem. Soc. 48:824–830 (1971).
fats of generally more complex composition and polymorphism 16. Adenier, H., M. Ollivon, R. Perron, and H. Chaveron, Le
that display simultaneous complex arrangements made from blanchiment gras. I. Observations et commentaires, Chocola-
terie Confiserie de France 315:7–14 (1975).
different classes of TAG in addition to a liquid fraction, as re-
17. Huyghebaert, A., and H. Hendrickx, Polymorphism of Cocoa
cently shown for anhydrous milk fat and its fractions (35), a sin- Butter, Shown by Differential Scanning Calorimetry, Lebensm.
gle denomination should be avoided. Wiss. Technol. 4:59–63 (1971).
On the other hand, these results show that, even for fats 18. Lovegren, N.V., M.S. Gline, and R.O. Feuge, Polymorphic
that display a simple composition, the interpretation of ther- Changes in Mixtures of Confectionery Fats, J. Am. Oil Chem.
Soc. 53:83–88 (1976).
mal analysis is difficult, especially when numerous polymor-
19. Merken, G.V., and S.V. Vaeck, Etude du polymorphisme du
phic transitions occur successively. In this respect, the use of beurre de cacao par calorimétrie DSC, Lebensm. Wiss. Technol.
X-ray (or neutron) diffraction/scattering techniques, prefer- 13:314–317 (1980).
ably recorded as a function of temperature, by allowing con- 20. Davis, T.R., and P.S. Dimick, Solidification of Cocoa Butter,
tinuous identification of the crystal structures formed, is ab- Proc. PMCA Prod. Conf. 40:104–108 (1986).
21. Schlichter-Aronhime, J., and N. Garti, in Crystallization and
solutely necessary for understanding these phenomena.
Polymorphism of Fats and Fatty Acids, Surfactant Science Se-
ries, Vol. 31, edited by N. Garti and K. Sato, Marcel Dekker
ACKNOWLEDGMENTS Inc., New York, 1988, pp. 363–393.
22. Schlichter-Aronhime, J., S. Sarig, and N. Garti, Reconsidera-
C. Loisel was supported by a grant from Association Nationale de la tion of Polymorphic Transformations in Cocoa Butter Using
Recherche Technique (A.N.R.T., Convention Cifre) and Groupe DSC, J. Am. Oil Chem. Soc. 65:1140–1143 (1988).

JAOCS, Vol. 75, no. 4 (1998)

 COCOA BUTTER POLYMORPHISM 439

23. Becker, K., Les causes du blanchiment gras par les matières 37. Fraisse, B., Automatisation, traitement du signal et recueil de
grasses du chocolat et des bonbons au chocolat, Rev. Int. Choc. données en diffraction X et analyse thermique. Exploitation,
1:250–254 (1957). analyse et représentation de données., Ph.D., Université de
24. van Malssen, K.F., Real-Time X-Ray Powder Diffraction Ap- Montpellier II, 1995.
plied to Cocoa Butter and Graphite Intercalates, Ph.D., Amster- 38. Grabielle-Madelmont, C., and R. Perron, Calorimetric Studies
dam University, 1994. on Phospholipid–Water Systems. I. DL-Dipalmitoylphos-
25. Davis, T.R., and P.S. Dimick, Crystals Formed During Cocoa phatidylcholine (DPPC) Water System, J. Colloid Interface Sci.
Butter Solidification, J. Am. Oil Chem. Soc. 66:1488–1493 95:471–482 (1983).
(1989). 39. Loisel, C., G. Lecq, G. Keller, and M. Ollivon, Dynamic Crys-
26. Davis, T.R., and P.S. Dimick, Lipid Composition of High-Melt- tallization of Dark Chocolate: Influence of Temperature and
ing Seed Crystals Formed During Cocoa Butter Solidification, Lipid Additives, J. Food Sci., in press.
Ibid. 66:1494–1498 (1989). 40. Adenier, H., H. Chaveron, and M. Ollivon, Mechanism of Fat
27. Chaiseri, S., and P.S. Dimick, Dynamic Crystallization of Cocoa Bloom Development on Chocolate, in Shelf Life Studies of
Butter. I. Characterization of Simple Lipids in Rapid- and Slow- Foods and Beverages; Chemical, Biological, Physical and Nu-
Nucleation Cocoa Butters and Their Seed Crystals, Ibid. 72: tritional Aspects, edited by G. Charalambous, Elsevier Science
1491–1496 (1995). Publishers B.V., 1993, pp. 353–389.
28. Chaiseri, S., and P.S. Dimick, Dynamic Crystallization of Cocoa 41. Ollivon, M., and R. Perron, Enthalpies and Entropies Measure-
Butter. II. Morphological, Thermal and Chemical Characteris- ments of Unstable Forms of Saturated Even Monoacid Triglyc-
tics During Crystal Growth, Ibid. 72:1497–1504 (1995). erides, Thermochim. Acta 53:183–194 (1982).
29. Cebula, D.J., K.M. Dilley, and K. Smith, Continuous Temper- 42. Rossell, J.B., Phase Diagrams of Triglyceride Systems, Adv.
ing Studies on Model Confectionery Systems, PMCA Produc- Lipids Res. 5:353–408 (1967).
tion Conference 45: 42–47 (1991). 43. Dimick, P.S., and M.D. Manning, Thermal and Compositional
30. Kellens, M., W. Meeussen, C. Riekel, and H. Reynaers, Time- Properties of Cocoa Butter During Static Crystallization, J. Am.
Resolved X-ray Diffraction Studies of the Polymorphic Behav- Oil Chem. Soc. 64:1663–1669 (1987).
iour of Tripalmitin Using Synchrotron Radiation, Chem. Phys. 44. Hayter, J.B., Physics of Amphiphiles: Micelles, Vesicles and
Lipids. 52:79–98 (1990). Microemulsions, Proceedings of the International School of
31. Lavigne, F., C. Bourgaux, and M. Ollivon, Phase Transition of Physics, edited by Corti and De Giorgio, Elsevier, Amsterdam,
Saturated Triglycerides, J. Phys. IV, Colloque C8, sup. au J. 1985, p. 59.
Phys. I. 3:137–140 (1993). 45. Small, D., Glycerides, The Physical Chemistry of Lipids, in
32. van Gelder, R.N.M.N., N. Hogdson, K.J. Roberts, A. Rossi, M. Handbook of Lipid Research IV, Plenum Press, New York,
Wells, M. Polgreen, and I. Smith, Crystallisation and Polymor- 1986.
phism in Cocoa Butter Fat: In-situ Studies Using Synchrotron 46. De Jong, S., and T.C. Van Soest, Crystal Structures and Melting
Radiation X-ray Diffraction, in Crystal Growth of Organic Ma- Points of Saturated Triglycerides in the β-2 Phase, Acta Crys-
terials, ACS Conference Proceedings Series, edited by A.S. My- tallogr.:1570–1583 (1978).
erson, 1996, pp. 209–215. 47. Timms, R.E., Phase Behavior of Fats and Their Mixtures, Prog.
33. Sato, K., in Advances in Applied Lipid Research, Vol. 2, edited Lipid Res. 23:1–38 (1984).
by F. Padley, JAI Press Inc., 1996, pp. 213–268. 48. van den Tempel, M., Crystallization in Dispersed Systems,
34. Keller, G., F. Lavigne, C. Loisel, M. Ollivon, and C. Bourgaux, Physico-Chimie des composés amphiphiles, Colloques na-
Investigation of the Complex Thermal Behavior of Fats. Com- tionaux du C.N.R.S. n° 938, edited by R. Perron and P. Bothorel,
bined DSC and X-ray Diffraction Techniques, J. Thermal Anal. 1979, pp. 261–264.
47:1–21 (1996). 49. Ollivon, M., and R. Perron, Fat Science, 107–116 (1985).
35. Lavigne, F., Polymorphisme et transitions de phases des Trigly- 50. Loisel, C., Physico-chimie du chocolat: cristallisation du beurre
cérides. Applications aux propriétés thermiques et structurales de cacao et proprietés structurales, Ph.D. Université Paris VII,
de la matière grasse laitière anhydre et de ses fractions, Ph.D. Paris XI et E.N.S.I.A., 1996.
Université Paris VII, Paris XI et E.N.S.I.A, 1995.
36. Ollivon, M., and R. Perron, Etude de mélanges binaires de
triglycérides dérivés des acides palmitique et stéarique, Chem.
Phys. Lipids, 25:395–414 (1979). [Received October 22, 1996; accepted September 19, 1997]

JAOCS, Vol. 75, no. 4 (1998)