J Am Oil Chem Soc (2019)

DOI 10.1002/aocs.12247

ORIGINAL ARTICLE

Crystallization of Cocoa Butter in Cocoa Powder

Paige A. Palmieri1 · Richard W. Hartel1

Received: 30 January 2019 / Revised: 12 May 2019 / Accepted: 15 May 2019

© 2019 AOCS

Abstract Cocoa powder quality is determined by its attributes of the cocoa powder as well as the conditions of
color, flavor, dispersion, and flow properties, which can be the tempering process.
controlled via tempering. Design of a cocoa powder tem-
pering profile, however, requires that the mechanism of Keywords Cocoa powder  Cocoa butter  Tempering 
cocoa butter crystallization in cocoa powder be fully under- Alkalization  Crystallization
stood. Low-fat (8–12%) and high-fat (20–24%) cocoas were
sourced from two commercial manufacturers at varying J Am Oil Chem Soc (2019).
degrees of alkalization and compared with two commercial
cocoa butters. Unrefined paired cocoa powders and cocoa
butters sampled from the hydraulic press were also evalu- Introduction
ated. Isothermal crystallization kinetics and polymorphism
of cocoa powders and cocoa butters were compared at Cocoa powder is a common raw ingredient in numerous
18, 21, and 24
C using a direct time-domain nuclear mag- food applications including beverages, finished baked
netic resonance method, differential scanning calorimetry, goods, and ice cream. It is manufactured by first grinding
and x-ray diffraction. Crystallization was also studied under winnowed and roasted cacao nibs into chocolate liquor.
dynamic tumbling conditions. It was found that cocoa but- That liquor is then passed into a hydraulic press where high
ter in cocoa powder was nucleated by the cocoa powder temperature and pressure combine to result in the expulsion
matrix and transitioned to higher-stability polymorphs more of cocoa butter. The dense press cake left behind contains
rapidly than bulk cocoa butters. High-fat cocoas also some amount of residual cocoa butter and can subsequently
exhibited enhanced crystallization kinetics relative to low- be ground to yield cocoa powder (Hartel et al., 2018). At
fat cocoas, showing that differences in the cocoa micro- several points during the manufacturing process, different
structure may influence crystallization behavior. Notably, cocoa products, nibs, liquor, or cocoa powder, may be
alkalization did not significantly affect the crystallization
treated with an aqueous solution of alkali salt. This process,
behavior of most cocoa powders. Finally, it was found that
called alkalization, promotes Maillard browning reactions
tumbling conditions led to crystallization of βV and that
that can be controlled by various process parameters to tai-
caking, especially of high-fat cocoas, could be reduced by
lor the flavor or color profile of a cocoa powder for a
a static low-temperature hold step prior to tumbling. Over-
desired application (Dyer, 2003).
all, these results demonstrated that crystallization of cocoa
butter in cocoa powder is influenced both by the intrinsic Fat content and alkalization not only define the sensory
attributes of the powder, but also its physical properties,
namely, shear flow and dispersibility (Petit et al., 2017;
* Richard W. Hartel Shittu and Lawal, 2007). As with other food powders, the
[email protected] amount of fat on the surface, its composition, and physical
1
Department of Food Science, University of Wisconsin-Madison,
state play a critical role in determining overall quality (Kim
1605 Linden Drive, Madison, WI 53706, USA et al., 2009). In the case of cocoa powder, preservation of

J Am Oil Chem Soc (2019)

 J Am Oil Chem Soc

powder flow is ensured by a highly controlled tempering Guittard Chocolate Company (Burlingame, CA, USA) and
process, the primary aim of which is to crystallize the cocoa one from Clasen Quality Chocolates (Madison, WI, USA).
butter in cocoa powder into the βV polymorph. When tem- Three cocoa cakes and three cocoa butters, also donated
pering is not done correctly, the powder’s brilliant hue is by Olam Cocoa/De Zaan (Rotterdam, Netherlands), were
dulled and it agglomerates into firm cakes that are difficult sourced separately from the above. These samples are den-
to break apart (Minifie, 1999). oted as “paired” cocoa powders and cocoa butters because
While the aim of the tempering process is to control they were sampled immediately after hydraulic pressing
crystallization of cocoa butter in cocoa powder, cocoa but- and are derived from the same lot. The aim with these sam-
ter crystallization has not been studied extensively in this ples was to isolate cocoa powders and cocoa butters with
context. In bulk, cocoa butter is known to be sensitive to a the same, or similar, composition so as to evaluate the
number of factors including crystallization temperature, the effect of alkalization and cocoa powder particles on crystal-
application of shear as well as compositional variation lization of cocoa butter with a major source of variability
resulting from geographic origin or processing (Chaiseri controlled for.
and Dimick, 1995; Dhonsi and Stapley, 2006). It has also All cocoa powders were stored at room temperature. To
been shown, more specifically, that accumulation of soaps limit sample variation due to fractionation, cocoa butters
in undeodorized, alkalized cocoa butters is correlated with were melted completely in a 40
C oven, stirred, and then
decreased crystallization rates as well as longer induction aliquoted into plastic bottles to be stored in a –20
C
times during isothermal crystallization (Foubert et al., freezer. Cocoa press cakes were ground in 10 g batches
2004). Cocoa butter crystallization is further influenced by using a coffee grinder, and stored at room temperature until
the presence of particle impurities; although the mechanism sampling.
is not entirely known, the addition of cocoa particles to a
cocoa butter system promotes nucleation (Bowser, 2006; Experimental Plan
Svanberg et al., 2011; Vogel, 2012). Within cocoa powder,
therefore, cocoa butter crystallization may be susceptible to Cocoa powders and cocoa press cakes were first character-
the particular parameters of the tempering process, compo- ized by powder pH and overall fat content using standard
sitional differences, alkalization, and the presence of cocoa methods (as described below). Also, solid fat content
particles, as well as the total fat content of the powder (SFC) profiles for commercial cocoa powders (paired cakes
itself. excluded) were constructed using a direct time-domain
The primary objective of this work was to investigate the nuclear magnetic resonance (TD-NMR) method to find
effects of the cocoa powder matrix, alkalization, and total SFC at different temperatures. Melting curves were gener-
fat content as well as temperature on the isothermal crystal- ated by heating the commercial powders in a differential
lization behavior of cocoa butter in cocoa powder. Methods scanning calorimeter (DSC). Polymorphism was deter-
traditionally used to study bulk cocoa butter crystallization mined using x-ray diffraction (XRD).
were applied to study crystallization in an assortment of Isothermal crystallization behavior of all sourced cocoa
commercial cocoa powders. Ultimately, these results may powders, press cakes, paired cocoa butters, and commercial
assist in optimization of industrial tempering protocols, and cocoa butters was analyzed using NMR at temperatures
thereby ward against quality defects. within the normal range of tempering conditions (18, 21,
and 24
C). From the resultant crystallization curves,
kinetic measures including the crystallization rate (K),
induction time (tind), crystallization half time (t1/2), and
Experimental Procedures SFC maximum were derived to provide a quantifiable
means of comparison across the sample type (cocoa butter
Materials vs. cocoa powder), fat content, and degree of alkalization.
To explain kinetic differences and variations in overall
Eleven commercial cocoa powders (Table 1) were selected curve shape, crystallizing samples were interrupted at vari-
to generate a sample set representative of different degrees ous points along the crystallization curve to evaluate the
of alkalization, fat contents, and manufacturing processes. polymorphic composition of the samples using both DSC
Powders were donated by Olam Cocoa/De Zaan (Rotterdam, and XRD.
Netherlands) and by Barry Callebaut (Eddystone, PA, Given the results of isothermal crystallization experi-
USA). All commercial powders were derived from beans of ments, a final set of trials was devised to more closely
West African origin and were stored under ambient conditions. model the crystallization behavior of cocoa powder during
Two commercial cocoa butters (CB1 and CB2) of unknown industrial tempering, but on a lab scale. A tempering drum,
geographic origin were used as points of comparison, one from designed to incorporate the shear and dynamic forces

J Am Oil Chem Soc (2019)

J Am Oil Chem Soc

Table 1 pH, total fat content, and peak melting temperatures of commercial and paired cocoa powdersa

Sample pH Fat content Peak melting T (
C)

Commercial powders Low-fat cocoas

Nat1b 5.66  0.02a 10.8 32.7  0.2a
Nat2 5.56  0.02b 8.4 34.3  0.2bc
MAlk1 6.82  0.05c 11.0 35.0  0.2b
MAlk2 6.90  0.05c 9.5 33.7  0.2cd
HAlk1 8.16  0.07d 10.0 33.8  0.2abcd
HAlk2 7.42  0.09e 11.5 33.3  0.1ad
HAlk3 7.89  0.06df 9.7 31.7  1.1abcd
HAlk4 7.81  0.05fg 9.5 33.2  0.1ad
Breakfast cocoas
Nat 6.05  0.01h 20.9 34.5  0.01c
MAlk 6.80  0.05c 22.0 34.4  0.2c
HAlk 7.72  0.03dfg 22.1 35.0  0.2bc
Paired cakes
Nat 5.56  0.03ab 8.6 N/A
MAlk 7.00  0.04c 10.2 N/A
HAlk 7.57  0.03e 10.7 N/A

Different letters denote statistical differences down columns as calculated using the standard errors shown to the right of the mean parameter
value.
a
An α value of 0.05 was used as a threshold for significance.
b
Nat, natural; MAlk, medium alkalized; and HAlk, heavily alkalized.

experienced by a cocoa powder during industrial temper- Powder Fat Content

ing, was utilized to temper cocoa powders using two dis-
tinct methods. The first method, in which warm cocoa Total fat content of all cocoa powder samples was deter-
powder was added to the tempering drum and immediately mined using AOAC method 963.15 (Helrich, 1990). The
tumbled, was meant to mimic a continuous tempering oper- method utilizes an acid hydrolysis pretreatment in 8 M
ation. A batch system inspired the second method in which boiling HCl followed by Soxhlet extraction with petroleum
warm cocoa was equilibrated statically at a constant tem- ether. Analyses were performed by Covance Laboratories
perature lower than that of the tempering process to follow. (Madison, WI, USA) and total fat content values were
In both cases, the tempering process was monitored using reported to one decimal place.
SFC analysis, DSC, and XRD. Once equilibrium had been
reached, caking behavior of the final product was evaluated SFC Measurement of Commercial Powders
visually.
Analyses were performed on a Minispec MQ20 NMR Ana-
lyzer (Bruker Optics, Inc., The Woodlands, TX, USA) cali-
Measurements brated daily with a liquid oil 0% solids standard and two
paraffin-based standards of 31.3% and 72.5% (Bruker,
Powder pH Woodlands, TX, USA). SFC measurements were made
using the direct method. For cocoa butters, the SFC was
pH of cocoa powders was determined using International reported directly. For cocoa powders, however, the contri-
Confectionery Association method 15/1972 (ADM Cocoa, bution of cocoa solids to the total reported SFC required
2006). Exactly 10 g of cocoa powder was measured into that a simple equation traditionally used to evaluate SFC of
300–500 mL glass beakers and dispersed in 100 mL of chocolate systems be applied (Padar et al., 2008;
boiling deionized water by gentle stirring. The slurries were Ziegleder, 1988):
cooled to room temperature, stirred once more, and pH was
ðN T − N M Þ
measured using a pH meter (Fischer Scientific, Waltham, SFCa ð%Þ = × 100 ð1Þ
ð100−N M Þ
MA, USA) calibrated to pH 4.0, 7.0, and 10.0 with stan-
dard solutions (LaMotte, Charlestown, MD, USA) daily. Here, SFCa is the adjusted SFC, NT is the proton contri-
The procedure was repeated in triplicate. bution of a crystallizing sample, and NM is the proton

J Am Oil Chem Soc (2019)

 J Am Oil Chem Soc

contribution of a completely melted standard equilibrated SFC equilibrium. The same tubes were used for each mea-
to the crystallization temperature of interest. Melted stan- surement. Approximately 15 s were required to remove the
dards were prepared by packing cocoa powders into clear tube, dry it, take a SFC reading, and return it to the water
shell vials (VWR International, Radnor, PA, USA) to a bath. For all cocoa powders and press cakes, the SFC at the
final weight of 0.45–0.60 g. Shell vials were then loaded first time point for a given temperature was set as NM
into 180 mm flat bottom borosilicate NMR tubes with a (Eq. 1). Crystallization experiments were repeated in tripli-
10 mm diameter (Fischer Scientific, Pittsburgh, PA, USA). cate, with two samples in each replicate (n = 6). SFC was
Standards were next melted using a thermal treatment of 80 reported as a raw percent for cocoa butters and as SFCa for


C for 30 min then 40
C for 15 min. Equilibration of cocoa powders.
melted standards to a given temperature varied depending Because SFC curves were highly variable in shape and
on the experimental design. In all cases, however, sampling curvature, a single equation could not be identified that
times were deemed short enough to precede any major would adequately fit all samples crystallized at all the tem-
crystallization events, but long enough to allow complete peratures evaluated. Therefore, parameters were determined
thermal equilibration. directly from the curves so that all results could be com-
To develop SFCa melting profiles of commercially tem- pared on a common basis. The parameters of interest
pered cocoa powders, NMR tubes were prepared and ana- included induction time (tind), crystallization half time
lyzed at 10, 18, 21, 24, 30, 35, and 40
C. Experimental (t1/2), SFC maximum (A), and crystallization rate (K). To
samples were equilibrated in a temperature-controlled water determine induction time and crystallization rate, the linear
bath for 1 hour at each testing temperature and were evalu- portion of the rapid increase in SFC was found (maximum
ated from low to high temperature. Triplicate samples of number of points that maintained linearity at a level of
each powder were measured. Melted standards were equili- R2 = 0.95). The slope was defined as the crystallization rate
brated for 2 min from 18–35
C and for 45 s at 10
C and while induction time was defined as the x-axis intercept.
the same sample was used to standardize for each testing Induction times could not be defined for samples crystal-
temperature. To minimize error incurred from accidental lized at 18
C, because the onset of crystallization was
crystallization, standards were analyzed from high to low almost immediate. Because cocoa butters exhibited a two-
temperature and were held at 40
C for at least 5 min before step crystallization mechanism at 18 and 21
C, points for
testing at the next temperature. The procedure was repeated linear regression were selected from the second growth
in duplicate and the SFC values for the standards were aver- phase and the induction time was defined as the first pla-
aged. The SFCa for a given sample at a given temperature teau intercept. A SFC maximum was defined as the average
was then derived using Eq. 1 where NM was the average SFC of the last two points of the plateau. To determine the
SFC of the melted standards for that sample (Eq. 1). inflection point, or crystallization half time, the slope of the
line passing through the points (x1, y1) and (x2, y2) such that
Isothermal Crystallization Curves of Cocoa Butters and y1 < ½ A < y2 was calculated. The inflection point was
Cocoa Powders then calculated using the point-slope form of the line.
Because commercial cocoa butters and LFHAlk1 could not
To evaluate the crystallization behavior of cocoa butters reach their SFC equilibria after 12 hours of crystallization
and cocoa powders under isothermal conditions, a TD- at 24
C, maximum SFC and crystallization half times for
NMR method was employed to follow SFC as a function these samples could not be determined.
of time at three temperatures. NMR tubes of commercial
cocoa powders were prepared as described previously. Dynamic Crystallization of Cocoa Powders
Cocoa butters were prepared using a different method.
Cocoa butters were first held at 80
C until melted, then To better model the crystallization behavior of cocoa butter
mixed thoroughly by inversion and pipetted into 1 mL clear in cocoa powder under industrial tempering conditions, a
shell vials. Shell vials were then transferred to NMR tubes. system in which tumbling forces and temperature could be
NMR tubes were held in an incubator at 80
C for 30 min controlled was devised. Based on the results of previous
then at 40
C for 15 min to equilibrate the samples to the experiments, which highlighted differences between low-fat
temperature of cocoa powder exiting the pin mill in a com- and high-fat cocoas, only Nat1 and the high-fat natural
mercial process. Samples were next transferred to a water cocoa were compared in these experiments. A plastic-
bath set to one of the three crystallization temperatures barreled rock tumbler (NSI International, Inc. New York,
(T = 18, 21, or 24
C) selected based on the range of tem- NY, USA) was selected as a tempering vessel. Approxi-
peratures encountered during a commercial cocoa powder- mately 175 g of assorted metal ball bearings sized from 3 to
tempering process. Measurements were taken at reasonable 6 mm in diameter (B/C Precision, Greendale, IN, USA)
intervals until the samples had reached an approximate were added to the barrel to properly weight the tumbler and

J Am Oil Chem Soc (2019)

J Am Oil Chem Soc

provide shear force during tumbling. The rock tumbler and Sample blanks were allocated into the corresponding tray.
ball bearings were held in an incubator set to the desired Samples were heated at a rate of 20
C min−1 to a final tem-
tempering temperature for at least 30 min prior to the start of perature of 65
C and peak melting temperatures were deter-
the experiment and remained in the incubator for the entire mined using an intrinsic function of the Pyris software. The
duration of the experiment to control for temperature fluctua- average of three replicates was reported.
tions. Approximately 20 g of cocoa powder was thermally For SFC crystallization curves, samples for DSC analysis
treated at 80
C for 30 min and at 40
C for 15 min then were taken at three separate time points—prior to the
loaded into the rock tumbler barrel for tempering. Two induction time, at the midpoint of the major phase of crys-
methods, a one-step and a three-step tempering method, were tal mass increase, and at the final time point in the crystalli-
developed to evaluate which factors might affect crystalliza- zation experiment—which were determined from kinetic
tion as well as caking behavior of cocoa powder during the analysis. Samples were obtained directly from vials sub-
tempering process. jected to the same thermal treatment and isothermal condi-
In the one-step method, an isothermal temperature of tions as previously devised in isothermal crystallization
21
C was maintained and tumbling was initiated immedi- experiments. Cocoa powders were crystallized in NMR
ately after loading the cocoa powder charge into the rock vials. Crystallized cocoa butters could not be easily
tumbler barrel. Every 5 min, tumbling was halted and a sam- removed from NMR vials, however, and were instead crys-
ple was removed for immediate SFC analysis. This sampling tallized in 20 mL scintillation vials (VWR International,
continued for 60 min by which time all samples had reached Radnor, PA, USA). Larger sample sizes, of about 2.15 g,
a plateau. In the three-step method, the tumbling system was than were used in isothermal crystallization experiments
pre-equilibrated to a temperature of 18
C. The heat-treated were also prepared to obtain enough sample for both DSC
charge was then added, stirred gently, and held statically for and subsequent XRD analyses. A separate vial was pre-
a period of time prior to initiating tumbling action. SFC anal- pared for each time point and DSC melting profiles were
ysis was again performed every 5 min to determine when obtained in duplicate.
the statically crystallized samples had entered into the phase After the designated time at the crystallization tempera-
of rapid crystallization. At this time, the tumbler was turned ture had elapsed, samples were rapidly quenched on dry ice
on. Tumbling proceeded for 5 min at 18
C and then the for at least 5 min to halt crystallization and limit polymor-
temperature was increased to 21
C for the remainder of the phic transition during sample preparation. After quenching,
experiment. For both experiments, the value of NM for cocoa powders were weighed directly into standard alumi-
LFNat and HFNat as determined at 21
C in SFCa melting num pans to final sample weights of approximately
profile experiments was used to determine SFCa. At the end 3.0–16.0 mg. Cocoa butters were scraped out of the sample
of 60 min of tumbling time, samples crystallized using both vials and ground using a mortar and pestle prechilled on
methods were taken for DSC and XRD analysis. A picture dry ice for at least 30 min. Cocoa butters were then ret-
of the rock tumbler barrel was also captured to provide urned to dry ice before being measured into hermetic alu-
visual documentation of overall caking behavior. Tumbling minum pans (DSC Consumables, Austin, MN, USA) to
experiments were conducted in triplicate. final sample weights of 3.0–10.0 mg. The remainder of the
ground cocoa butter was reserved on dry ice for XRD.
Differential Scanning Calorimetry Sample pans of cocoa powders and cocoa butters were pre-
pared atop a small metal screw prechilled on dry ice to min-
DSC was used in all experiments to evaluate the peak melt- imize sample changes due to temperature fluctuations in the
ing temperatures of tempered cocoa powders or crystalliz- lab environment. After preparation, all samples were ana-
ing samples. A DSC 8500 (Perkin Elmer, Waltham, MA, lyzed using an initial oven temperature of −20
C and a
USA) outfitted with a –100
C Intercooler 2 (PolyScience, heating regimen of 20
C min−1 from −20 to 65
C.
Niles, IL, USA) was used for all analyses. Prior to use, the Samples obtained from dynamic crystallization experi-
DSC was calibrated using an indium standard run in dupli- ments were prepared at room temperature and loaded into
cate. It was verified that the onset of melting reported by standard aluminum pans. Due to the complexity of some
the Pyris software was 0.5
C from the reported melting melting profiles obtained during analysis of isothermally crys-
temperature of indium (Tm = 156.60
C). tallizing samples using an initial oven temperature of −20
C,
To assess the state of cocoa butter in untreated commer- an oven temperature of 20
C was used for these analyses.
cially tempered cocoa powders, standard aluminum pans
(Perkin Elmer, Norwalk, CT, USA) were prepared at room Powder X-Ray Diffraction
temperature and loaded to final sample weights of about
approximately 6.0–16.0 mg. Pans were then sealed and trans- Powder XRD was used in conjunction with DSC to vali-
ferred to the DSC oven preset to a temperature of −20
C. date the polymorphic composition of crystalline samples

J Am Oil Chem Soc (2019)

 J Am Oil Chem Soc

received from manufacturers or generated experimentally. cocoas than natural cocoas. Though the method of alka-
Isothermally crystallized samples for powder XRD were lization is unknown, it is possible that the variability of
collected during the final plateau phase as well as during measured pH for these samples is a function of the degree
the phase of rapid crystallization for two select cocoa but- of penetration of the alkali solution into the nibs. In other
ters (one commercial and one paired) to evaluate the pres- words, solids closer to the surface might have a higher pH
ence of unstable intermediates. Cocoa powders could not than solids in the nib interior. Importantly, no statistical
be analyzed by XRD at any point except during the final overlap was observed between natural, medium-alkalized,
plateau due to low overall sample crystallinity and high and heavily alkalized cocoas in either the commercial or
background resulting from amorphous materials. Residual paired categories.
sample from DSC analysis of isothermally crystallized Although powders are categorized into two levels of fat
samples was preserved on dry ice for XRD. Residual sam- content, specified from 10–12% and 22–24% fat, a number
ple from dynamic crystallization experiments taken after of cocoas evaluated fell outside of the specification range.
60 min of tumbling time and commercially tempered For the purposes of this study, it was deemed advantageous
cocoas were maintained at room temperature prior to XRD. to include these cocoas within the data set because they
Cocoa powders were analyzed by loading 0.5 g of sam- provided the opportunity to study crystallization behavior
ple into Teflon sample holders and pressing firmly to yield of cocoa powders over a wider range of fat contents than
a completely flat surface (Siemens Industrial Automation, previously anticipated.
Inc. Madison, WI, USA). Approximately 0.75 g of pow-
dered cocoa butter was used for analysis and packed simi- SFCa Melting Profiles
larly. Diffraction patterns were collected at room
temperature using a Bruker SMART APEX2 diffractometer SFCa melting profiles of commercially tempered cocoa
equipped with a Cu Kα tube (Bruker Optics, Inc. The powders were measured to study the effects of alkalization,
Woodlands, TX, USA). Data were collected from 18.5–25 fat content, and the cocoa powder matrix on melting behav-
two theta using an increment of 0.0167
for all experi- ior of cocoa butter in cocoa powder and on cocoa powder
ments. Patterns were acquired at an exposure time of storage stability. Melting profiles for select representative
2–2.5 s for cocoa powders and 1–2 s for cocoa butters. samples are shown in Fig. 1. Both low-fat natural cocoas as
well as the two heavily alkalized cocoas with the highest
and lowest SFC profile are depicted (Fig. 1a). In Fig. 1b,
Statistics
the high-fat cocoa with the lowest SFC profile is compared
to the low-fat cocoa with the highest SFC profile. Overall,
Statistical significance of data collected from each experi-
similar profile shapes were observed for all samples. At
ment was assessed using a Welch’s t-test, which assumes
temperatures below 24
C, measured SFCa was relatively
unequal variation and a normal distribution with the data
constant showing that cocoa butter in cocoa powder was
set. From the t-statistic, a P value was derived and evalu-
stable over the range of 10–24
C. Above 24
C, however,
ated at a threshold of α = 0.05 to determine significance.
a significant amount of melting occurred with the greatest
change in SFC observed between 30 and 35
C. Results of
DSC and XRD analysis suggested that melting in this range
Results and Discussion likely corresponded to the melting of the βV polymorph
(Table 1). Between 35 and 40
C, however, there was still
Powder Chemical Characterization a decrease in total SFC in some samples, potentially
corresponding to the melting of β crystals composed of
Commercially tempered and paired cocoa powders were higher melting TAG. Based on these results, it can be stated
first characterized using simple chemical methods, pH, and that cocoa powders held at a storage temperature above
total fat content analysis, in an attempt to provide a link ambient conditions are prone to melting. If storage temper-
between chemical attributes and physical behavior. All atures exceed this threshold for extended periods, the
cocoa powders fell within the accepted range of pH speci- melted cocoa butter may travel through the porous network
fied for the cocoa type—natural, medium alkalized, or of the cocoa particle and accumulate on the surface where
heavily alkalized (Table 1). The spread of pH for heavily it can recrystallize leading to caking and an overall loss of
alkalized commercial cocoas was much greater than the quality (Jacquot et al., 2016). Therefore, maintaining stor-
range observed in the other cocoa categories. This demon- age temperatures at or below ambient conditions is critical
strates that the diversity of heavily alkalized cocoas was to preserving cocoa quality.
well represented by the sample set. Additionally, the stan- With regard to the effect of alkalization, no significant
dard error in pH determination was greater for alkalized differences in maximum SFC value or overall profile shape

J Am Oil Chem Soc (2019)

J Am Oil Chem Soc

(a) (b)
90 90
LFNat1 LFCP
80 80
LFNat2 HFCP
70 LFHAlk2 70

LFHAlk3
60 60

SFCa (%)
50
SFCa (%)

50

40 40

30 30

20 20

10 10

0 0
10 20 30 40 10 20 30 40
Temperature (ᵒC) Temperature (ᵒC)

Fig. 1 SFCa melting profiles of (a) select natural (Nat) and heavily alkalized (HAlk) cocoa powders and of (b) representative low-fat (LF) and
high-fat (HF) cocoa powders (CP)

were consistently observed (Fig. 1a). Therefore, there was valuable insights into the relationship between particle
no significant effect of alkalization on melting behavior of microstructure and the physical behavior of cocoa butter in
cocoa powders. Though there was little variation in the cocoa powder.
overall melting profile, it did appear, however, that there
were differences in maximum SFCa at both 24 and 30
C SFC Crystallization Curves
between high-fat and low-fat cocoas (Fig. 1b). At these
temperatures, all high-fat cocoas had a significantly higher SFC crystallization curves of cocoa butters and cocoa pow-
maximum SFCa than any low-fat cocoa evaluated. Because ders demonstrated that the mechanism of crystallization of
there were no statistical differences in peak melting temper- cocoa butter in cocoa powder was fundamentally different
ature between low-fat or high-fat cocoas, this finding could than that in bulk cocoa butter (Fig. 2). Crystallization of
not be attributed to a difference in cocoa butter crystal sta- cocoa butters was complex; at 18 and 21
C, crystallization
bility (Table 1). began immediately and appeared to be divided into two
Though not compared in this study, previous publica- phases of crystal mass increase before plateauing at longer
tions of SFC melting profiles of cocoa butters have shown times. The separation of these two crystallization phases
that cocoa butters, irrespective of origin, have higher SFC was more apparent at 21
C due to the presence of an inter-
values at 10
C than all cocoa powders—85% compared to mediate plateau region. Though crystallization at 18
C did
75% (Hartel et al., 2018). In other words, cocoa butter crys- appear to slow down between 15 and 20 min, there was no
tallized in cocoa powder cannot achieve the same overall distinctive point at which crystallization halted completely
crystallinity as cocoa butter crystallized in bulk. It is possi- as observed at 21
C. Unlike at the other temperatures,
ble this difference between bulk cocoa butter and cocoa crystallization at 24
C occurred gradually over time and
butter in cocoa powder may be attributed to capillary forces via a one-step mechanism. The samples crystallized at this
acting on the entrapped cocoa butter. The difference in temperature also did not appear to fully reach equilibrium
SFCa maximum between high-fat and low-fat cocoa pow- even after 12 hours of sampling. Notably, similar SFC pro-
der might further suggest that the overall impact of these files for the studied temperature range were obtained by
capillary forces on crystal content of cocoa powers is other researchers (Dewettinck et al., 2004). All cocoa pow-
related to the physical aspects of the individual cocoa parti- ders, on the other hand, crystallized in one step at all tem-
cles. Although it is known that higher pressures and longer peratures; at no point prior to the major crystallization
press times are used to manufacture low-fat cocoa powders, event did crystallization slow down or plateau.
the effects of higher pressures on cocoa powder microstruc- Evaluation of crystallization kinetic data not only
ture have not been studied. Such a study might provide pointed to further differences between the crystallization of

J Am Oil Chem Soc (2019)

 J Am Oil Chem Soc

Low-fat High-fat Cocoa butter

90 90 90
80 80 80
70 70 70
60 60 60

SFCa (%)
SFCa (%)

SFC (%)
50 50 50
18 ° C
40 40 40
30 30 30
20 20 20
10 10 10
0 0 0
0 50 100 150 200 250 0 50 100 150 200 250 0 50 100 150 200 250
Time (min) Time (min) Time (min)
70 70 70

60 60 60

50 50 50
SFCa (%)

SFCa (%)

40

SFC (%)
40 40
21 ° C 30 30 30

20 20 20

10 10 10

0 0 0
0 60 120 180 240 300 360 0 60 120 180 240 300 360 0 60 120 180 240 300 360
Time (min) Time (min) Time (min)

50 50 50
45 45 45
40 40 40
35 35 35
30 30 30
SFCa (%)

SFC (%)
SFCa (%)

25 25 25
24 ° C 20 20
20
15 15 15
10 10 10

5 5 5

0 0 0
0 100 200 300 400 500 600 0 100 200 300 400 500 600 0 100 200 300 400 500 600
Time (min) Time (min) Time (min)

Fig. 2 Solid fat content (SFC) and adjusted SFC (SFCa) crystallization curves of commercial low-fat cocoa powders, high-fat cocoa powders,
and cocoa butters, respectively at 18, 21, and 24
C. Natural cocoas ( ), medium-alkalized cocoas ( ) and heavily alkalized samples ( : 1–2, ;
3–4) are denoted and samples of the same type are distinguished by closed and opened symbols

cocoa butter in bulk and in cocoa powder, but also to dif- the same fat content level were found to be statistically sim-
ferences in the physical behavior of cocoa butter in low-fat ilar, only the kinetics for a representative sample from each
and high-fat cocoa powder matrices (Fig. 3). As most crys- cocoa powder group is shown. Bulk cocoa butter crystalli-
tallization kinetic parameters associated with samples of zation rates (K) diminished rapidly with increased

J Am Oil Chem Soc (2019)

J Am Oil Chem Soc

(a) (b)
3 CB1 400 CB1
CB2 CB2
350
2.5 HF HF
LF 300 LF

K (SFC min-1) 2

t ind (min)
250

1.5 200

150
1
100
0.5
50

0 0
17 19 21 23 25 17 19 21 23 25
Temperature (°C) Temperature (°C)
(c) (d)
180
CB1 CB1
90
160 CB2
CB2
140 HF 80 HF
LF LF
120 70
t 1/2 (min)

100
A (%)

60
80
50
60
40
40

20 30

0 20
17 19 21 23 25 17 19 21 23 25
Temperature (°C) Temperature (°C)

Fig. 3 Effect of temperature on (a) crystallization rate (K), (b) induction time (tind), (c) crystallization half time (t1/2), and (d) maximum SFC
(A) for representative ( ) low-fat (LF) and ( ) high-fat (HF) cocoas and both ( ) commercial cocoa butters (CB)

temperature to the point of being equal to or less than that all temperatures, low-fat cocoas not only demonstrated
of cocoa powders. Therefore, it is possible that the interac- lower maximum SFC values than high-fat cocoas, but they
tion between the cocoa powder particle and cocoa butter is also crystallized at approximately half the rate and had lon-
capable of promoting crystallization of cocoa butter at ger induction times to crystallization (Fig. 3). Without fur-
higher temperatures. Cocoa butters also exhibited longer ther work characterizing the microstructural differences
induction times and crystallization half times than cocoa between these materials, however, a mechanism for this
powders at all temperatures, reiterating the nucleating abil- behavior cannot be put forth.
ity of the cocoa powder particle discussed in previous pub- Uniquely, high-fat cocoas exhibited an intermediate
lications. Interestingly, there was not a significant maximum crystallization rate at 21
C, which was signifi-
difference in maximum SFC of cocoa butters and high-fat cantly greater than the rates of crystallization calculated at
cocoas at 18 and 21
C, indicating that differences in over- 18 or 24
C (Fig. 3). Thermodynamically, this behavior is
all crystallinity, as seen in SFCa melting profile experi- contrary to the expectation that the crystallization rate
ments, must arise under different crystallization conditions. should decrease with increasing temperature due to a
The same could not be said for low-fat cocoas, however. At decreased driving force. This surprising trend indicates that

J Am Oil Chem Soc (2019)

 J Am Oil Chem Soc

more research must be conducted not only regarding the were apparent (Table 2). At 21
C, the natural cocoa but-
cocoa particle structure, but also the dynamics of this struc- ter exhibited a crystallization rate that was statistically
ture with regard to thermal fluctuations. faster than the heavily alkalized cocoa butter, but similar
Despite previous results suggesting that alkalization can to the medium-alkalized cocoa butter. Because the
inhibit cocoa butter crystallization, only a few significant dif- heavily alkalized cocoa butter was manufactured from a
ferences were observed in this work (Foubert et al., 2004). different bean blend, we cannot conclude that this differ-
Only one low-fat heavily alkalized powder (HAlk1) consis- ence is purely related to the effect of alkalization. At the
tently exhibited poor crystallization ability. At temperatures same temperature, however, the induction time of the
above 18
C, this cocoa powder demonstrated longer induc- natural cocoa butter was statistically shorter than both
tion times, slower crystallization rates, longer crystallization alkalized samples. Although significant differences were
half times, and lower maximum SFC values than any other observed in kinetics of cocoa butters at this temperature,
cocoa powder evaluated. At 24
C, this was also the only cocoa powders behaved statistically similarly in all
cocoa powder that did not reach SFC equilibrium within the kinetic measures except one; the natural cocoa powder
experimental time. Given the powder’s high degree of alka- had a statistically shorter induction time than both alka-
lization, it was the most heavily alkalized of the powders lized samples at 24
C. Overall, these results show that
studied, it is possible that this sample contained a significant the effect of alkalization on crystallization of cocoa
amount of soaps that interfered with crystallization despite butter is temperature dependent and that inhibition of
the nucleation capacity of the cocoa powder itself. crystallization of cocoa butter in cocoa powder is only
Among the paired cocoa press cakes and cocoa but- significant at higher temperatures than are relevant for
ters, only a few significant differences in kinetic rates cocoa butter in bulk.

Table 2 Crystallization kinetic parameters at 18, 21, and 24
C for paired cocoa butters and cocoa powders sampled from the hydraulic pressa

Sample K (SFC min−1) tind (min) t1/2 (min) A (SFC %)

18
C Cocoa powders
Natb 1.18  0.05a — 18  1a 83.4  0.4a
MAlk 1.25  0.06ab — 20  1ab 80.9  0.8ab
HAlk 1.38  0.05abc — 21  0ac 77.7  0.1b
Cocoa butters
Nat 1.51  0.05bc — 24  2ad 85.6  0.2c
MAlk 1.81  0.10c — 23  1bcd 84.1  0.6ac
HAlk 1.59  0.05c — 23  0bd 83.8  0.1ab
21
C Cocoa powders
Nat 1.18  0.13ab 32  2a 61  2a 65.4  2.2a
MAlk 1.16  0.09ab 32  2a 59  2ab 60.1  3.5ab
HAlk 1.04  0.04ab 28  1a 60  3abc 58.7  3.5ab
Cocoa butters
Nat 1.75  0.03c 39  0a 52  1b 71.6  0.2b
MAlk 1.46  0.08ac 50  2c 70  2c 70.5  0.1a
HAlk 1.40  0.03a 55  0c 67  4abc 69.7  0.2a
24
C Cocoa powders
Nat 0.18  0.02ab 35  3a 131  6a 34.9  2.7ac
MAlk 0.31  0.05ab 62  5b 123  3a 35.8  3.5abc
HAlk 0.30  0.04b 65  5b 125  2a 35.4  1.7c
Cocoa butters
Nat 0.10  0.02a 334  56c 528  47b 45.9  1.9b
MAlk 0.15  0.01ab 320  48c 512  47b 44.7  4.2abc
HAlk 0.11  0.00a 263  31c 466  42b 43.7  0.9ab

Different letters denote statistical differences down columns as calculated using the standard errors shown to the right of the mean parameter
value.
a
An α value of 0.05 was used as a threshold for significance.
b
Nat, natural; MAlk, medium alkalized; and HAlk, heavily alkalized.

J Am Oil Chem Soc (2019)

J Am Oil Chem Soc

Polymorphic Composition of Isothermally Crystallized broad melting peak had split into two distinct peaks at
Samples about 19 and 32
C. The results of XRD indicated that β0
was the most abundant species at both the 18 and 21
C
To understand the differences in the SFC curve shape crystallization half times (Fig. 5). Contrary to previous
between cocoa butters and cocoa powders, DSC and XRD findings, there was no reliable indication of the presence
were employed to monitor change in polymorphic compo- of the α polymorph (Dewettinck et al., 2004; Marangoni
sition over time. DSC melting profiles of a representative and McGauley, 2003; Van Malssen et al., 1999). It is pos-
cocoa butter, low-fat cocoa powder, and high-fat cocoa sible either that a polymorphic transition from α to β0
powder sampled during each phase of crystallization at 21 occurred despite best efforts to quench crystallization or


C are shown in Fig. 4. All samples evaluated at all tem- that the absence of temperature control during pattern
peratures followed the same general pattern in the DSC acquisition resulted in the melting of any α present in the
profile as the representative scans portray. Prior to the sample. By the end of crystallization, however, it was
induction period, cocoa butters exhibited a very broad apparent from XRD that β0 was present almost exclusively
melting peak, which appeared to be comprised of several regardless of the results of DSC analysis that pointed
species with different melting temperatures (Fig. 4a). As toward the presence of β.
time at the crystallization temperature increased, the peaks DSC scans of low-fat cocoa powders sampled during
shifted to higher temperatures reflecting an increase in the pre-induction period showed an initial broad peak
crystal stability. By the end of the crystallization time, the spanning the temperatures of 15–21
C (Fig. 4b). This

(a) (b)
75 38
70 37
65 36
Heat flow endo up (mW)

Heat flow endo up (mW)

60 35
34
55
33
50
32
45 31
40 30
35 29

30 28
–20 –10 0 10 20 30 40 50 60 –20 –10 0 10 20 30 40 50 60
Temperature (ºC) Temperature (ºC)

(c)
46
44
Heat flow endo up (mW)

42
40
38
36
34
32
30
28
–20 –10 0 10 20 30 40 50 60
Temperature (ºC)

Fig. 4 DSC thermograms of a representative cocoa butter (a), low-fat cocoa powder (b), and high-fat cocoa powder (c) showing the melting pro-
files of samples taken during the pre-induction period (—), at the crystallization half time (– –), and at equilibrium (• • •)

J Am Oil Chem Soc (2019)

 J Am Oil Chem Soc

peak decreased in magnitude over time, and in some in the pre-induction period, whereas small amounts of crys-
cases, narrowed as the sample transitioned to more stable tals melting in the range of 24–28
C were present for both
polymorphs. Often, a small secondary peak was observed high-fat and low-fat cocoa powders at this time. Importantly,
at temperatures of about 24–28
C, indicative of small it was determined by XRD that, despite the difference in the
amounts of higher stability crystals present prior to the final melting point at equilibrium between cocoa butters and
main phase of crystal mass increase. Notably, the area cocoa powders, all samples were crystallized in the β0 poly-
under the curve spanning the range of 24–28
C grew with morph (Fig. 6). Taken together, these results show that cocoa
increasing crystallization time. High-fat cocoas exhibited a butter in cocoa powder transitions to the β0 polymorph more
melting profile that shared features of both cocoa butters rapidly than bulk cocoa butter. This finding would also
and cocoa powders (Fig. 4c). In samples taken during the explain why a one-step crystallization mechanism was
pre-induction period, a single peak at about 19
C appeared observed for cocoa powders isothermally crystallized at all
with two shoulder regions. A larger shoulder that trailed temperatures. Since XRD could only be performed once the
into lower temperatures resembled that seen in cocoa butter crystallizing cocoa powders had reached equilibrium, how-
and a small shoulder extending into the range of 24–28
C ever, this mechanism could not be confirmed.
aligned with the distinct peak observed in this range in Data obtained in this study showed no contraindication of
melting profiles of low-fat cocoas. The crystal fraction in previous theories regarding the mechanism of cocoa butter
this melting range also grew with increasing time at the crystallization under isothermal conditions. Although the α
crystallization temperature. polymorph could not be reliably captured by DSC or XRD,
Perhaps, the most significant difference between the melt- SFC crystallization curves mirrored the previous data and
ing profiles of both high-fat and low-fat cocoa powders and clearly indicated that crystallization at 18 and 21
C
cocoa butters was how the area under the curve at the highest occurred in two distinct phases separated by a lag time that
observed melting point changed with sampling time. For was longer at 21
C than at 18
C (Dewettinck et al., 2004;
cocoa butters, no crystals that melted at 35
C were present Marangoni and McGauley, 2003; Van Malssen et al., 1999).

100,000
– 4.31

– 4.17

– 3.66
– 4.00
– 4.60

– 4.39
– 4.37

– 3.80
– 3.77
– 3.96

– 3.90

90,000

80,000

70,000 24 °C end

60,000
Intensity

50,000

40,000

30,000

21 °C end
20,000 21 °C mid
18 °C end
10,000

0 18 °C mid
18.5 19.5 20.5 21.5 22.5 23.5 24.5
2 Theta

Fig. 5 X-ray diffraction patterns of a representative cocoa butter sampled at various temperatures and times during isothermal crystallization.
Peak locations are indicated by d value above

J Am Oil Chem Soc (2019)

J Am Oil Chem Soc

(a) (b)
8000
18,000
7000
16,000

–4.17

–3.96
–3.90

–3.77

–3.66

–3.96

–3.66
–3.90
–3.81
–4.27
–4.39

–4.31
–4.39
–4.60

–4.60

–4.37

–4.17
6000
14,000

5000 12,000
Intensity

Intensity
4000
10,000
24 ᵒ C
8000
3000
6000 24 ᵒ C
21 ᵒ C
2000
4000
21 ᵒ C
1000 18 ᵒ C
2000
18 ᵒ C
0 0
18.5 19.5 20.5 21.5 22.5 23.5 24.5 18.5 19.5 20.5 21.5 22.5 23.5 24.5
2 Theta 2 Theta

Fig. 6 X-ray diffraction patterns of a representative low-fat (a) and high-fat (b) cocoa powder sampled at the end point of isothermal crystalliza-
tion at 18, 21, and 24
C. Peak locations are indicated by d-value above

It was found in those works that the phases most likely cor- and the data reported, it appeared that more stable poly-
responded to nucleation in α and then transformation from α morphs, here β0 , were crystallized more rapidly in cocoa
to β0 . At 24
C, crystallization proceeded in one step, likely powder. The difference in the melting profile shape of high-
via direct nucleation of β0 , and increased more rapidly once fat and low-fat cocoas also points toward a difference in
a critical concentration of nuclei was present in the system. their polymorphic behavior. Because the presence of α could
Cocoa powders, on the other hand, did not seem to follow not be tracked reliably, however, it is uncertain whether
this mechanism; at 18 and 21
C, a lag phase was not appar- cocoa powder, low-fat or high-fat, nucleated directly in β0 ,
ent. Given the results of the SFC crystallization experiments or transitioned morerapidly from α to β0 .

(a) (b)
70 70

60 60

50 50

40 40
SFCa (%)

SFCa (%)

30 30
One-step One-step
20 Three-step 20
Three-step
10 10

0 0
0 10 20 30 40 50 60 0 10 20 30 40 50 60
Time (min) Time (min)

Fig. 7 Adjusted solid fat content (SFCa) as a function of time for a low-fat natural (a) and high-fat natural (b) cocoa powder tempered using two
methods. The dotted line distinguishes the static precrystallization period employed in the three-step method from the first tumbling phase

J Am Oil Chem Soc (2019)

 J Am Oil Chem Soc

50,000 tumbling environment followed similar trends to that

– 4.60
45,000
observed in isothermal crystallization experiments (Fig. 7).
40,000
Using the three-step method, which employed a static pre-
35,000
crystallization step, the high-fat cocoa had a shorter induc-
Intensity

30,000
25,000 tion time than the low-fat cocoa. This was expected given
20,000 that this step closely resembled the static, isothermal condi-

– 3.77

– 3.66
– 3.90
– 4.00
15,000 tions of the crystallization kinetics experiments. Also in
10,000
accordance with previous results, the high-fat cocoa crys-
5000
0
tallized more rapidly, regardless of the tumbling method
18.5 19.5 20.5 21.5 22.5 23.5 24.5 employed, and reached a higher maximum SFC. In contrast
2 Theta
with the results of isothermal experiments, however, all
samples were crystallized in the βV polymorph by the end
Fig. 8 Cocoa powder crystallized in the βV polymorph after 60 min of the experiment irrespective of the method used (Fig. 8).
of tumbling. Peak locations are indicated by d-value above
This demonstrates that the application of the tumbling force
promotes crystallization and polymorphic transformation of
Dynamic Crystallization of Cocoa Powders cocoa butter in cocoa powder and that it is essential for
proper tempering of cocoa powder on normal industrial
Given that differences in crystallization kinetics were most timescales.
significantly related to differences between high-fat and In comparing the one-step and three-step methods, there
low-fat cocoa powders, only one natural cocoa from each was no significant difference in the final maximum SFC,
group was evaluated in these experiments. SFC crystalliza- the approximate time to the final plateau, or the final poly-
tion profiles of high-fat and low-fat cocoa powders in a morph for a given cocoa. Caking behavior, on the other

Fig. 9 Caking behavior of a low-fat natural cocoa powder tempered using a one-step method (a) and a three-step method (b), and of a high-fat
natural cocoa powder tempered using a one-step method (c) and a three-step method (d)

J Am Oil Chem Soc (2019)

J Am Oil Chem Soc

hand, varied significantly especially for the high-fat cocoa heavy alkalization. More work is required to be conclusive,
(Fig. 9). Caking, as visually approximated by the amount however. Furthermore, it was demonstrated that tempering
of cocoa that remained adhered to the metal ball bearings processes must include a dynamic component and that a
after tumbling, was minimal for the low-fat cocoa powder static precrystallization step may be beneficial to ensuring
and did not appear to differ when using either method. In high-fat cocoa powder quality post-tempering. Ultimately,
contrast, the high-fat cocoa, when tempered using the one- a complete understanding of the variables affecting cocoa
step method, caked substantially, but when tempered with powder tempering might allow for optimization of temper-
the three-step method was primarily free flowing. Because ing process parameters to account for the intrinsic crystalli-
caking in fat-containing powders is correlated with the zation properties of the cocoa powder. Control of
amount of liquid fat on the particle surface, it is likely that crystallization might then afford better control of cocoa
the precrystallization step was responsible for crystallizing quality.
fat on the exterior of the particle (Kim et al., 2009). By the
same rationale, the low-fat cocoa likely exhibited a similar Acknowledgments Samples were donated by Olam/De Zaan
caking behavior using both tempering methods because the Cocoa, Barry Callebaut, Clasen Quality Chocolates, and Guittard
Chocolate Company.
total fat content of the powder and, therefore, the total
amount of surface fat were below some threshold necessary
Conflict of Interest The authors declare that they have no conflict
for caking to occur.
of interest.
Overall, these results indicate that use of a low tempera-
ture, static precrystallization step may be an effective way
to reduce caking and improve overall quality in industrially
tempered high-fat cocoas. Not only is the method effective, References
but it is also relatively efficient. Based on the results from
ADM Cocoa (2006) The De Zaan Cocoa manual. Rotterdam,
SFC analysis, a dramatic reduction in caking of high-fat Netherlands: Author.
cocoa powders can be achieved by precrystallizing only Bowser, A. (2006) Crystallization of Cocoa butter. The Manufactur-
one-sixth of the total crystallizable fat in the sample. Addi- ing Confectioner, 86:115–118.
tionally, though the induction time to crystallization is lon- Chaiseri, S., & Dimick, P. S. (1995) Dynamic crystallization of cocoa
ger when employing a precrystallization step, samples butter. I. Characterization of simple lipids in rapid- and slow-
nucleating cocoa butters and their seed crystals. Journal of the
crystallized using either method reach their maximum SFC American Oil Chemists’ Society, 72:1491–1496.
at approximately the same time. This indicates that the Dewettinck, K., Foubert, I., Basiura, M., & Goderis, B. (2004) Phase
more complex three-step method may not require longer behavior of cocoa butter in a two-step isothermal crystallization.
processing times. Though the tumbling method devised Crystal Growth & Design, 4:1295–1302.
Dhonsi, D., & Stapley, A. G. F. (2006) The effect of shear rate, tem-
sought to model industrial tempering in a lab-scale setting, perature, sugar and emulsifier on the tempering of cocoa butter.
pilot trials on large-scale manufacturing equipment must be Journal of Food Engineering, 77:936–942.
conducted to understand the applicability of these results Dyer, B. (2003) Alkalized cocoa powders. The Manufacturing Con-
on the commercial scale. fectioner, 83:47–54.
Foubert, I., Vanrolleghem, P. A., Thas, O., & Dewettinck, K. (2004)
Influence of chemical composition on the isothermal cocoa butter
crystallization. Journal of Food Science, 9:E478–E487.
Hartel, R. W., von Elbe, J. H., & Hofberger, R. (2018) Confectionery
Conclusions science and technology. Madison, Switzerland: Springer.
Helrich, K. (1990) AOAC official methods of analysis (Vol. 1, 15th
ed.). Arlington, VA: AOCS Press.
It was the aim of this work to fully characterize the factors
Jacquot, C., Petit, J., Michaux, F., Chávez Montes, E., Dupas, J.,
affecting crystallization behavior of cocoa butter in cocoa Girard, V., … Gaiani, C. (2016) Cocoa powder surface composition
powder. These factors included intrinsic properties (the during aging: A focus on fat. Powder Technology, 292:195–202.
cocoa powder matrix itself, powder fat content, and degree Kim, E. H.-J., Chen, X. D., & Pearce, D. (2009) Surface composition
of alkalization) as well as processing parameters (tempera- of industrial spray-dried milk powders. 3. Changes in the surface
composition during long-term storage. Journal of Food Engineer-
ture and dynamic forces). Crystallization behavior was a ing, 94:182–191.
function of each of the variables evaluated. Interactions Marangoni, A. G., & McGauley, S. E. (2003) Relationship between
between the cocoa powder particle and cocoa butter in the crystallization behavior and structure in cocoa butter. Crystal
cocoa powder matrix resulted in more rapid nucleation and Growth & Design, 3:95–108.
the appearance of more stable polymorphs at earlier crystal- Minifie, B. (1999) Chocolate, cocoa and confectionery technology
(3rd ed.). Gaithersburg, MD: Aspen.
lization times than in bulk cocoa butters. It was also found
Padar, S., Jeelani, S. A. K., & Windhab, E. J. (2008) Crystallization
that cocoa particle structure may be the most important fac- kinetics of cocoa fat systems: Experiments and modeling. Journal
tor governing crystallization, except in cases of extremely of the American Oil Chemists’ Society, 85:1115–1126.

J Am Oil Chem Soc (2019)

 J Am Oil Chem Soc

Petit, J., Michaux, F., Jacquot, C., Chávez Montes, E., Dupas, J., Van Malssen, K., Langevelde, A., Peschar, R., & Schenk, H. (1999)
Girard, V., … Gaiani, C. (2017) Storage-induced caking of cocoa Phase behavior and extended phase scheme of static cocoa butter
powder. Journal of Food Engineering, 199:42–53. investigated with real-time X-ray powder diffraction. Journal of the
Shittu, T. A., & Lawal, M. O. (2007) Factors affecting instant American Oil Chemists’ Society, 76:669–676.
properties of powdered cocoa beverages. Food Chemistry, 100: Vogel, S. (2012) The effects of origin and additives on the crystalliza-
91–98. tion onset temperature of cocoa butter. Madison, WI: University of
Svanberg, L., Ahrne, L., Loren, N., & Windhab, E. (2011) Effect of Wisconsin.
sugar, cocoa particles and lecithin on cocoa butter crystallization in Ziegleder, G. (1988) Kristallisation von Kakaobutter unter statischen
seeded and non-seeded chocolate model systems. Journal of Food und dynamischen Bedingungen (DSC, Rheomeeter). Suesswaren,
Engineering, 104:70–80. 12:487–493.

J Am Oil Chem Soc (2019)