What the machines before the conche do to cocoa, by measurement: how the kind of roaster changes the bean, why a ball mill and a roll refiner do not make the same chocolate, and what alkalising costs in flavanols.
17 studies2007 to 2026Each checked against its published abstract
What these studies add up to
The kind of roaster matters, not only the temperature. A fluidised bed roasted twelve times faster than an oven and left nibs more porous; microwave roasting gave a chocolate tasters could tell from the convection-roasted one.
Roasting rearranges aroma more than it creates it: in one study the same compounds were present before and after, in different amounts.
On polyphenols the roasting studies disagree. One found the best-tasting roast had the least, one found a rise before the fall, and one found hot and short kinder than cool and long.
Ball mills and roll refiners give different spreads of particle size from the same recipe, and so different flow. No study compared how the two taste.
Alkalising removes flavanols in proportion to how far it is taken, about 60% of the flavonoids measured in one study, and does most of its work in the first minute.
Re-tuning a factory ball mill's speed and time raised its output by 43%
What refining time and agitator shaft speed minimise refining time and energy use in a ball mill refiner for chocolate?
How
Experiments on a working ball mill refiner following a central composite design with refining time and agitator shaft speed as factors, a second-order model, and a control run at the calculated optimum. The abstract does not give the number of runs.
Finding
Before the experiments the mill ran at 70 rpm for 55 min. The calculated optimum was 58 rpm and 38.5 min. The new conditions raised output from 109 kg/h to 156 kg/h, a 43% increase in productivity, and a control experiment confirmed the predicted responses. The most significant model coefficients were those for energy consumption, iron content and particle size.
Does not establish: That slower, shorter ball milling is better everywhere. Iron content of the chocolate was among the responses most affected by the settings, so speed and time are not free choices.
Limits: One mill and one chocolate recipe. The optimum is specific to that machine.
Roasting made no new aroma compounds in Criollo cocoa; it changed their amounts
Frauendorfer and Schieberle, 2008. Journal of Agricultural and Food Chemistry. doi:10.1021/jf802098f
Question
Which key aroma compounds change when fermented Criollo cocoa beans are roasted?
How
Comparative aroma extraction dilution analysis of unroasted and roasted Criollo cocoa beans, followed by quantitation of 30 odorants by stable isotope dilution assays and calculation of odour activity values. One bean type; the abstract gives no number of batches.
Finding
The analysis revealed 42 aroma compounds, the same in unroasted and roasted beans but differing in intensity. Concentrations exceeded the odour threshold for 22 compounds in unroasted and 27 in roasted beans. The Strecker aldehydes 3-methylbutanal and phenylacetaldehyde and 4-hydroxy-2,5-dimethyl-3(2H)-furanone increased strongly. 3-Methylbutanoic acid, ethyl 2-methylbutanoate and 2-phenylethanol were already present in unroasted fermented beans and did not increase.
Does not establish: That roast flavour comes entirely from the roaster. Several contributors to roasted cocoa aroma were already present after fermentation.
Limits: One cocoa type and, as far as the abstract states, one roasting condition. The roasting equipment and conditions are not given in the abstract.
The more a cocoa powder was alkalised, the less flavanol it held, in a straight line
Miller and colleagues, 2008. Journal of Agricultural and Food Chemistry. doi:10.1021/jf801670p
Question
How does the degree of alkalisation affect the antioxidant and flavanol content of commercial cocoa powders?
How
Survey of representative commercial natural cocoa powders and lightly, medium and heavily alkalised powders, grouped by extractable pH: natural 5.3-5.8, lightly treated 6.50-7.20, medium 7.21-7.60, heavily treated 7.61 and higher. The abstract does not give the number of powders.
Finding
Natural powders had the highest ORAC, total polyphenols and flavanols, and these showed a linear decrease as extractable pH increased. Average total flavanol contents were 34.6 +/- 6.8 mg/g for natural cocoas, 13.8 +/- 7.3 mg/g for lightly processed, 7.8 +/- 4.0 mg/g for medium processed and 3.9 +/- 1.8 mg/g for heavily processed powders.
Does not establish: That alkalised cocoa has no flavanols, or that the measured loss changes any health outcome. The study measured composition only.
Limits: A cross-sectional survey of commercial powders made from different beans, not the same cocoa before and after treatment. The abstract does not give the number of powders in each group.
Alkalising removed about 60% of the flavonoids measured in cocoa powder
Andres-Lacueva and colleagues, 2008. Journal of Agricultural and Food Chemistry. doi:10.1021/jf0728754
Question
What are the flavanol and flavonol contents of cocoa powder products, and how does alkalisation change them?
How
Analysis of major brands of cocoa powder products on the Spanish market for monomeric flavanols and flavonols, with a study of the effect of the alkalinisation treatment on the original flavonoid content. The abstract does not give the number of products.
Finding
The alkalinisation treatment resulted in 60% loss of the mean total flavonoid content. (-)-Epicatechin declined more (67%) than (+)-catechin (38%), which the authors attribute probably to epimerisation into (-)-catechin. Quercetin showed the highest loss (86%). A decline was also confirmed for di-, tri- and tetrameric procyanidins.
Does not establish: That the catechin left after alkalisation is as useful as what was lost. The authors note the remaining catechin may be in a less bioavailable form, but did not test absorption.
Limits: Monomers and a few flavonols were quantified; total procyanidins were not. The link to bioavailability is inferred, not measured.
Can the particle size distribution and flow properties of ball-mill milk chocolate be controlled to resemble those of roller-refined chocolate?
How
Milk chocolate samples produced at laboratory and pilot scale by ball milling, with separate grinding of components, blending of fine and coarse fractions, and blending of roller-refined with ball-milled samples. Particle size distribution and flow properties were measured. The number of samples is not given.
Finding
Pre-grinding milk and cocoa finely and sugar coarsely did not work well for practical reasons. Blending fine and coarse fractions of the same composition gave bimodal distributions like those of roller-refined products, and the lowest flow parameters came from a mix of 63 % coarse and 37 % fine particles. One-third ball-milled chocolate could be blended into roller-refined chocolate without considerably changing flow properties.
Does not establish: That a ball mill alone reproduces roll-refined chocolate. The match was reached only by an additional blending step.
Limits: Flow and particle size only; taste is not reported in the abstract. Milk chocolate only.
How do roasting temperature and time affect flavanols, proanthocyanidins and antioxidant activity of cocoa beans?
How
Laboratory study. Cocoa beans were roasted at three temperatures (125, 135 and 145 °C) for different times to reach moisture contents of about 2 g per 100 g. Flavanols, proanthocyanidins and antioxidant activity by three methods (TPI, FRAP, TRAP) were measured. No sample size is given.
Finding
The rates of flavanol and total proanthocyanidin loss increased with roasting temperature. With final moisture equal, high temperature-short time processes minimised proanthocyanidin loss. The middle temperature (135 °C) gave the highest TPI and FRAP values and the highest temperature (145 °C) the lowest TPI. Low temperature-long time roasting maximised chain-breaking activity by the TRAP method.
Does not establish: That one roast profile is best for antioxidants. The three antioxidant methods favoured different conditions.
Limits: Comparisons are made at a fixed moisture end point, so time and temperature were not varied independently. Three temperatures within a 20 degree span were tested.
How do ball size, milling time and lecithin concentration affect dark chocolate made with a ball mill coupled to a liquefier, compared with conventionally made chocolate?
How
Laboratory-scale study of dark chocolate produced by a ball mill coupled with a liquefier device, varying ball size, milling time and lecithin concentration, with cocoa mass or cocoa powder as cocoa source. Reference chocolates were made with a planetary mixer, roll-refiner and conche. No sample size is given.
Finding
Ball size, milling time and their interaction were the most decisive variables in milling. Moisture and free fat content were comparable regardless of processing method. Casson yield values of the alternatively processed chocolates were higher than those of conventionally processed ones, while Casson viscosities were relatively similar.
Does not establish: That ball-milled chocolate is equivalent to roll-refined and conched chocolate. Flow at low shear differed, and taste was not compared.
Limits: The abstract gives no figures. Flavour was not assessed here, and the equipment is a research set-up proposed for small producers.
How do alkalisation temperature, alkali concentration and time affect the colour, pH and phytochemical content of cocoa powder and the aroma of cocoa drinks?
How
Laboratory study of cocoa powder alkalised at 60, 70, 80, 90 or 100 °C with NaOH at 0.59, 1.17, 2.34 or 3.59% w/w of cocoa powder for 1 and 10 min, with HPLC-UV for non-volatiles and HS-SPME-GC-MS for aroma of the corresponding drinks.
Finding
Major changes occurred during the first minute of alkalisation. Higher temperature and alkali concentration generally reduced epicatechin and lightness (L*), and pH followed alkali concentration. Red and yellow colour components and theobromine were not significantly affected by temperature or alkali concentration. Higher temperature did not affect volatile concentrations, while more alkali reduced certain chemical classes.
Does not establish: That alkalisation removes theobromine. Theobromine levels were not significantly affected here.
Limits: Powder was alkalised in the laboratory with sodium hydroxide only, and aroma was measured on drinks by instrument, not by a panel.
Can microwave roasting of cocoa beans reach the same roasting degree, microstructure and aroma as convection roasting?
How
Laboratory study. Convection roasting was varied in time (10–50 min) and temperature (110–160 °C); microwave roasting in time (7–17 min) and power input (180–450 W). Roasting degree was judged by moisture, colour and the tetramethylpyrazine/trimethylpyrazine ratio, microstructure by cryo-SEM, and aroma by HS-SPME–GC–MS. No sample size is given.
Finding
With suitable time and power, microwave-roasted beans had a more pronounced brown colour and a lower tetramethylpyrazine/trimethylpyrazine ratio at a similar moisture content to convection-roasted beans. Differences in the number and size of macropores followed the temperature or power applied, not the technique. The aroma profile indicated a more intense cocoa aroma after microwave roasting.
Does not establish: That microwave roasting works at factory scale, or that the chocolate made from these beans tastes better. Neither was tested in this paper.
Limits: Aroma was measured instrumentally on beans, not tasted in chocolate. The abstract does not state the bean origin, batch size or number of replicates.
Fluidised bed roasting was twelve times faster and left nibs more porous
Peña-Correa and colleagues, 2023. Journal of the American Oil Chemists' Society. doi:10.1002/aocs.12696
Question
How do forced convective oven roasting and fluidised bed roasting differ in what they do to the microstructure of cocoa nibs and to the quality of the cocoa butter obtained from them?
How
Laboratory comparison of two roasting techniques for cocoa nibs, a forced convective oven and a fluidised bed, with X-ray tomography of pore structure and analysis of the cocoa butter. The abstract gives no sample size or batch size.
Finding
Fluidised bed roasting was 12 times faster than oven roasting. Total porosity of unroasted nibs was 8.5 ± 2.0% (vol/vol); it doubled on oven roasting and tripled on fluidised bed roasting. Fluidised bed nibs had the lowest densities and the highest cocoa butter yield, and their butter had more pyrazines and 3-methylbutanal and less hydroperoxide.
Does not establish: That industrial fluidised bed roasters give better chocolate than drum roasters. The comparison was with a laboratory oven, and no finished chocolate was tasted.
Limits: One research group, nibs rather than whole beans, and no scale, replication or bean origin stated in the abstract. The energy and carbon advantages are described as recognised, not measured here.
How is the flavour of dark chocolate made from microwave-roasted cocoa beans perceived by a trained panel and by consumers, compared with chocolate from convection-roasted beans?
How
70% dark chocolate from beans microwave roasted at 600 W for 35 min was compared with 70% dark chocolate from beans convectively roasted at 130 °C for 30 min. Physical properties were measured, a trained panel performed triangle tests with 27 judgements in total, and consumers assessed the chocolates (n = 112 for microwave, n = 100 for convection).
Finding
Colour, hardness, melting and flow behaviour did not differ significantly (p > 0.05). The triangle tests showed the two chocolates were distinguishable (d'-value = 1.62). Consumers cited cocoa aroma significantly more often for the microwave-roasted chocolate. Preference and willingness to buy were higher for it but not significantly at a 5% level. The energy saving from microwave roasting was estimated at 75%.
Does not establish: That consumers prefer microwave-roasted chocolate. The preference difference was not statistically significant.
Limits: One pair of roasting conditions and one chocolate recipe. The energy figure is an estimate, and the abstract does not say how it was derived.
How do roasting temperature and time affect alkylpyrazines and other properties of cocoa powder?
How
Laboratory study varying roasting temperature and time, with alkylpyrazines measured by gas chromatography-mass spectrometry and colour, polyphenols, chemical properties and sensory attributes also assessed. The abstract does not list the full set of conditions or the sample size.
Finding
The properties changed significantly with roasting time and temperature. Cocoa powder roasted at 140 °C for 40 min had the highest browning index, the highest tetramethylpyrazine to trimethylpyrazine ratio and the highest sensory score, and the lowest polyphenol content of the samples.
Does not establish: That 140 °C for 40 min is the best roast for cocoa in general. It was the best of the conditions tried on this material, and it was also the condition that lost the most polyphenols.
Limits: The abstract reports only the one best-scoring condition and gives no figures for the size of the changes. Sensory panel details are not stated.
How does polyphenol content change through the steps of chocolate production, and in particular during roasting at different maximum temperatures?
How
Samples were followed through all steps of chocolate production, with 3 roasting cycles having 80, 100 and 130 °C as maximum temperature. Total polyphenols, antioxidant power, epicatechin content and epicatechin mean degree of polymerisation were measured. The abstract gives no sample size or scale.
Finding
Total polyphenol content and antioxidant power followed a bell-shaped curve during roasting, an increase followed by a decrease, at all three temperatures. At 100 °C the higher polyphenol content was found just after roasting. Epicatechin content followed a similar trend, while mean degree of polymerisation moved in the opposite direction.
Does not establish: That roasting simply destroys polyphenols in proportion to heat. This study measured a rise before the fall.
Limits: The abstract reports trends without figures. The authors' explanation, that roasting frees epicatechin from oligomers, is offered as likely and was not shown directly.
Can FT-NIR spectroscopy with chemometrics give a routine measure of shell content in cocoa nibs and masses?
How
51 fermented cocoa samples of different varieties from 14 cocoa growing countries, crop years 2012-2017. Admixtures of nibs with 0-20% shell were analysed by FT-NIR spectroscopy and modelled by support vector machine regression. Industrial cocoa masses were also predicted and compared with a liquid chromatography-mass spectrometry method.
Finding
The models predicted shell content over 0-20% with an RMSE of 2.05% and an R2 of 0.88, and over 0-10% with an RMSE of 1.70% and an R2 of 0.72. Predictions for several industrial cocoa masses showed good consistency with the mass spectrometry method.
Does not establish: How well any winnower removes shell. The paper validates a way of measuring residual shell, not a machine's separation efficiency.
Limits: Calibration used laboratory admixtures. An error of this size is large relative to low shell contents.
How do alkali concentration, temperature and time of alkalisation affect the properties of low-fat cocoa powder?
How
Laboratory response-surface study (central composite design) with NaOH concentration of 3.0-6.0 g/100 mL, alkalisation temperature of 60-90 °C and time of 20-40 min, measuring colour, pH, powder characteristics, volatiles, total polyphenol content and antioxidant activity.
Finding
Significant models were identified for L*, a*/b*, pH, total polyphenol content and DPPH antioxidant activity (p < 0.05). Alkali concentration had a more significant effect than temperature or time, especially on the alkalisation indicators. Total polyphenols and antioxidant activity correlated strongly with colour. The optimum for maximising a*/b* was 5.3 %, 84 °C and 35.7 min.
Does not establish: That the reported optimum is best for the product overall. It optimises one colour ratio, and colour and polyphenol content moved together.
Limits: The optimum maximises redness of colour only. The abstract does not give the polyphenol loss at that optimum.
Roasting in superheated steam left fewer total phenolics and more of some procyanidins
Navare and colleagues, 2026. Journal of the Science of Food and Agriculture. doi:10.1002/jsfa.70417
Question
Does roasting cocoa beans in superheated steam instead of hot air change their in vitro phospholipase A2 inhibitory activity and polyphenol composition?
How
Laboratory study. Beans were roasted at 150 °C (10, 20, 30 min), 175 °C (8, 15 min) and 200 °C (7 min) using superheated steam or hot air. Extracts were tested for inhibition of phospholipase A2 in vitro and analysed for polyphenols. No batch size is given.
Finding
The IC50 against phospholipase A2 of extracts from steam-roasted beans at 175 °C was 50% lower than for comparable hot-air beans. Total phenolic content in hot-air beans was 20% to 33% higher than in steam-roasted beans across treatments, yet proanthocyanidin tetramer and pentamer in beans steam roasted at 175 °C for 8 min were 23% and 36% higher than in hot-air beans.
Does not establish: That chocolate from steam-roasted beans reduces inflammation in people. No human or animal was studied.
Limits: An enzyme assay in a test tube is described by the authors as a surrogate for anti-inflammatory activity. Flavour was not assessed in the abstract.
How do refining method, particle size and fat content together affect the flow and friction properties of molten chocolate compounds?
How
Laboratory measurement of rheological and tribological properties of molten dark and milk chocolate compounds with fat contents of 26-36% and particle sizes of 19-40 μm, made by roller refining or dry grinding. The number of samples is not given.
Finding
Yield stress and plastic viscosity decreased exponentially with increasing particle size. Roller-refined compounds, with bimodal and broader particle size distributions and more irregular shapes, had lower yield stress and viscosity than dry-ground samples. The friction coefficient also decayed exponentially with particle size, plateauing beyond 32 μm.
Does not establish: That people can feel the difference in the mouth. Friction was measured on an instrument, and no tasting is reported.
Limits: Compound coatings, not chocolate made with cocoa butter alone. No sensory panel is reported in the abstract.
We would like to set two Google Analytics cookies to count page views, and to count them in our own cookie-free measurement as well. They do not identify you and nothing is shared with advertisers. Decline and the site works exactly the same — see cookies for what each one does.