Chromatographic Profiling of Artisanal Chocolates: Volatile Composition and Methylxanthines as Markers of Cocoa Content

TL;DR

This study uses chemical analysis to authenticate artisanal chocolates and verify their cocoa content using volatile compounds and methylxanthines.

Contribution

The study introduces a reliable workflow combining chromatography and chemometrics for chocolate authentication and fraud prevention.

Findings

Methylxanthines like theobromine and caffeine strongly correlate with declared cocoa content in chocolates.

72 volatile compounds were identified, contributing to flavor descriptors like roasted, nutty, and fruity.

Principal component analysis effectively separated low- and high-cocoa chocolates using theobromine as a key marker.

Abstract

The increasing demand for artisanal chocolates highlights the importance of reliable analytical strategies for quality assurance and product authentication. In this study, 45 Brazilian artisanal chocolates (36%–100% cocoa) were characterized by integrated chromatographic and chemometric analyses. Seventeen representative samples were evaluated for volatile compounds using HS‐SPME/GC‐MS, and methylxanthines were quantified by HPLC‐DAD after optimization of an ultrasound‐assisted liquid–liquid extraction method. The procedure achieved recoveries of 92% ± 10% for theobromine and 95% ± 3% for caffeine, meeting international validation criteria. In total, 72 volatile compounds were identified, mainly acids, esters, pyrazines, and aldehydes, associated with descriptors such as roasted, nutty, floral, and fruity. Theobromine (1.158–21.033 g kg−1) and caffeine (0.058–1.997 g kg−1) concentrations showed strong positive correlations with declared cocoa content. Principal component analysis revealed clear separation between low‐ and high‐cocoa chocolates, with theobromine as the main discriminating variable. The results demonstrate that methylxanthines are robust chemical markers for chocolate authentication and classification, and that the combined use of GC‐MS, HPLC, and multivariate analysis provides a reliable workflow for quality control, traceability, and prevention of commercial fraud in the chocolate industry.

The combined use of chromatographic and chemometric approaches demonstrated in this study can be applied in food laboratories to improve the authentication, classification, and quality monitoring of chocolates. The optimized analytical methods for volatile and methylxanthine profiling are suitable for both artisanal and industrial products, supporting standardization, traceability, and prevention of commercial fraud in the chocolate industry.

Figures

Number of occurrences of each class of VOCs in the analyzed samples.

Total ion chromatograms (TIC) of artisanal chocolates with (A) 44% cocoa, (B) 60% cocoa, and (C) 70% cocoa.

Representative HPLC‐DAD chromatogram of a fortified artisanal chocolate sample (70% cocoa), showing separation of TEO and CAF under optimized conditions.

Representative HPLC‐DAD chromatograms of artisanal chocolates with (A) 44% cocoa, (B) 60% cocoa, and (C) 70% cocoa.

Linear regressions between cocoa content (%) and methylxanthine concentrations in chocolates: (A) TEO and (B) CAF.

Score (A) and loading (B) plots for PCA applied to the concentrations of TEO and CAF in chocolate samples.

Tables

TABLE 1. Volatile compounds identified in all samples of artisanal chocolates from Espírito Santo and their odor descriptors.

Compound	Odor descriptor	Reference
Acetic acid	Sour, pungent	(Rodriguez‐Campos et al. 2011)
Isobutyric acid	Rancid	(Quelal et al. 2023)
Benzeneacetic acid	Honey, floral, sweet	(Deuscher et al. 2020)
Dodecanoic acid	Fatty, waxy	(Deuscher et al. 2020)
Hexadecanoic acid	Fatty, waxy	(Rodriguez‐Campos et al. 2011)
Ethyl acetate	Fruity, sweet	(Cemin et al. 2022)
Phenethyl acetate	Honey, floral, sweet	(Quelal et al. 2023)
3‐methylbutanal	Fruity, malt	(Cemin et al. 2022)
Acetoin	Buttery	(Braga et al. 2018)
Trimethylpyrazine	Toasty, nuts	(Oliveira et al. 2021)
Caffeine	Bitter	(Menezes et al. 2016)
2,3‐dihydro‐3,5‐dihydroxy‐6‐methyl‐4H‐pyran‐4‐one	Caramel, sweet	(Deuscher et al. 2020)

TABLE 2. Validation parameters for the HPLC‐DAD method applied to the determination of TEO and CAF in artisanal chocolate samples.

Analyte	Linear range (mg L−1)	R 2	t‐test	LOD (g kg−1)	LOQ (g kg−1)	REC (%)
TEO	0.200–1.200	0.9933	0.377	0.021	0.065	92 ± 10
CAF	0.150–0.900	0.9990	0.378	0.010	0.029	95 ± 3

TABLE 3. Concentrations of TEO and CAF in artisanal chocolate samples, grouped according to cocoa content.

Cocoa content (%)	Compound	Minimum	Maximum	Mean ± SD
30–65	TEO	1.16	12.29	5.86 ± 3.40
	CAF	0.06	0.98	0.45 ± 0.32
70–100	TEO	4.74	21.03	12.43 ± 3.60
	CAF	0.15	1.99	1.08 ± 0.46

Full text

Introduction

1

Chocolate is a complex food matrix whose chemical composition is strongly influenced by the botanical origin of cocoa beans, post‐harvest practices, and processing conditions, including fermentation, drying, roasting, and conching. These factors collectively shape not only the sensory attributes of chocolate, but also its nutritional profile and chemical markers associated with quality and authenticity (Afoakwa 2016). The chemical composition of cocoa and the transformations it undergoes during post‐harvest processing have been investigated for several decades, with early studies already describing the major classes of compounds present in cocoa beans and their evolution during fermentation and curing (Forsyth and Quesnel 1963). In recent years, the global chocolate market has experienced a growing demand for products with higher cocoa content and for chocolates classified as artisanal or bean‐to‐bar, which are commonly perceived by consumers as premium products associated with origin, craftsmanship, and distinctive sensory characteristics (Afoakwa 2016; Beckett 2009).

Artisanal or bean‐to‐bar chocolates are generally produced in small batches, often by a single producer who controls most stages of the production chain, from cocoa bean selection to final chocolate formulation. This production model contrasts with industrial chocolate manufacturing and is frequently associated with minimal formulation, reduced use of additives, and greater emphasis on cocoa origin and processing transparency. Despite the increasing market relevance of these products, the lack of standardized definitions and the premium value attributed to artisanal chocolates raise concerns regarding authenticity, label compliance, and potential misrepresentation of cocoa content (Torres‐Moreno et al. 2021; Perez et al. 2021; Carrillo et al. 2014).

From a chemical perspective, chocolate composition has been widely investigated through the determination of key compound classes, including lipids, sugars, alkaloids, and volatile organic compounds (VOCs). Among these, VOCs play a central role in defining chocolate aroma and flavor, being largely formed during fermentation and roasting through enzymatic reactions, lipid oxidation, and non‐enzymatic pathways such as Maillard and Strecker reactions (Afoakwa et al. 2007; Goya et al. 2022). Studies have reported the identification of hundreds of VOCs in chocolates, encompassing acids, aldehydes, alcohols, ketones, esters, pyrazines, and lactones, which collectively contribute to sensory complexity and consumer perception (Rodriguez‐Campos et al. 2011; Marseglia et al. 2020).

Although VOC profiling is highly informative for characterizing aroma, sensory attributes, and processing effects in chocolate, its direct application as a quantitative indicator of cocoa content is limited due to the strong influence of technological variables, such as fermentation degree and roasting conditions. Consequently, VOCs are more appropriately interpreted as qualitative or semi‐quantitative descriptors of processing history and sensory identity rather than as direct proxies for cocoa solids content.

In contrast, methylxanthines, particularly theobromine (TEO) and caffeine (CAF), are naturally occurring alkaloids in cocoa beans and represent chemically stable constituents that remain largely preserved throughout chocolate processing. TEO is the predominant methylxanthine in cocoa, while CAF is present at lower concentrations, and their levels have been consistently associated with cocoa solids content in chocolates and cocoa‐derived products (Menezes et al. 2016). Due to their stability and direct botanical origin, these compounds have been proposed as reliable chemical markers for evaluating cocoa content, product authenticity, and label conformity.

Chromatographic techniques are widely applied in complex food matrices, offering selectivity, speed, and robustness for the identification and quantification of compounds. In chocolate, headspace solid‐phase microextraction (HS‐SPME) coupled with gas chromatography–mass spectrometry (HS‐SPME/GC‐MS) has proven highly effective for detecting sensory‐relevant VOCs. Likewise, high‐performance liquid chromatography coupled with diode array detection (HPLC‐DAD) is well established for the sensitive determination of methylxanthines (Torres‐Moreno et al. 2021; Nascimento et al. 2024). To support the interpretation of complex chromatographic datasets, multivariate statistical tools such as principal component analysis (PCA), hierarchical cluster analysis (HCA), and partial least squares–discriminant analysis (PLS‐DA) have been widely employed in chocolate studies to explore patterns and relationships among chemical variables.

Despite the extensive literature addressing either VOC profiling or methylxanthine determination in chocolates, studies integrating both approaches within a single analytical framework, particularly for artisanal chocolates, remain limited. Moreover, few investigations have explicitly combined chromatographic characterization of volatile compounds, as indicators of sensory and processing attributes, with the quantitative determination of methylxanthines as markers of cocoa content.

The central hypothesis of this study is that methylxanthines, specifically TEO and CAF, can serve as reliable chemical markers of cocoa content in artisanal chocolates, while VOCs provide complementary information on compositional and sensory characteristics related to processing and origin. Accordingly, this work aims to perform a chromatographic profiling of artisanal chocolates produced in Espírito Santo, Brazil, by combining HS‐SPME/GC‐MS analysis of VOCs with HPLC‐DAD quantification of methylxanthines. This integrated approach seeks to enhance the chemical characterization of artisanal chocolates by combining complementary chromatographic analyses of volatile compounds and methylxanthines.

Materials and Methods

2

Sampling

2.1

A total of 45 artisanal chocolate samples, with cocoa contents ranging from 36% to 100%, were obtained from artisanal and family‐run factories located in the northern region of Espírito Santo, Brazil. This cocoa content range reflects the products commercially available from the selected producers at the time of sampling. The cocoa content classification was based on the information reported on the product labels by the manufacturers, and only products clearly labeled within this range were included in the study. No independent analytical determination of cocoa content was performed prior to sample selection.

The original packaging was removed, and the samples were manually fragmented using a stainless‐steel grater to obtain a homogeneous particle size, transferred to sterile containers, and stored at −10°C until analysis.

A detailed list of samples analyzed by HS‐SPME/GC‐MS and HPLC‐DAD is provided in Table S1 (Supporting Information).

Reagents and Solutions

2.2

Ultrapure water (18.2 MΩ·cm, PURELAB Ultra system, Elga, UK) and dimethyl sulfoxide (DMSO, ≥ 99.9%, Neon, Brazil) were used for solution preparation. Working solutions were prepared at 1.2 mg mL^−1^ for TEO and 1.0 mg mL^−1^ for CAF. Analytical standards with a purity of ≥ 99.4% for TEO (Sigma‐Aldrich, St. Louis, MO, USA) and ≥ 99.5% for CAF (Jilin Shulan Synthetic Pharmaceutical Co. Ltd., Jilin, China) were employed. The certified reference material SRM 2384–Baking Chocolate (NIST, Gaithersburg, MD, USA) was used to validate the extraction method and compare TEO and CAF concentrations during the chromatographic procedure.

Instrumentation

2.3

The VOCs were identified using a gas chromatograph coupled to a quadrupole mass spectrometer (GC‐MS QP2010, Shimadzu, Kyoto, Japan). Methylxanthine quantification was carried out on a HPLC‐DAD system (Vanquish Flex, Thermo Fisher Scientific, Waltham, USA). For the extraction step, a minicentrifuge (K14‐0602, Kasvi, Curitiba, Brazil) and an ultrasonic bath (Q3.0/40, Eco‐Sonics, Indaiatuba, Brazil) were used.

Identification of VOCs by HS‐SPME/GC‐MS

2.4

The extraction and injection conditions were established based on preliminary assays, guided by conditions commonly reported in the literature for HS‐SPME/GC‐MS analysis of cocoa and other food matrices (Torres‐Moreno et al. 2021; Rodriguez‐Campos et al. 2011; Oliveira et al. 2021; Marseglia et al. 2020). Initial tests evaluated fiber performance, extraction temperature, and exposure time, and the final conditions represent an adapted protocol in relation to those reported in the literature, selected to improve analyte release and chromatographic reproducibility for the analyzed chocolate samples.

VOCs were extracted by HS‐SPME using a 50/30 DVB/CAR/PDMS fiber (Supelco, Bellefonte, PA, USA). An aliquot of 2 g of the sample was transferred to 20 mL headspace vials sealed with magnetic caps and equilibrated at 95°C for 15 min. The fiber was then exposed to the sample headspace under constant agitation for 30 min and subsequently introduced into the GC injector. The syringe temperature was maintained at 100°C with an injection speed of 50 µL s^−1^.

Chromatographic separation was achieved on an OV‐WAX capillary column (30 m × 0.25 mm × 0.25 µm). The oven program was set as follows: 40°C for 2 min, then increased at 5°C min^−1^ to 230°C, and held for 8 min. Helium (99.999%) was used as the carrier gas at a constant flow of 1.0 mL min^−1^. Samples were injected at 240°C in split mode with a purge flow of 3.0 mL min^−1^. The chromatographic run lasted 48 min.

The quadrupole mass spectrometer was operated under electron ionization at 70 eV, with an ion source temperature of 230°C and an interface temperature of 220°C. Compound identification was performed by comparison of the mass spectra with those from the NIST spectral library, considering a minimum similarity match of 85%. A detailed description of the HS‐SPME/GC‐MS parameters is presented in Table S2 of Supporting Information.

Ultrasound‐Assisted Liquid‐Liquid Extraction

2.5

The ultrasound‐assisted liquid‐liquid extraction (US‐LLE) method was adapted from (Carrillo et al. 2014). The extraction procedure was designed according to a 2^3^ full factorial design. The evaluated variables included centrifugation time, solvent concentration, and solvent pH. For the optimization of the US‐LLE procedure, two representative chocolate samples containing 44% and 70% cocoa were selected. These samples were chosen to encompass distinct levels of cocoa solids and matrix composition within the range investigated in this study, allowing the extraction conditions to be optimized considering different cocoa contents. The same samples were used throughout the factorial design experiments to ensure consistency during method optimization.

The experimental variables were assessed at three coded levels (−1, 0, and +1) using a 2^3^ full factorial design with three replicates at the central point, totaling 11 experiments. The tested variables were centrifugation time (2, 4, and 6 min), concentration of the extracting solution (40, 60, and 80%), and pH of the extracting solution (5, 7, and 9). Data analysis was performed using Statistica software, version 14.1.0.8 (StatSoft Inc., Tulsa, OK, USA).

Samples were removed from the freezer, allowed to reach room temperature, homogenized, quartered, and weighed (approximately 100 mg). They were then transferred to 2.0 mL microcentrifuge tubes. To reduce matrix effects, defatting was carried out by adding 2.0 mL of n‐hexane (Merck, Darmstadt, Germany), repeated three times, followed by centrifugation at 6400 rpm according to the experimental design. After each step, the n‐hexane was discarded.

Analyte extraction was performed at room temperature with 1.5 mL of an aqueous 2‐propanol solution (PanReac AppliChem, Darmstadt, Germany), prepared at different concentrations and pH values according to the factorial design. The mixture was sonicated in an ultrasonic bath operating at 40 kHz and approximately 110 W RMS for 60 min and subsequently stored under refrigeration for three days to facilitate separation of the lipid fraction, which was removed with a Pasteur pipette. After this step, the samples were centrifuged again under the same conditions as in the initial stage.

The resulting supernatant was filtered through a 0.45 µm PTFE syringe filter, transferred to 2.0 mL vials, and stored at −10°C until analysis. The experimental variables and their respective levels used in the 2^3^ full factorial design for method optimization are summarized in Table S3 (Supporting Information).

To evaluate the efficiency of the defatting procedure and to verify the presence of fatty acid methyl esters (FAMEs) after extraction, the n‐hexane fractions obtained from 10 randomly selected samples were collected. These samples were subjected to alkaline hydrolysis followed by methylation and subsequent GC‐MS analysis.

Between 1 and 3 mg of sample were transferred to 2.0 mL vials, and 300 µL of 0.5 mol L^−1^ KOH in methanol were added. The reaction mixture was maintained at 60°C for 30 min to promote hydrolysis of the esterified lipids. Subsequently, 300 µL of BF_3_ in methanol (Sigma‐Aldrich, St. Louis, MO, USA) was added, and the mixture was kept at 60°C for an additional 30 min to convert the fatty acids into their corresponding methyl esters.

After methylation, 300 µL of water and 400 µL of n‐hexane were added, followed by gentle agitation to facilitate phase separation. The organic phase was then collected and used for GC‐MS analysis. A detailed description of the GC‐MS operating conditions is provided in Table S4 (Supporting Information).

Determination and Quantification of Organic Compounds by HPLC‐DAD

2.6

Chromatographic separation was performed on a Luna(2) C18 reversed‐phase column (250 mm × 4.6 mm, 5 µm; Phenomenex, Torrance, CA, USA). The mobile phase consisted of acetonitrile (solvent A; J.T. Baker, Phillipsburg, NJ, USA) and water acidified with 0.1% v/v acetic acid (solvent B; Neon, São Paulo, Brazil), applied in a gradient program as described in Table S5 (Supporting Information). The injection volume was 3 µL, the column was maintained at 40°C, and the flow rate was set to 0.9 mL min^−1^. The DAD was operated over the 250–280 nm range, with detection performed at 276 nm, a characteristic wavelength for methylxanthines, ensuring adequate selectivity and sensitivity for the quantification of these analytes.

Method Validation

2.7

The optimized method was validated according to guidelines commonly applied in chromatographic analyses of food matrices. Linearity was assessed using calibration curves with six concentration levels, each prepared in triplicate. Limits of detection (LOD) and quantification (LOQ) were estimated from signal‐to‐noise (S/N) ratios of 3 and 10, respectively, considering the standard deviation of blank responses and the slope of the calibration curves. Accuracy was evaluated through recovery assays in fortified samples, calculated as the ratio between the measured and expected concentrations. Precision was determined by repeatability (intra‐day) and intermediate precision (inter‐day), expressed as relative standard deviation (RSD, %).

Statistical Analysis

2.8

Prior to the application of parametric statistical tests, data distribution was evaluated to verify the assumptions of normality. Data from the US‐LLE optimization experiments were evaluated by ANOVA, and regression models were fitted to assess the significance of the factors and their interactions.

For the chocolate sample analyses, descriptive statistics, correlation tests, and regression analyses were performed. Pearson correlation coefficients were calculated between cocoa content and methylxanthine concentrations, and linear regression models were fitted to estimate the relationship between these variables.

PCA was applied as an exploratory multivariate tool considering TEO and CAF concentrations, and the first two principal components were retained for interpretation.

All statistical analyses were performed using Data Science Workbench (Statistica) 14.1.0 (Cloud Software Group, Palo Alto, USA) and OriginPro 2025b Learning Edition (OriginLab Corporation, Northampton, MA, USA). A significance level of 95% (p < 0.05) was adopted.

Results and Discussion

3

Identification of VOCs

3.1

VOC analysis was conducted on a subset of 17 artisanal chocolate samples. These samples were selected to represent the variability in cocoa content and production characteristics relevant to this study. The objective of VOC analysis was to provide compositional and aromatic characterization rather than exhaustive coverage of all samples. This subset was considered sufficient and produced profiles consistent with the study objectives.

In total, 72 VOCs were identified in the analyzed samples. These compounds were classified as carboxylic acids, alcohols, aldehydes, ketones, esters, lactones, methylxanthines, pyrazines, pyranones, pyrroles, and terpenes. The main VOCs, together with their chemical classification and odor descriptors, are presented in Table S6 (Supporting Information). This finding is consistent with previous reports describing more than 500 different VOCs in chocolates, reflecting the chemical complexity of this matrix (Britto De Andrade et al. 2021; Braga et al. 2018; Hernandez and Rutledge 1994).

During cocoa bean processing, reactions such as Maillard reactions, Strecker degradation, aldol condensations, polymerization, and cyclization contribute to the formation of these compounds. These reactions are also directly associated with the development of chocolate sensory properties (Whitfield and Mottram 1992; Ducki et al. 2008; Afoakwa et al. 2008). Figure 1 summarizes the number of occurrences of each chemical class identified in the analyzed samples.

Number of occurrences of each class of VOCs in the analyzed samples.

As shown in Figure 1, carboxylic acids represented the predominant class among the identified VOCs, a result that agrees with findings reported by Cemin et al. (2022). According to these authors, the predominance of this class is associated with cocoa fermentation, during which native bean sugars are degraded, promoting the formation of aroma compounds that enhance chocolate sensory complexity. Fermentation generally leads to an increase in acids and esters, accompanied by a reduction in alcohols, aldehydes, and ketones, resulting in more complex aromatic profiles with floral and fruity notes.

Several linear carboxylic acids were identified in the analyzed samples, including acetic, propanoic, butanoic, hexanoic, heptanoic, octanoic, nonanoic, decanoic, dodecanoic, tetradecanoic, pentadecanoic, and hexadecanoic acids. These compounds are largely attributed to cocoa butter lipolysis, which releases free fatty acids, followed by oxidation of medium‐ and long‐chain fatty acids (Rodriguez‐Campos et al. 2011; Shahidi and Hossain 2022; Tape et al. 2023).

Branched‐chain acids such as 2‐methylpropanoic and 3‐methylbutanoic acids are associated with the oxidation of their corresponding Strecker aldehydes. These aldehydes are formed from the degradation of branched‐chain amino acids during Maillard reactions. Model and applied food studies have shown that Strecker‐type reactions can generate both aldehydes and their corresponding acids through oxygen‐dependent pathways (Hofmann et al. 2000; Smit et al. 2009; Bueno et al. 2018; Marrufo‐Curtido et al. 2022).

Isobutyric acid, also detected in the analyzed chocolates, is associated with undesirable sensory descriptors, such as rancid, fatty, and spoiled butter notes. Rodríguez‐Campos et al. (2011) reported that this compound may significantly impact chocolate sensory quality even at low concentrations. Quelal et al. (2023) further indicated that elevated levels of isobutyric acid can impair sensory perception by masking desirable aromas, such as fruity, nutty, and caramel notes.

Cemin et al. (2022) reported that chocolates produced with cocoa from Espírito Santo exhibited the lowest mean peak areas of isobutyric acid (0.39 ± 0.08) when compared with samples from other regions, such as Côte d'Ivoire (2.02 ± 0.40) and Rondônia (1.54 ± 0.26). This difference has been attributed to fermentation conditions and post‐harvest practices, indicating that although isobutyric acid is present in chocolates from Espírito Santo, its concentration is lower than that observed in products from other origins.

Cemin et al. (2022) also identified (Z)‐2‐heptenal and 2‐pentylfuran in chocolates produced with cocoa from Espírito Santo, compounds associated with green and moldy notes that may increase consumer rejection. However, the presence of these compounds was not confirmed in the present study.

Compounds such as trimethylpyrazine and acetoin, which are formed during roasting and are responsible for toasted, buttery, and nutty aromas, were also identified (Braga et al. 2018; Crafack et al. 2014; Marseglia et al. 2020). National studies have reported strong positive correlations between trimethylpyrazine and chocolate flavor intensity, suggesting its potential use as an aromatic marker for chocolates with higher cocoa content (Cemin et al. 2022; Oliveira et al. 2021).

Among the 72 VOCs identified across the analyzed samples, twelve compounds were consistently detected in all samples (100% occurrence), regardless of origin or cocoa content. These compounds, which are common to all analyzed samples, were: 2,3‐dihydro‐3,5‐dihydroxy‐6‐methyl‐4H‐pyran‐4‐one, isobutyric acid, 3‐methylbutanal, lauric acid, palmitic acid, 2,3,5‐trimethylpyrazine, acetoin, phenethyl acetate, acetic acid, benzenoacetic acid, CAF, and ethyl acetate. Table 1 lists these compounds along with their odor descriptors reported in the literature. When evaluated in conjunction with their sensory descriptors, this subset allows the delineation of the aromatic identity common to the analyzed chocolates.

The main sensory descriptors associated with the VOCs composition of the samples included roasted, nutty, buttery, floral, sweet, slightly sweet, fruity, and bitter. Figure 2 presents the total ion chromatogram (TIC) of a representative artisanal chocolate sample, illustrating the complexity of the volatile profile and the diversity of compounds detected by HS‐SPME/GC‐MS. Acids such as acetic and hexadecanoic acids (Rodríguez‐Campos et al. 2011), as well as isobutyric acid (Quelal et al. 2023), indicate sensory contributions related to fermentation. Floral and sweet characteristics were associated with compounds such as phenylacetic acid and dodecanoic acid (Deuscher et al. 2020), as well as phenethyl acetate (Quelal et al. 2023). Aldehydes and esters, including 3‐methylbutanal and ethyl acetate (Cemin et al. 2022), contributed fruity and malty nuances, while acetoin imparted buttery notes (Braga et al. 2018).

Total ion chromatograms (TIC) of artisanal chocolates with (A) 44% cocoa, (B) 60% cocoa, and (C) 70% cocoa.

Processing‐related markers were also identified, such as trimethylpyrazine, associated with toasted and nutty aromas (Oliveira et al. 2021), and 2,3‐dihydro‐3,5‐dihydroxy‐6‐methyl‐4H‐pyran‐4‐one, linked to caramel and sweet notes (Deuscher et al. 2020). CAF, known for its bitter taste (Menezes et al. 2016), contributed to the gustatory intensity of the product. Together, these compounds indicate that although the chocolates share elements commonly found in Brazilian chocolates from other origins, those produced in Espírito Santo exhibit a distinctive aromatic arrangement combining sweetness, fruitiness, and toasted notes, reflecting both regional cocoa characteristics and processing practices.

The diversity of compounds observed confirms the complexity of the aromatic matrix of artisanal chocolate and its dependence on production processes, in agreement with previous studies (Crafack et al. 2014; Marseglia et al. 2020; Braga et al. 2018). Acetic acid, resulting from acetic fermentation, was present in all samples and increased proportionally with cocoa content, a behavior attributed to the higher activity of acetic bacteria in intense or prolonged fermentations (Utrilla‐Vázquez et al. 2020; Marseglia et al. 2020). Similarly, CAF showed an increase in peak areas with increasing cocoa content, corroborating that the concentration of methylxanthines is directly related to the proportion of cocoa in the final product (Afifah et al. 2025; Menezes et al. 2016).

Although CAF was detected by HS‐SPME/GC‐MS, TEO was not observed; this behavior is consistent with the markedly lower volatility of TEO, which severely limits its partitioning into the headspace, in contrast to CAF. Physicochemical data reported in public databases indicate that the vapor pressure of TEO is several orders of magnitude lower than that of CAF, supporting its poor suitability for headspace‐based extraction techniques. Moreover, the determination of TEO has long been recognized as more appropriate for liquid‐phase chromatographic techniques, as established in studies on methylxanthine analysis in chocolates (Afoakwa et al. 2008; Bispo et al. 2002).

Overall, these results demonstrate that the VOC composition of artisanal chocolates from Espírito Santo reflects the biochemical reactions occurring during cocoa fermentation and roasting. The application of GC‐MS proved effective for profiling and discriminating samples, suggesting its use as a tool for quality control and the commercial valorization of origin‐labeled chocolates.

Ultrasound‐Assisted Liquid‐Liquid Extraction

3.2

The use of chocolate samples with 44% and 70% cocoa for extraction optimization allowed the evaluation of the US‐LLE performance under matrices with distinct cocoa solid contents. This strategy ensured that the optimized conditions were suitable for chocolates with both moderate and high cocoa percentages, covering a broad compositional variability in terms of lipid content, methylxanthine levels, and matrix complexity. Consequently, the optimized extraction procedure was considered applicable to the full set of samples analyzed in this study.

The full factorial design indicated that centrifugation for 4 min, combined with the use of a 60% 2‐propanol extracting solution (pH 7), maximized methylxanthine recovery, resulting in greater efficiency and lower data dispersion.

Regression models evaluated by ANOVA demonstrated statistically significant relationships between the process variables and the analytical responses. The coefficients of determination (R ^2^) ranged from 0.666 to 0.823, indicating good model performance and agreement with previous studies. These values are consistent with previous studies applying factorial designs to the extraction of methylxanthines from cocoa husks, seeds, and chocolates, which reported statistically significant models (p < 0.05) with high coefficients of determination. In particular, solvent composition and pH were identified as critical factors for extraction efficiency, findings that are consistent with Box–Behnken and Doehlert models applied to cocoa by‐products (Lordêlo Nascimento et al. 2025; Pagliari et al. 2022; Mellinas et al. 2020). For CAF_70%_, the concentration of the extracting solution showed a statistically significant effect (p = 0.041963 < 0.05; F = 8.703449), a result also reported in ultrasound‐assisted methods, in which operational variables directly influence TEO and CAF recovery (Peralta‐Jiménez and Cañizares‐Macías 2013).

Recovery was evaluated as the ratio between measured and expected concentrations in CRM‐fortified samples. This procedure is recommended for estimating recovery, associated uncertainty, and ensuring metrological traceability of the method (Coordenação Geral de Acreditação 2020). In the assays performed, average recoveries of 92% ± 10% for TEO and 95% ± 3% for CAF were observed, values consistent with acceptance ranges and with the expected precision for the evaluated matrix (AOAC International 2023; Coordenação Geral de Acreditação 2020). These results are in line with validation studies for methylxanthines in cocoa/chocolate, which report recoveries close to or higher than 95% with RSDs generally below 3%–4%, corroborating the suitability of the extraction procedure and the robustness of the method for the investigated working range (Aresta et al. 2005; Caudle et al. 2001).

The degreasing step was applied to remove the lipid fraction from the chocolate matrix, thereby minimizing potential interferences and matrix effects during HPLC‐DAD analysis. To assess the efficiency of this procedure, the n‐hexane fractions recovered from a subset of samples were subjected to GC–MS analysis after alkaline hydrolysis and methylation. The results revealed the presence of several FAMEs, including myristic, palmitic, stearic, oleic, linoleic, arachidic, and behenic acids, among others.

Cocoa butter is known to be predominantly composed of saturated and monounsaturated fatty acids, particularly palmitic, stearic, and oleic acids, which together account for more than 90% of its lipid fraction (Alotaibi et al. 2024; de Kobi et al. 2024). Therefore, the detection of these compounds in the discarded n‐hexane fraction confirms the effective removal of lipids from the chocolate matrix. Moreover, the conversion of fatty acids into their corresponding methyl esters prior to GC‐MS analysis is consistent with established methodologies for lipid characterization in cocoa and cocoa‐derived products, including studies on composition, authenticity, and adulteration (Alotaibi et al. 2024; Maurer and Rodriguez‐Saona 2013).

The identification of FAMEs in the discarded fraction confirms the effectiveness of the degreasing step. This result indicates that the final extracts were free from lipid interference and suitable for accurate methylxanthine determination by HPLC‐DAD.

Determination and Quantification of Organic Compounds by HPLC‐DAD

3.3

Chromatographic Method Validation

3.3.1

The evaluation of separation conditions was performed based on the resolution values of chromatographic peaks (Rs), calculated from retention times and baseline widths of adjacent peaks (Fanali et al. 2023). Rs values greater than 1.5 are generally considered indicative of adequate separation (Lundanes and Reubsaet 2013).

In the optimized procedure, an Rs value of 1.8 was obtained, confirming the efficiency of the chromatographic separation. The observed retention times were 4.26 min for TEO and 4.55 min for CAF, as illustrated in Figure 3.

Representative HPLC‐DAD chromatogram of a fortified artisanal chocolate sample (70% cocoa), showing separation of TEO and CAF under optimized conditions.

Method validation parameters were evaluated according to the criteria summarized in Table 2. TEO and CAF showed linear responses with coefficients of determination (R ^2^) higher than 0.99.

The mean peak areas obtained for TEO and CAF were 3.09 × 10^6^ and 2.37 × 10^6^, with standard deviations of 1.30 × 10^6^ and 1.42 × 10^6^, respectively, indicating comparable variability between the analytes.

The 95% confidence intervals overlapped, and Student's t‐test did not reveal a significant difference between the means (p = 0.377; Welch's p = 0.378), confirming the reproducibility of the method under the evaluated conditions.

LOD and LOQ were calculated from the S/N ratio, considering the standard deviation of the blank and the slope of the calibration curve, in accordance with recommendations for analytical method validation (Coordenação Geral de Acreditação 2020).

The obtained values met acceptance criteria, confirming adequate sensitivity and reliability for TEO and CAF determination in the analyzed matrix.

Quantification of Methylxanthines in Artisanal Chocolates

3.3.2

A total of 45 commercial chocolate samples with different cocoa contents were analyzed for methylxanthines. TEO concentrations ranged from 1.158 to 21.033 g kg^−1^, while CAF ranged from 0.058 to 1.997 g kg^−1^ for CAF.

To evaluate the influence of cocoa content, samples were grouped into two classes: 30%–65% and 70%–100% cocoa. As summarized in Table 3, higher‐cocoa chocolates presented mean values of 12.43 g kg^−1^ for TEO and 1.08 g kg^−1^ for CAF. Lower‐cocoa chocolates showed mean values of 5.86 g kg^−1^ and 0.45 g kg^−1^, respectively.

These results correspond to approximately a twofold increase in TEO and a threefold increase in CAF in higher‐cocoa chocolates, consistent with the natural methylxanthine composition of cocoa.

Representative HPLC‐DAD chromatograms of chocolates containing 44%, 60%, and 70% cocoa are shown in Figure 4, illustrating the proportional increase in TEO and CAF concentrations with cocoa content.

Representative HPLC‐DAD chromatograms of artisanal chocolates with (A) 44% cocoa, (B) 60% cocoa, and (C) 70% cocoa.

Analysis of the reference material SRM 2384 yielded concentrations of 11.901 ± 3.699 g kg^−1^ for TEO and 0.619 ± 0.205 g kg^−1^ for CAF. When compared with the values reported by Nascimento et al. 2020 for the same material (TEO: 10.804 ± 1.026 g kg^−1^; CAF: 1.058 ± 0.087 g kg^−1^), agreement was observed for TEO within the uncertainties, whereas CAF showed a lower value.

This difference may be related to sample preparation and handling, as well as to analyte instability over time, as previously reported by NIST (2020).

Pearson's correlation indicated positive associations between cocoa content and methylxanthine concentrations: r = 0.836 for TEO and r = 0.811 for CAF. Linear regression confirmed this trend (p < 10^−10^).

For TEO, the model estimated an average increase of 0.248 g kg^−1^ for each percentage point of cocoa content (R ^2^ = 0.692), while for CAF the increase was 0.027 g kg^−1^ (R ^2^ = 0.649), as shown in Figure 5.

Linear regressions between cocoa content (%) and methylxanthine concentrations in chocolates: (A) TEO and (B) CAF.

An increase from 30% to 70% cocoa corresponds, on average, to an increment of approximately 9.9 g kg^−1^ of TEO and 1.10 g kg^−1^ of CAF.

To further investigate the relationship between methylxanthine composition and cocoa content, PCA was applied using the concentrations of TEO and CAF measured in the samples as variables. Samples were grouped into two classes: Class A (36%–65% cocoa) and Class B (70%–100% cocoa).

The PCA scores plot (Figure 6A) showed clear separation between these groups, with higher‐cocoa samples located in the positive region of PC1 and lower‐cocoa samples in the negative region. Because the PCA model was built using two correlated variables, the first two principal components accounted for 100% of the total variance, allowing straightforward interpretation of the data structure.

Score (A) and loading (B) plots for PCA applied to the concentrations of TEO and CAF in chocolate samples.

The positive and similar loadings (Figure 6B) indicate that PC1 represents the joint variation of methylxanthines, with predominance of TEO. This result agrees with previous studies describing TEO as the main cocoa alkaloid and directly proportional to cocoa solids (Carrillo et al. 2014).

Although present at lower concentrations, CAF showed a similar trend. Literature reports CAF concentrations between 0.6 and 1.5 mg g^−1^ in chocolates containing 70%–85% cocoa, which are higher than values observed in chocolates with lower cocoa content.

The second principal component (PC2), which accounted for 6.10% of the variability, likely reflects subtler differences related to the TEO/CAF ratio, processing conditions, or genetic variability among cocoa varieties.

The literature highlights the TEO/CAF ratio as a useful parameter for classifying cocoa beans and chocolates, allowing discrimination by geographical origin or genotype (Carrillo et al. 2014).

Overall, these results reinforce the relevance of methylxanthines as chemical markers for the characterization, classification, and compliance assessment of chocolates. The predominance of TEO observed in this study is consistent with reports for Latin American and West African cocoa, supporting the robustness of TEO and CAF as universal indicators in cocoa and chocolate authentication and quality control strategies (Carrillo et al. 2014; Calva‐Estrada et al. 2020; Afoakwa et al. 2013).

Although this study provides relevant insights into the organic composition of artisanal chocolates, future studies may benefit from expanded sample sets, quantitative analysis of volatile compounds, and the inclusion of samples from other producing regions.

Conclusion

4

This study presents a targeted chromatographic characterization of artisanal chocolates from Espírito Santo, Brazil, combining HS‐SPME/GC‐MS analysis of VOCs with HPLC‐DAD quantification of methylxanthines. A total of 72 VOCs were identified across the analyzed samples, reflecting compositional variability associated with fermentation and roasting processes. Among these, twelve compounds were detected in all samples, regardless of origin or cocoa content, defining a set of aroma‐related constituents common to the chocolates analyzed. The VOC data were interpreted qualitatively, providing complementary information on processing and sensory attributes.

The optimized ultrasound‐assisted liquid–liquid extraction method enabled reliable quantification of TEO and CAF, whose concentrations showed a strong positive correlation with declared cocoa content. Chocolates with higher cocoa percentages exhibited substantially higher methylxanthine levels, and PCA, clearly differentiated samples according to cocoa content based on these compounds. Overall, the results demonstrate that TEO and caffeine are robust chemical markers for cocoa content verification and authentication in artisanal chocolates, while VOC profiling supports qualitative characterization related to processing and aroma.

Author Contributions

Joseline Altissimo: investigation, writing – original draft, methodology, formal analysis, validation. Ian Cardoso: formal analysis, methodology. Bruno Quirino Araújo: methodology, investigation, formal analysis. Alan R. Pereira: formal analysis. João Victor M. de Almeida: formal analysis. Nayara A. dos Santos: formal analysis, investigation. Vinicius Câmara Costa: conceptualization, investigation. Maria Tereza Weitzel Dias Carneiro Lima: conceptualization, funding acquisition, writing – review and editing, project administration.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supplementary Table S1‐S6: jfds70979‐sup‐0001‐TableS1‐S6.docx