Classification of Coffees from Different Origins by Chemical Sensor Technology

Significance of the Topic

The rapid and reliable classification of coffee varieties by geographical origin is essential for quality control, product consistency and the development of tailored blends in the coffee industry. Traditional steam distillation extraction–GC–MS methods provide robust discrimination but require extended analysis times. The integration of fast headspace sampling with mass spectrometry as a chemical sensor offers a streamlined alternative for industrial and research environments aiming to combine speed with analytical performance.

Objectives and Study Overview

This study evaluated a static headspace–chemical sensor (S-HS-ChemSensor) platform to classify six roasted coffee varieties (three Arabica and three Robusta) from distinct origins. The goal was to test whether this fast, fully automated approach could reproduce or improve upon prior SDE-GC-MS classification results, while greatly reducing sample throughput time.

Used Instrumentation

• Agilent 6890/5973N GC-MS with high-temperature PONA column (250 °C)
• Gerstel MultiPurpose Sampler MPS 2 for automated static headspace sampling
• Pirouette 3.02 software for pattern recognition (HCA, PCA, SIMCA, k-NN)

Methodology

Roasted and ground coffee samples (2 g in 10-mL vials, ten replicates per origin) were incubated at 80 °C for 60 min. A 2.5-mL headspace aliquot was injected into the GC-MS in split mode (1:30). The column operated at 250 °C to avoid separation, delivering the total ion fingerprint (m/z 40–180). After removing common interferences (CO₂, column bleed, high-mass noise), a 60×141 data matrix was generated for multivariate analysis.

Main Results and Discussion

Hierarchical cluster analysis (HCA) of the processed fingerprints grouped the six coffee origins into distinct clusters with high reproducibility. Principal component analysis (PCA) captured over 96% of variance in the first two components and clearly separated Arabica and Robusta along PC1. Loading analysis linked Arabica samples to acid-related ions and Robusta to characteristic furanyl and alkyl fragments. Classification models based on k-nearest neighbors (1-NN) and SIMCA yielded zero misclassifications, demonstrating the method’s accuracy despite excluding lower-volatility phenolic markers.

Benefits and Practical Applications

• Analysis time reduced from 4 h to 1 h per sample.
• Fully automated on-line sampling enhances laboratory throughput.
• Flexible configuration allows switching between GC-MS profiling and ChemSensor mode.
• Reliable origin authentication supports quality assurance and blend optimization.

Future Trends and Applications

Advances may include integration of alternate rapid extraction techniques (e.g., SPME), miniaturized sensor arrays, and machine-learning algorithms for real-time quality monitoring. Expanding the sensor library to capture phenolic and high-boiling volatiles could further refine discrimination among closely related coffee cultivars.

Conclusion

The S-HS-ChemSensor approach provides a fast, robust alternative to traditional SDE-GC-MS for coffee origin classification. It delivers high accuracy with full automation and minimal sample preparation, making it attractive for industrial quality control and research applications.

References

• [1] I. Dirinck, I. Van Leuven, P. Dirinck, Czech J. Food Sci. 2000, 18, 50–51.
• [2] I. Dirinck, I. Van Leuven, P. Dirinck, in: Proceedings of the 11th European Conference on Food Chemistry, RSC, Cambridge 2001, p. 248.
• [3] I. Dirinck, I. Van Leuven, P. Dirinck, in: Proceedings of the 19th International Conference on Coffee Science (ASIC’01), chemistry section, item 15.