ABSTRACT
Process control with high time resolution is essential to maintain high product quality in coffee roasting. However, analytical techniques for quality assurance or measurements of desired coffee properties are often labor-intensive and can only be conducted after dropping the coffee beans. Resonance-enhanced multi-photon ionization time-of-flight mass spectrometry (REMPI-TOFMS) at 248 nm and 266 nm was applied to analyze the composition of the roast gas from small-scale Arabica coffee roasting. Coffee beans were dropped after different roasting times, ground and analyzed by Colorette to obtain the roast degree. Additionally, the antioxidant capacity of the coffee brew was determined by Folin-Ciocalteu (FC) assay. Models for the prediction of Colorette and FC values from REMPI mass spectra were constructed by partial least squares (PLS) regression. REMPI-TOFMS enables the prediction of Colorette values with a root-mean-square error in prediction (RMSEP) below 5 for both wavelengths. FC values could be predicted using REMPI at 248 nm with an RMSEP of 80.3 gallic acid equivalents (GA-eq) mg L-1, while REMPI at 266 nm resulted in RMSEP of 151 GA-eq mg L-1. Finally, the prediction of Colorette and FC value at 5 s time resolution were demonstrated with online measurements.
ABSTRACT
Precise controlling and monitoring the status of the coffee roasting process is essential for consistent product quality and optimization toward targeted coffee properties. In small-scale roasting experiments, the chemical composition of the roasting off-gas was analyzed by online single-photon ionization time-of-flight mass spectrometry (SPI-TOFMS) at 118 nm with 5 s time resolution. Subsequently, mass spectra at the drop of the coffee beans were combined with off-line measurements of roast degree, described by color value "Colorette", and the antioxidant capacity, obtained from the Folin-Ciocalteu (FC) assay, in an explanatory projection on latent structure regression model. While the roast degree gives an indication of the coffee flavor, antioxidants in brewed coffee are regarded as beneficial for human health. Colorette and FC values could be derived from the SPI mass spectra with root-mean-square errors from Monte Carlo cross-validation of 6.0 and 139 mg of gallic acid equiv L-1, respectively, and explained covariance (R2CV) better than 89%. Finally, the regression models were applied to the SPI mass spectra over the entire roast to demonstrate the predictive ability for online process control in real time.