The feasibility of Attenuated Total Reflectance Fourier Transform Infrared (ATR FTIR)spectroscopy and chemometric modeling for analysis of trace compounds is studied on the example of 5-(hydroxymethyl)furfural (HMF)in honey up to a concentration of 110 ppm, and compared to results in literature. Randomized repeated measurements of the calibration samples were carried out to prevent analytical artifacts. A partial least squares (PLS)model was calculated for quantitative determination. The considered spectral range has significant influence on the PLS model. Furthermore, averaging of spectra can improve the model quality due to the elimination of statistical noise. A low root mean square error of prediction (RMSEP)of 13 ppm was achieved. Hence, the results indicate that trace compound determination with ATR FTIR spectroscopy is possible, although calibration and modeling efforts are higher compared to standard applications of ATR FTIR spectroscopy and chemometrics.
