RESUMEN
We investigated caffeine and l-theanine, quality characteristics for camellia sinensis, in milled and ground black tea samples with near-infrared (NIR) spectroscopy giving a direct comparison between the performances of benchtop and handheld NIR spectrometers. The constructed partial least squares regression (PLSR) models for all spectrometers were validated by test-set-validation and according to the obtained root mean square errors of prediction (RMSEP). The performances of the spectrometers were as follows: The benchtop spectrometer NIRFlex N-500 (Büchi, Flawil, Switzerland) showed the best results for milled samples with a RMSEP of 0.14% for caffeine and 0.12% for l-theanine. For the ground samples, a RMSEP of 0.17% for caffeine and 0.12% for l-theanine was gained. While the handheld spectrometers MicroNIR 2200 (Viavi Solutions (former: JDS Uniphase Corporation), Milpitas, USA) and the microPHAZIR (Thermo Fisher Scientific, Waltham, USA) both provided good results for the prediction of caffeine in milled samples (RMSEP of 0.22% and 0.26%), only the microPHAZIR was able to satisfactorily determine the caffeine content in ground samples (RMSEP of 0.28%). The investigation of l-theanine with handheld spectrometers did not lead to convincing results, since R2 was 0.75 for milled samples while ground samples could not be calculated. Decisive differences were concluded in how different NIR instruments capture the chemical information on caffeine vs. l-theanine. The handheld spectrometers manifested limited applicability to l-theanine. Deeper insight was obtained through the detailed NIR band assignments of caffeine and l-theanine derived from quantum mechanical simulation. Narrow working spectral region of handhelds omits the characteristic absorption bands of l-theanine. Therefore, information on l-theanine content measured by the evaluated miniaturized spectrometers is insufficient to enable its effective quantification. In contrast, the most characteristic NIR absorption of caffeine matches the working spectral regions of the handheld NIR spectrometers, hence their performance is comparable with the benchtop device.
RESUMEN
This study examined the applicability of near-infrared (NIR) spectroscopy coupled with multivariate data analysis (MVA) to determine the ideal harvest time of Verbena officinalis. NIR analyses were performed non-invasively on the fresh plant material based on the quantification of the key constituents verbenalin and verbascoside. Vibrational spectroscopic measurements were performed applying a conventional NIR benchtop device as well as a laboratory independent handheld NIR spectrometer. A novel high performance liquid chromatography (HPLC) method was applied as a reference method. For both instruments partial least squares (PLS) regression models were established performing cross validations (CV) and test-set validations (TSV). Quality parameters obtained for the benchtop device revealed that the newly established NIR method enabled reliable quantifications of the main compounds verbenalin and verbascoside related to the dried and fresh plant material. The results of the miniaturised spectrometer revealed that accurate quantitative calibration models could be developed for verbascoside achieving a comparable prediction power to the benchtop device. PLS models for verbenalin were less precise suggesting the application of portable devices including a different spectral range and resolution. The work demonstrated the feasibility of NIR vibrational spectroscopy performing direct measurements on pharmaceutically relevant fresh plant material enabling a quick and simple determination of the ideal harvest time of Verbena officinalis.