Pathlength effects on NIR Transflectance performance for polar and apolar liquid food analysis

Near-infrared (NIR) transflectance spectroscopy is widely used across various applications, including food analysis, yet the effect of accessory pathlength on spectral quality and subsequent chemometric performance has not been systematically assessed. This study evaluates how pathlengths (1, 2, 4, and 6 mm) influence spectral characteristics and partial least squares regression (PLSR) accuracy in two contrasting liquid food systems: an apolar matrix-tree-picked olive oil (TPOO) adulterated with ground-picked olive oil (GPOO)-and a polar matrix-vodka spiked with methanol. Spectra were acquired using both a benchtop FT-NIR spectrometer (10,000-4000 cm^-1) and a handheld NIR device (7407-3921 cm^-1). In the apolar oil system, longer pathlengths (4-6 mm) enhanced CH overtone bands, with the 6 mm pathlength delivering the highest prediction accuracy among those evaluated (1-6 mm) for %GPOO (SEP = 0.56 %, r = 1.000, RPD = 52.6). In the polar methanol-vodka system, strong OH absorption caused early saturation, making the 1 mm pathlength optimal within the tested pathlength range of 1-6 mm; removing the 5300-5000 cm^-1 region improved model accuracy (SEP = 0.036 %, r = 0.999, RPD = 47.2). While handheld NIR measurements followed the same pathlength trends, their predictive performance was lower due to narrower spectral range, lower resolution, and reduced signal-to-noise ratio. These findings offer practical guidance within the tested interval: in lipid-rich, apolar matrices, intermediate-to-long pathlengths (4-6 mm) maximize informative signal capture, whereas in water-rich, polar matrices, short pathlengths (1 mm) prevent spectral saturation and preserve detail, ensuring robust chemometric models.