Iterative optimization technology (IOT), a method that predicts the component composition from only the infrared (IR) spectra of the pure components and mixtures by using Beer's law, has been proposed to reduce the number of calibration samples for process analytical technology in the pharmaceutical industry. However, IOT cannot be applied to mixtures that have wavelength regions where Beer's law does not hold, such as liquid mixtures. The objective of this study is to apply IOT to liquid mixtures to realize a calibration-minimum method. We propose a novel calibration-minimum method that formulates spectral changes by polynomials of the mole fractions considering reasonable boundary conditions for online monitoring. The prediction ability of the proposed method was verified by three case studies: two binary mixtures and one ternary mixture. The model selection strategy, conditions for calibration, and estimation of missing pure component spectra are also discussed. This research represents a step towards advanced calibration-minimum methods. (C) 2018 Elsevier Ltd. All rights reserved.