Composition dependence of the dielectric and optical properties of (HfZrO4)((1-x)) (SiO2)((x)) (0 <= x <= 0.2) gate dielectric thin films, grown on Si(100) by the atomic layer deposition method, was investigated by means of reflection electron energy loss spectroscopy (REELS). The quantitative analysis of REELS spectra was carried out by using the QUASES-XS-REELS and QUEELS-epsilon(k,omega)-REELS software, which determine the dielectric function and optical properties through an analysis of experimental REELS in terms of a simulated energy loss function (ELF). For HfZrO4, the ELF showed peaks in the vicinity of 10.5, 15.6, 18, 21.5, 26.7, 34.5, 39.5, 46.5 and 57 eV. For HfZr-silicates with low SiO2 concentration (x = 0.10 and 0.15), the peak positions were similar to those of HfZrO4, but for x = 0.20, the number of peaks were reduced and are at 14.8, 23.5, 34.5, 40.5, and 46.5 eV. These peaks originate from the superposition among d electron states of Zr,f electron states of Hf and p electron states of Si. The strength of the peak at 46.5 eV decreased as we increased the amount of SiO2 in the compounds, which indicates that it is due to excitation between Hf 4f and Zr 3d electron states. Changes in the complex dielectric function, optical properties and band gap related to the SiO2 concentration in the films were discussed systematically. In addition, the inelastic mean free path (IMFP) was also calculated from the determined dielectric function. The IMFP of the HfZr-silicates increased with increasing SiO2 content and with increasing primary energy. This method is of high importance in terms of determining the dielectric and optical properties and inelastic mean free path from REELS spectra. The advantage of the method applied here is that no information on the morphology and components of the sample is required. This means that this method presented here turned out to be a convenient and efficient tool for investigation of optical properties and inelastic mean free path of ultrathin high-k alloy materials. (C) 2016 Elsevier B.V. All rights reserved.