Chalcogenide glass has been considered to be a promising optical material for infrared (IR) transmission and nonlinear optics because of its favorable physical properties such as wide IR transparent windows, high linear and nonlinear refractive indices, and tunable photosensitivity. In many optical designs and practical applications, the refractive index (n) and optical bandgap (Eg) are two important parameters. Aiming to evaluate the composition dependence of the n and Eg in Ge-As-S chalcogenide glasses, a series of glasses with different stoichiometric characteristics are synthesized in quartz tubes under vacuum by the melt quenching technique. The structure, n and Eg of the glass are investigated by Raman spectroscopy, ellipsometry, and diffused reflectance spectroscopy, respectively.To eliminate thermal effects on the measured Raman spectra, the data are corrected by the Bose-Einstein thermal factor. Raman spectrum analyses indicate that Ge-As-S glass has a continuous network structure with interconnected [GeS4] tetrahedra and [AsS3] pyramids forming the backbone. When S amount is excess, S chains or S8 rings emerge. When S amount is deficient, As4S4/As4S3 molecules are formed, and even a large number of As-As/Ge-Ge homopolar bonds appear in the structure. The n values at different wavelengths are obtained by fitting the ellipsometry data with the Sellmeier dispersion model. The values of molar refractivity (Ri) of Ge, As and S elements are evaluated by using the measured n and density (d) of the investigated glass. The optimal values of Ri at 2-10 m for each element are RGe=9.83-10.42 cm3/mol, RAs=11.72-11.87 cm3/mol, and RS=7.78-7.86 cm3/mol, respectively; and the values decrease with increasing wavelength. The n of Ge-As-S glass is well quantitatively correlated to the d and the Ri of constituent elements, so that its value can be predicted or tailored within 1% deviation. A method to determine reliable Eg of a glass is proposed based on diffuse reflectance spectrum (DRS) of glass powders. To determine Eg of a glass, the absorption coefficient () is required to be as low as ～104 cm-1. For a 1-mm-thick bulk glass, the detection limit of a spectrophotometer is typically 100 cm-1. To obtain a reasonable Eg, the sample thickness used for the measurement must be less than 10 m. Such a thin glass sample is difficult to prepare. In comparison, DRS of glass powers measured using a spectrophotometer is able to provide valid absorption data in a 104 cm-1 range required for Eg determination. In this proposed method, the Kubelka-Munk function F(R), which is proportional to of the glass, is calculated from the measured DRS on the glass powders. The F(R) is calibrated by using the DRS of a glass (e.g. As2S3) with a known Eg. Using the same F(R) absorbance value, Eg of the Ge-As-S glass is determined based on DRS of powders measured under the same condition. The Eg of Ge-As-S glass is broadly correlated to the average bond energy of the glass. The glass containing more S atoms tends to show a higher average bond energy, and therefore exhibits a larger Eg.