The paper presents an original method developed by the authors for determining water content in highly polymerized rhyolite and rhyodacite glasses in melt inclusions. The method involves simultaneous determination of water content by two techniques: electron probe X-ray microanalysis (EPMA) and Raman spectroscopy, which are applied to mutual verify the results. The Raman spectroscopic technique was calibrated using a set of standard reference synthetic glasses of haplogranitic composition (SGS), containing 1.8 to 5.9 wt % water. The calibration was verified using a set of reference natural rhyolite obsidian samples (NRS) with water contents of 4.7 to 9.9 wt %. Two ratios were chosen as the calibrated parameters: (1) the area of the water and hydroxyl bands in the range of 2900-3800 cm(-1) to the area of the silicate band vibration (Si-O and Al-O) in the range of 850-1200 cm(-1) (A(w)/A(s)) and (2) the ratio of the intensities of the water band in the 3550-3560 cm(-1) region and the 480 cm(-1) band in the silicate vibration region (I-3550/I-480). As a result, the following calibration equations were acquired: C-H2O = 0.8458 A(w)/A(s) (sigma) +/- 0.17 wt % and C-H2O = 11.494I(3550)/I-480(sigma) +/- 0.21 wt %. Comparison of water concentrations in standard glasses determined by Raman spectroscopy using the A(w)/A(s) and I-3550/I-480 ratios and measured by X-ray microprobe analysis, showed a good consistency. Electron probe X-ray microanalysis was used to determine water content in glasses by recalculating the excess oxygen content. The method was tested on unheated glassy melt inclusions in minerals from dacitic and rhyolitic pumice from caldera eruption deposits of Okataina volcanic center, New Zealand; Kurile Lake, Kamchatka Peninsula; Vetrovoy Isthmus; and Lvinaya Past Bay of Iturup Island in the Kuriles Islands. The ability of using different A(w)/A(s) or I-3550/I-480 ratios makes it possible to minimize the effect of the overlaps of host-mineral bands in the low-frequency part of the Raman spectrum of the melt inclusions. We found out that the usage of the I-3550/I-480 ratio for melt inclusions in pyroxenes produces more reliable results than the use of the A(w)/A(s) ratio. At the same time, the A(w)/A(s) ratio is better to use for inclusions in quartz and plagioclase. Water content in the inclusions varies from 0.5 to 7.9 wt %. The proposed method can be routinely used for reliable measurements of water content in the glasses of melt inclusions no smaller than 10 mu m, at water contents no lower than 0.5 wt %.
