PyIRoGlass: An open-source, Bayesian MCMC algorithm for fitting baselines to FTIR spectra of basaltic-andesitic glasses

Abstract

Quantifying volatile concentrations in magmas is critical for understanding magma storage, phase equilibria, and eruption processes. We present PyIRoGlass, an open-source Python package for quantifying concentrations of H2O and CO2 species in the transmission FTIR spectra of basaltic to andesitic glasses. We leverage a dataset of natural melt inclusions and back-arc basin basalts with volatiles below detection to delineate the fundamental shape and variability of the baseline underlying the CO32- and H2Om, 1635 peaks, in the mid-infrared region. All Beer-Lambert Law parameters are examined to quantify associated uncertainties. PyIRoGlass employs Bayesian inference and Markov Chain Monte Carlo sampling to fit all probable baselines and peaks, solving for best-fit parameters and capturing covariance to offer robust uncertainty estimates. Results from PyIRoGlass agree with independent analyses of experimental devolatilized glasses (within 6 %) and interlaboratory standards (10 % for H2O, 6 % for CO2). We determine new molar absorptivities for basalts, εH2Ot,3550 = 63.03 ± 4.47 L/mol · cm and εCO2−3,1515,1430 = 303.44 ± 9.20 L/mol · cm; we additionally update the composition-dependent parameterizations of molar absorptivities, with their uncertainties, for all H2O and CO2 species peaks. The open-source nature of PyIRoGlass ensures its adaptability and evolution as more data become available.

Keywords

• ftir
• open-source
• python
• bayesian inference
• markov chain monte carlo
• volatiles