Sensitivity study of chemistry in AGB outflows using chemical kinetics

Abstract

Asymptotic giant branch (AGB) stars shed a significant amount of their mass in the form of a stellar wind, creating a vast circumstellar envelope (CSE). Owing to the ideal combination of relatively high densities and cool temperatures, CSEs serve as rich astrochemical laboratories. While the chemical structure of AGB outflows has been modelled and analysed in detail for specific physical setups, there is a lack of understanding regarding the impact of changes in the physical environment on chemical abundances. A systematic sensitivity study is necessary to comprehend the nuances in the physical parameter space, given the complexity of the chemistry. This is crucial for estimating uncertainties associated with simulations and observations. In this work, we present the first sensitivity study of the impact of varying outflow densities and temperature profiles on the chemistry. With the use of a chemical kinetics model, we report on the uncertainty in abundances, given a specific uncertainty on the physical parameters. In addition, we analyse the molecular envelope extent of parent species and compare our findings to observational studies. Mapping the impact of differences in physical parameters throughout the CSE on the chemistry is a strong aid to observational studies.

MACE: A Machine-learning Approach to Chemistry Emulation

“…A one-dimensional (1D) representation of an AGB star's circumstellar envelope is used as the astrochemical environment for the development of MACE. Since this environment is already well studied and understood (e.g., Millar et al 2000;Li et al 2016;Maes et al 2023), MACE can be benchmarked against this well-studied case. To generate training data, we use a version of the publicly available CSE model of the UMIST Database for Astrochemistry, 5 which calculates the chemical abundances as a function of the distance from the star (using a modified form of Equation (2)).…”

“…To be efficient with data and computational resources, we reuse the grid of chemical models from the sensitivity analysis performed by Maes et al (2023) as training data for MACE. The grid was constructed to span a broad range of CSE parameters found in AGB stars, based on observations.…”

“…To build the training data set from the grid of chemistry models from Maes et al (2023), the density ρ and temperature T are calculated as given by Equations (4) and (5), respectively. The two remaining parameters ξ and A V depend indirectly on ρ.…”

“…For the purpose of this research, the radii are converted to time, using the expansion velocity of the corresponding simulations. Adapted from Maes et al (2023). ( * ) For M  we used the values 1 × 10 −p , 2 × 10 −p , and 5 × 10 −p , with p ä [5, 6, 7, 8]; see also Figure 14.…”

“…For lower-density models, a systematic offset is noticed in the predicted abundances by MACE (see, e.g., left panel of Figure 19 in Appendix C). This is due to the effect of the CO self-shielding, which depends on the velocity of the outflow and can affect the abundances significantly (Maes et al 2023). However, since a MACE model receives the density as input (Equation ( 11)) and not the mass-loss rate and expansion velocity separately (contrary to the classical model), this chemical subtlety is not grasped by MACE.…”