Maja Jabłońska
SPICE: Differentiable Synthetic Spectra Modelling for Stars with Surface Inhomogeneities
May 12, 2025
When: May 12, 2025 3:30PM
Where:
LCO Downstairs Conference Room
Maja Jabłońska
Research School of Astronomy & Astrophysics, Mt. Stromlo Observatory
In this talk, I will present SPICE, a JAX-based spectroscopic simulation framework designed for efficient and fast generation of synthetic stellar spectra for stars with surface inhomogeneities. SPICE incorporates a transformer-based machine learning model for synthesizing spectra across arbitrary wavelength ranges and parameter spaces, significantly reducing computational costs while preserving physical accuracy.
The framework features a 3D stellar mesh modelling system that captures complex surface phenomena, including stellar spots, temperature and abundance gradients, pulsations, and binary occultation effects. This representation enables the studies of spectroscopic and photometric signatures of stellar activity and binarity.
SPICE emphasizes flexibility, allowing researchers to incorporate custom spectral grids, implement additional physical effects, and define surface properties according to specific research requirements.
The framework's differentiability paves the way for inference with gradient-based methods and facilitates the development of machine-learning models for parameter inference across various stellar phenomena.
