Dartmouth Stellar Evolution Emulator (DSEE) 1: Generative Stellar Evolution Model Database

TL;DR

The Dartmouth Stellar Evolution Emulator (DSEE) is a fast, probabilistic, generative model trained on extensive stellar evolution data, enabling high-fidelity, uncertainty-aware stellar modeling and analysis.

Contribution

DSEE introduces a flow-based generative model for stellar evolution, unifying track and isochrone construction with continuous physics interpolation and uncertainty quantification.

Findings

DSEE achieves high fidelity in HR diagram validation.

It provides orders-of-magnitude speedups over direct models.

Integrated into CONF1DENCE, it enables uncertainty-aware stellar analysis.

Abstract

We present the Dartmouth Stellar Evolution Emulator (DSEE), a flow-based stellar evolution model emulator trained on a comprehensive database comprising over eight million evolutionary tracks that vary across twenty input-physics dimensions and span broad ranges in mass and composition. DSEE learns phase-conditioned stellar state snapshots, unifying track and isochrone construction as marginals of one generative model. It delivers continuous interpolation across high-dimensional physics, probabilistic predictions with calibrated credible intervals, and orders-of-magnitude speedups over direct modeling. Validation against current stellar evolution models shows high fidelity across the HR diagrams, while distributional tests recover the full distributions obtained from brute-force Monte Carlo sampling. To broaden impact, DSEE is integrated into the open-source CONF1DENCE package, enabling fast, end-to-end creation of stellar tracks and isochrones. CONF1DENCE includes the ability to make uncertainty-aware age determinations for clusters taking into account observational effects. CONF1DENCE replaces bespoke, fixed-physics grids with a generative, physics-marginalized emulator, setting a practical new standard for stellar modeling and enabling survey-scale analyses with rigorous uncertainty.