Dynamo

Towards realistic stellar obserables

Dynamo is a Python package for simulating stellar light curves and spectra, combining detailed models of stellar evolution, surface activity, planetary transits, and noise. It is based on several exisiting open-source projects that were modified and combined to create stronger link between stellar evolution and stellar observables.

Features

• Stellar Surface Simulation: Models the emergence and evolution of star spots and active regions using configurable activity cycles, latitude distributions, and spot evolution laws. This part is based on the Butterpy package.
• Grid-based lightcurve and spectra: Generates synthetic spectra and light curves, including the effects of limb darkening, rotational broadening and spot evolution. This is done in a grid approach and motivated by the Starsim package.
• Stellar Evolution Parameters: Interpolates stellar parameters (e.g., mass, radius, temperature, luminosity) from evolutionary models. To get stellar evolution parameters we use kiauhoku package.

Using Dynamo we can easily link between stellar evolution parameters (Temperature, mass, logg, etc.) and magnetic-driven spots. Such connection enable the creation of simulated light curve and spectra of the same star, which can serve as a simulated dataset for multi-modal machine learning models for example.

Installation

1. Clone the repository:

   git clone https://github.com/IlayMalinyak/Dynamo.git
   cd Dynamo
   

2. Install dependencies:

   pip install -r requirements.txt
   

Usage

Simulating a Star

You can configure and run a stellar simulation by editing the starsim.conf file and using the main classes in the Dynamo package. Example:

from Dynamo.star import Star

# Initialize a star from configuration
star = Star(conf_file_path='starsim.conf')

# Set up simulation times and wavelength grid
import numpy as np
times = np.linspace(0, 30, 1000)  # 30 days, 1000 points
wavelengths = np.linspace(4000, 9500, 1000)  # Angstroms

# generate spots map
sm.generate_spot_map(ndays=1000)
# Run the forward simulation
star.compute_forward(t=times, wv_array=wavelengths)

# Access results (e.g., light curve, spectra)
# star.results, star.spot_map, etc.

Example: Generating a Dataset of Light Curves and Spectra

You can generate a full dataset of physically-consistent light curves and spectra using the script dataset/create_data.py. This script samples stellar parameters, simulates surface activity, and produces time series and spectra for each synthetic star. Note-in dataset/create_data.py we assume some priors that might be non-trivial (for example age-activity and spot-activity relationships). You might want to edit them for your use case

Command-Line Usage

python dataset/create_data.py \
    --models_root /path/to/stellar/models \
    --dataset_dir dataset \
    --plot_dir images \
    --plot_every 100 \
    --num_simulations 1000 \
    --ndays 1000 \
    --n_cpu 4

this would create 1000 simulations of 1000 days each. The arguments are:

• --models_root: Path to the directory containing stellar evolution models.
• --dataset_dir: Output directory for the generated dataset (default: dataset).
• --plot_dir: Directory to save diagnostic plots (default: images).
• --plot_every: Save a plot every N simulations (default: 100).
• --num_simulations: Number of stars (light curves/spectra) to generate (default: 1000).
• --ndays: Number of days to simulate for each light curve (default: 1000).
• --n_cpu: Number of CPU cores to use for parallel processing (default: 1).

Configuration

Simulation parameters are controlled via the starsim.conf file, which includes sections for:

• Stellar properties (mass, radius, temperature, rotation, etc.)
• Spot and faculae properties
• Limb darkening
• Planetary parameters
• Noise characteristics

See the comments in starsim.conf for detailed descriptions of each parameter.

License

This project is licensed under the MIT License. See LICENSE for details.

Acknowledgments

• PHOENIX stellar atmosphere models
• Butterpy
• Starsim
• kiauhoku

Contact

ilay.kamai@campus.technion.ac.il