Python codes to perform simple stellar classifications given a set of input observables. The code is based on the work described in Huber et al. 2017, Berger et al. 2020 and Berger et al. 2023 - please cite these papers when using the code. Additional references to be cited depending on the use-case:
- MESA grid: Choi et al. 2016
- PARSEC grid: Bressan et al. 2012
- Extinction map: Bovy et al. 2016 and mwdust dependencies
-
Download
mwdust(see https://github.com/jobovy/mwdust). -
pip install the code:
pip install isoclassify
-
If the install succeeded,
isoclassify -h
should output
usage: isoclassify [-h] {run,batch,multiproc,scrape-output} ... optional arguments: -h, --help show this help message and exit subcommands: {run,batch,multiproc,scrape-output} run run isoclassify -
Optional: Set an environment variable for a path to store the MESA models downloaded in step 5. Otherwise, skip this step and use the default location (
~/.isoclassify).export ISOCLASSIFY=/path/to/data/dir -
Download models into the
~/.isoclassifydirectory created upon installation or the directory specified in step 4.Bolometric correction grid:
https://drive.google.com/file/d/1vbgKqrghAWJrDJnRBLMlv--J8VSVUOpi/view?usp=sharingMESA models:
https://drive.google.com/drive/folders/1GC81YxBvMF2wu4TdL0vItvdYv-zsaLI1?usp=sharingParsec models:
https://drive.google.com/file/d/1gURKVsL5jPxWZ08ipzALXiUnX7lsjoGe/view?usp=sharing
Derives posterior distributions for stellar parameters (teff, logg, feh, rad, mass, rho, lum, age, distance, av) through direct intregration of isochrones, given any set of observables (photometry, spectroscopy, parallax, asteroseismology) and priors. Can fit for extinction or use reddening map, depending on input parameters. The Parsec grid has been amended for M dwarfs using empirical relations by Mann et al. 2019; see Berger et al. 2023 for details.
Uses bolometric corrections and extinction maps to derive stellar parameters using direct physical relations. Masses and densities are only calculated for cool stars within the empirical relations by Mann et al. 2019 (to derive masses for any star, use grid modeling mode described above). Options are:
-
RA/DEC + Photometry + (Spectroscopy) + Parallax -> Teff, R, L, distance, Av (+ M & rho for cool stars). Uses Monte-Carlo method to implement exponentially decreasing volume density prior with a length scale of 1.35 kpc (Astraatmadja & Bailer-Jones 2016)
-
RA/DEC + Asteroseismology + Spectroscopy -> logg, rho, rad, mass, lum, distance, Av.
Bolometric corrections are interpolated in (Teff, logg, FeH, Av) from the MIST grid (http://waps.cfa.harvard.edu/MIST/model_grids.html)
ra ... Right Ascension (degrees)
dec ... Declination (degrees)
band ... Photometric band to use for absolute magnitude (see nomenclature below)
dust ... Extinction map to use (none,allsky,green19)
grid ... Stellar evolution grid to use (mesa,parsec)
dmag ... Contrast for secondary star (mag). Must be in same bandpass as "band"
plx ... Parallax (arcseconds)
teff ... Effective Temperature (K)
logg ... Surface Gravity (cgs)
feh ... Metallicity (dex)
lum ... Luminosity (solar)
bmag ... Johnson B (mag)
vmag ... Johnson V (mag)
btmag ... Tycho B (mag)
vtmag ... Tycho V (mag)
gmag ... Sloan g (mag)
rmag ... Sloan r (mag)
imag ... Sloan i (mag)
zmag ... Sloan z (mag)
gamag ... Gaia G (mag)
bpmag ... Gaia Bp (mag)
rpmag ... Gaia Rp (mag)
jmag ... 2MASS J (mag)
hmag ... 2MASS H (mag)
kmag ... 2MASS Ks (mag)
numax ... Frequency of maximum power (muHz)
dnu ... Asteroseismic frequency of maximum power (muHz)
The preferred mode to run isoclassify is through a command line interface (CLI), which allows single star processing as well as batch processing of many stars.
isoclassify run <mode> <star name> --csv <csv file> --outdir <output directory> --plot <plotmode> <mode>direct or grid<star name>tells isoclassify which row of to pull the input parameters<output directory>will contain the isoclassify output files<csv file>contains as columns parameters that are passed to isoclassify<plotmode>tells isoclassify whether pop interactive plotsshow, save to a pngsave-png, or to not plot at allnone.
The file examples/example.csv includes a few example stars that cover various use cases, e.g.:
mkdir -p output/sol # make sure output directory exists
isoclassify run grid sol --csv examples/example.csv --outdir output/sol --plot showThe CLI also makes parallel processing easy. To run all stars in the example.csv file:
# generate batch scripts
isoclassify batch direct examples/example.csv -o output > isoclassify.tot
# make executable and run script
chmod +x isoclassify.tot
./isoclassify.tot
# Alternatively, run with GNU parallel to use multiple cores
parallel :::: isoclassify.tot
# Combine outputs into one CSV file
isoclassify scrape-output 'output/*/output.csv' output.csvYou can also run python-based multiprocessing through joblib and memory-mapping to reduce RAM overhead.
isoclassify multiproc <mode> <num processes> <input csv> <output csv> --baseoutdir <base output directory> --plot <plot mode><mode>direct or grid<num processes>number of processes to run at a time. -1 uses all processors, including hyperthreading for machines that have it (num cores * 2). -1 should be used with caution depending upon the # of processors and RAM in your machine, as a single process can easily consume ~4GB of RAM by itself<input csv>location and name of csv file containing all stars you want to run<output csv>location and name of csv file, scraping all results from folders in<base output directory>location of individual stellar output folders (default = ./)<plot mode>tells isoclassify whether to save to a pngsave-png(default), save to a pdfsave-pdf, or to not plot at allnone. Does not useshowas above due to issues with multiprocessing functionality.
This will run all stars within the designated csv file and produce an output csv file, so there no need to use the scrape-output function detailed above in combination with this script.
Note: isoclassify is not natively setup to run in ipython notebooks or be imported with other scripts. The example directory includes notebooks constructed by hacking pre-defined functions that allows direct interaction with posteriors. The notebooks are provided for guidance only and are not supported with the latest version of the code.
In the Gaia era, a typical use-case for isoclassify is the derivation of mass, radius, luminosity, density and age for a single star with existing constraints on Teff, log(g) metallicity from spectroscopy, a parallax from Gaia, and photometry from various surveys (Gaia, 2MASS, etc). The recommended procedure is to first use direct-mode to determine the luminosity directly through the bolometric flux and distance. Using pi Men in the example.csv file, this would look like:
# make sure output directory exists
mkdir -p output/piMendirect
isoclassify run direct piMendirect --csv examples/example.csv --outdir output/piMendirect --plot showHere, the bolometric flux is estimated through 2MASS K-band (less sensitive to extinction) and the bolometric correction grid.
Using the luminosity from the direct method as an input to grid-mode, together with the spectroscopic Teff and metallicity, we can then derive the remaining parameters (log(g) is not needed as input, since spectroscopic constraints are less reliable than the evolutionary state constraint from the Gaia luminosity):
# make sure output directory exists
mkdir -p output/piMengrid
isoclassify run grid piMengrid --csv examples/example.csv --outdir output/piMengrid --plot showNote that the uncertainties reported by isoclassify are precisions intrinsic to the grid that is used. When comparing these values to other methods, systematic errors should be considered. These can be estimated by running the same star through both grids (mesa and parsec) or following the guidance described in Tayar et al. 2022.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent