CoolStar model repository

This server stores a few stellar models that were calculated at University of California San Diego for the purposes of CoolStar lab led by Adam Burgasser. While the models are stored here primarily for our own convenience, we also provide a web interface (this site) to share them with the community. Despite investing our best efforts into the accuracy of the content, we are aware of a few inconsistencies that are yet to be resolved. As such, please feel free to join the discussion if you have an application for our models or means to test their accuracy.

The rest of this page will briefly summarize the software we used in our calculations and list a few references. The models themselves are available through the main menu.

ATLAS atmospheres

ATLAS is one of the earliest atmosphere modelling codes introduced in 1960s by Robert Kurucz for LTE, plane-parallel atmospheres. A detailed description of the package can be found in SAO Special Report #309. ATLAS has seen significant improvements throughout the years by both Robert Kurucz and a number of other contributors. At the moment of writing, two versions of ATLAS are actively used: ATLAS9 and ATLAS12. Both versions in their latest installments include the following features:

While originally written in Fortran IV, most of the ATLAS code has been ported to modern releases of GNU Fortran (see details in Sbordone+2004) and can be straightforwardly compiled and used on most Linux and Windows setups. The source code of both ATLAS9 and ATLAS12 as well as satellite packages can be downloaded from Fiorella Castelli's website. The website also contains download links for necessary databases and example run scripts.

Both ATLAS9 and ATLAS12 work with low-resolution spectra on every iteration, which are assumed sufficient to compute the atmospheric structure. The high-resolution output spectrum cannot be computed by either version directly and is, instead, handled by a third program called SYNTHE (also distributed under the link above). ATLAS12 uses the so-called opacity sampling method, whereby the spectrum is sampled during the calculation at every optical depth in the atmosphere. ATLAS9, instead, relies on opacity distribution functions (ODFs) that can be computed beforehand and require the assumption of constant spectrum throughout the atmosphere. A fourth software package called DFSYNTHE is typically used to compute the ODFs for any given set of abundances. While inevitably introducing some hopefully small physical errors, ATLAS9 has a tremendous advantage over ATLAS12 in speed, which is crucial for extensive grids. Further details on SYNTHE, ATLAS9 and ATLAS12 can be found in Kurucz 2014.

For CoolStar purposes, only ATLAS9 was used alongside SYNTHE and DFSYNTHE as necessary. On our setup, all three packages are dispatched by a single publically available (GitHub) Python-based pipeline. While the pipeline has been successfully tested on both Linux and Windows machines, users of the latter may experience problems with some of the source code that has not been fully ported to GNU Fortran. Intel Fortran compilers serve as a suitable alternative and their Windows versions are distributed as a part of the Intel Parallel Studio.

It must be noted that both ATLAS9 and ATLAS12 experience convergence issues at temperatures below 4000 K and fail almost every time below 3500 K, which is primarily due to incomplete or absent treatment of complex molecules. At colder temperatures, other codes, such as PHOENIX, must be used.

PHOENIX atmospheres

PHOENIX is a much more recent and incredibly versatile stellar modelling code, presented in Hauschildt+1996. The lead developer of the project is Peter H. Hauschildt and the rest of the development team are credited on the project website. Our installation of PHOENIX was configured by Derek Homeier - one of the members of the PHOENIX collaboration - who kindly shared the settings on his setup and provided assistance in troubleshooting issues. Below we provide a brief description of this setup.

Contrary to ATLAS9, PHOENIX performs direct opacity sampling, i.e. opacities are calculated at all wavelengths in every layer on every iteration. Calculating the final high-resolution emergent spectrum is programmatically indistinguishable from a routine opacity sampling during a regular iteration, which removes the need for a separate spectrum synthesis software such as SYNTHE. All calculations are carried out in spherical geometry (as opposed to the plane-parallel simplification made by ATLAS) with depth-dependent gravity. The radius of the star is estimated from published isochrones.

PHOENIX is able to simulate chemical processes throughout the atmosphere including molecular formation, ionization and condensation of dust particles. Over 500 species are considered based on thermochemical data from JANAF as well as other sources. The results are expressed in terms of partial pressures of every species. We use CSPPRESS - a satellite code from the PHOENIX suite - to pre-tabulate partial pressures at a range of overall gas pressures and temperatures. The tables produced by CSPPRESS can then be used by PHOENIX on every iteration to reconstruct the equation of state in all layers of the atmosphere.

PHOENIX features extensive databases of atomic and molecular lines compiled from various sources. After the first iteration, the databases are searched for lines that may have a non-negligible effect on the radiation field. All lines that meet such threshold are then considered when sampling opacities in each layer on every iteration. Depending on the strength of each line, Gaussian or Voigt opacity profiles may be used. The turbulent velocity characterizing the profiles is estimated dynamically using the relationship from Gebran+2004. NLTE treatment is available for selected atomic species. PHOENIX uses precomputed collisional and radiative cross-sections to set up the NLTE rates equations which are subsequently solved using the accelerated lambda iteration technique.

Convective energy transfer is modelled using the Böhm-Vitense mixing length theory. The mixing length is determined dynamically using the data from hydrodynamic simulations in Ludwig+1999. Convection-induced adiabatic gradients in temperature are allowed for selected atomic and molecular species. Non-equilibrium chemistry due to rapid convective transport is modelled for certain molecules (e.g. see Visscher+2010).

Condensation of dust clouds is modelled for a number of liquid and crystalline constituents, including multiple orientations for birefringent crystals such as corundum. Our setup uses both equilibrium and settling cloud models depending on the effective temperature of the atmosphere. In the former case, the clouds are assumed to remain at the depth where they originally formed. In the latter case, more appropriate at cooler temperatures, gravitational settling of dust is modelled by calculating the expected fractions of dust species to sediment out of every layer of the atmosphere. In both cases, dust precipitation depletes parent gaseous species, reducing their opacities, and contributes additional opacity of its own, unless it "rains out" and settles below the spectrum-forming region of the atmosphere. Dust grains are assumed to follow a log-normal distribution of size with no porosity. Dust clouds are assumed to cover the entire surface area of the star.