BOSSBrain

Introduction

|Doppler| [#f1]_ is a general-purpose stellar radial velocity determination software. It uses a forward-modeling approach, convolving a model spectrum to the resolution or Line Spread Function (LSF) of the observed spectrum. |Doppler| can be used with a high-resolution model of the The Cannon (Casey et al. (2016)) and also of The Payne (Ting et al. (2019)), both machine-learning approaches to modeling stellar spectra. Doppler can determine the radial velocity (RV) and stellar parameters for a spectrum of any wavelength (3000-18000A) and resolution (R<20,000 at the blue end and 120,000 at the red end) with minimal setup.

The current set of three Cannon models cover temperatures of 3,500K to 60,000K with 3-parameter (Teff, logg, [Fe/H]) and radial velocity. The current Payne model covers temperatures of 3,500K to 6,000K with 33 labels (Teff, logg, Vmicro, [C/H], [N/H], [O/H], [Na/H], [Mg/H], [Al/H], [Si/H], [P/H], [S/H], [K/H], [Ca/H], [Ti/H], [V/H], [Cr/H], [Mn/H], [Fe/H], [Co/H], [Ni/H], [Cu/H], [Ni/H], [Cu/H], [Ge/H], [Ce/H], [Nd/H], [Ba/H], [Eu/H], [La/H], [Y/H], [Sc/H], [Zr/H], [Pr/H], [Yb/H]) as well as radial velocity, rotational velocity and macrotubulence.

|Doppler| also has the ability to simultaneously fit (“jointfit”) multiple spectra of a star, with a single set of stellar parameters and elemental abundances and separate radial velocities for each spectrum.

Description

|Doppler| fits spectra using a multi-step approach to zero-in on the best solution.

The default, multi-step approach using the Cannon is:

  1. Get initial RV using cross-correlation with rough sampling of Teff/logg/[Fe/H] parameter space.

  2. Get improved Cannon stellar parameters using initial RV.

  3. Improved RV using better Cannon template.

  4. Improved Cannon stellar parameters.

  5. Full least-squares fitting of all stellar parameters and RV.

  6. Run fine-grid in RV using template from previous step

  7. Run MCMC (if requested).

The approach with the Payne is:

  1. Get initial RV using cross-correlation with rough sampling of Teff/logg/[Fe/H]/[alpha/Fe] parameter space.

  2. Least-squares fitting of all desired Payne labels and RV, using best-fit of previous step as initial guess.

  3. Run fine-grid in RV using best-fit template from previous step.

  4. Run MCMC (if requested).

When jointfit is used,

  1. Run regular |Doppler| fit on each spectrum separately.

  2. Find weighted mean of all labels and Vhelio.

  3. Fit all spectra simultaneously determining one set of labels and a separate RV for each spectrum.

BOSSBrain can be called from python directly or the command-line script bossbrain can be used.

Examples

Index