SAPP is a fully automated python code which culminates different types of observation data (photometry, parallax, asteroseismology and spectroscopy) in order to determine fundamental and atmospheric parameters.
This pipeline can be used in two different modes: Full or Lite
- Full: Determines parameters from a bayesian framework which calculates and combines 3D probability distribtutions.
- Lite: Determines parameters from the Spectroscopy module combined with asteroseismic/surface gravity data
SAPP is developed with Python 3.7
SAPP requires the following python packages:
- numpy 1.18.5
- scipy 1.5.0
- astropy 3.2.3
- matplotlib 3.2.2
The current README needs to be updated inline with the newest commits - M. Gent
There are 3 steps for installation, assuming the above packages have been installed:
-
Clone the repository in the directory you want.
-
Download stellar evolution models via https://keeper.mpdl.mpg.de/d/5162ad5abd954ae8a457/ into directory:
SAPP-WEAVE/Input_data/GARSTEC_stellar_evolution_models/evo_tracksThere are two memory maps listed in the keeper photometry uses photometry_stellar_track_mmap_v2.npy and asteroseismology uses photometry_stellar_track_mmap.npy in the PLATO setup.
Please read the paper at:
https://www.aanda.org/articles/aa/pdf/2022/02/aa40863-21.pdf
You must absolutely have stellar evolution models memory map within the directory
SAPP-WEAVE/Input_data/GARSTEC_stellar_evolution_models/evo_tracksCurrently the code can only run evo tracks tailored to GARSTEC stellar evolution models.
Due to GitHub's file number upload limit, the memory maps are stored on a private keeper (see Installation) or otherwise contact me.
There are four folders which concern Output data:
- Stars_Lhood_combined_spec_phot
Spectral and photometric PDF grids, statistics as well as best fit parameters are saved in this directory. Each star/field has a given folder.
- Stars_Lhood_photometry
Photometric and astrometric PDF grids are saved in this directory. Each star/field has a given folder.
- Stars_Lhood_spectroscopy
Spectroscopy best fit parameters, associated uncertainties and covariance matrix are saved in this directory. Each star/field has a given folder.
- Stars_Lhood_asteroseismology
Asteroseismic PDF grids/Logg values are saved in this directory. Each star/field has a given folder.
- test_spectra_emasks
Spectra error masks which have been created to be used in this pipeline.
There are 3 stellar folders representing typical FGK stars in a given temperature and close to solar metallicity.
These are $\delta$ Eri (low temperature), the Sun (median temperature), and Procyon (high temperature).
This contains all the scripts necessary to run the pipeline as well as extra.
-
main.py: is the main script used to run the pipeline in different ways, it should be the only script you will have to edit besides SAPP_best_spec_payne_v1p1.py in the directory SAPP_spectroscopy.
-
SAPP_spectroscopy: Spectroscopic module with all the tools to process spectra such as RV correction, convolving, cleaning, continuum normalisation as well as calculating best fit values.
-
SAPP_stellar_evolution_scripts: This contains any script used to calculate the photometry and asteroseismology grids. These are currently calculated using the same module.
-
SAPP_tools: This folder contains functions which are commonly used throughout the pipelines use and therefore exists for easy access. This also includes querying script for extinction values.
The script in directory
Source_scripts/SAPP_stellar_evolution_scripts/photometry_asteroseismology_lhood_TEST.pyCalculates the photometric grid. The input data required is located in directory
Input_data/photometry_asteroseismology_observation_data/PLATO_benchmark_stars/PLATO_bmk_phot_data/PLATO_photometry.csvThe header is defined in the later section Bayesian Framework. Essentially there are 8 photometric bands in total, 5 non-Gaia i.e. B, V (Johnsons), H, J, Ks (2MASS)
and 3 Gaia i.e. G, Bp, Rp. Included with these are parallaxes from Gaia DR2 (now DR3 is released) and Simbad with associated errors, magnitude extinction from Gaia (if it exists),
and reddening, E(B-V), queried from the tool Stilism. All of which are used to calculate the photometric chi2
grids for photometry.
There are four grids which can be calculated; Magnitude (Gaia and Non-Gaia); Colour (Gaia and Non-Gaia).
For the colour grid, only two colours are used, these are V-K and Bp-Rp, this is to ensure the independence of Gaussian factors i.e. zero covariance is assumed.
Each grid is calculated by running through stellar evolution models, calculating a chi2 value via the equation
$\chi^2$ = $\sum_i \frac{(X_i-\mu)^2}{\sigma^2}$
where i represents the model points,
N.B. Photometric uncertainties are calculated by combining the band uncertainty, distance modulus uncertainty (propagated from parallax), and uncertainty in extinction (propagated from reddening). If the parallax measurement has no value i.e. is a NaN, the fiducial value is 20$%$. If the magnitude measurement is a NaN, the fiducial value is 0.075 magnitude. If the reddening uncertainty measurement is a NaN, the fiducial is 15$%$ of E(B-V). It is unlikely that the values will be NaN, however the option is there incase there are none.
The output of the photometry module is as follows:
[Teff,logg,[Fe/H],non_gaia_mag,gaia_mag,non_gaia_col,gaia_col,Asteroseismolog,Age,Mass,Radius,Age_step,Luminosity]
The units are:
K, dex, dex, --, --, --, --, --, Gyrs, Msol, Rsol, Gyrs, Lsol
The grid is a 13 column array of parameters and chi2 values.
The Asteroseismic module is similar in structure and the core process which creates the grid of chi2 values is currently the same script for the photometric values.
Calculates the asteroseismic grid. The input data required is located in directory
Input_data/photometry_asteroseismology_observation_data/PLATO_benchmark_stars/PLATO_bmk_phot_data/Seismology_calculation/PLATO_stars_seism.txtThe header is defined in the later section Bayesian Framework. There are two
asteroseismic quantities that the module uses, these are
There is only one grid which is calculated which combined these two quantities with associated errors.
Each grid is calculated by running through stellar evolution models, calculating a chi2 value via the same equation as in photometric module.
N.B. Asteroseismic uncertainties are calculated by combining the input variables uncertainty and the corresponding solar value uncertainty.
This is because the solar asteroseismic
The output of the asteroseismic module is as follows:
[Teff,logg,[Fe/H],non_gaia_mag,gaia_mag,non_gaia_col,gaia_col,Asteroseismolog,Age,Mass,Radius,Age_step,Luminosity]
The units are:
K, dex, dex, --, --, --, --, --, Gyrs, Msol, Rsol, Gyrs, Lsol
The grid is a 13 column array of parameters and chi2 values.
The central script required to perform spectroscopy is called SAPP_best_spec_payne_v1p1.py.
This is located in the directory:
Source_scripts/SAPP_spectroscopy/The best way to explain how to use spectroscopy is with an example!
First, you need to store your observation spectra in the directory
Input_data/spectroscopy_observation_data/You can create your own directory to store one or a list of spectra.
Second, you need a Neural Network trained grid of model spectra made by
The Payne. The training script is currently not in SAPP's architecture, however
we already have a trained file already loaded. This is called NN_results_RrelsigN20.npz
and it is located in the directory
Input_data/spectroscopy_model_data/Payne_input_data/This python binary file contains all you need in order to produce a model spectra
in the Gaia ESO Giraffe hr10 format with R=20,000 and wavelength range
5329 $\AA$ to 5616 $\AA$.
Specifically it contains the weights and bias parameters required for the Neural Network as well as the limits of the grid it was trained on.
This training grid consists of an 8 Dimensional parameter space with the following variables
Teff,logg,[Fe/H],Vmic,Vbrd,[Mg/Fe],[Ti/Fe],[Mn/Fe]
Effective temperature, surface gravity, iron metallicity, microturbulence, velocity broadening (this is effectively vsini), magnesium, titanium and manganese metallicity.
There is only two functions you have to have to worry about:
- find_best_val()
- "read_spectra_func()"
find_best_val() processes the spectra in several ways, in short it can continuum normalise, calculate radial velocity correction, convolve the observation to a lower resolution and match the wavelength scale to the grid of models.
"read_spectra_func()" is in quote marks as this function can vary in structure, The user should tailor this function specifically the spectra they would like to analyse.
For example, already in the script there are 3 functions to read the following spectra:
- Gaia ESO Giraffe fits files,
read_fits_GES_GIR_spectra - Generic text files,
read_txt_spectra - HARPS/UVES fits files
read_fits_spectra
These all follow similar input which is the variable spec_path
and output wavelength,flux,error_med,rv_shift,rv_shift_err,snr_star .
At the very least these outputs should exist, it is possible for rv_shift and
rv_shift_err to be NaN values. This is because sometimes the RV correction is
not provided. However the SNR of the star must be known, even a 1st order
approximation is good enough (For HARPS/UVES you can set the SNR to be 300/200)
respectively.
The input arguments for find_best_val() are described within the doc string of the function. Below we will go through what all these inputs mean and learn what you can do with this module
error_mask_index = 0
emask_kw_instrument = "HARPS" # can be "HARPS" or "UVES"
emask_kw_teff = "solar" # can be "solar","teff_varying"
error_mask_recreate_bool = False # if this is set to True, then emask_kw_teff defaults to "stellar"
error_mask_recreate_arr = [error_mask_recreate_bool,emask_kw_instrument,emask_kw_teff]
error_map_use_bool = True
cont_norm_bool = False
rv_shift_recalc = [False,-100,100,0.05]
conv_instrument_bool = False
input_spec_resolution = 18000
numax_iter_bool = False
niter_numax_MAX = 5
numax_input_arr = [3170,159,niter_numax_MAX]
recalc_metals_bool = False
feh_recalc_fix_bool = False
recalc_metals_inp = [5770,4.44,0,feh_recalc_fix_bool]
ind_spec_arr = [[spec_path,\
error_map_spec_path,\
error_mask_index,\
error_mask_recreate_arr,\
error_map_use_bool,\
cont_norm_bool,\
rv_shift_recalc,\
conv_instrument_bool,\
input_spec_resolution,\
numax_iter_bool,\
numax_input_arr,\
recalc_metals_bool,\
recalc_metals_inp,\
logg_fix_bool,\
logg_fix_input_arr],stellar_names[inp_index],spec_type]
find_best_val(ind_spec_arr)The spectra chosen is a 18 sco HARPS spectra which has been continuum normalised and convolved from R = 118,000 to 20,000 already. Hence why
cont_norm_bool = False
conv_instrument_bool = FalseIf you would like to test out these processes, there is a unprocessed HARPS
spectra of 18 sco in the spectroscopy_observation_data directory.
The python code above is the simplest case of processing, finding the best fit parameters of a spectra using an error mask.
An error mask is a way of introducing model uncertainties to the fitting routine. Grab the cleanest spectra you have of the star, set the reference parameters and you will create a mask of residuals representing how different the best possible model is from a real observation.
For this example you are giving error_map_spec_path the same path as spec_path,
this is because with only one spectra this will have to be your cleanest one.
The variable error_mask_index refers to a list of reference parameters
located in directory:
Input_data/Reference_data/PLATO_stars_lit_params.txtHere we have already made a list of reference parameters for our PLATO benchmark stars.
The first star in the list is 18 sco, hence why error_mask_index = 0 in this
example. If you would like to change this then you just have to edit this parameter list.
However, if you are not sure what the reference parameters are for the star (likely for a big survey),
then you can still use the error mask, just set error_mask_recreate_bool = False. This will then check
to see what temperature the given star is closest to of our three error mask candidates:
Note, the instrument must be specified, the current options are HARPS or UVES, this is because as well as model uncertainties, the instrument itself provides uncertainties of measurement, thus an error mask has been created for each instrument in each temperature range.
emask_kw_instrument = "HARPS" # can be "HARPS" or "UVES"
emask_kw_teff = "solar" # can be "solar","teff_varying"
error_mask_recreate_bool = False # if this is set to True, then emask_kw_teff defaults to "stellar"
error_mask_recreate_arr = [error_mask_recreate_bool,emask_kw_instrument,emask_kw_teff]Like above, you choose the instrument and whether the emask is just solar or Teff varying. If you
set error_mask_recreate_bool = True however, the stellar emask will be assumed i.e. a mask
with all of the reference parameters assumed from the Input_data directory.
We reconmend if you are working with Solar-type stars to use the instrument HARPS.
This fits file did not come with a radial velocity value, therefore we must calculate it. The "read_spectra function" will give rv_shift as a NaN value, the code will recognise this and start the procedure for RV correction.
Our RV correction function is called rv_cross_corelation_no_fft and derives
from PyAstronomy's RV Cross Correlation function with some minor changes.
To perfom RV correction, you will require at least one "template" to compare your observations to.
If your star is a subgiant/dwarf Main-Sequence star similar to our Sun then having a solar template will be good enough. We have already provided a spectral template in the directory
Input_data/spectroscopy_model_data/Sun_model_R5e5_TS_hr10/We produced a R ~ 500,000 solar model using TurboSpectrum with the hr10
wavelength coverage. Due to the high sampling, this had to be done in sections
and therefore we created a function stitch_regions to stitch the model together.
Now all the RV correction function requires is for you to specify the intended Resolution, RV limits and a step size.
The parameter input_spec_resolution is where you define the resolution, it
should match the spectral resolution of the trained grid.
The array rv_shift_recalc exists to allow you to decided whether you would
like to re-calculate the rv correction or not, as some files do provide this
information, you may want to re-calculate it. The first parameter in the array
you will set to True or False. The rest are rv_min, rv_max, and drv i.e. the limits
for the RV correction calculation and step size (in Km/s).
This of course will affect the computation time however I normally pick between -100 and 100 Km/s for the limits with 0.05 Km/s as the step size. If you would like to RV correct the spectra as a pre-process, the RV correction function can be found in the directory:
Source_scripts/SAPP_tools/Running this code in the the most basic format will produce something like the following output
SNR 396
BEST fit parameters = [ 5.77216 4.37047 0.02535 1.04010 4.12242 0.02084 0.00929 -0.03453]
parameters ERROR upper = [ 0.00630 0.00842 0.00400 0.01466 0.12393 0.00561 0.00711 0.00615]
parameters ERROR lower = [ 0.00630 0.00842 0.00400 0.01466 0.12393 0.00561 0.00711 0.00615]The SNR of the spectra, then the best fit parameters alongside the respective errors. NOTE that the order of the parameters follows the order described above with effective temperature, surface gravity and [Fe/H] as the first three parameters. It was found that fitting Teff in units of K/1000 was easier for the training, hence why you see 5.77216 and not 5772.16.
Already we can see that the temperature estimation is good but the logg
estimation could be better. This is where the numax_iter_bool variable
comes into play.
Our Lite mode is the use of Spectroscopy with Asteroseismic data. This is how
we do it, you must set numax_iter_bool = True and fill out the
parameters niter_numax_MAX and numax_input_arr. For 18 Sco we
have already filled out the information for you.
The parameter niter_numax_MAX defines the maximum number of iterations
for the logg refinement. 5 is good enough for most iterations and this goes into
the parameter numax_input_arr where the first two elements are
the $\Delta\nu_{max}$ alongside its respective uncertainty.
The process itself is quite simple and can be understood from the code. In short
the code first makes its best guess of all the parameters, which will result in
the exact same output as above. It then uses the effective temperature combined
with $\Delta\nu_{max}$ to calculate a new logg value via the scaling
relationship below:
$\nu_{max}$ = $\nu_\odot\frac{g}{g_\odot}\left\sqrt{\frac{T_{eff,\odot}}{T_{eff}}]\right}$
This logg estimate is then fixed while all the other parameters are re-calculated, this gives a new effective temperature and so a new logg value can be calculated. This iterative process runs for as many loops as the user gives it.
Once the process is finished, the code finds where the process has a temperature value oscillating around some central value within 10 K. The process would never fully converge, however once we get to the oscillating region we consider it finished.
The last way to use the code directly relates to how it is used in the Full mode which uses the Bayesian framework.
This is concerning the booleans recalc_metals_bool and feh_recalc_fix_bool.
The Full mode only explores the first 3 dimensions of the spectroscopy PDF while keeping the other 5 parameters at their best fit solution. Once we combine the different PDFs alongside priors to create an overall posterior the rest of the parameters need to be recalculated based on the new best fit solution.
Therefore recalc_metals_bool = True will tell the code to look at the
parameters given in the array recalc_metals_inp and fix the first two/three
parameters while re-calculating the rest. The effective temperature and logg
are always fixed, however I give the the user the choice to not fix [Fe/H] as
this is an abundance and you might want to self-consistently re-calculate them all.
The inputs we have chosen are the Solar values, as 18 sco is known as the Solar
twin. NOTE by setting recalc_metals_bool = True this will negate the nu_max
iterative process since both Teff and logg are being fixed.
The output of the code in this mode with feh_recalc_fix_bool = False is shown below:
SNR 396
BEST fit parameters = [ 5.77030 4.43961 0.01079 1.05678 4.27973 -0.00909 0.02424 -0.02365]
parameters ERROR upper = [ 0.00637 0.00828 0.00420 0.01495 0.12486 0.00554 0.00713 0.00617]
parameters ERROR lower = [ 0.00637 0.00828 0.00420 0.01495 0.12486 0.00554 0.00713 0.00617]And what follows is if feh_recalc_fix_bool = True:
SNR 396
BEST fit parameters = [ 5.77030 4.43961 0.00038 1.09698 4.12245 0.00072 0.03523 -0.00777]
parameters ERROR upper = [ 0.00645 0.00843 0.00419 0.01484 0.12359 0.00551 0.00723 0.00618]
parameters ERROR lower = [ 0.00645 0.00843 0.00419 0.01484 0.12359 0.00551 0.00723 0.00618]If you would like to run the FULL bayesian scheme then you should look at the script located in the following directory:
Source_scripts/main.pyInside this script you will find several functions, of which only one concerns
you with respect to input, main_func_multi_obs().
Firstly, you need to understand the inputs for the full mode, these are split up into 3 sections
- Photometry data
- Asteroseismology data
- Spectroscopy data
These are loaded in from the Input_data directory and the headers are defined
within the main_FULL.py script.
You should see something like this
Input_phot_data = np.loadtxt('../Input_data/photometry_asteroseismology_observation_data/PLATO_benchmark_stars/PLATO_bmk_phot_data/PLATO_photometry.csv',delimiter = ',',dtype=str)
"""
0 ; source_id (Literature name of the star)
1 ; G
2 ; eG
3 ; Bp
4 ; eBp
5 ; Rp
6 ; eRp
7 ; Gaia Parallax (mas)
8 ; eGaia Parallax (mas)
9 ; AG [mag], Gaia Extinction
10 ; Gaia DR2 ids
11 ; B (Johnson filter) [mag]
12 ; eB (Johnson filter) [mag]
13 ; V (Johnson filter) [mag]
14 ; eV (Johnson filter) [mag]
15 ; H (2MASS filter) [mag]
16 ; eH (2MASS filter) [mag]
17 ; J (2MASS filter) [mag]
18 ; eJ (2MASS filter) [mag]
19 ; K (2MASS filter) [mag]
20 ; eK (2MASS filter) [mag]
21 ; SIMBAD Parallax (mas)
22 ; eSIMBAD Parallax (mas)
23 ; E(B-V), reddening value from stilism
24 ; E(B-V)_upp, upper reddening uncertainty from stilism
25 ; E(B-V)_low, lower reddening uncertainty from stilism
"""
Input_ast_data = np.loadtxt('../Input_data/photometry_asteroseismology_observation_data/PLATO_benchmark_stars/Seismology_calculation/PLATO_stars_seism_copy.txt',delimiter = ',',dtype=str)
"""
0 ; source_id (Literature name of the star)
1 ; nu_max, maximum frequency [microHz]
2 ; err nu_max pos, upper uncertainty
3 ; err nu_max neg, lower uncertainty
4 ; d_nu, large separation frequency [microHz]
5 ; err d_nu pos, upper uncertainty [microHz]
6 ; err d_nu neg, lower uncertainty [microHz]
"""
Input_spec_data = np.loadtxt('../Input_data/spectroscopy_observation_data/spectra_list.txt',delimiter=',',dtype=str)
"""
0 ; source_id (Literature name of the star)
1 ; spectra type (e.g. HARPS or UVES_cont_norm_convolved)
2 ; spectra path
"""The observation data needs to match these headers EXACTLY.
Once the input data is loaded you need to make some decisons based on how the spectra are processed (just like in the spectroscopy section).
emask_kw_instrument = "HARPS" # can be "HARPS" or "UVES"
emask_kw_teff = "solar" # can be "solar","teff_varying"
error_mask_recreate_bool = False # if this is set to True, then emask_kw_teff defaults to "stellar"
error_mask_recreate_arr = [error_mask_recreate_bool,emask_kw_instrument,emask_kw_teff]
error_map_use_bool = True
cont_norm_bool = False
rv_shift_recalc = [False,-100,100,0.05]
conv_instrument_bool = False
input_spec_resolution = 20000
numax_iter_bool = False
niter_numax_MAX = 5
recalc_metals_bool = False
feh_recalc_fix_bool = False
logg_fix_bool = False
spec_fix_values = [5770,4.44,0]To understand these settings better, read the spectroscopy documentation above.
The only input for main_func_multi_obs() is inp_index, which refers to the
index of a list of stars organised by your input data.
NOTE: all input data including reference data should be in the same order (by rows) with the exact same identifier, its how cross correlation works.
After deciding the spectra treatment, you decide on the limits and central values of the core parameter space, Teff, logg and [Fe/H]. This will dictate how much of the parameter space you want to explore which will be later used to create the spectroscopy grid.
The default limits are
Teff_resolution = 250 # K
logg_resolution = 0.3 # dex
feh_resolution = 0.3 # dexAnd the central values are decided in two ways
- phot_ast_central_values_bool = True, means that the best fit values calculated from spectroscopy will be used as a good idea/place to start.
- phot_ast_central_values_bool = False, means that you decide where the centre of the grid will be
phot_ast_central_values_bool = False # if set to True, values below will be used
phot_ast_central_values_arr = [5900,4,0] # these values are at the centre of the photometry gridThe next set of inputs concerns spectroscopy
spec_type = Input_spec_data[:,1][correlation_arr[inp_index][spec_obs_number]]\
+ "_" + str(spec_obs_number)
spec_path = Input_spec_data[:,2][correlation_arr[inp_index][spec_obs_number]]
error_map_spec_path = Input_spec_data[:,2][correlation_arr[inp_index][spec_obs_number]]
error_mask_index = inp_index
nu_max = Input_ast_data[inp_index][1]
nu_max_err = Input_ast_data[inp_index][2]
numax_input_arr = [float(nu_max),float(nu_max_err),niter_numax_MAX]
recalc_metals_inp = [spec_fix_values[0],spec_fix_values[1],spec_fix_values[2],feh_recalc_fix_bool]
logg_fix_load = np.loadtxt(f"../Input_data/photometry_asteroseismology_observation_data/PLATO_benchmark_stars/Seismology_calculation/seismology_lhood_results/{stellar_filename}_seismic_logg.txt",dtype=str)
logg_fix_input_arr = [float(logg_fix_load[1]),float(logg_fix_load[2]),float(logg_fix_load[3])]You shouldn't need to edit any of these, unless you would like to fix the first
3 parameters in spectroscopy, if you want to do this, pick what values you want in
the variable recalc_metals_inp and set feh_recalc_fix_bool=True outside the main_func_multi_obs().
Note that the observation number relates to multiple observations of spectroscopy ONLY.
There are multiple stages in this code that requires all the modules that have been discussed, the order of stages is reflected in the order of these bool parameters:
best_spec_bool = True
phot_ast_track_bool = False
spec_track_bool = False
bayes_scheme_bool = FalseThe first parameter will activate the spectroscopy module only, calculating best fit parameters with the conditions and processes that have been selected to treat the spectra. These parameters, degrees of freedom, and covariance matrix for each observation of each star are saved in the directory
Output_data/Stars_Lhood_spectroscopy/multiobs/stellar_nameThe second parameter will activate the photometry and asteroseismology module, creating the grids and saving the tracks in the following directory
Output_data/Stars_Lhood_photometry/stellar_nameN.B. Photometry is only run for the first spectral observation of a given star. This is because multiple observations (unless the spectra are bad) will give similar best fit parameters and therefore central values. The limits currently provided should be large enough to take into account any variation in these values. Thus, computation time is saved.
The third parameter runs once the photometry-asteroseismology grids are completed, these grids are loaded into the workspace, and are used to calculate the spectroscopy grid. This is done by comparing the best fit spectroscopy parameters Teff, logg, [Fe/H], to the values tabulated in the grids, the errors used are ones calculated from the spectroscopy module.
The parameter chi_2_red_bool decides whether a reduced chi2 will be calculated or not. If False
then the chi2 values for all grids are normalised by their minimum chi2, thus rendering their maximum likelihood value
to be 1. If you would like spectroscopic covariance to be included (I would reconmend, it costs nothing in terms
of memory and computation) then set the parameter spec_covariance_bool to True.
The fourth and final parameter runs the bayesian scheme, where all of the chi2 grids are loaded up,
converted to PDFs and combined to make the final PDF with priors included. The priors currently are
IMF which is calculated from the tabulated Mass and Age_step which is used to reduce the probability of false
PMS stellar ages. Once the grids are created, best fit parameters and associated errors are calculated by
fitting to the combined PDF in each parameter space. See Log_prob_distributions_1D function for more detail on how
this is done.
These values are then combined with the best fit spectroscopic values which were fixed (i.e. Vmic, Vbrd, abundances) alongside the uncertainties and then saved in the directory
Output_data/Stars_Lhood_combined_spec_phot/multiobs/stellar_nameN.B. if the probability distributions of for example spectroscopy and photometry are separated enough i.e. they do not overlap in a given parameter, then the fitting function will not be able to fit it and will output a NaN value instead. This can happen!
Once you have run it entirey, you can of course set some of these steps to False to save re-calculation, for example if you want to recalc spectroscopy but photometry already exists, set everything to False except spectroscopy.
The errors and best fit parameters are in separate files.
Lastly, there are two naming conventions which are used in the filenames of several of the results, these are
- mode_kw
- extra_save_string
mode_kw must be defined, this is to give you can indication of what "mode" of the code you ran, e.g. "FULL" would be the full process, bayesian and all. "LITE_numax" is the spectroscopic module but with error mask and nu_max prior applied, "MASK" is the spectroscopic process with just the mask.
extra_save_string can be nothing i.e. "", but I typically use it to indicate what type of error mask I am using or if I've medelled with some part of the source code like "_gaia_mags_only". something like that.
You can run the code on a single star such as below:
if __name__ == '__main__':
inp_index = 15
main_func_multi_obs(inp_index)where inp_index refers to the star's source id in the input photometry data, asteroseismology data, and reference data. E.g. the 16th star in the list (counting from 0) is KIC3427720.
If you would like to run multiple stars at once, the code can be parallelised by the number of stars:
num_processor = 8
inp_index_list = np.arange(len(stellar_names))
if __name__ == '__main__':
#
p = mp.Pool(num_processor) # Pool distributes tasks to available processors using a FIFO schedulling (First In, First Out)
p.map(main_func_multi_obs,inp_index_list) # You pass it a list of the inputs and it'll iterate through them
p.terminate() # Terminates process Now, try to run the full code! If there are any issues, please contact me at