The Spectral Separation Routine

The package is separated into multiple modules. The functions associated with the spectral separation are located in the module

spectral_separation_routine

Here, the main two functions that can be used for spectral separation are called spectral_separation_routine and spectral_separation_routine_multiple_intervals. The last function is a wrapper on the first, that can also split the input spectra into multiple different intervals.

spectral_separation_routine()

spectral_separation_routine requires 8 input arguments:

• flux_collection (numpy array shape (data.size, n_spectra) or (data.size, n_orders, n_spectra)). This 2- or 3-dimensional array should include all observed normalized spectra to be analysed.
• flux_templateA (numpy array shape (data.size, ). 1-dimensional array that includes normalized spectrum of a stellar template to use for the first stellar component.
• flux_templateB (numpy array shape (data.size, ). Same as before, but for the secondary component.
• wavelength (numpy array shape (data.size, ). 1-dimensional array of wavelength values for the spectra. All spectra must have been previously resampled to the same wavelength grid, equi-spaced in velocity. This can be done with functions in the spectral_processing_functions module.
• options (RoutineOptions object). This object includes all options and some extra parameters needed by the parent routine.
• sep_comp_options (SeparateComponentsOptions). This object includes options and extra parameters which should be passed down to the child routine ‘separate_component_spectra’, which handles the actual spectral separation when radial velocities are supplied.
• rv_options (RadialVelocityOptions). This object includes options and extra parameters needed by the other child routine ‘recalculate_RVs’, which handles radial velocity calculation given separated component spectra.
• RV_guess_collection (numpy array shape (n_spectra, 2) or (n_orders, n_spectra, 2)). Initial guesses for radial velocities of both components (component A is (:, 0) if ndim=2 or (:, :, 0) if ndim=3). These must be provided for both components in order to ensure convergence.

Most optional arguments are encapsulated in the *Options objects. The only three direct optional arguments that can be supplied are

• mask_collection_orders (numpy array shape (data.size, n_orders, n_spectra)). This must be a boolean array that indicates the wavelength indices of each spectral order if flux_collection.ndim == 3 (see note below).
• period (float), Default = None. Orbital period of the system. Must be supplied along with time_values if eclipse masking is to be done during RV calculation, e.g. if one of the components should not be subtracted from the spectrum at specific phases.
• time_values (numpy array shape (n_spectra, )), Default = None. Timestamps of all the spectra. np.mod(time_values, period)/period should range from 0 to 1. Must be supplied if eclipse masking is to be done, e.g. if one of the components should not be subtracted from the spectrum at specific phases.

Note on ndims and spectral orders

It is currently NOT possible to supply multiple spectral orders to the function and gain merged separated spectra as output in a convenient way. This was possible previously, but the underlying functions for MERGED spectra has been continously developed while the order-to-order methods have NOT. The functions still appears to support this, but DO NOT USE IT.

RV_guess_collection can have either shape (n_orders, n_spectra, 2) or (n_spectra, 2) in this case. If the latter, each order is assumed to have the same initial RV guess, which is generally a decent assumption. (AS MENTIONED, DO NOT USE THE SHAPE (n_orders, n_spectra, 2)).

spectral_separation_routine_multiple_intervals()

The wrapper function spectral_separation_routine_multiple_intervals is similar, but requires 9 input arguments instead, and has 3 optional arguments. These are, in order:

• wavelength (shape (data.size, )). The full wavelength interval supplied.
• wavelength_intervals (list of tuples (float a, float b)). Each tuple in the list defines a single independent instance that the routine should be run, on the wavelength interval (a, b). Providing the argument [ (4500, 5000)] would thus tell the function to perform spectral separation once, with data only in the wavelength interval 4500-5000 Ångström.
• flux_collection (shape (data.size, n_spectra)).
• flux_templateA
• flux_templateB
• options
• sep_comp_options
• rv_options
• RV_guess_collection
• buffer_size (float), Default: None. This will be described in more detail in the Routine options section with the buffer_mask parameter. It defines the size in Ångström of a small buffer of data attached around the edges of each interval, that can be used to avoid effects of data-wrapping on both the separated spectra and the broadening function calculation. None results in no attached buffer.
• period (float), Default: None.
• time_values (np.ndarray shape (n_spectra, )). Default: None.

Next page: Routine options