Entangled in disentangling

A crucial step in our parameter extraction of binary and triple stars was spectral disentangling and/or spectral separation. A detailed discussion of the methodology can be found in a review by Hadrava,2008 https://arxiv.org/abs/0909.0172.

In our work, used two codes, <p id="small-caps">fd3</p> http://sail.zpf.fer.hr/fdbinary/ and <p id="small-caps">disentangling_shift_and_add</p> https://github.com/TomerShenar/Disentangling_Shift_And_Add to disentangle spectra with resolution between 28000 to 67000. We also showed that it is possible to extract spectra with small sample spectra. Some of the important considerations that we learnt from our work are listed below:

• Stability of spectrograph and reduction method is important. With low-mass CHTs, we sometimes work with a tertiary light fraction of ∼ 0.1 or lower. In such cases, small biases in measurements are propagated and also amplified through the disentangling method https://ui.adsabs.harvard.edu/abs/2008A%26A...482.1031H/abstract. Such biases can be prominent in echelle spectra where order-wise disentangling can be used to avoid biases due to complex reduction methods like blaze correction.

In our work, we found FEROS and HARPS spectra to be stable. These are the best spectra to be used for disentangling.

• Post-processing is helpful in cases of low spectra. While with a clean set of spectra, we can get clean disentangled spectra, most of the time we have to reject few available observations due to superimposed spectral lines, low SNR, emission lines etc. In this case, post-processing of disentangled spectra is needed. While the cleaning of the bias can be done by simply subtracting the bias signal, it is the modelling of the signal which can be tricky.

• Disentangle in super-imposed segments. A good practice is to break the spectra into segments where the end of two segments superimpose with 2-3 common lines. This is a good way to compare line depths for bias cleaning and the convergence of the disentangling routine.