At the time of writing, over 2 000 000 variable stars are catalogued in the Variable Star Index (VSX), which is currently ran by the American Association of Variable Stars Observers 1. Due to the expensive cost of spectroscopic studies, these catalog submission studies adopt phenomenological approaches rather than physical ones. These analysis, classify stars by parameters such as its period $P$, initial epoch $T_0$ and magnitude range $m_{\text{max}} - m_{\text{min}}$. Variability type classification is performed manually by visual inspection of the light curve or phase plot, where the variability cycles are displayed with higher resolution.
In addition, the amateur community can easily contribute to variable star detection since said community obtains several CCD images for a variety of purposes (astrophotography, comet tracking, etc.). In this work we present VarStar Detect, a Python library dedicated to the semi-automatic detection of stellar variability. This library is thought to be the basis code of the VarStar Detect program (with implemented GUI) which remains to be written. This program will perform aperture photometry of all targets in the FOV automatically and present possible stellar variable candidates for later visual inspection by the astronomer. Through such program usage, the effort and time consumed in mining variability in the astronomer's images will be significantly reduced, encouraging astronomers to look for variability in their data for VSX submission. This presentation describes the functionality and assessment of the VarStar Detect package available on PyPI2 for installation.
For variable star detection VarStar Detect must accomplish four tasks:
1. Fit a trigonometric polynomial of degree $s$ ($P_s$) to the light curve [1] (using the weighted least squares method) and calculate its corresponding reduced $\chi^2$ parameter ($\chi_r^2$).
$$P_s(t) = C_0 + \sum_{i=1}^{s} \left( C_{i,1} \cdot \cos(i\omega(t - T_0)) + C_{i,2} \cdot \sin(i\omega(t - T_0)) \right) \tag{1}$$$$\chi_r^2 = \frac{\sum_{j=1}^n \left(\frac{y_j - P_s(t_j)}{\sigma(y_j)}\right)^2}{n - 2s - 1} \tag{2}$$2. Choose the trigonometric polynomial fit which satisfies $\chi_r^2 \to 1$.
3. Calculate the amplitude of the polynomial fit.
$$Amp = \max(P_s(t)) - \min(P_s(t)) \qquad \text{where } t \in [0, 30] \tag{3}$$4. Determine if star is variable using the amplitude test.
The purpose of the amplitude test is to decide if the star in question is a potential variable candidate that deserves to be visually inspected. In essence, it nominates all stars which have an amplitude greater or equal than the threshold amplitude as potential variables. The question comes when selecting the threshold amplitude, which will depend on the equipment used and quality of the night sky. The threshold measures the smallest amplitude detected before the equipment measures just noise. Our work does not go into explaining the physical meaning of this threshold; we encourage further investigations to do so. In testing of the code, an empirical approach was used to determine the value of threshold for the data used.
Finally, to test the program's functionality, the program as a whole was tested on Sector 1 TESS PDCSAP data, downloaded with the data_download function designed and available in the python repository. The program was proved to be 96.45 % efficient, detecting a total of 163 variable stars inside the imported sector 1 database.