Parameter Optimzation of Transiting Exoplanet Data

K M Shariat Ullah*, Touhidul Islam Shuvo†, Dr. Md Rasedujjaman‡

Department of of Electrical and Electronic Engineering, Shahjalal University of Science and Technology, Sylhet
Email: *[Link]@[Link] † touhidul81@[Link] ‡ mrased-eee@[Link]

Background Results

Planets outside of our solar system that orbit stars other than our Sun are known as (a) (c) (e)
extrasolar planets, or exoplanets. The first discovery of an exoplanet orbiting a Sun-
like star, Pegasi 51b was made in 1995. [1] Since then 5700 exoplanets have been
discovered till now. More than 4600 exoplanets out of them were discovered using a
method called the transit method. [2]

(b)
(d)

Parameter Initial Guess Nelder-Mead Optimization MCMC Optimization Literature Value

Figure: The visual brightness of a star decreases when a transit occures. The +0.11
0.1 0.047 0.04 -0.60
brightness comes back to normal after the transit.
+0.99
5.0 7.22 6.74 -0.84
A transit occurs when an exoplanet passes in front of its host star and blocks out +1.44
some of the star's light. Upon studying the light curve of the star, a plot of brightness 90.00 89.84 89.38 -0.69
+1.08
(flux) vs time, a dip in the brightness of the star can be observed. This method of 2746.13 2748.30 2748.26 -0.95
detecting an orbiting body around any star is known as the transit method. [3] +0.45
2.14587 2.13185 2.13 -0.99
+0.29 +0.43
In this research, we model the light curve to initially guess the planetary parameters. [0.1, 0.3] [0.087, 0.346] [0.30-1.17 0.36-1.10 ]
These parameter values were then optimized using the Nelder-Mead algorithm and
Markov Chain Monte Carlo (MCMC) framework.

Methodology Discussion, Future Work and Acknowledgement

The exoplanet TIC 284475976 was chosen for experimentation purposes. The data The phase folded data was then fitted using the batman pipeline. The pipeline uses the The optimized values given in the Table above (See Figure d) for the given TIC id agrees with the literature value. This method has less
was fetched from the Transiting Exoplanet Survey Seattleite (TESS). [4] The data Mandel-Agol formulation to model the light curve. [6] Inside the pipeline, quadratic limb time complexity and is easy to use over the conventional method. But it has its own limitations. The developed pipeline doesn’t perform
was further cleaned removing the NaN values. The outliers of the data were darkening was implemented with two co-efficients u1 and u2 where the limb darkening effect well when there are multiple planets and stars. The orbits of the planets become non-Keplerian producing a transit timing variation
removed using the Huber Loss function. The cleaned data has some discontinuities is described as (TTV). As our future work, we’ll devise a pipeline to optimize parameters for TTV. The window size of the filters used in the noise
due to the maintenance of the instuments can be seen in Figure (a). reduction process also needs to be well furnished for that type of optimization.

The period of the planet P and the transit time at inferior conjuction t0 was For his guidance throughout the project, we owe a debt of gratitude to our supervisor, Dr. Md. Rasedujjaman. Additionally, we are
determined from the periodogram of the data. A periodogram in the context of The initial guess for the two limb darkening co-efficients were taken to be 0.1 and 0.3 as this is grateful to Dr. Jerome de Leon of the University of Tokyo for sharing his expertise in optimizing the parameters and for providing us
exoplanets, is the power spectrum of the data. The classical Schuster periodogram found to be the most frequently found. [7] The initial guess for the semi major axis is taken to with suggestions for the future direction of our study.
can be calculated using the following expression of Fourier Transform [5] - be 5 AU as this is the most frequently found value of a for transiting exoplanets. [7] The
phased folded data upon properly modeled revealses the initial guesses for the six different
parameters of the exoplanet. We used the Nelder-Mead algorithm to optimize these parameter Reference
values.
[1] M. Mayor and D. Queloz, “A Jupiter-mass companion to a solar-type star,” Nature, vol. 378, pp. 355–359, Nov. 1995.
To find the uncertainty levels in the optimized parameters, we first devise a Bayesian [2] NASA Exoplanet Archive. Available at: [Link]
The periodogram for the chosen id can be seen in Figure (b). Period at maximum framework with the likelihood to be normally distributed against some mean value and some [3] Joshua Winn, “Transits and occultations,” arXiv preprint, 2010.
power is the best guess for the period of the planet and transit time at maximum standard deviation from the mean. For digital computational purposes, we take the log [4] S. W. Yee et al., “The TESS Grand Unified Hot Jupiter Survey. I. Ten TESS Planets,” The Astronomical Journal, vol. 164, no. 2, p.
power is the best guess for t0. Using these two parameters the data was then phase likelihood function described as below - 70, Aug. 2022, doi: 10.3847/1538-3881/ac73ff.
folded using the following formula -
[5] J. T. VanderPlas, “Understanding the Lomb–Scargle Periodogram,” The Astrophysical Journal Supplement Series, vol. 236, no. 1, p.
16, May 2018, doi: 10.3847/1538-4365/aab766.
[6] K. Mandel and E. Agol, “Analytic Lightcurves for Planetary Transit Searches,” The Astrophysical Journal, vol. 580, no. 2, Nov.
2002, doi: 10.1086/345520.
[7] N. Espinoza and A. Jordán, “Limb darkening and exoplanets: testing stellar model atmospheres and identifying biases in transit
parameters,” Monthly Notices of the Royal Astronomical Society, vol. 450, no. 2, pp. 1879–1899, Jun. 2015, doi: 10.1093/mnras/stv744.