New Exoplanet Discovered! Deep Learning Enables the Fastest, Most Sensitive Detections to Date

AJAS · 2022 Physics

David Chang

Thumbnail supplied by the source for New Exoplanet Discovered! Deep Learning Enables the Fastest, Most Sensitive Detections to Date

Overview

The NASA Kepler mission is humanity’s most extensive survey of sun-like stars in the Milky Way for exoplanets. Even today, Kepler’s dataset is at the forefront of discovery; methods for vetting new exoplanets are rapidly being improved with the introduction of cutting-edge technology. This research project’s engineering objective is to implement deep learning and GPU fast-folding techniques to produce the fastest, most accurate, and most sensitive exoplanet detections to date. A convolutional neural network was trained on artificial transit data, and software was developed to efficiently analyze Kepler lightcurves. This methodology successfully identified a previously undiscovered exoplanet candidate of 1.7 Earth radii. After thorough data analysis, including an independent verification with the Box Least Squares method and transit curve fitting, candidacy was confirmed and relevant parameters (such as radius, period, and duration) were derived. In the future, this innovative technique will be applied to all ~100,000 Kepler lightcurves and those from other space missions (TESS). It is a framework for the discovery of new exoplanets that will one day illuminate the Kepler mission’s greatest aspiration – calculating the occurrence rate of Earth-like planets throughout the Milky Way.

From the student

Round and smooth—perfect. My four-year-old eyes were entranced as the spinning glass sphere rolled across the dining table. Just before it would fall, I scooped it into my palm and studied it. The outer surface was a thin, translucent layer of pale green—pristine and cool to the touch while inviting me to peek inside. From marbles to globes, I have always enjoyed observing spherical objects. My obsession got to the point where I would carry bouncy ball s in my pockets at all times

But the greatest ball of all is something I discovered stargazing with my 80-year-old grandfather. He was a physics researcher who committed his life and career to ionized plasmas, a form of matter not commonly observable on Earth. He wasn't famous but clearly loved his work. Whatever question I asked, my grandfather would have an explanation. One day, as we looked up into the night sky, I wondered out loud, “Why are the stars not pointed like the ones I draw on paper?” “Because they’re actually like balls.” I smiled.

In high school, my own fascination for physics blossomed. From twirling stoppers in the air to learning about uniform circular motion and crashing carts together during the collisions unit—every lab was engaging. My favorite topic was Kepler’s laws and gravitation, which showed me how physics extends to an unimaginable scale beyond Earth and sparked my burning curiosity for astrophysics.

That summer, I sought out an opportunity in astrophysics research and met my mentor, professor Jian Ge. He introduced me to the study of exoplanets, which was at the cusp of a data science breakthrough. Machine learning is a relatively new technology in the field with implementation only beginning a few years ago. Searching through lightcurves in existing space telescope data for exoplanets is an area that would especially benefit from the precise, automated classification that neural networks enabled. I made it a personal goal to explore this amazing field of research, and contribute to a fresh framework for exoplanet detection that will forever change our understanding of life in the universe. After months of hard work, I discovered a likely genuine, new exoplanet.

Through research, my simple interest in spheres evolved into deeply exploring astronomy. But I also came to realize that research is an endless adventure I can forever pursue. As distant planets carry on orbiting their host stars, as the world constantly rotates through day and night, and as the marbles keep rolling across the dining table, I hope to continue reshaping my understanding of the universe.

From the student

References: https://drive.google.com/file/d/1L7L21HcTsjJT2NxpqufW-_CX0v3VUQkr/view?usp=sharing

Acknowledgments:

I would like to thank my research advisor Dr. Ellsworth, Mr. Queenan, and the research program at High Technology High School for introducing me to scientific research and actively supporting me with my work. I also greatly appreciate Dr. Jian Ge and other mentors/student researchers at the Science Talent Training Center, who provided the scientific expertise I needed throughout the project. Finally, a thank you to my family and friends for always talking to and encouraging me along the way.