Background and Motivation: For the 400 million years following the Big Bang, the universe was extremely hot and dense. As the universe expanded, it started to cool down until it eventually reached temperatures cool enough to create neutral hydrogen atoms. This period of time is known as recombination. Quickly following recombination was the “cosmic dark ages,” which got its name because no stars had yet formed. During this time, the universe had only 21cm emissions coming from the neutral hydrogen atoms and radiation emitting from the CMB (Cosmic Microwave Background). Over the course of the following 500 Million years, gravitational collapse allowed for the first stars and galaxies to form. The radiation from these galaxies ionized the neutral hydrogen atoms in a time period called the Epoch of Reionization (EoR). To observe the physics of the EoR, researchers use 21 cm cosmology. There is a specific method to observing and studying the EoR because the first galaxies were distant and small. Learning about the EoR is important because it gives us insight into what the early universe was like and how early structure in the universe evolved. Because of their distance and size, they have weak cosmological signals, which is a barrier for most instruments. A primary tool for examining the EoR is interferometry. [1]

Interferometry is a method of measurement by the interference of sound waves or light waves. In radio astronomy, interferometry is useful in regards to its measurement of light waves specifically. Interferometers allow us to see a larger field of view than a single dish telescope does. This is due to the antenna separation, which allows for high resolution imaging with a larger field of view. Applying interferometry in radio astronomy allows astronomers to gather much more precise images than even the largest single dish telescopes. 

Interferometry is a very common practice in radio astronomy and cosmology experiments. Telescopes like the Hydrogen Epoch of Reionization Array (HERA) and the Murchison Widefield Array (MWA) are interferometers rather than the traditional single dish telescope design. The interferometer telescope consists of a large number of antennas that are connected to a central computer which takes the signal from every antenna and correlates it to the signals from the other antennas. The Interferometer does this to observe emission from the intergalactic hydrogen of the EoR.[2]

When working with a large number of components like the antennas in HERA, there are many problems that arise. The first issue is that there are only a handful of people who are overseeing the project as well as ensuring the interferometer is working properly. Very quickly, many projects reach the point in which the amount of hardware components that need maintenance vastly outweighs the amount of people working with the system. This makes it difficult to operate on each antenna and make sure they are all working properly together. This is vital because if an antenna stops working correctly, then the data is inaccurate, rendering it useless. With an interferometer like HERA every component needs to be aligned for the system to work properly. If antenna alignment becomes imbalanced then the data being gathered by the telescope becomes inaccurate. My proposed project will work to combat these issues by creating an automated system using machine learning techniques. The automated system will allow us to know which antennas are not working properly, so we do not have to manually inspect each antenna. It will also check that data volume is under control and that the antennas are properly aligned and working in sync with each other. With this automated system I will be working with interferometers made up of multiple antennas like HERA or MWA. My research will be beneficial to these types of telescopes because of the amount of components that need to be managed and the lack of resources to properly maintain them.

Research Project: For my research, I will be working with the HERA (Hydrogen Epoch of Reionization Array) telescope. HERA is a radio interferometer. The telescope is designed to have 350 antennas with a total collecting area of  54,000m2.[3] I propose to use machine learning techniques to help solve these technical hardware issues. Doing so will be very useful to enable efficient calibration and commissioning.

Phase I: I will visually analyze any deficiencies within the antennas and identify any issues or problems that arise. After I do this, I will document a summary of any outliers in the data, then I will plan a method for resolving the issue and making the data usable again. 

Phase II: Using the information from my initial analysis, I will use machine learning techniques trained to the data analysed by eye, that properly identifies which antennas are malfunctioning and what data should be removed. This algorithm will be programmed to allow the antennas to work and collect data in the proper way.

Phase III: The final step will be implementation. I will begin this process by addressing small groups of the antennas, seeing their performance with my automated system, and if any other issues arise. I will continue this process until all of the antennas are being commissioned by the automated system. The main part of this step will be focused on optimization. I will be making sure the system is efficient in both time and memory usage.

Intellectual merit: This proposal will advance the field of interferometry by ensuring that the components of interferometers are easier to maintain. My algorithm will also identify the limitations of the computing power. The more antennas that are added, the higher the demand is for computing power. In addition to functioning technology, this project will also ensure that the future research data will be accurate and precise. When I worked with HERA, I assessed antenna functionality by eye. This process can be very intimidating and complex for inexperienced students. In addition to the immediate impact, my research will also grant many opportunities to individuals who are new to the field of radio astronomy.

Broader impacts: My work will make interferometry and radio astronomy more accessible to students who may feel intimidated or are unaware. Making radio astronomy more accessible will create opportunities for students to expose themselves to the subject. As time progresses, I will incorporate my work into existing research experience programs (CHAMP, CAMPARE, REU) that help prepare undergraduates for research. I will help make these programs more accessible to new students without prior experience in radio astronomy. The University of Washington offers a plethora of outreach programs dedicated to educating individuals from all backgrounds. As a graduate student, I will seek the opportunity to work with the “Astronomy for Sight Impaired” program and help teach astronomy through the use of braille and large print astronomy textbooks aimed at sight impaired learners. I will also work with the UW Planetarium to help educate the public and spark interest in the subject.

[1] Furlanetto, S. R., S. P. Oh, and F. H. Briggs, 2006: Phys. Rep., 433, 181–301 

[2] Fomalont, E. B. and M. C. H. Wright, 1974: Interferometry and Aperture Synthesis . p. 127.

[3] Deboer, D. R., Parsons, A. R., Aguirre, J. E., et al., 2017, PASP, 129, 045001