| Summary: | Thesis: S.M., Massachusetts Institute of Technology, Department of Physics, February, 2020 === Cataloged from PDF version of thesis. === Includes bibliographical references (pages 35-36). === A cm line of hydrogen hyperfine structure gives us an opportunity to study the formation of the first stars [11, galaxies 121 and black holes [3] as well as the standard cosmological model in the Dark Ages and the Epoch of Reionization via both the power spectrum and direct imaging [4]. In order to get three-dimensional images from calibrated inteferometer data, we develop a mapping algorithm called Direct Inversion Matrix (DIM) (151, [6]) that has many advantages over other existing algorithms. However, DIM tends to get computationally expensive. As data grow, the computational time and memory costs grow and become intractable for large instruments such as HERA-350. Through simulation, I will demonstrate the performance of DIM, in terms of computational speed and required memory, as well as the quality and fidelity of maps when tuning various parameters such as validpixel threshold that determines which pixels will be mapped, resolution, LST range and regularization matrix. This reveals which parameters tend to best relieve the computational burden of DIM while still maintaining sufficient data quality to achieve our scientific goal. To save computational time and reduce memory cost, I also show how to utilize parallel computing and other computational tricks for DIM. Finally, I explain the framework of the algorithm and also describe the implementation of DIM with the HERA IDR2.2 data [7] and show the preliminary results. === by Jianshu Li. === S.M. === S.M. Massachusetts Institute of Technology, Department of Physics
|
|---|
