Foreground separation and signal extraction method development for the Cosmological HI intensity mapping experiments.
A redshifted signal of the neutral hydrogen from the Epoch Of Reionization (EoR) should be detectable at the low frequencies 50 – 200 MHz. This signal is very weak having an amplitude of the tens of millikelvins, mK. The main challenge was to develop a technique to extract the cosmological HI signal in the presence of the strong and varying Galactic foregrounds like a synchrotron and a free-free emission.
Maximum Entropy Method in spectral domain was used to perform a foreground separation and signal extraction using the simulated low-frequency data. Several optimization solutions were tested including the SciPy and PyTorch.
A new user-friendly Python code was developed on the base of the old F90 program. The interactive Jupyter notebooks for the data simulation and signal extraction were developed on top of the Python core functions. The benchmarking showed that the Python-based code works faster that the old F90 version due to the accelerated nature of NumPy/SciPy functions.
ToDo list includes sharing the code via GitHub. The paper on this project is also planned and the work on this paper is ongoing.
Anything else you would like to convey to the EC project office & reviewer
There were several talks and presentations describing the method and demonstrating the results of the simulations. Some of them were
Dr. Vladislav Stolyarov