Spatial variations in polarized synchrotron emission in the QUIJOTE MFI data using neural networks

Polarized synchrotron emission from ultra-relativistic electrons spiraling the Galactic magnetic field has recently become one of the most relevant emissions in the interstellar medium because the quality of low-frequency observations has improved. One recent experiment designed to explore this emission is QUIJOTE. We study the spatial variations in the synchrotron emission in QUIJOTE MFI data by dividing the sky into physically separated regions. For this task, we first used a novel component-separation method based on artificial neural networks to clean the synchrotron maps. After training the network with simulations, we fit EE and BB spectra by assuming a power-law model. Then, we estimated the index α_S, the amplitude, and the ratio of the B and E amplitudes. When analyzing the real data, we found a clear spatial variation in the synchrotron properties throughout the sky at 11 GHz, consistent with previous analyses. We obtained a steeper index in the Galactic plane of α_ S ^ EE = -3.10 ± 0.30 and α_ S ^ BB = -3.10 ± 0.28 and a flatter index at high Galactic latitudes of α_ S ^ EE = -3.05 ± 0.16 and α_ S ^ B = -2.98 ± 0.23. We found average values throughout the sky of α_ S ^ EE = -3.04 ± 0.18 and α_ S ^ BB = -3.00 ± 0.26. Furthermore, after obtaining an average value of A_ S ^ EE = 3.31 ± 0.17 μ K^ 2 and A_ S ^ BB = 0.93 ± 0.04 μ K^ 2 , we estimated a ratio of the B and E amplitudes of A_ S ^ BB /A_ S ^ EE = 0.28 ± 0.06. Based on the results, we conclude that although neural networks appear to be valuable methods for application to real observations of the interstellar medium, in future QUIJOTE MFI2 data, combined analyses with , WMAP, and/or CBASS data are mandatory to reduce the noise contamination from QUIJOTE-estimated maps and then improve the accuracy of the estimations. Planck