Colloquium Abstract - Pal - 2026Apr10
April 10, 2026
11:00am Mountain
Arpan Pal (NCRA)
When Clusters Collide and Telescope Misbehaves: From Chaotic Signals to High-Performance Pipelines
Abstract
When galaxy clusters collide at thousands of kilometers per second, they create cosmic particle accelerators spanning millions of light-years. These collisions drive shocks that accelerate particles and amplify magnetic fields, producing diffuse radio emission that reveals the otherwise invisible structure of these extreme environments. I will present low-frequency polarization observations of merging galaxy clusters using the upgraded Giant Metrewave Radio Telescope (uGMRT), complemented by X-ray and high-frequency radio data from the VLA and MeerKAT. Surprisingly, we found that low-frequency polarization does not follow high-frequency predictions. The observed depolarization is inconsistent with single-component magnetic field models. In fact, we report the first detection of linear polarization below 1 GHz that requires a multicomponent magnetic field distribution, challenging conventional shock-compression expectations.
Extracting reliable polarization signals was not straightforward. Mysterious, time-variable instrumental effects initially appeared astrophysical. I spent months systematically investigating the telescope’s signal chain, testing individual components from feeds to correlators and eliminating potential sources one by one, before identifying an unexpected instrumental culprit responsible for the anomalous signals. This detective work highlights how instrumental systematics can convincingly masquerade as cosmic phenomena, and I will walk through this technical adventure.
To handle the massive data volumes from modern radio surveys, we developed Charizard, a fast radio imaging pipeline that uses aggressive parallelization strategies to process terabytes of interferometric data orders of magnitude faster than traditional approaches. This speed enables iterative analysis, better characterization of systematics, and the study of larger statistical samples of merger systems. I will describe the design journey of Charizard and also introduce a Simulation-Based Inference framework, which accelerates QU-fitting by a factor of 1000 and lays the groundwork for 3D RM CLEAN of complex, depolarized sources.
Local Host: Preshanth Jagannathan

Connect with NRAO