Sifting for astrophysical gems in the spectral domain
Traditional time-domain searches for pulsars and highly temporally variable sources are computationally prohibitive with modern radio telescopes such as the Murchison Widefield Array (MWA) because of the large numbers of images and pixels that must be searched. However, several classes of exotic and compact objects can reveal their existence by the fact that their intensity fluctuations vary both in frequency and time, potentially obviating the need for such data-intensive searches.
We have devised a way to detect such sources by examining the differences between spectra of pixels in pairs of images taken days to weeks apart. The process of image subtraction removes imaging artefacts common to both datasets, while preserving evidence of any temporal and spectral variability that may be present. Pixels of non-variable sources exhibit the statistical properties of thermal noise, so the intensity fluctuations of their difference spectra are normally-distributed. However, pulsars are sufficiently compact that they exhibit propagation effects that skew their intensity distribution to be highly non-gaussian. Other compact systems exhibit intrinsically highly bursty and frequency-structured emission that is also non-gaussian. By testing for non-gaussianity in the spectral domain, this method circumvents the excessive computational demands that have plagued previous searches for pulsars and certain transient radio sources with widefield radio interferometers.
In this project you will implement this technique on several years’ worth of fortnightly monitoring of the Galactic Plane and Galactic Bulge regions, where there is a high density of compact stellar-mass systems including pulsars and other exotic radio transients. The bright background synchrotron emission from this region means that standard image-plane searches for transient events are limited by confusion noise and high sky temperatures. Moreover, deconvolution of such a complex region is a challenging computational problem in and of itself. Using the technique of image differencing we can bypass these problems, subtracting away the vast majority of the non-variable emission to leave behind only the most interesting compact variable sources, which you will aim to detect using our new algorithm, and subsequently characterise with follow-up multi-frequency observations. While this new technique is extremely powerful, it is computationally intensive, and you will need to make use of the computational resources available locally at the Pawsey supercomputing centre.