Erik Osinga

Pushing LOFAR to the absolute limit

LOFAR is the only radio telescope in the world that can observe 'decametre' wavelengths (i.e. <10 to 30 MHz) with high resolution. In fact, telescopes on Earth cannot observe below 10 MHz, because the ionosphere starts to block all radiation. However, even above 10 MHz the ionosphere severely distorts radio waves, and it has been very challenging to calibrate these effects.

I spent part of my PhD on producing science-quality images at the lowest possible frequencies, and made it work! Below you can see an image of the cluster Abell 2256 at extremely low frequencies. The image has more than 10x higher resolution and sensitivity than previous images we had of the low-frequency radio sky, opening up a new window of opportunities for discovery.

Detecting cluster magnetic fields through depolarisation

Magnetic fields are quite a big mystery on cosmic scales, while they are important in many astrophysical processes such as the large-scale formation of clusters. To probe magnetic fields in clusters, we can use polarised radio sources because their polarisation angle will rotate proportionally to the magnetic field strength along the line of sight. Because we don't have infinite resolution, different rotation along different sight-lines causes the radio sources to de-polarise. The depolarisation is expected to be larger closer to the centre of the cluster, where the magnetic field strength is generally higher.

However, this is hard to detect observationally, because only a small fraction of radio sources is actually polarised. So, in this study I stacked VLA observations of 124 clusters to find this depolarisation signal. Impressively, due to the large sample, this was the first time that we see a clear depolarisation signal in galaxy clusters! I could even compare the depolarisation with models to put constraints on the average magnetic field strengths and correlation scales in clusters. Below is a plot where you can see the depolarisation (where 0 is completely depolarised) against radius, indeed showing most depolarisation near the centre of the clusters.

Looking for cluster radio emission in the deepest ever LOFAR images

In April 2021, the deepest ever LOFAR images were released. So of course I looked through them to check for large scale radio emission in all of the galaxy clusters in those images. I found a radio halo in a fairly low mass cluster, of which only a few are known so far. I was also able to put some really deep upper limits on clusters with non-detections. The radio halo is visualised in the white contours below, overlaid on an optical image where you can see the cluster of galaxies.

Large Scale Alignment in LOFAR Radio Galaxies

In my very first publication, I tested the hypothesis that angles of radio galaxies are randomly oriented. Interestingly, some studies have found that polarisation angles of quasars and jet directions of radio sources are not randomly distributed throughout the Universe, but show correlations on unrealistically large scales. To test this with LOFAR, I extracted a sample of 6795 double-lobed radio galaxies from the 325,694 low frequency radio sources in the first data release of the LOFAR Two-metre Sky Survey.

I performed various statistical tests and found that there was indeed a significant non-uniformity at angular scales of around 4 to 5 degrees, meaning that the orientation of the sources is correlated on these large scales. Interestingly, the effect was mainly found in the brightest sources [Bin 3 in the plot below]. However, when taking the distance of the sources into account to compute their 3D positions, the signal disappeared. The fact that we saw alignment in 2D but not in 3D pointed towards the cause being unknown systematics or biases in the LOFAR survey and not real spooky alignment.

Observing experience

Leiden University

My thesis was titled Data compression for weak lensing studies with the upcoming Euclid mission.

I used an information maximising neural network (IMNN) to find an optimal compression function for a Euclid-like weak-lensing survey. Over the next decade, "Stage IV" lensing surveys such as Euclid will revolutionize our view of the Universe by providing some of the tightest constraints on cosmological parameters. To accomplish this, these surveys will detect billions of galaxies. A crucial step in the inference of cosmological parameters is compressing this large volume of data to a manageable amount of statistical summaries. I found that the IMNN provided an informative mapping from highly dimensional data to just a few summary statistics that can be used to to infer accurate posterior distributions without assuming the form of the likelihood function. You can find all the details here. Below is an example of the posterior of the matter density of the Universe found by the network (blue), where the dotted line is the correct answer.