Piercing the dusty veil of hyper-luminous infrared galaxies: Sub-arcsecond 144 MHz ILT observations of HLIRGs in the Lockman Hole

Context. Hyper-luminous infrared galaxies (HLIRGs) are among the most extreme systems in the Universe. With infrared (IR) luminosities of L-IR > 10(13) L-circle dot they can have IR-derived star formation rates (SFRs) exceeding 10(3) M(circle dot)yr(-1). Theoretical models have a hard time reproducing the observed number densities of such extreme star-forming systems. It is known that at least part of the population harbours active galactic nuclei (AGNs), but their prevalence and relative contribution to the IR output is still debated. Assessing this is further complicated by the heavy dust obscuration. Aims. We aim to investigate the HLIRG population in the Lockman Hole field to assess whether they are truly highly star-bursting systems or whether notable AGN activity is present. A substantial AGN population could help resolve the tension between the HLIRG number densities obtained from observations and predicted by galaxy formation models by lowering the amount of truly extremely star-forming galaxies. Methods. Starting from a highly complete Herschel-selected sample, we made use of recent wide-field sub-arcsecond 144MHz International LOFAR Telescope (ILT) observations of the Lockman Hole field to probe AGN activity in HLIRGs in a dust-unobscured way. AGN presence was deduced through means of the brightness temperature (T-b). Brightness temperature measurements were made to determine the fraction of HLIRGs harbouring a radio AGN. This identification was then compared to the classification based on spectral energy distribution (SED) fitting based identification from the LOFAR Deep Fields project, the radio-excess q and IRAC infrared colours. Results. We detect 33% of previously identified HLIRGs at sub-arcsecond resolution. All but one of the detected sources is found to exceed the T-b threshold for pure star formation, showing 98% of detections to contain a radio AGN, even though lower-resolution observations had classified them as star-forming galaxies (SFGs). The remaining source is concluded to be consistent with having no AGN activity. All of the sources in our sample that were previously classified as radio AGNs, radio-quiet AGNs or those that were unclassified were detected as high-T-b objects (16% of the total sample or 47% of the detected sub-arcsecond detections). In addition, we identify AGNs through T-b measurements in 20% of sources that were classified as SFGs through SED fitting, raising the overall fraction of AGNs in the total sample from 16% to 32%. AGNs identified through brightness temperature measurements are also found to be more likely to be mid-IR AGNs. Conclusions. 98% of our detected sources are found to likely host radio-AGNs, raising the number of identified AGNs among the HLIRG population in this field from 16% to 32%. This increased number of AGNs is not sufficient to bring observations and predictions of HLIRG number densities in agreement, however. Even at cosmic noon around z similar to 2, where the tension is lowest, it remains at a factor of just a few. The identification of radio AGNs in supposed SFGs highlights the value of high-resolution radio observations in studying dusty objects such as HLIRGs. Broad consistency is seen between T-b and the other AGN indicators, and the observed relation between SFR and T-b is seen as indicative of co-evolution between stellar mass build-up and black hole growth.