Final Report Summary - HISTORYNU (The HI Story of Galaxy Evolution in the Nearby Universe)
The main focus of the project was science preparation for HI surveys with the new APERTIF phased array feed (PAF) system on the Westerbork Synthesis Radio Telescope (WSRT) providing a 37 fold expanded field of view and hence a huge increase in survey speed. Simply put: to observe an area of ~ six square degrees now takes a day instead of a month. This preparation focused on producing important elements of an automated analysis pipeline and using ongoing research projects to explore and focus the science goals and help an optimal preparation and execution of the HI imaging surveys planned with APERTIF.
Analysis preparation focused on source finding and source characterisation and interactive visualisation of 3D data. Two contributions under this ERC grant can already be identified as having been important. One is the contribution to improving the robustness and reliability of 3D methods to find HI emission objects in synthesis data cubes (providing 2D positional and 1D spectral information). This was part of an international collaboration and has led to a now well performing and widely used source finding system SoFiA (https://github.com/SoFiA-Admin/SoFiA) that had its third stable release in May 2018 and is moving to a next stable version to be issues early 2019. SoFiA now has a well established reputation and has begun to acquire a large user base.
The other concerns the visualisation of, and the analysis and modelling of HI (neutral hydrogen) in galaxies using synthesis spectral line data. The analysis can be greatly enhanced when coupled to quantitative and interactive 3D and 2D visualisation. A very powerful package has been developed within the existing environment of 3DSlicer (http://slicer.org a package originally for medical imaging and still mostly used for this purpose) which allows interactive visualisation and modelling of 3D radio astronomical data sets taking on board the full 3D astronomical coordinate and physical units system. The package, called SlicerAstro (https://github.com/Punzo/SlicerAstro/wiki) is an extension to 3DSlicer, but is also downloadable as a custom built 3DSlicer+SlicerAstro binary executable for Linux and MacOS platforms. SlicerAstro allows flexible 3D/2D visualisation of spectral line data, manipulation of the data (including 3D segmentation, construction of data slices, calculation of moment images) and a direct inter-comparison between kinematic models of the data and data itself. It also allows direct visualisation and comparison of data from different sources (i.e. at different wavelengths).
In addition non-parametric methods of measuring asymmetries in the distributions of HI emission in galaxies (2D) have been examined and appear to be quite sensitive to signal-to-noise, resolution and galaxy orientation. In particular the signal-to-noise is an important effect, which has been ignored in studies to date. Methods to correct for the noise bias have been designed and tested. The software will be made available as a GitHub repository. From examining model galaxies it has become clear that determination of the asymmetries in 3D, i.e. taking into account both morphological and kinematic asymmetries, is much more robust, also in the presence of noise. This entails a simple modification of the 2D software and will also be made available at the GitHub repository.
Members of the research group have contributed to both the testing and the improvement of the Apertif data calibration and imaging pipeline Apercal (https://github.com/apertif/apercal). The improvements largely focused on introducing parallel processing to speed up the processing speed. This is possible because in principle each Apertif compound beam is a separate data set and is processed separately. Apercal is a python based system making use of Casacore (https://github.com/casacore/casacore) Casa (https://casa.nrao.edu) and Miriad (https://www.atnf.csiro.au/computing/software/miriad/) for the data processing.
Science exploration involved both exploring state-of-the-art cosmological simulations as well as the reduction and analysis of existing HI observations. Comparing the simulations with observations will be increasingly important. A comparison of simulations of a single Milky Way type galaxy with differing amounts of supernova feedback with the properties of the HI and hot gas in the corona (halo) of the Milky Way has shown that considerable feedback, driving gas from the disk into the halo is required to explain the distribution and kinematics of HI emission and O VI absorbers in the Milky Way corona. A detailed study of the HI content of galaxies in groups of different total mass, i.e. different environments in terms of galaxy density, has shown that a mix of different physical processes shape the HI content (and presumably also the HI distributions) of galaxies: galaxy interactions, gas removal by tidal and ram pressure stripping, gas accretion from the cosmic web and gas removal and redistribution by the induces star formation in the galaxies. The simulations suggest that the most common cause of H I removal is satellite– satellite interactions, followed by ram pressure and tidal stripping. Ram pressure stripping is the most common mechanism that disturbs the H I morphology of galaxies at z=0, though this concerns mostly galaxies that are well within the viral radius of a group.
Ongoing HI projects have led to a number of results. Investigation of the HI Mass Function (describing the volume densities of galaxies of differing HI mass) in the Ursa Major cluster shows that the number of low HI mass galaxies is considerably smaller that found in two major surveys, HIPASS with the Parkes telescope and ALFALFA with the Arecibo telescope, of large areas/volumes. This is not yet understood. The provision should be made that the statistics at the low HI mass end is quite uncertain, though it can be clearly shown that based on the HI Mass Functions of the HIPASS and ALFALFA surveys the number of low HI mass objects in Ursa Major should have been at least a factor of four higher. Large blind volumes are required to have a definitive answer to the question of whether and how the low mass end of the HI Mass Function depends on environment.
A comparison of the HI and CO distributions with the distributions of star formation derived from UV and IR imaging data in 29 galaxies shows that all galaxies exhibit more or less the same relationship between gas surface density and star formation density. The star formation process apparently is self regulated and not affected by whether a galaxy is undisturbed, mildly disturbed or clearly interacting. This confirms the outcome of other studies of the star formation in galaxies.
A real pilot for deep HI observations with future telescopes (in particular MeerKAT and SKA) is the Chiles project aimed at a 1000 hour integration with the JVLA on the Cosmos field. Involvement in this effort has proven useful for knowledge transfer between this project and Apertif HI surveys. Analysis of the first 178 hours of JVLA data has led to the discovery of a gas rich galaxy at z=0.376 (ApJ 824, 1F, 2016). An investigation of the galaxies in the redshift ranges 0.11-0.13 and 0.16-0.18 and their relation to the large scale structure in the observed volume has been completed. Most of the detected (and hence relatively HI rich) galaxies lie close to the large scale structure filaments and spines in the volume and many show signs of disturbed morphologies. This study made use of SoFiA for locating the detections and made use of DisPerSE (http://www2.iap.fr/users/sousbie/disperse.html) to characterise the large scale structure. These methodologies proved to be very useful and will probably be used widely in future HI survey projects.
In addition one can use the Dressler the Dressler–Shectman technique for finding kinematic substructure in the large scale distribution of galaxies for a comparison with HI properties of galaxies. This was done successfully in the Antlia cluster using KAT-7 data (MNRAS 452,1617, 2015) and will also be applied to other data under investigation: the Westerbork Coma Survey (Serra, van der Hulst et al.), a 24 pointing mosaic with the WSRT, the deepest blind HI survey of the Coma cluster to date and similar blind survey projects in the Pisces-Perseus supercluster region.
Science exploration also included designing an optimal pointing grid for the Apertif imaging surveys. This included optimising between the size of the area covered and minimal sensitivity variation over the covered area. Key to this are both the layout of the compound beams on the sky (the Apertif footprint) and the layout of the grid of pointings covering the northern sky between declinations of 30 and 75 degrees.
Analysis preparation focused on source finding and source characterisation and interactive visualisation of 3D data. Two contributions under this ERC grant can already be identified as having been important. One is the contribution to improving the robustness and reliability of 3D methods to find HI emission objects in synthesis data cubes (providing 2D positional and 1D spectral information). This was part of an international collaboration and has led to a now well performing and widely used source finding system SoFiA (https://github.com/SoFiA-Admin/SoFiA) that had its third stable release in May 2018 and is moving to a next stable version to be issues early 2019. SoFiA now has a well established reputation and has begun to acquire a large user base.
The other concerns the visualisation of, and the analysis and modelling of HI (neutral hydrogen) in galaxies using synthesis spectral line data. The analysis can be greatly enhanced when coupled to quantitative and interactive 3D and 2D visualisation. A very powerful package has been developed within the existing environment of 3DSlicer (http://slicer.org a package originally for medical imaging and still mostly used for this purpose) which allows interactive visualisation and modelling of 3D radio astronomical data sets taking on board the full 3D astronomical coordinate and physical units system. The package, called SlicerAstro (https://github.com/Punzo/SlicerAstro/wiki) is an extension to 3DSlicer, but is also downloadable as a custom built 3DSlicer+SlicerAstro binary executable for Linux and MacOS platforms. SlicerAstro allows flexible 3D/2D visualisation of spectral line data, manipulation of the data (including 3D segmentation, construction of data slices, calculation of moment images) and a direct inter-comparison between kinematic models of the data and data itself. It also allows direct visualisation and comparison of data from different sources (i.e. at different wavelengths).
In addition non-parametric methods of measuring asymmetries in the distributions of HI emission in galaxies (2D) have been examined and appear to be quite sensitive to signal-to-noise, resolution and galaxy orientation. In particular the signal-to-noise is an important effect, which has been ignored in studies to date. Methods to correct for the noise bias have been designed and tested. The software will be made available as a GitHub repository. From examining model galaxies it has become clear that determination of the asymmetries in 3D, i.e. taking into account both morphological and kinematic asymmetries, is much more robust, also in the presence of noise. This entails a simple modification of the 2D software and will also be made available at the GitHub repository.
Members of the research group have contributed to both the testing and the improvement of the Apertif data calibration and imaging pipeline Apercal (https://github.com/apertif/apercal). The improvements largely focused on introducing parallel processing to speed up the processing speed. This is possible because in principle each Apertif compound beam is a separate data set and is processed separately. Apercal is a python based system making use of Casacore (https://github.com/casacore/casacore) Casa (https://casa.nrao.edu) and Miriad (https://www.atnf.csiro.au/computing/software/miriad/) for the data processing.
Science exploration involved both exploring state-of-the-art cosmological simulations as well as the reduction and analysis of existing HI observations. Comparing the simulations with observations will be increasingly important. A comparison of simulations of a single Milky Way type galaxy with differing amounts of supernova feedback with the properties of the HI and hot gas in the corona (halo) of the Milky Way has shown that considerable feedback, driving gas from the disk into the halo is required to explain the distribution and kinematics of HI emission and O VI absorbers in the Milky Way corona. A detailed study of the HI content of galaxies in groups of different total mass, i.e. different environments in terms of galaxy density, has shown that a mix of different physical processes shape the HI content (and presumably also the HI distributions) of galaxies: galaxy interactions, gas removal by tidal and ram pressure stripping, gas accretion from the cosmic web and gas removal and redistribution by the induces star formation in the galaxies. The simulations suggest that the most common cause of H I removal is satellite– satellite interactions, followed by ram pressure and tidal stripping. Ram pressure stripping is the most common mechanism that disturbs the H I morphology of galaxies at z=0, though this concerns mostly galaxies that are well within the viral radius of a group.
Ongoing HI projects have led to a number of results. Investigation of the HI Mass Function (describing the volume densities of galaxies of differing HI mass) in the Ursa Major cluster shows that the number of low HI mass galaxies is considerably smaller that found in two major surveys, HIPASS with the Parkes telescope and ALFALFA with the Arecibo telescope, of large areas/volumes. This is not yet understood. The provision should be made that the statistics at the low HI mass end is quite uncertain, though it can be clearly shown that based on the HI Mass Functions of the HIPASS and ALFALFA surveys the number of low HI mass objects in Ursa Major should have been at least a factor of four higher. Large blind volumes are required to have a definitive answer to the question of whether and how the low mass end of the HI Mass Function depends on environment.
A comparison of the HI and CO distributions with the distributions of star formation derived from UV and IR imaging data in 29 galaxies shows that all galaxies exhibit more or less the same relationship between gas surface density and star formation density. The star formation process apparently is self regulated and not affected by whether a galaxy is undisturbed, mildly disturbed or clearly interacting. This confirms the outcome of other studies of the star formation in galaxies.
A real pilot for deep HI observations with future telescopes (in particular MeerKAT and SKA) is the Chiles project aimed at a 1000 hour integration with the JVLA on the Cosmos field. Involvement in this effort has proven useful for knowledge transfer between this project and Apertif HI surveys. Analysis of the first 178 hours of JVLA data has led to the discovery of a gas rich galaxy at z=0.376 (ApJ 824, 1F, 2016). An investigation of the galaxies in the redshift ranges 0.11-0.13 and 0.16-0.18 and their relation to the large scale structure in the observed volume has been completed. Most of the detected (and hence relatively HI rich) galaxies lie close to the large scale structure filaments and spines in the volume and many show signs of disturbed morphologies. This study made use of SoFiA for locating the detections and made use of DisPerSE (http://www2.iap.fr/users/sousbie/disperse.html) to characterise the large scale structure. These methodologies proved to be very useful and will probably be used widely in future HI survey projects.
In addition one can use the Dressler the Dressler–Shectman technique for finding kinematic substructure in the large scale distribution of galaxies for a comparison with HI properties of galaxies. This was done successfully in the Antlia cluster using KAT-7 data (MNRAS 452,1617, 2015) and will also be applied to other data under investigation: the Westerbork Coma Survey (Serra, van der Hulst et al.), a 24 pointing mosaic with the WSRT, the deepest blind HI survey of the Coma cluster to date and similar blind survey projects in the Pisces-Perseus supercluster region.
Science exploration also included designing an optimal pointing grid for the Apertif imaging surveys. This included optimising between the size of the area covered and minimal sensitivity variation over the covered area. Key to this are both the layout of the compound beams on the sky (the Apertif footprint) and the layout of the grid of pointings covering the northern sky between declinations of 30 and 75 degrees.