(PhD) Sparsity-based methods for multi-frequency, rotation measure synthesis deconvolution

Supervisors: Prof O. Smirnov (osmirnov@gmail.com),

Dr Julien N. Girard (RATT/Rhodes & AIM/CEA-Saclay) (jgirard@ska.ac.za / julien.girard@cea.fr)

Recent progress in compressive sensing (CS) methods has yielded a number of deconvolution algorithms based on the sparsity assumption of the sky that seem far better at recovering extended emission than CLEAN and its multiscale variations. For example, the MORESANE (Dabbech et al. 2015, arXiv:1412.5387) algorithm as implemented and released by J. Kenyon (Rhodes PhD student, see http://github.com/ratt-ru/PyMORESANE) has already  given us the deepest-ever images of Cyg A in S-band (using JVLA data). Other promising algorithms include PURIFY (Carrillo et al. 2014) and SASIR (Garsden at al. 2015, arXiv:1406.7242 -- one of the key developers of the latter did a postdoc at Rhodes/SKA SA and remains a collaborator of the project). New work by Ferrari et al. (2015, arXiv:1504.06847, same team as MORESANE) and Jiang et al. (2017, arXiv:1703.02650, same team as SASIR) have extended previous into the multi-frequency regime (and other teams are working on similar extension). We propose to develop, jointly with the Ferrari et al. team (and possibly other teams, e.g. using Morphological Component Analysis at AIM and more generally, within the SKA-France consortium), a sparsity-based deconvolution method for full-polarization multi-frequency rotation measure (RM) synthesis.
Relevance to MeerKAT and the SKA. MeerKAT and SKA1 will enable wide-band radio observations to unprecedented sensitivity. A lot more faint extended and polarized emission will be detectable, but the wide frequency band makes it crucial to incorporate the frequency and polarization axis into the deconvolution process if the full potential of these instruments is to be realized.

(MSc/PhD) Bayesian Inference for Polarimetric Calibration

Supervisors: Prof O. Smirnov (osmirnov@gmail.com),

Recent work on BIRO (Bayesian Inference for Radio Observations) by Lochner at al. (2015) and Natarajan et al. (in prep.) has demonstrated the promise of Bayesian inference techniques in the domain of radio interferometric calibration. We now propose to apply BIRO to polarization calibration. Accurate polarization calibration in radio interferometry is a difficult problem, as the instrumental polarization leakage terms interact with the time-variable gain terms, and also need to be disentangled from intrinsic source polarization. The matrix-based radio interferometry measurement equation (RIME) provides a complete mathematical model for forward modelling of this process; inverting the model to go from observed visibilities back to polarization properties is a different matter. The Bayesian approach
seems ideally suited to this, its particular strength being the ability to map out correlations between instrumental and scientific parameters while imposing prior constraints.
Relevance to MeerKAT and the SKA. A number of key MeerKAT and SKA1 surveys will rely on accurate polarimetric calibration to achieve their science goals. The problem is especially severe with the linearly polarized feeds employed by KAT-7, MeerKAT and SKA1. Urgent progress on this problem is required.

(PhD) Understanding the Limits of Interferometric Techniques for Epoch of Reionisation Detection

Supervisors: Prof O. Smirnov  (osmirnov@gmail.com),
Prof F.B. Abdalla  (UCL/Rhodes) (fba@star.ucl.ac.uk)

The Epoch of Reionisation (EoR) is the next frontier for the cosmologist. The faint signal arising from this epoch can be measured by radio interferometers at low frequencies capturing the redshifted signal from the 21cm line. However, an outstanding issue not clearly resolved by any current EoR experiment is that calibration issues could plague this signal and prevent a clear detection from taking place. In particular, it is possible that direction

dependent effects (DDEs) will change the nature of this signal and smear any possible detection. We propose to investigate how DDEs affect this cosmological signal, and whether the statistical properties of this signal are maintained after DDEs are calibrated out, and more specifically if the magnitude of the fluctuations is maintained. The student would obtain simulations of the 21cm signal, pass them through a radio interferometry simulator, and obtain visibilities which would be corrupted with a direction dependent signal. This signal would be calibrated with DD-solution algorithms such as SAGECAL or StefCal, and the signal would be remeasured and compared to the original input. The result of this research would place clear constrains on our ability to measure the EoR signal using current and

future interferometers, and inform the calibration strategy of future EoR experiments, and

(PhD/MSc) Analyzing The Impact Of Ghosts On Future Cosmological Radio Surveys

Supervisors: Prof O. Smirnov  (osmirnov@gmail.com),
Prof F.B. Abdalla  (UCL/Rhodes) (fba@star.ucl.ac.uk)

Ghosts are calibration artefacts that are always present in radio maps (even if only below the noise) that are calibrated with incomplete sky models, as would be the case for any blind survey. While the presence of ghosts and the mechanism by which they form has been conclusively established (see recent work by Grobler, Smirnov et al.), their impact on the science performance of future deep surveys needs to be studied. Ghosts show up in specific configurations determined by the interferometer layout, and In principle affect the statistical distribution of galaxies in an image. The aim of this project is to obtain a scientific measure of how much ghosts affect the power spectrum in the intensity mapping of galaxies and to determine if they are serious enough to put a show-stopper in cosmological measurements of the galaxy distribution. The student would attempt to calibrate radio data with incomplete models and hence obtain images with ghosts and obtain a measurement of the n-point correlation functions to measure cosmological parameters. These measurements would be biased due to the presence of calibration ghosts. The magnitude of the measurements would yield information about how well calibration needs to be done (i.e. how deep the sky models need to be) in order to obtain correlations of galaxies which are not corrupted by the presence of ghosts. Other related observational techniques such as intensity mapping. this is directly relevant to SKA1 and SKA2. The results of this work will inform the calibration strategies for the next generation of deep radio surveys with JVLA,

MeerKAT and SKA1.

(PhD)  Kinematics and dynamics of resolved MALS emission line systems

Supervisors:

Dr G. I. G. Józsa (gigjozsast@gmail.com)

Prof O. Smirnov (osmirnov@gmail.com)

Dr N. Gupta (ngupta@iucaa.in)

Dr P. Kamphuis(kamphuis@astro.rub.de)

The MeerKAT Absorption Line Survey (MALS) will target ~1100 pointings centered at bright radio sources and will be sensitive to detect cold atomic and molecular gas in absorption lines of HI and OH towards >25000 compact radio sources. It will hence become the largest systematic absorption line survey to date targeting several science goals, from characterization of the  interstellar medium of distant galaxies to investigating the variations of fundamental constants of physics.

MALS is at the same time a moderately sensitive HI blind emission line survey. The targeted background continuum sources are at high redshift and therefore, relative to the local structure, randomly distributed across the sky. In emission only local gas is detected (because of a lack of sensitivity to detect gas at high redshift), making MALS a blind HI emission line survey.

In this project, the PhD candidate will analyse the many serendipitously detected resolved emission line objects in the MALS survey. The first step is to provide an automized method in the MALS pipeline to perform a tilted-ring analysis of sufficiently extended sources (up to several 100 galaxies with a diameter of > 5 beams) as detected in MALS. With this we will quantify the HI rotation curves, the spin orientation, and the HI distribution in MALS emission line objects. The PhD candidate can rely on ample experience in RATT to perform this task, and single software components for first tests are already at our hands (TiRiFiC and FAT). The candidate will then test if a full Bayesian inference tool to derive tilted-ring models, as currently being developed at RATT, can be implemented in the pipeline. In a second step the resulting parametrisations will be analysed to investigate statistical trends for the observed galaxies:  how do mass-distribution and rotation curves relate to the environment, the host's Hubble type, colour, mass, luminosity? How does the disk's shape relate to the environment? Are certain types of galaxies more disturbed than others? But we also want to answer the question how absorption systems relate to galaxies in general: do they originate from gas associated with the extended disk or clouds outside the galaxy? MALS is the first MeerKAT survey that enables us to address these questions, and it might be followed by other systematic blind surveys that will profit from the experience gained.  This project is hence an exciting opportunity to enter the topic of statistics in resolved HI emission line systems in large radio surveys.

(PhD/MSc)  Characteristics of HI-dominated systems from WSRT and MeerKAT surveys

Supervisors:

Dr G. I. G. Józsa (gigjozsast@gmail.com)

Prof O. Smirnov (osmirnov@gmail.com)

Large blind HI surveys like HIPASS and ALFALFA have revealed hat most neutral gas in the local universe is bound to stellar systems and there are only few local objects whose mass in dominated by neutral gas. Less than 1% of all HI detections in ALFALFA cannot be associated with a stellar system and there are only a few systems that contain extreme gas mass-stellar luminosity ratios MHI/L* of more than 20 M/L. While single-dish surveys are suitable to identify such objects their HI nature had so far to be investigated by pointed surveys with radio telescopes. This will change with blind interferometric radio surveys like WALLABY with ASKAP, but also MALS with MeerKAT, which is on the brink of collecting its first data. The systems can roughly be separated into tidal systems, in which gas is separated from a host galaxy through interactions, intergalactic clouds without obvious origin, extreme low-surface brightness galaxies, an extreme galactic population, and truly "Dark Galaxies", systems consisting solely of Dark Matter halos and neutral hydrogen. To date, no such Dark Galaxy has yet been identified, while the ALFALFA catalogue contains at least one very intriguing system in which several extreme low-surface brightness galaxies share the same volume, coined HI-bearing Ultra-Diffuse Galaxies.

In all these cases, the neutral hydrogen traces extreme physical conditions: how can an (apparently) isolated cloud survive against the ionizing UV background? Is an isolated cloud in fact only part of huge tidal features or even primordial HI shared by a group of galaxies? How long can tidal features grow?

In this thesis, the candidate will look in detail at a small number (~10) of WSRT observations of HI dominated galaxies to investigate their physics. The sample has been selected from a catalogue of (almost) Dark systems in ALFALFA. Once these have been studied, the candidate will attempt to characterize the detection fraction of dark systems in the MALS survey with MeerKAT to collect ancillary observations of the most intriguing cases. This thesis will prepare the analysis of such systems which is expected to appear in larger number in blind SKA- andSKA progenitor surveys.

(PhD)  Blue Early Type Galaxies with MeerKAT and other telescopes

Supervisors: Dr G. I. G. Józsa (gigjozsast@gmail.com)

Prof O. Smirnov (osmirnov@gmail.com)

 Dr O. I. Wong (ivy.wong@uwa.edu.au)

In this study, we will embark on a multi-wavelength campaign to study the nature of Blue Early Type Galaxies.

Normally, galaxy either belongs to the population of blue, gas-rich spiral galaxies or it is a red, gas-poor elliptical galaxy. To understand cosmic evolution it is crucial to understand the evolution into this bimodal distribution, and to understand whether and how transitions from one type to the other take place. A clue might be provided by (rare) intermediate or mixed types.

Blue Early Type Galaxies (BETGs) appear blue, which is a signature of star formation, either ongoing or very recent, and yet their optical shape is that of an elliptical galaxy. We have argued that they are a fast transition type from spiral to elliptical. In that picture star formation is shut down rapidly, potentially induced by nuclear activity. This is supported by interferometric observations showing that with increasing stellar age the cool neutral gas is found further away from the galaxies as if being pushed out. Other studies claim that BETGs would have recently acquired gas and re-start (some) star formation.

To solve the question what BETGs are, multiwavelength studies of a larger sample are required. We hence propose a PhD project to conduct a targeted survey of  a moderate number of (~35) Blue Early Type Galaxies with MeerKAT, other radio telescopes (HI and continuum to study HI structure and kinematics and nuclear activity), and optical telescopes (deep photometry to identify past interactions and the star formation history, spectroscopy to characterise nuclear and star forming activity). The definition of the sample will be informed by recent HI observations with the Arecibo radio telescope to select an HI-bearing sample. The student’s task is, to perform observations, data reduction, and analysis. Such a study has the potential to strongly support or disprove the hypothesis of gas-removal by kinetic feedback or gas accretion by gas capture.

(PhD/MSc) Classifying and interpreting marginally resolved or unresolved detections in HI surveys

Supervisors: Dr G. I. G. Józsa (gigjozsast@gmail.com)

Future blind and targeted HI surveys with MeerKAT, and other SKA progenitor surveys will provide a large amount of resolved detections of galaxies in the HI emission line, for which a more detailed analysis of the rotation curve, the orientation of the disk, and other parameters is possible. The vast majority of the detected galaxies, however, are unresolved or marginally resolved, having a diameter between 1 and 3 resolution elements.

While the basic parameters, the line width and the total flux already allow to investigate the rotational amplitude (in combination with ancillary data) and the HI mass, the data contain, at least on a statistical basis, much more information. For example, the shape of the rotation curve determines whether the profile is centrally peaked (for e.g. solid-body rotation curves) or whether it has the typical two-horned shape (for flat rotation curves). At the same time, the shape can also be influenced by other parameters, physical and geometrical. The centrally peaked profile would also turn up for a single rotating ring (a physical property) or for a more face-on orientation (a purely geometrical property). These parameters are hence correlated.

But it should be possible to infer some more information and break these degeneracies using anillary data (the optical luminosity, e.g., will also inform about the rotational amplitude via the Tully-Fisher-Relation, which we can use to get information about the shape of the rotation curve). Also, if more than one resolution element is available, this can possibly be used to break degeneracies even better.

The candidate will have the task to investigate how much information is really contained in HI spectra and marginally resolved data cubes. One way to do this is using simulated data using one of our simulation tools (TiRiFiC), but also available data of well-resolved galaxies, studying degeneracies between galaxy parameters (describing well-resolved galaxies) when interpreting spectra. Once this is accomplished, we will turn to real surveys with MeerKAT and other telescopes (WALLABY, MALS, to which we have direct access) to interpret the spectra of marginally resolved galaxies in those surveys.

This is a project for a candidate who is familiar with statistical methods. If successful, it can be applied to a vast number of detections in future surveys.

(PhD/MSc) Studying the Epoch of Reionization and Cosmic Dawn with HERA

Supervisors: Prof. Gianni Bernardi (giannibernardi75@gmail.com)

Prof. Oleg Smirnov (osmirnov@gmail.com)

Our current understanding of the cosmic evolution from the epoch of Hydrogen recombination (300000 years after the Big Bang) till the first billion years is very poorly known from an observational perspective, although it must have seen the growth of the initial density perturbations via gravitational attraction into the first stars and galaxies.

The 21cm transition from neutral Hydrogen promises to be the best observational probe of such cosmic time and has driven the construction of the new generation of low frequency radio arrays. The HI Epoch of Reionization Array (HERA) is the next generation of such low frequency arrays and is a partnership between several US, UK and South African institutions. The first 80 HERA dishes have now been built, with 331 elements expected to be completed by the end of 2019, becoming, by then, the most sensitive 21-cm telescope in the pre-SKA era.

We are seeking for an ambitious Msc/PhD candidate who is willing to become an observational 21cm cosmologist and is ready to face the many challenges that such field presents. The candidate will be able to participate in the HERA commissioning activities as well as its data analysis towards measurements of the 21-cm signal. In particular, the candidate is expected to work on the crucial topic of foreground subtraction and 21-cm power spectrum estimation. As HERA is a collaborative effort, the candidate is expected to work together with the US and UK partners.

An ideal candidate will have a good background in physics and mathematics, the willingness to learn about advanced radio interferometry, statistics, signal processing techniques and cosmology, but, mostly, the desire to contribute to the 21-cm cosmology revolution

(MSc)  Facet-based primary beam corrections for HERA imaging

Supervisors:        Prof. Oleg Smirnov (osmirnov@gmail.com)

Prof. Gianni Bernardi (giannibernardi75@gmail.com)

Dr Cyril Tasse (cyril.tasse@obspm.fr)

Although designed primarily as a power spectrum experiment, HERA can also be used as an all-sky imaging instrument. Imaging with HERA is important for several reasons: as a cross-check of the calibration and instrumental stability, as an aid to bright foreground subtraction, and as a low-frequency all-sky monitor of transient events.

The main difficulty in imaging HERA data is that it is a transit instrument: sources drift through the primary beam of the instrument, and a continuous direction-dependent full-Stokes correction is required to account for this effect.

This project aims to leverage the new DDFacet imager (developed jointly between Rhodes, SKA SA and Paris Observatory) to develop a facet-based imaging strategy for HERA using full polarization corrections, and to demonstrate it on HERA data.

(MSc/PhD) Observations of diffuse radio emission in galaxy clusters

Supervisors: Prof. Gianni Bernardi (giannibernardi75@gmail.com)

Prof. Oleg Smirnov (osmirnov@gmail.com)

Dr. Tiziana Venturi (tventuri@inaf.ira.it)

Galaxy clusters are the largest gravitationally bound structures and emit electromagnetic radiation at almost all wavelength. In particular, some of them host diffuse, Mpc-scale, low-surface brightness radio emission that has been the subject of intense investigation over the last two decades. Developments in recent observations and theoretical models suggest that such structures are created by merging events, when magnetic fields are amplified and particles re-accelerated to relativistic energies but additional observations are required to further test models.

The candidate will have the opportunity to work on MWA observations of the cluster sample observed by Bernardi et al. (2016) with KAT-7. The aim of the KAT-7 observations was to assess the presence of radio halos in a sample of mass-selected nearby cluster in order to provide a crucial test of radio halo formation models. The MWA follow up of the KAT-7 sample is aimed at characterizing the low frequency (~200 MHz) properties of the radio emission, searching, in particular, for steep spectrum radio halos predicted from re-acceleration models (Cassano & Brunetti 2005).

Observations of diffuse emission from galaxy clusters are also ideal laboratories to test advanced calibration (i.e. direction dependent calibration) and imaging techniques (PyMORESANE, DDfacet, WSClean) as they require a careful analysis to extract the low brightness surface structures.

An ideal candidate will have the interest in developing a radio observational expertise, familiarizing/developing calibration and imaging pipelines and the desire to work on a prominent field (galaxy clusters) that will be relevant for MeerKAT observations.

Finally we note that, for a PhD project, that candidate may expect to have access to MeerKAT cluster observations from their second year.

(PhD/MSc) Peering through the Cosmic Dawn

Supervisors: Prof. Gianni Bernardi (giannibernardi75@gmail.com)

Prof. Oleg Smirnov (osmirnov@gmail.com)

Our current understanding of the cosmic evolution from the epoch of Hydrogen recombination (300000 years after the Big Bang) till the first billion years is very poorly known from an observational perspective, although it must have seen the growth of the initial density perturbations via gravitational attraction into the first stars and galaxies.

The 21-cm transition from neutral Hydrogen promises to be the best observational probe of such cosmic time and has driven the construction of the new generation of low frequency radio instruments. The Large aperture Experiment to detect the Dark Ages (LEDA) was built to attempt the detection of the sky-averaged 21-cm from the Cosmic Dawn at 15 < z < 36, before widespread reionization took place, in order to understand the formation of the first luminous structures in the Universe.

The candidate will work on the analysis of LEDA radiometric data, with the aim to calibrate them and extract the 21-cm signal. Depending upon the candidate's inclinations he could work on the analysis of LEDA interferometric data, with the goal of calibrating them and extract the 21-cm power spectrum.

An ideal candidate will have a good background in physics and mathematics, the willingness to learn about advanced radio interferometry, statistics, signal processing techniques and cosmology, but, mostly, the desire to contribute to the 21-cm cosmology revolution.

(PhD) Development of new calibration techniques for variable sources using the MeerKAT/KAT7 observations

Supervisors: Dr. S.K. Sirothia (sirothia@gmail.com)

Prof. Oleg Smirnov (osmirnov@gmail.com)

The project is divided into multiple parts
(i)  Develop a method to identify variable sources while mapping using the existing multi-epoch data of MeerKAT AR1. Example of Extragalactic variable sources are blazars and galactic variable sources are pulsars .
(ii) Study the impact of these variability on calibration solutions and dynamic range of the map. These sources will directly affect direction dependent solving techniques, whereby the source variability will get attributed to instrumental and beam related errors, thereby also spoiling flux density measurements of sources in their vicinity.
(iii) Subsequently, develop new calibration techniques taking into account the variability of such sources and improve dynamic range of map.

During first year method will be developed to identify variable sources. Flux density, time series related study of extragalactic variable sources, like blazars and galactic variable sources like pulsars using the existing multi-epoch data of MeerKAT AR1 to establish and quantify the effect. During second year we study the impact of these variability on calibration solutions and dynamic range of the map and develop new calibration techniques.
During third year special emphasis will on
(a) develop new calibration techniques taking into account the variability of such sources
(b) subsequently, high dynamic range images will be made taking into account the variability of such sources.

Multi epoch data of selected fields from MeerKAT AR1/KAT7 archive will be used. Some of the datasets have specifically been obtained by proposer for such studies. No new observations and proposals are being required for this purpose.

Relevance of the research to the research priority areas of MeerKAT and SKA: Presence of variable sources in any field impact calibration solutions and dynamic range of map. In addition, the direction dependent solving techniques gives undesired results, whereby the source variability will get attributed to instrumental and beam related errors, and even result in spoiling the flux density of sources in the immediate vicinity of such variable sources. The project attempts to address issues so that the flux density accuracy and dynamic range of the images from MeerKAT and SKA can be improved.


(MSc) Developing advanced beam models of MeerKAT antennas using interferometer data


Supervisors: Dr. S.K. Sirothia (sirothia@gmail.com)

Prof. Oleg Smirnov (osmirnov@gmail.com)


The image is obtained by combining visibilities from various antennas having varied beams. A nominal primary beam correction is applied to the image, to get the true image. However, the interferometric image beam could be widely different from individual antenna beams. Techniques will be developed to measure the combined interferometer beam using the data used for mapping, and errors will be quantified as compared to individual antenna beams.

During the first year the candidate will learn the analysis techniques for making MeerKAT AR1 images. Also a framework will be developed to determine beam using the data, using sources and noise maps of the field of view. During second year the analysis technique developed for determination of beam will applied on actual cases and results compared with those obtained from the conventional methods of beam measurements.

Test observations using MeerKAT
 AR1 will be used for the purpose of determination of primary beams of individual antennas and the combined interferometer beam, simultaneously with the observations.

Relevance of the research to the research priority areas of MeerKAT and SKA: Accurate primary beam correction is essential for making wide field maps and surveys. Conventionally, individual antenna beams are separately measured and their models are used for primary beam corrections while mapping. The project aims to determine the primary beams of individual antennas using the visibility data during the observation in interest. The method could improve flux density accuracy from interferometric arrays like MeerKAT.

(MSc) Locating stationary Radio Frequency Interference (RFI) sources MeerKAT data


Supervisors: Dr. S.K. Sirothia (sirothia@gmail.com)

Prof. Oleg Smirnov (osmirnov@gmail.com)

Radio Frequency Interference (RFI) is one of the major challenges in MeerKAT data analysis. Currently, flagging of data is mostly used during analysis to minimize the effects of RFI.

This project aims to locate the nearby stationary RFI sources in the existing MeerKAT interferometric archival data. We intend to use the near field Electromagnetic equations and interferometric techniques to get the location of these active RFI sources and determine their statistical properties.

Multi epoch data of selected fields from MeerKAT-AR1 archive will be used. Some of the datasets have specifically been obtained by proposer for such studies. No new observations and proposals are required for this purpose.

(PhD) A proposed design for Baseline-dependent correlators using Farrow filters combined with the raised cosine

Supervisors: Prof. Oleg Smirnov (osmirnov@gmail.com)

Dr Marcellin Atemkeng (m.atemkeng@gmail.com)

                                                Dr Griffin Foster (griffin.foster@gmail.com)

Baseline-dependent averaging (BDA) is a potential uv-data compression scheme that could

be adopted for next generation radio telescopes such as MeerKAT and the future SKA. A

large body of work has already been done regarding BDA applied to MeerKAT, VLA, EVN

and LOFAR (e.g. Atemkeng et al., MNRAS 2016; C. Tasse et al., A&A, submitted; Atemkeng

et al., MNRAS, in prep.). Currently, BDA can only be used post-correlation and not in

real-time. The main aim of this work is to focus on a correlator design that implements BDA

in real time correlations.

This project will use finite impulse response filters (FIR) (e.g. the raised cosine) and Farrow

filters to implement a low pass and decimation filters to design a possible correlator

work-flow for BDA. If a long baseline is down sampled, then the decimation phase filters will

consist of interpolating the signal by some factor and decimate with an offset by taking the

baseline-dependent delay parameters into account: this is equivalent to decompression on

these type of long baselines and this will preserve uv-bins resolution. If a short baseline is

oversampled, then the decimation phase filters will resample the data according to the

variable delays: this is equivalent to compression.

A naive implementation would be very inefficient given that most of the decimation filters’

band will not contribute to the correlator output. An ideal FIR filter will be the one that uses

as few samples as possible, has good alias rejection, and is computationally efficient. This

motivates the use of Farrow filters for this project.

MeerKAT is already producing a large volume of data. The data rate of the future SKA will

be several orders of magnitude higher, and will increase even more for any SKA surveys

that adopt multiple phase tracking (e.g. AVN). New data compression algorithms and

storage systems are key techniques to handle these volume of data. A large field of view

(FOV) is essential for the SKA, and the current compression method i.e., averaging in the

correlator suffers from limited FOV and compression factor due to decorrelation/smearing.

BDA has been shown (in post-correlation) to be a potential compression method: we need to

investigate and design BDA in real life correlator architectures.

BSc Hons - Computation of closure relations for imaging in the absence of calibrators

Supervisors - Prof. Oleg Smirnov (osmirnov@gmail.com)

Dr. Landman Bester (landman.bester@gmail.com)

The lack of good calibrators for extreme imaging projects such as the Event Horizon Telescope necessitates a rethinking of the traditional imaging pipeline. One way of generating an initial sky model that as largely independent of instrumental and atmospheric systematic errors is to use phase and amplitude closure relations to design the likelihood function used during imaging. Such a likelihood function can be used in conjunction with regularisers (priors) which leverage sparsity to yield a convex optimisation problem. This way the first imaging step is essentially guaranteed to find a good initial model with which to perform self calibration.

The primary aim of this project would be to implement an algorithm which computes the closure phases and amplitudes for the visibilities corresponding to an arbitrary array configuration. A secondary aim would be to investigate how these relations could be used to design statistically robust likelihood functions that can be plugged into existing imaging algorithms. Apart from their usefulness in imaging in the absence of calibrators, the closure relationships break the degeneracy between the imaging and calibration problems (to some extent) and could therefore be useful in sampling based approaches which aim to quantify image uncertainty.    

(MSc/PhD) Identifying PSF patterns around sources using deep learning techniques

Supervisors: Prof Oleg Smirnov (osmirnov@gmail.com)

Dr Marcellin Atemkeng (m.atemkeng@gmail.com)

New radio interferometers rely on very large volume of data to be transferred between various parts of data processing over distances that are often several kilometers. A typical example is the MeerKAT telescope. Its primary beam at the full-width half maximum and at 1.4 GHz covers a field of view (FOV) across ~2.6 deg. At this frequency, and for about 15 s time and 0.84 MHz frequency bins resolution, sources are degraded towards the edges of the FOV. To retrieve the total FOV the data should be correlated using high time and spectral resolution, however, this leads to high data volume. Any radio interferometric data compression method that is based on averaging is detrimental to the image fidelity because the PSF is distorted and attenuated (smearing). Radio interferometric algorithm for source finding relies on the clean image. The clean image is produced by a deconvolution algorithm, which means that if the clean algorithm does not perform accurately then the source finding algorithm will fail. Clean algorithms are regarded as an iterative approach, where the brightest pixel value is found step by step and subtracted from the image until the loop reaches a given threshold. During each step, the brightest pixel is considered to be a source, which may not be the case, however, the brightest pixel could be an artifact or a ghost source that does not have the PSF structure. There have been efforts to detect sources and calibration artifacts in an automated fashion using machine learning methods (e.g., Aniyan et al, in prep.). This gives the

potential for application of machine learning based methods for similar problems in the discussed context. Finding the PSF structure around an object in the dirty image is quite intuitive and was impossible in the past since the PSF is position-dependent and expensive to compute iteratively for each location in the dirty image where there is a brighter pixel. A framework to accurately compute these position-dependent PSFs with fewer computations is established in Atemkeng et al, in prep.

This work will use machine learning algorithms and the position-dependent algorithms to find the structure of a given position-dependent PSF at the location of a potential source (the source located at the brightest pixel in the dirty image). If the structure matches (for a given probabilistic trust) then the algorithm should classify this source as a true source that should be deconvolved, if not the source is an artefact that should not be included in the clean image/restored map.

(MSc/PhD) Using machine learning techniques to derive calibration priors from MeerKAT telescope sensor data

Supervisors: Prof Oleg Smirnov (osmirnov@gmail.com)

MeerKAT’s control and monitoring system (CAM) routinely collects large amounts of environmental and telescope sensor metadata alongside any given observation. This is also the case for other radio observatories. Currently, very little of this metadata is actually used in subsequent science data calibration (apart from setting initial sensitivity weights, diagnosing error conditions, etc.), primarily because the relationships between specific sensors and observed calibration parameters are highly non-linear and very poorly understood. Current telescope calibration procedures therefore proceed from scratch, without reference to sensor metadata.

Recent MSc work by S. Zitha, based on observational data and metadata from the KAT-7 telescope, showed that machine learning (ML) techniques can be used to determine these relationships empirically, and that sensor data can (sometimes) be used to predict calibration parameters (in particular, telescope gains). Potentially, this allows for (a) cross-checks on traditional calibration techniques, (b) improved calibration estimates at the beginning of the self-calibration cycle, (c) fewer calibrator visits.

This project aims to extend this work to the MeerKAT telescope, using MeerKAT commissioning data, and to develop a practical tool that can provide initial calibration estimates to data processing pipelines.

(MSc) The compact radio source population towards the Galactic Centre

Supervisors: Prof Ian Heywood (ianh@astro.ox.ac.uk)

Prof Oleg Smirnov (osmirnov@gmail.com)

This project will make use of the groundbreaking MeerKAT observations of the Galactic Centre. The radio waves penetrate the high levels of dust in this region, revealing many bright and mysterious radio structures, along with a large number of very compact features. Most stars are too faint at radio wavelengths to be routinely detected, so the true point-like sources (as opposed to things that are merely small in angular scale) will be a combination of external galaxies very distant from our own, as well as exotic objects in our own galaxy such as pulsars, and stellar-mass black holes in interacting binary pairs. Re-imaging of the MeerKAT galactic centre data in two frequency chunks will provide an initial classification for many of these object based on their spectral shape, and for this project additional radio observations using the Karl G. Jansky Very Large Array are also available. These are in the most extended A-configuration, so much of the diffuse emission seen in the MeerKAT image is resolved out, leaving only the compact features. Multi-wavelength data can also be used as a further discriminator, and potentially interesting objects, many of which will have never before been seen, can be identified for follow-up observations. The student undertaking this project will develop radio interferometry skills and knowledge of many different astrophysical phenomena, leading to the production the most comprehensive catalogue to-date describing the properties of compact sources towards the centre of the Milky Way.

(MSc) MeerKAT’s unobscured view of star formation in the early Universe

Supervisors: Prof Ian Heywood (ianh@astro.ox.ac.uk)

Prof Oleg Smirnov (osmirnov@gmail.com)

Abell S1063 is a distant cluster of galaxies, with a total mass of about 100 trillion times that of the Sun. This huge concentration of mass acts as a strong gravitational lens, magnifying the galaxy population behind the cluster. This magnification boosts the apparent brightness of the background galaxies, allowing the detection of objects that are intrinsically very faint, and would otherwise be non-detectable using reasonable amounts of telescope time. This project will exploit recent, deep MeerKAT observations of this cluster to search for distant galaxies with high magnification factors, and investigate the faint population of star-forming galaxies during the peak epoch of star-formation in the Universe. Abell S1063 is one of the Hubble Frontier Fields, and thus has a wealth of publicly available optical and infrared data, as well as high-quality lens magnification models to aid this analysis. Higher frequency radio observations from the Australia Telescope Compact Array in support of this project will also be used.

(MSc) Secondary Ghosts in Radio Interferometry

Supervisors: Prof Ian Heywood (ianh@astro.ox.ac.uk)

Dr. Sphesihle Makhathini (sphemakh@gmail.com)

Calibration of radio interferometer data involves constructing a model of the sky and then solving for the instrumental parameters that minimise the difference between this model and the observed measurements. Incomplete sky models during self-calibration have been shown to lead to the appearance of patterns of ‘ghost’ sources in the image, spurious features which can be mistaken for genuine astrophysical emission. Dish arrays such as MeerKAT typically make use of a strong, compact calibrator source which is used to calibrate the array at regular intervals prior to self-calibration of the target field. However because of its wide field of view and high sensitivity, additional sources can be detected in these calibrator fields, breaking the standard assumption that the calibrator is an isolated point source. The use of an incomplete calibrator model has been seen to impart ghost sources into the target field in early MeerKAT data for the first time. This project will use observations and simulations to investigate this phenomenon and quantify its effects on MeerKAT observations. The project will directly influence the calibration strategy for MeerKAT and SKA-MID.