Research Career

January 2005 – april 2012


A summary of my postdoctoral research at the University of Oxford – Department of Physics, Liverpool John Moores University – Astrophysics Research Institute, and Lawrence Livermore National Laboratory – Institute of Geophysics & Planetary Physics.



Galaxies like our own Milky Way typically form in groups and make up a small fraction of the web-like structure of our Universe. Resembling the complex, network of Neurons in our brain, the “cosmic web” contains clusters (or nodes) and filaments of galaxies stretching across the Universe for millions of lightyears.

My research focussed on galaxy clusters – the most massive gravitationally bound structures in the Universe; understanding how cluster galaxies evolve and what that tells us about the physics of the Universe. More specifically, I worked on the structure of dwarf & massive galaxies and their stellar populations, to understand how they evolve in the cluster environment. I was also involved in the first discoveries of ultra-compact dwarfs (UCDs) in the Fornax, Virgo and Coma galaxy clusters; a new galaxy type; “mini” galaxy discovered deep in the cluster core. .

“Hubble isn’t just a satellite, it’s about humanities quest for knowledge”
— John M. Grunsfield, American Physicist, former NASA Astronaut & Hubble Hugger

Scientific Programming & Data Analysis

The majority of my astronomy research was done using IDL; UNIX languages; (e.g. Bash, AWK) and shell scripts (e.g C–shell for wrappers and batch jobs); the PGPLOT graphics subroutine library; and domain specific software such as SExtractor (source detection and extraction), IRAF (primarily STScI packages), Supermongo, and database programs like TOPCAT.

  • Mathematical programming techniques include identifying trends and correlations in data, constructing and fitting multivariate models, contour and surface analyses, advanced statistical analysis, and robust error estimation

  • Astronomy instrument specific software: HST/ACS multidrizzle reduction, WIYN/Hydra software, AAT/2dFdR & UKST/6dFdR reduction software, Magellan/COSMOS software, Keck/HIRES Redux (IDL based) software, Keck/DEIMOS, OSIRIS and NIRC2 pipeline reduction software.

  • Web development using HTML & CSS

Angular Momentum vs. The Cores of Galaxies

Astrophysics Group, Department of Physics, Oxford University – UK

November 2011 – May 2012

Scientific Research GOalS:


In November 2011, I moved to the University of Oxford to work with Roger Davies and Davor Krajnovic on a follow-up observations for the ATLAS3D Galaxy Project. The project is a multi-wavelength survey of a complete sample of early-type (typically elliptical) galaxies with numerical simulations and semi-analytic modeling of galaxy formation. The project aims to quantify the global stellar kinematics and dynamics and relate this to their formation and evolution.

The goal of my research was to determine the internal, three-dimensional structure of ~130 ATLAS3D galaxies and to explore correlations with their known kinematic properties. The modelling was done using high-resolution archival data from the Hubble Space Telescope's (HST) Advanced Camera for Surveys (WFC/ACS) and Wide-Field Planetary Camera (WFPC2/WFC+PC).

Prior to this work, the project team had focussed on understanding the kinematic properties – by looking at how the stars move within the galaxy, and their molecular and neutral gas distributions – the fuel for making stars. The data enables the ATLAS3D team to explore the underlying physics, specifically the angular momentum of a galaxy – or how fast or slow it rotates.

The analysis was based on data from the powerful SAURON optical integral-field spectrograph on the William Herschel Telescope (WHT), radio observations with the Westerbork Radio Synthesis Telescope (WRST) & IRAM 30m telescope, and millimeter observations using the Combined Array for Research in Millimeter-wave Astronomy (CARMA).

Why is this important? Dichotomies of physical parameters offer safe anchor points to which one can tie theoretical scenarios of galaxy evolution. Previous analyses of imaging and kinematic data showed that (to first order) there are two types of elliptical galaxies: luminous, slow rotating ellipticals with round or boxy isophotes, and faint, fast rotating ellipticals with disky isophotes.

Data Analysis & Processing Techniques:

The Hubble Space Telescope (HST) images were used to probe the internal structure, specifically, whether or not the Atlas3D galaxies had cores (flat centres, or deficits of light). Usually cores are found only in massive galaxies (highest central densities) and only in systems with a stellar mass (measured via dynamical models) greater than 8 × 1010 M. But when you look at the SAURON kinematic propertes of galaxies (e.g. angular momentum) there is clear trend that cores are indeed found in slow rotators, while power-laws (or excesses of light) are found in fast rotators, independent of how luminous (or how massive) a galaxy is.

  • A custom data analysis pipeline was developed using the Space Telescope's STSDAS IRAF packages and various IDL astro libraries. IDL is vectorized, numerical, and interactive, and is commonly used for interactive processing of large geopspatial, medical and astronomy datasets. The syntax includes many constructs from Fortran and some from C.
  • The CCD images are processed to remove bad data, e.g. defects on the CCD such as null or spurious values. Since the fitting algorithms can be biased spurious data (e.g. foreground stars, bad pixels on the CCD), it's important that the data be cleaned properly, and that the properties of the data are well understood (e.g. signal-to-noise, error properties, instrument calibration).
  • The analysis involves fitting 2D intensity contours or ellipses across a 2D image of the galaxy. The intensity, I(Φ) is azimuthally sampled along each ellipse, described by the semimajor axis length from the ellipse centre, position angle and ellipticity, ε. As the fitting proceeds the semimajor axis is increased logarithmically such that the proceeding ellipse has a semimajor axis which is 10% larger. The best-fitting parameters for each ellipse (ellipse centre, and ε) are determined by minimizing the sum of squares of the residuals between the data and the first two moments in the Fourier expansion.
  • A surface brightness profile is then constructed from the fitted ellipses (see images below).
  • For each galaxy a Nuker Law, was fit to each galaxy profile. This an empirical model; a double power-law involving three unknown, or fitted parameters. These provide physical constraints for each galaxy, which can then be used in theoretical models.
  • Once the fit parameters were determined, linear regression analysis was used to identify and explore correlations, and test hypotheses.

Challenges with the Data:

The data processing and cleaning was relatively straightforward. The bulk of the manual effort went into compiling a comparative dataset. This is not always trivial. The HST archival images were taken with different camera – each having a different CCD response, pixel resolution, position relative to the CCD centre, and with different filters – meaning that the you are sampling different types and fractions of lights from different stars.

Determining the appropriate fitting model (or rather parameterisation) was also a challenge. There are various arguments for using Nuker (5 parameter), Sersic (3–parameter) or core-Sersic (6-parameter) fits. In most cases the latter is arguably the "best" choice. Our reasoning for choosing Nuker profiles is explained in the paper. One reason was to ensure internal consistency with data from the literature. The underlying problem is that nearly all of the fitting formulas have no physical foundation, so derving insight into physical processes is difficult.


Following publication we were awarded further HST observations (GO-13324), and a German Space Agency research grant. Dr. Ugur Ural was employed at the Liebniz Astrophysik Institut Potsdam (AIP), to analyse the follow-up data. During this period she had an opportunity to work with me at Swinburne University. Ugur now works as a data scientist at Climate Analytics.

The HST/ACS Coma Cluster Treasury Survey

Astrophysics Research Institute, Liverpool John moores University – UK  |  April 2009 – November 2011

Scientific Research goals:

The Coma Cluster is one of the best studied galaxy clusters in the universe and it was one of the first environments that showed evidence for "missing", or dark matter. Its mean distance makes it the perfect laboratory for studying galaxy clusters. 

In April 2009, I moved to LJMU's Astrophysics Research Institute (ARI) to join the HST/ACS Coma Cluster Treasury Survey (PI: Prof David Carter). One of only two Treasury Surveys awarded in Cycle–15 (2006), it is a deep two-passband imaging survey of one of the nearest rich and dense clusters of galaxies.

The survey was designed to image both the dense cluster core and an infall region. The images contain thousands of galaxies down to apparent magnitudes g=27.5 and I=26.6 mag. In addition, numerous supporting ground-based and space-based observing campaigns were also taken at other wavelengths. 

The main scientific objectives were:

Papers, Published Data & Talks:

Data Analysis & Processing Techniques:

During my time at the ARI worked on a number of datasets, primarily the Hubble Space Telescope (HST) images from the survey, and follow-up spectroscopic observations taken with the DEIMOS spectrograph on the 10m Keck-II Telescope. 

  • Spectroscopic observations provide astronomers with a measure of the velocity dispersion or average motion of the stars within a galaxy (and consequently their dynamical masses), as well as a determination of how far away the galaxy is, or rather, how fast the galaxy is moving away. For Coma Cluster galaxies, using DEIMOS is a real challenge. Ideally you want to use a grating that gives you the highest velocity resolution, but this makes it difficult to analyse the faintest galaxies. These also happen to be the least well understood, and therefore arguably the most interesting. Not only do you need to know about the potential target galaxies, you need to have a really good understanding of the spectrograph, the pros and cons of the various gratings, how it's calibrated, but how residual optical effects and CCD defects affect the raw data. The data processing is significantly time consuming. Data was processed using the standard IDL pipeline routines, however since we used a non-standard instrument setup, we had tweak the wavelength calibration, by which I mean wrangling the data pipeline. In addition, night sky atmospheric emission lines coincide with the absorption line features used to measure velocity dispersions. Removing these emission lines without affecting the signal from the galaxy is not trivial.

Challenges with the Data:

The HST/ACS Coma Cluster Treasury Survey suffered from HST/ACS failure in January 2007, with only 28% of the survey fields observed. The camera failed due to a short circuit in its backup power supply, and was out of commission for two years. The Wide Field Channel (WFC) was returned to service by STS-125 in May 2009, however HST Operations was not able to reschedule "lost" observations. 

Searches for UCDs & Globular Clusters in M31 & Virgo

Institute of Geophysics & Planetary Physics (IGPP), Lawrence Livermore National Lab (LLNL) – USA  |   January 2005 – January 2009

PhD Research

Astrophysics group, department of physics , University of Melbourne – AU  |  Awarded January 2005

Galaxy Formation & Destruction in the Fornax Cluster – PhD Thesis Abstract

While much of modern astronomical research is focused on the question of galaxy formation, a growing body of evidence suggests that galaxy destruction may be a more common occurrence than previously thought, playing a vital role in the evolution of present day galaxies. In this thesis, wide-field imaging and spectroscopic observations of the Fornax Cluster provide the opportunity to conduct a detailed study of the formation and destruction of galaxies in the cluster. Particular emphasis was placed on the large population of dwarf galaxies, generally too faint to observe at high redshifts as well as a new type of cluster galaxy -- ultra-compact dwarf galaxies (UCD) -- which were discovered recently in the cores of both the Fornax and Virgo galaxy clusters. The formation of these enigmatic objects is intimately connected to processes within the cluster environment.

Multicolour B, V & I wide-field imaging of the cluster was taken with the CTIO Curtis Schmidt Telescope and used to explore the wide variety of cluster galaxies.  Deeper and higher resolution multicolour u' g' r' i' z' imaging of the core of the cluster is used to explore the UCD and globular cluster population associated with the central cluster galaxy, NGC 1399. Wide-field spectroscopy of cluster galaxies, taken with the 6dF multi-fibre spectrograph on the UKST, complement the optical images.

Based on their photometric and spectroscopic properties, a model for the formation of UCDs is investigated. This model is called the ``galaxy threshing'' scenario, whereby cluster dE,Ns are tidally stripped of their stellar envelopes by the cluster tidal field as they orbit bright cluster galaxies. The remaining nucleus of the dE,N results in a UCD. The colours and metallicities of the UCDs suggest they are an older stellar population not unlike globular clusters. The UCDs lie well off the surface brightness and colour-magnitude relations for early-type galaxies, consistent with the ``galaxy-threshing'' hypothesis.

The well-established empirical surface-brightness magnitude relation for normal elliptical and dwarf galaxies is used to explore the structure and evolution of cluster galaxies. Despite a lack of physical understanding, the relation also provides one of the most fundamental means of classifying galaxies. The traditional method of classifying galaxies according to their surface brightness is investigated and its ability to distinguish between cluster and background galaxies is tested. This is the first investigation of the relation for such a wide range of galaxy types, all observed using the same instrument. The evolution of cluster galaxies is further explored using metallicities derived from absorption line indices.