I work on a variety of topics related to the late stages of stellar evolution and the nebulae surrounding evolved stars. Here are some of the things I've been working on recently. You can also see what I've been studying on the sky map.
One thing we did with neat was to investigate how non-Gaussian measurement uncertainties affect the results of nebular abundance analyses. We incorporated the probability distributions suggested by Rola and Pelat (1994), who found that in noisy spectra, line flux measurements have log normal probability distributions and are strongly biased upwards. We found that assuming their probability distributions reduced the estimated uncertainty on the derived quantities. However, other authors have questioned whether these measurement biases really exist so we used tens of thousands of line flux measurements from the Sloan Digital Sky Survey to investigate further. We found little evidence for strong upward biases caused by measurement, but the results clearly show that the probability distribution of any ratio becomes highly non-Gaussian when the measurements are noisy. We couldn't reproduce the results of Rola and Pelat, finding only a small upward bias in flux measurements with one line fitting approach, and a downward bias with another.
I've released alfa — Automated Line Fitting Algorithm — a code which automated the measurements of line fluxes. It does so using a genetic algorithm to optimise a large number of parameters simultaneously, and rather than search for possible emission features, measuring them and later attempting to identify them, it takes the approach of starting from a list of emission lines known or suspected to be present in the relevant wavelength range, and calculating the fluxes required to best fit the observed spectrum. Thus, line identification is not separate from fitting, and without any interaction you get a list of emission line fluxes from your data. I'm currently applying alfa to very large datasets from MUSE, and in conjunction with neat (see below) it takes a few hours to measure several million emission lines and then extract dozens of maps of physical quantities from a reduced data cube.
In ionised nebulae, heavy element abundances derived from recombination lines exceed those derived from collisionally excited lines typically by a factor of 2-3, and sometimes by vastly more. There's strong evidence that a small amount of cold hydrogen-deficient gas embedded in the hot H-rich gas could give rise to this discrepancy, but where that cold gas came from is not yet well understood. However, there's growing evidence that objects with the highest abundance discrepancies are associated with close binary central stars.
One such object in NGC 6778, which I've studied as part of a team with colleagues at ESO and the IAC. The nebula is known to have a binary central star with an orbital period of only 3.7 hours, one of the shortest known, and we obtained high quality new spectra to determine its abundances. We measured emission line fluxes using alfa and calculated the abundances using neat, and we found that it has an abundance discrepancy of a factor of about 20, one of the largest known. We're now working on measuring abundances in more nebulae with close binary central stars to better understand how these phenomena are linked.
My compilation of abundance discrepancy measurements from the literature clearly shows the striking link between binarity and abundance discrepancies.
Studies of the very distant universe show that galaxies which are less than a billion years old contain huge amounts of dust. The most likely source of this dust is massive stars, because low mass stars would not evolve quickly enough to become dust producers. But in the nearby universe, many studies of supernova remnants find that the quantity of dust forming in the supernova ejecta within the first few years after the explosion is much lower than you'd expect, if supernovae were to be the dominant source of dust in the distant universe.
A key object in unravelling this puzzle is SN1987A, which was unexpectedly detected in Herschel observations, implying that a large amount of dust has formed in its ejecta. Using an extensive compilation of archival data, I constructed radiative transfer models of the expanding supernova ejecta to determine when and where this dust formed. I found that the majority of it must have formed many years after the supernova exploded, and that very large dust grains must be present now, 28 years after the supernova exploded.
I was part of a team which carried out a study of a red supergiant in the massive galactic star cluster Westerlund 1. The project was led by Nick Wright, and the star attracted our attention because it was found to be surrounded by ionised gas in images from the VPHAS survey. Red supergiants don't emit any ionising photons, so it's possible that the star has an unseen hot companion, or that the UV field from the hot OB stars in the cluster is sufficient to ionise the ejecta from the RSG. We have obtained follow-up observations with FLAMES and VIMOS on the VLT, firstly to determine the kinematics of the gas, and then to measure its chemical abundances. These data will constrain when the material was ejected and how it is ionised.
I lead the development of a code, neat, to facilitate calculation of abundances in photoionised nebulae. We have fully automated time-consuming processes, and neat can give you temperatures, densities, ionic and total abundances for deep line lists in a fraction of a second. neat propagates uncertainties using a Monte Carlo technique, which is robust even when fractional uncertainties on line fluxes are large, as is often the case in astronomical spectra. It also accounts for non-Gaussian probability distributions, and we found that adopting a more realistic probability distribution for weak lines results in reduced statistical uncertainties on abundances.
We are currently carrying out a large investigation into systematic effects in abundance determinations, using our code to assess the relative magnitude of systematic effects such as choice of interstellar reddening parameterisation, atomic data, and ionisation correction scheme.