My research initially focused on the study of massive stars. I have investigated the importance of mass loss and duplicity in stellar evolution (Eldridge & Tout, 2004; Eldridge & Vink, 2006; Eldridge, Izzard & Tout, 2008). I have actively collaborated with many astronomers to model progenitors of core-collapse supernovae and compare my models with observations (e.g. Eldridge et al. 2007, Pastorello et al., 2007; Mattila et al., 2008; Stancliffe & Eldridge, 2009,Fraser et al., in prep). I have also written a review on supernova progenitors (Eldridge, 2008) for the Royal Society. This work culminated in accurately identifying the progenitors of type IIP supernovae as red supergiants (Smartt et al., 2009).

An offshoot of my research played a key role in understanding the evolution of stars near the minimum mass for a core-collapse supernovae. I showed that these super-AGB stars are unlikely to be the supernova progenitors as previously suspected (Eldridge, Mattila & Smartt, 2007). Recently there have been more observed supernova progenitors near this minimum mass with luminosities less than those expected from stellar models. By varying the model physics I have shown that the luminosities are achievable but only by varying uncertain physics within the stellar models, such as increasing the carbon burning rate (Fraser et al., in prep).

My research on core-collapse supernovae has led to studies into the related events of long gamma-ray bursts (GRBs). My investigations began with attempts to model the circumstellar medium around the progenitor stars in order to link them to the circumburst media inferred from the GRB afterglows (Eldridge et al., 2006; Eldridge, 2007). Due to the current plethora of new observations this work has now expanded to studies of the host galaxies of long and short GRBs.

More recently I have adapted the Cambridge STARS code to create detailed stellar models of binary stars. I have shown that binary systems can naturally answer many discrepancies between observations and model predictions of resolved stellar populations and supernova rates. This is not surprising because at least 50% of massive stars are in binaries. Binary interactions cause increased mass loss and mass transfer, and this gives rise to many different avenues for stellar evolution (e.g. Eldridge, Izzard & Tout, 2008). Because of the variety of possible evolutionary paths, the importance of binaries is often underestimated.

As one simple example of how important binary models can be, I recently determined a new age for the closest Wolf-Rayet star to the Sun in the Gamma Velorum binary system (Eldridge, 2009). The new age of 5.5 Myrs exceeds that estimated from single-star models by 2 Myr (almost doubling the age of this star). Crucially, this brings the age of the Wolf-Rayet star in line with surrounding low-mass stars. This suggests that it is coeval with them and formed as part of the same starburst. This allows a better understanding of the evolution of this region and highlights the inaccurate conclusions that arise when relying on single star models alone.

My results (Eldridge et al., 2008) have been compared, by others, to observations and a better agreement is typically found with my binary models. For example, Boissier & Prantzos (2009) found a better agreement to the relative rates of SN types with metallicity and Brinchmann et al. (2008) achieved a better fit to observations of unresolved stellar populations in the Sloan Digital Sky Survey (SDSS) galaxies. The latter study inspired me to combine my stellar evolution models with libraries of synthetic atmosphere spectra to create a unique Binary Population and Spectral Synthesis code (BPASS, Eldridge & Stanway, 2009; www.bpass.org.uk).

While similar codes (such as starburst99) exist BPASS has five important features, each of which set it apart from other codes and in combination make it the cutting edge:

1. First, and most important, is the inclusion of binary evolution in modelling the stellar populations. The general effect of binaries is to cause a population of stars to look bluer at an older age than predicted by single-star models.
2. Detailed stellar evolution models are used rather than an approximate rapid population synthesis method.
3. I use only theoretical model spectra in my syntheses with as few empirical inputs as possible to create completely synthetic models to compare with observations.
4. I use Cloudy (Ferland et al., 1998) to determine the nebular emission. This means I model not only the stars in detail but also the surrounding gas.
5. The code is easily adaptable to determine the input physical parameters required to match observations. I have invested a great amount of time and effort to create BPASS. The development is now complete and it is at the centre of much of my current and future research.