Dynamics of external galaxies: stellar kinematics, black holes, and machine learning modelling

Thesis

Spectroscopy has played an important role in many of the major discoveries in astronomy over the last century. Integral field spectroscopy has continued this trend. Originating in the late 1980s, it is now a fully mature technology and a staple component of modern observing facilities.

I showcase the power of integral field spectroscopy by using MUSE and OASIS data to measure the supermassive black hole mass in the early-type galaxy M87. M87 is the poster child for black hole studies, but presents a number of challenges in the reduction and analysis of the data. To overcome these issues, we pioneer several new techniques for handling the dynamical modelling, as well as determining key parameters such as the PSF and central stellar distribution.

From here, I turn my attention to some of the details of integral field kinematics and dynamical modelling. Making the most out of integral field data typically requires two steps. The first step is to measure the spatially resolved line-of-sight velocity distribution (LOSVD) using a spectral fitting code. This involves a number of choices, from the choice of stellar library to the parameterisation of the LOSVD. The second step is to perform dynamical modelling using these extracted kinematics.

I investigate the impact these choices make at two levels using a high-resolution N-body simulation. First, the choice of the parameterisation of the LOSVD is typically done using a truncated Gauss-Hermite series. Different choices of truncation, that is, different numbers of moments, can induce systematics in the recovered kinematics. At another level, template mismatch and the specific choice of stellar library can induce systematic offsets in the kinematics which impact dynamical models. I investigate each of these independently, and determine some of the best practices for fitting stellar kinematics.

Using these stellar kinematics, I perform JAM, axisymmetric Schwarzschild, and triaxial Schwarzschild modelling and describe the systematic differences between these methods and how well they are able to recover the true black hole mass.

The role of systematics in dynamical models is also important. Most dynamical modelling methods make a number of assumptions which don't perfectly hold for real galaxies. In the penultimate chapter, I discuss an interesting new direction, where some of these assumptions can be broken using machine learning models which replace traditional dynamical models. I conclude with a very optimistic picture of the future of this field.

Authors: Simon, DA

• DOI: 10.5287/ora-44mkkxabd
• Type of award: DPhil
• Level of award: Doctoral
• Awarding institution: University of Oxford
• Language: English
• Keywords: galaxy kinematics; machine learning; galaxy dynamics; black hole; jeans modelling; dynamical modelling
• Subjects: Astrophysics