Observations of galaxies by integral-field spectrographs (IFS) provide detailed information on both their spatial and spectral characteristics. The ratio of a galaxy's rotational velocity to its dispersion is a key parameter used to classify disk galaxies, particularly at high-redshift where spatial resolution is lower. Additionally, galaxy rotation curves and flux maps can be combined with stellar mass-to-light ratios to study dark matter abundances (Binney and Tremaine, 2008). Therefore extracting kinematic information from IFS data is extremely useful, though it is not without its difficulties. Beam smearing is caused by an observation's point spread function (PSF) and spatially blurs information. This means data from a single galaxy location is spread over numerous spatial pixels in derived kinematic maps and has the effect of decreasing the magnitude of observed rotational velocities and increasing velocity dispersions (Johnson et al. 2017). This effect is more pronounced in high redshift observations where the ratio of PSF to galaxy angular size is greater.
A number of forward modelling codes currently exist to extract kinematic information whilst accounting for the effect of beam smearing in 2D kinematic maps, such as GalPaK3D (Bouch et al. 2015), 3DBarolo (Di Teodoro and Fraternali, 2015) and Blobby3D (Varidel et al. 2019). However these often rely on making assumptions of underlying galaxy kinematics by comparing observations to 3D galaxy models which are altered and recomputed at each step, making them computationally expensive. Therefore, we are adapting the existing emission-line fitting code ROHSA (Marchal et al. 2019) developed by RSAA Postdoctoral Fellow Antoine Marchal which does not rely on assumptions of an underlying model. ROHSA fits the spectra of each pixel in an IFS observation with a Gaussian model by altering maps of amplitude, peak wavelength and dispersion through a gradient-descent approach. An additional regularisation term set by the user ensures adjacent pixels in these maps are sufficiently similar, thus producing smooth kinematic maps, reducing the effect of noise. However ROHSA was originally developed to be fit radio observations of HI 21-cm emission line data and does not account for beam smearing effects caused by the PSF. Therefore, during the first project of my PhD I have been working on including this functionality.
An initial fit to galaxy kinematics is obtained by running of the original ROHSA code. We then make use of the fact that observed kinematics are the result of intrinsic galaxy kinematics being convolved with the PSF of an observation. We can convolve our initial fit with the observation's PSF and calculate the difference between the convolved fit and observed data, the cost. By using a using a gradient-descent method to iteratively alter each Gaussian in the initial fit and then re-convolving, the cost can be reduced. Once the method converges, the convolved Gaussian fit and observed data should closely match, meaning the underlying Gaussians well describe the intrinsic kinematics of the galaxy, without the effect of beam smearing. In order to test the adapted code, we have first been fitting mock observations of idealised disks created with the KinMS package (Davis et al. 2013), as we can compare the true intrinsic model to the best-fit of our code. The results so far have been promising, as demonstrated in the included figure. Fitting to real data will be in the near future!
Plot Caption: A fit to a mock disk observation using the adapted ROHSA code. To create this observation the mock disk was convolved with a PSF to mimic real data. The FWHM of this PSF is shown in the first panel in white.
The top three panels show the output Gaussian parameters returned by the adapted code, which should closely match the intrinsic parameters of the galaxy (ie. the mock galaxy before any PSF convolution), the outline of which is traced in black. To illustrate the accuracy of the fit, the bottom three panels plot the galaxy kinematics measured along its kinematic axis (shown in purple in the top panels). In these plots the observed galaxy parameters are plotted as black dashed lines and the original ROHSA fit to the observed galaxy in dashed red, with the two matching very well. The solid blue line is then the resulting fit from the adapted ROHSA code. This should ideally match the intrinsic galaxy parameters, which is plotted in solid black. The two solid lines agree quite closely, suggesting the adapted code can extract kinematics without the effects of beam-smearing.