91̽��

Abstract:

Vertically thickened bars, observed in the form of boxy/peanut (B/P) bulges, are found in the majority of massive barred disc galaxies in the local Universe, including our own. B/P bulges indicate that their host bars have suffered violent bending instabilities driven by anisotropic velocity distributions. We investigate for the first time how the frequency of B/P bulges in barred galaxies evolves from z = 1 to z ≈ 0, using a large sample of non-edge-on galaxies with masses M* > 1010 M☉, selected from the HST COSMOS survey. We find the observed fraction increases from 0+−3060 per cent at z = 1 to 37.8+−5541 per cent at z = 0.2. We account for problems identifying B/P bulges in galaxies with low inclinations and unfavourable bar orientations, and due to redshift-dependent observational biases with the help of a sample from the Sloan Digital Sky Survey, matched in resolution, rest-frame band, signal-to-noise ratio and stellar mass and analysed in the same fashion. From this, we estimate that the true fraction of barred galaxies with B/P bulges increases from ∼10 per cent at z ≈ 1 to ∼ 70 per cent at z = 0. In agreement with previous results for nearby galaxies, we find a strong dependence of the presence of a B/P bulge on galaxy stellar mass. This trend is observed in both local and high-redshift galaxies, indicating that it is an important indicator of vertical instabilities across a large fraction of the age of the Universe. We propose that galaxy formation processes regulate the thickness of galaxy discs, which in turn affect which galaxies experience violent bending instabilities of the bar.

Abstract:

We use Bayesian convolutional neural networks and a novel generative model of Galaxy Zoo volunteer responses to infer posteriors for the visual morphology of galaxies. Bayesian CNN can learn from galaxy images with uncertain labels and then, for previously unlabelled galaxies, predict the probability of each possible label. Our posteriors are well-calibrated (e.g. for predicting bars, we achieve coverage errors of 11.8 per cent within a vote fraction deviation of 0.2) and hence are reliable for practical use. Further, using our posteriors, we apply the active learning strategy BALD to request volunteer responses for the subset of galaxies which, if labelled, would be most informative for training our network. We show that training our Bayesian CNNs using active learning requires up to 35–60 per cent fewer labelled galaxies, depending on the morphological feature being classified. By combining human and machine intelligence, Galaxy zoo will be able to classify surveys of any conceivable scale on a time-scale of weeks, providing massive and detailed morphology catalogues to 91̽�� research into galaxy evolution.

Abstract: