Day 1762

Tuesday.  Wednesday.

2009.12872

Machines learn to infer stellar parameters just by looking at a large number of spectra

Sedaghat, et al

Machine learning has been widely applied to clearly defined problems of astronomy and astrophysics. However, deep learning and its conceptual differences to classical machine learning have been largely overlooked in these fields. The broad hypothesis behind our work is that letting the abundant real astrophysical data speak for itself, with minimal supervision and no labels, can reveal interesting patterns which may facilitate discovery of novel physical relationships. Here as the first step, we seek to interpret the representations a deep convolutional neural network chooses to learn, and find correlations in them with current physical understanding. We train an encoder-decoder architecture on the self-supervised auxiliary task of reconstruction to allow it to learn general representations without bias towards any specific task. By exerting weak disentanglement at the information bottleneck of the network, we implicitly enforce interpretability in the learned features. We develop two independent statistical and information-theoretical methods for finding the number of learned informative features, as well as measuring their true correlation with astrophysical validation labels. As a case study, we apply this method to a dataset of ~270000 stellar spectra, each of which comprising ~300000 dimensions. We find that the network clearly assigns specific nodes to estimate (notions of) parameters such as radial velocity and effective temperature without being asked to do so, all in a completely physics-agnostic process. This supports the first part of our hypothesis. Moreover, we find with high confidence that there are ~4 more independently informative dimensions that do not show a direct correlation with our validation parameters, presenting potential room for future studies.

2009.12915

Presence of a fundamental acceleration scale in galaxy clusters

Edmonds, et al

An acceleration scale of order $10^{-10}\mathrm{m/s^2}$ is implicit in the baryonic Tully-Fisher and baryonic Faber-Jackson relations, independently of any theoretical preference or bias. We show that the existence of this scale in the baryonic Faber-Jackson relation is most apparent when data from pressure supported systems of vastly different scales including globular clusters, elliptical galaxies, and galaxy clusters are analyzed together. This suggests the relevance of the acceleration scale $10^{-10}\mathrm{m/s^2}$ to structure formation processes at many different length scales and could be pointing to a heretofore unknown property of dark matter.

2009.13539

The best place and time to live in the Milky Way

Spinelli, et al

Among the most powerful cosmic events, supernovae (SNe) and gamma-ray bursts (GRBs) can be highly disruptive for life: their radiation can be harmful for biota or induce extinction by removing most of the protective atmospheric ozone layer on terrestrial planets. Nearby high-energy transient astrophysical events have been proposed as possible triggers of mass extinctions on Earth. We aim at assessing the habitability of the Milky Way (MW) along its cosmic history against potentially disruptive astrophysical transients with the scope of identifying the safest places and epochs within our Galaxy. We also test the hypothesis that long GRBs had a leading role in the late Ordovician mass extinction event (~440 Myrs ago). We characterise the habitability of the MW along its cosmic history as a function of galactocentric distance of terrestrial planets. We estimate the dangerous effects of transient astrophysical events (long/short GRBs and SNe) with a model which binds their rate to the specific star formation and metallicity evolution within the Galaxy along its cosmic history. Our model also accounts for the probability of forming terrestrial planets around FGK and M stars. Until ~6 billion years ago the outskirts of the Galaxy were the safest places to live, despite the relatively low density of terrestrial planets. In the last ~4 billion years, regions between 2 and 8 kpc from the center, featuring a higher density of terrestrial planets, became the best places for a relatively safer biotic life growth. We confirm the hypothesis that one long GRB had a leading role in the late Ordovician mass extinction event. In the last 500 Myrs, the safest galactic region is comprised between 2 and 8 kpc from the center of the MW, whereas the outskirts of the Galaxy have been sterilized by 2-5 long GRBs.

2009.13918

Sensitivity of solar wind mass flux to coronal temperature

Stansby, et al

Solar wind models predict that the mass flux carried away from the Sun in the solar wind should be extremely sensitive to the temperature in the corona, where the solar wind is accelerated. We perform a direct test of this prediction in coronal holes and active regions, using a combination of in-situ and remote sensing observations. For coronal holes, a 50% increase in temperature from 0.8 MK to 1.2 MK is associated with a tripling of the coronal mass flux. At temperatures over 2 MK, within active regions, this trend is maintained, with a four-fold increase in temperature corresponding to a 200-fold increase in coronal mass flux.

2009.14016

A simple method of producing images of SDSS spectra in a free spreadsheet program

Falcone

Using Google Sheets, I develop a method to easily reproduce thousands of images of SDSS spectra so that they may be studied in only a fraction of the time it would otherwise take. This method may be helpful in projects requiring large samples of SDSS objects with spectra, and is described in a step-by-step manner so that it is accessible to everyone.

2009.14066

TheHaloMod: an online calculator for the halo model

Murray, et al

The halo model is a successful framework for describing the distribution of matter in the Universe -- from weak lensing observables to galaxy 2-point correlation functions. We review the basic formulation of the halo model and several of its components in the context of galaxy two-point statistics, developing a coherent framework for its application. We use this framework to motivate the presentation of a new Python tool for simple and efficient calculation of halo model quantities, and their extension to galaxy statistics via a \textit{halo occupation distribution}, called \halomod. This tool is efficient, simple to use, comprehensive and importantly provides a great deal of flexibility in terms of custom extensions. This Python tool is complemented by a new web-application at https://thehalomod.app that supports the generation of many halo model quantities directly from the browser -- useful for educators, students, theorists and observers.