Deriving star cluster parameters with convolutional neural networks

Context. Convolutional neural networks (CNNs) have been established as the go-to method for fast object detection and classification of natural images. This opens the door for astrophysical parameter inference on the exponentially increasing amount of sky survey data. Until now, star cluster analysis was based on integral or resolved stellar photometry, which limit the amount of information that can be extracted from individual pixels of cluster images. Aims. We aim to create a CNN capable of inferring star cluster evolutionary, structural, and environmental parameters from multiband images and to demonstrate its capabilities in discriminating genuine clusters from galactic stellar backgrounds. Methods. A CNN based on the deep residual network (ResNet) architecture was created and trained to infer cluster ages, masses, sizes, and extinctions with respect to the degeneracies between them. Mock clusters placed on M 83 Hubble Space Telescope images utilizing three photometric passbands (F336W, F438W, and F814W) were used. The CNN is also capable of predicting the likelihood of the presence of a cluster in an image and quantifying its visibility (S/N). Results. The CNN was tested on mock images of artificial clusters and has demonstrated reliable inference results for clusters of ages ≲100 Myr, extinctions AV between 0 and 3 mag, masses between 3 × 103 and 3 × 105 M⊙, and sizes between 0.04 and 0.4 arcsec at the distance of the M 83 galaxy. Real M 83 galaxy cluster parameter inference tests were performed with objects taken from previous studies and have demonstrated consistent results.

The specific angular momenta of superthin galaxies: Cue to their origin?

ABSTRACT Superthin galaxies are low surface brightness (LSB) bulgeless disc galaxies having stellar discs with unusually high planar-to-vertical axes ratio b/a > 10–20, the formation and evolution of which is not well understood. We calculate the specific angular momenta of a sample of six superthins and nine other bulgeless LSBs using stellar photometry, atomic hydrogen (HI) surface density, and high-resolution HI rotation curves available in the literature. We find that the stellar specific angular mometum js, and hence the stellar disc size given by the exponential stellar disc scale length RD, of three superthins and seven LSBs lie above the 95.4 ${{\ \rm per\ cent}}$ confidence band of the js − Vrot regression line for ordinary bulgeless disc galaxies, Vrot being the asymptotic rotational velocity. Further, we find that superthins and LSBs have higher js and RD values for a given value of stellar mass Ms at high values of statistical significance, compared to ordinary disc galaxies. Therefore, we conclude, a superthin is may be distinguished by a characteristically larger disc size which could possibly explain the origin of its large planar-to-vertical axes ratio. Interestingly, we find that the median spin parameter $\lambda = \frac{ j_{\rm {stars}} }{ {\sqrt{(}2) V_{\rm {vir}} R_{\rm {vir}} } }$, Vvir and Rvir being the virial velocity and virial radius of the galaxy respectively, is 0.13 ± 0.01 for superthin galaxies which is an order of magnitude higher than those of LSBs and ordinary disc galaxies, which may have important implications for the existence of superthin stellar discs in these LSB galaxies.

J-PLUS: A wide-field multi-band study of the M 15 globular cluster

Context. As a consequence of internal and external dynamical processes, Galactic globular clusters (GCs) have properties that vary radially. Wide-field observations covering the entire projected area of GCs out to their tidal radii (rtidal) can therefore give crucial information on these important relics of the Milky Way formation era. Aims. The Javalambre Photometric Local Universe Survey (J-PLUS) provides wide field-of-view (2 deg2) images in 12 narrow, intermediate and broad-band filters optimized for stellar photometry. Here we have applied J-PLUS data for the first time for the study of Galactic GCs using science verification data obtained for the very metal-poor ([Fe/H] ≈−2.3) GC M 15 located at ~10 kpc from the Sun. Previous studies based on spectroscopy found evidence of multiple stellar populations (MPs) through their different abundances of C, N, O, and Na. Our J-PLUS data provide low-resolution spectral energy distributions covering the near-UV to the near-IR, allowing us to instead search for MPs based on pseudo-spectral fitting diagnostics. Methods. We have built and discussed the stellar radial density profile (RDP) and surface brightness profiles (SBPs) reaching up to rtidal. Since J-PLUS FoV is larger than M 15’s rtidal, the field contamination can be properly taken into account. We also demonstrated the power of J-PLUS unique filter system by showing colour-magnitude diagrams (CMDs) using different filter combinations and for different cluster regions. Results. J-PLUS photometric quality and depth are good enough to reach the upper end of M 15’s main-sequence. CMDs based on the colours (u − z) and (J0378 − J0861) are found to be particularly useful to search for splits in the sequences formed by the upper red giant branch (RGB) and asymptotic giant branch (AGB) stars. We interpret these split sequences as evidence for the presence of MPs. Furthermore, we show that the (u − z) × (J0378 − g) colour–colour diagram allows us to distinguish clearly between field and M 15 stars, which is important to minimize the sample contamination. Conclusions. The J-PLUS filter combinations (u − z) and (J0378 − J0861), which are sensitive to metal abundances, are able to distinguish different sequences in the upper RGB and AGB regions of the CMD of M 15, showing the feasibility of identifying MPs without the need of spectroscopy. This demonstrates that the J-PLUS survey will have sufficient spatial coverage and spectral resolution to perform a large statistical study of GCs through multi-band photometry in the coming years.

Estimating stellar ages and metallicities from parallaxes and broadband photometry: successes and shortcomings

A deep understanding of the Milky Way galaxy, its formation and evolution requires observations of huge numbers of stars. Stellar photometry, therefore, provides an economical method to obtain intrinsic stellar parameters. With the addition of distance information – a prospect made real for more than a billion stars with the second Gaia data release – deriving reliable ages from photometry is a possibility. We have developed a Bayesian method that generates 2D probability maps of a star’s age and metallicity from photometry and parallax using isochrones. Our synthetic tests show that including a near-UV passband enables us to break the degeneracy between a star’s age and metallicity for certain evolutionary stages. It is possible to find well-constrained ages and metallicities for turn-off and sub-giant stars with colours including a U band and a parallax with uncertainty less than ∼20%. Metallicities alone are possible for the main sequence and giant branch. We find good agreement with the literature when we apply our method to the Gaia benchmark stars, particularly for turn-off and young stars. Further tests on the old open cluster NGC 188, however, reveal significant limitations in the stellar isochrones. The ages derived for the cluster stars vary with evolutionary stage, such that turn-off ages disagree with those on the sub-giant branch, and metallicities vary significantly throughout. Furthermore, the parameters vary appreciably depending on which colour combinations are used in the derivation. We identify the causes of these mismatches and show that improvements are needed in the modelling of giant branch stars and in the creation and calibration of synthetic near-UV photometry. Our results warn against applying isochrone fitting indiscriminately. In particular, the uncertainty on the stellar models should be quantitatively taken into account. Further efforts to improve the models will result in significant advancements in our ability to study the Galaxy.

Deriving star cluster parameters with convolutional neural networks

Context. Convolutional neural networks (CNNs) have been proven to perform fast classification and detection on natural images and have the potential to infer astrophysical parameters on the exponentially increasing amount of sky-survey imaging data. The inference pipeline can be trained either from real human-annotated data or simulated mock observations. Until now, star cluster analysis was based on integral or individual resolved stellar photometry. This limits the amount of information that can be extracted from cluster images. Aims. We aim to develop a CNN-based algorithm capable of simultaneously deriving ages, masses, and sizes of star clusters directly from multi-band images. We also aim to demonstrate CNN capabilities on low-mass semi-resolved star clusters in a low-signal-to-noise-ratio regime. Methods. A CNN was constructed based on the deep residual network (ResNet) architecture and trained on simulated images of star clusters with various ages, masses, and sizes. To provide realistic backgrounds, M 31 star fields taken from The Panchromatic Hubble Andromeda Treasury (PHAT) survey were added to the mock cluster images. Results. The proposed CNN was verified on mock images of artificial clusters and has demonstrated high precision and no significant bias for clusters of ages ≲3 Gyr and masses between 250 and 4000 M⊙. The pipeline is end-to-end, starting from input images all the way to the inferred parameters; no hand-coded steps have to be performed: estimates of parameters are provided by the neural network in one inferential step from raw images.

New developments in aerosol measurements using stellar photometry

The idea of using stellar photometry for atmospheric monitoring for optical experiments in highenergy astrophysics is seemingly straightforward, but reaching high precision of the order of 0.01 in the determination of the vertical aerosol optical depth (VAOD) has proven difficult. Wide-field photometry over a large span of altitudes allows a fast determination of VAOD independently of the absolute calibration of the system, while providing this calibration as a useful by-product. Using several years of data taken by the FRAM (F/(Ph)otometric Robotic Atmospheric Monitor) telescope at the Pierre Auger Observatory in Argentina and about a year of data taken by a similar instrument deployed at the planned future Southern site of the Cherenkov Telescope Array in Chile, we have developed methods to improve the precision of this measurement technique towards and possibly beyond the 0.01 mark. Detailed laboratory measurements of the response of the whole system to both the spectrum and intensity of incoming light have proven indispensable in this analysis as the usual assumption of linearity of the CCD detectors is not valid anymore for the conditions of the observations.

IAU Commission 25, and the development of early photometric systems

The International Astronomical Union was conceived in 1918, and was formed one year later in Brussels. One of the 32 initial Commissions was the Committee on Stellar Photometry that later on became IAU Commission 25 Astronomical Photometry and Polarimetry, and since 2015 Commission B6 with the same name. The initial functions to be exercised by the Committee were (a)to advise in the matter of notation, nomenclature, definitions, conventions, etc., and(b)to plan and execute investigations requiring the cooperation of several observers or institutions.The basic philosophy was that IAU Commission 25 was to be an advisory body, rather than a decision-making committee that imposes its regulations. This position was reconfirmed at the 10th IAU General Assembly in 1958.From the early days on, the Commission members engaged in the teaching of the principles of photometric measurement – either via the Commission meetings and the ensuing reports, or via external means, such as lectures and publications. The topics of instruction dealt with absorption of light in the atmosphere, the modification imposed by the character of the receiving apparatus, the unequal response of different receivers to a same stimulus, and variations in the data-recorder response from one experiment to another.From the 1930s on it was suggested that IAU Commission 25 takes responsibility in matters of standard stars, standard filters and standard calibration methods.During the first half-century since its foundation, Commission 25 was an active forum for discussions on the basic principles of astronomical photometry, including the associated problems of transformability of magnitudes and colour indices from one instrumental configuration to another. During the second half-century of its existence, the Commission has served as a sort of news agency reporting on the developments in detector engineering, filter technology and data reduction. All along the Commission members were committed to accuracy and precision, a struggle that was primarily driven by the jumps forward in performance and sensitivity of every new detector that was introduced.The development over one century shows that the Commission was continuously touching on the philosophy of precise measurement, where accurate measuring – for a select group of pioneers – was an end in itself.This presentation looks back on the opinions of key players in the photometric standardisation debate, and briefly presents two case studies that illustrate the illusionary accuracy reached over a century in determining, as Commission member Ralph Allan Sampson put it, “a detail like magnitude”.

A new method to derive star formation histories in dwarf galaxies

We present a new method to derive 2D star formation histories in dwarf irregular galaxies. Based on multicolor stellar photometry data we have found that in the Leo A galaxy during the last ∽400 Myr star formation was propagating according to the inside-out scenario. Star-forming regions have spread strongly asymmetrically from the center and their present day distribution correlates well with the Hi surface density maps.