This is an in-person meeting, as it focus on the practical use of statistics and machine learning in academic research. We will supply all the necessary guidelines through Astronomical problems.
Check the website for further information. Registration closes on March 24!
Getting observing time needs both a good proposal and some … luck! Although I have been into many proposals (with some of them written by me actually) I did not have the excitement to get it as PI. The time has finally come now!!
At the current observing period of ESO (P111, running 1 April 2023 – 30 September 2023) we managed to get ~18 hours to use FORS2 to observe a set of three dwarf galaxies (IC 1613, Pegasus DIG, and Phoenix Dwarf), part of the ASSESS galaxy sample. The difference, with respect to our previous successful proposals both at ESO and GTC, is that we will use the predictions of the machine-learning classifier we developed for this project.
Now I have to work on the Phase-2 material, which consists of the preimaging OBs first, and then the mask designing (based on the images).
A proceedings paper from IAUS 366 that took place virtually back in October 2021 (for which I had another poster contribution) was finally published at the end of 2022. It summarizes a collective work led by Michaela on B[e] Supergiants and Yellow Hypergiants, two massive star phases where we observe episodic mass loss.
Environments of evolved massive stars: evidence for episodic mass ejections
M. Kraus, L. S. Cidale, M. L. Arias, A. F. Torres, I. Kolka, G. Maravelias, D. H. Nickeler, W. Glatzel and T. Liimets
The post-main sequence evolutionary path of massive stars comprises various transition phases, in which the stars shed large amounts of material into their environments. Our studies focus on two of them: B[e] supergiants and yellow hypergiants, for which we investigate the structure and dynamics within their environments. We find that each B[e] supergiant is surrounded by a unique set of rings or arc-like structures. These structures are either stable over time or they display high variability, including expansion and dilution. In contrast, yellow hypergiants are embedded in multiple shells of gas and dust. These objects are famous for their outburst activity. Moreover, the dynamics in their extended atmospheres imply an enhanced pulsation activity prior to outburst. The physical mechanism(s) leading to episodic mass ejections in these two types of stars is still uncertain. We propose that strange-mode instabilities, excited in the inflated envelopes of these objects, play a significant role.
Figure 1. Real parts (= pulsation periods, left panel) and the imaginary parts (right panel) of the eigenfrequencies, which are normalized to the global free-fall time. Positive imaginary parts correspond to damped modes, and negative ones to unstable modes. The computations have been performed for T eff = 7000 K and log L/L = 5.7, matching the observed values of ρ Cas.
The biggest issue when dealing with multiple projects is how to keep track of the various Python and other packages’ versions you are using in each project. It is not uncommon to update something and get something else broken…
The solution is virtual environments. And for these there are few options such as virtualenv and conda (among others). So far I have postponed (and not with good results) the use of environments. But finally, I took the necessary time to investigate pros and cons and make a decision. I found the following guides from WhiteBox exceptional well written and useful:
I am actually excited about this. But not because I am part of the award… I am excited because this award recognizes the effort put by some amateurs in order to produce not only scientifically useful images/data but document their effort, share it with both the amateur and professional communities and contribute to the advance of our knowledge by analyzing and publishing their results. I am honored that I could help Manos to achieve that. I strongly believe that it is a worthy recognition of his more than two decades of contribution in Astronomy and the community in general.
Of course, kudos to Javier – without his expertise and guidance this work wouldn’t have materialized.
The two main driving forces behind the work: Javier Peralta (left) and Manos Kardasis (right) holding the Javier Gorosabel award.
Properties of luminous red supergiant stars in the Magellanic Clouds
S. de Wit, A.Z. Bonanos, F. Tramper, M. Yang, G. Maravelias, K. Boutsia, N. Britavskiy, and E. Zapartas
There is evidence that some red supergiants (RSGs) experience short lived phases of extreme mass loss, producing copious amounts of dust. These episodic outburst phases help to strip the hydrogen envelope of evolved massive stars, drastically affecting their evolution. However, to date, the observational data of episodic mass loss is limited. This paper aims to derive surface properties of a spectroscopic sample of fourteen dusty sources in the Magellanic Clouds using the Baade telescope. These properties may be used for future spectral energy distribution fitting studies to measure the mass loss rates from present circumstellar dust expelled from the star through outbursts. We apply MARCS models to obtain the effective temperature (Teff) and extinction (AV) from the optical TiO bands. We use a χ2 routine to determine the best fit model to the obtained spectra. We compute the Teff using empirical photometric relations and compare this to our modelled Teff. We have identified a new yellow supergiant and spectroscopically confirmed eight new RSGs and one bright giant in the Magellanic Clouds. Additionally, we observed a supergiant B[e] star and found that the spectral type has changed compared to previous classifications, confirming that the spectral type is variable over decades. For the RSGs, we obtained the surface and global properties, as well as the extinction AV. Our method has picked up eight new, luminous RSGs. Despite selecting dusty RSGs, we find values for AV that are not as high as expected given the circumstellar extinction of these evolved stars. The most remarkable object from the sample, LMC3, is an extremely massive and luminous evolved massive star and may be grouped amongst the largest and most luminous RSGs known in the Large Magellanic Cloud (log(L∗/L⊙)∼5.5 and R=1400 R⊙).
Fig. 9: Top: HRD indicating the locations of our LMC targets with inverted red triangles. The Teff for all data points was derived through the TiO method. Smaller light grey squares and stars are objects from Levesque et al. (2006) and Davies et al. (2013), respectively. For the two outliers we have extended the uncertainty assuming a shift of 0.3 mag in the K−band (dotted vertical error bar) instead of only the propagated uncertainty, to visualize the effect of intrinsic variability. The colour map represents the central 12C mass fraction, while the nodes on the track again indicate a step of 104 years
This is actually a preview of what will follow after the first paper of the machine-learning classifier. We put it into action to get predictions for a number of galaxies and we start exploring the results. Of more interest is the fractions of the populations with metallicity, although a more detailed study is needed to take care of all caveats.
Using machine learning to investigate the populations of dusty evolved stars in various metallicities
Grigoris Maravelias, Alceste Z. Bonanos, Frank Tramper, Stephan de Wit, Ming Yang, Paolo Bonfini, Emmanuel Zapartas, Konstantinos Antoniadis, Evangelia Christodoulou, Gonzalo Muñoz-Sanchez
Mass loss is a key property to understand stellar evolution and in particular for low-metallicity environments. Our knowledge has improved dramatically over the last decades both for single and binary evolutionary models. However, episodic mass loss although definitely present observationally, is not included in the models, while its role is currently undetermined. A major hindrance is the lack of large enough samples of classified stars. We attempted to address this by applying an ensemble machine-learning approach using color indices (from IR/Spitzer and optical/Pan-STARRS photometry) as features and combining the probabilities from three different algorithms. We trained on M31 and M33 sources with known spectral classification, which we grouped into Blue/Yellow/Red/B[e] Supergiants, Luminous Blue Variables, classical Wolf-Rayet and background galaxies/AGNs. We then applied the classifier to about one million Spitzer point sources from 25 nearby galaxies, spanning a range of metallicites (1/15 to ∼3 Z⊙). Equipped with spectral classifications we investigated the occurrence of these populations with metallicity.
The fractions, of the predicted class members over the total sample size for each galaxy, with metallicity.
JWST delivered its first images (publicly) today! That was the best news to get as an astronomer for two reasons. The one is of course the scientific reasons, as it is one of the most important telescopes sent to space and perhaps the most challenging mission to launch and deploy at that distance (L1). There should have been already tens (or hundreds ??) of articles with respect to this part.
However, there is another important reason. Everything went finally smooth and we have a working telescope. That means that Astronomy will not dismay! Imagine if all this effort and funding of 10 billion dollars would fail… I cannot and I do not want to! Thankfully this did not happen and hopefully this will motivate more funding towards Astronomy that should be reflected to jobs also….
So let’s enjoy and cheer about this new era in Astronomy!
Webb’s First Deep Field, the image of galaxy cluster SMACS 0723 (Credit: NASA)
In March I got invited to give a talk for the Thüringer Landessternwarte Tautenburg group. Back then June 21 looked like a far date in a relative relaxed time period. I was so wrong… they are so hectic days!
Nevertheless, I managed to prepare a talk entitled: “The ASSESS classifier: a machine-learning tool to uncover populations of evolved massive stars in nearby galaxies”. I tried to prepare a more general introduction to the massive stars and their evolution, along with the issues we are facing as far as mass loss. I highlighted the importance of the ASSESS project and the provide a short description of what is ASSESS all about. Then I described the machine-learning classifier we developed and our first results from this.
Although pressing, at the end I really enjoyed this talk!
Last night I was working on the laptop when naturally it asked for some power. Typically, when you plug in the power is should charge the battery, right? Well… no, it didn’t do that at all!
And then dread start developing! What is wrong? Is it the plug, the connector or even worse the port? As it was late I called it a day and went to bed. Today, I started by trying another power cable and finally another charger … in vain! Then, Tassos K. told me the obvious “Why you don’t look it up, somebody should have written about it“.
