maravelias.info

Although the summer has finished long ago, only now I got some time to update on summer activities, i.e. a number of conferences I attended and contributed to.

Since 2018, I have been working in an automated classifier for massive stars in nearby galaxies, using photometric datasets. These have been produced by my colleagues within the ASSESS team, an ERC project led by Alceste Bonanos at the National Observatory of Athens, and I have been responsible to develop a machine learning method to achieve this. We have made a lot of progress and we have reached to the point that the results are almost final (working now on the Maravelias et al. paper). So, this work has been presented in:

1. A poster presentation at the Supernova Remnants II, Chania, Greece, 3-8 June 2019,
   as “Identifying massive stars in nearby galaxies, in a smart way”
2. A talk, done by Frank Tramper due to my unavailability to attend the 14th Hellenic Astronomical Conference, Volos, Greece, 8-11 July 2019,
   as “Automated classification of massive stars in nearby galaxies”
3. A talk at the Computational Intelligence in Remote Sensing and Astrophysics, FORTH workshop, Heraklion, Greece, 17-19 July 2019,
   as “An automated classifier of massive stars in nearby galaxies”
4. A remote talk for the ASTROSTAT 2nd Consortium meeting, Boston, USA, 18-19 July 2019,
   as “Towards an automated classifier of massive stars in nearby galaxies”

Grigoris Maravelias, Alceste Z. Bonanos, Ming Yang, Frank Tramper, Stephan A. S. de Wit, Paolo Bonfini

Abstract:
Current photometric surveys can provide us with multiwavelength measurements for a vast numbers of stars in many nearby galaxies. Although the majority of these stars are evolved luminous stars (e.g. Wolf-Rayet, Blue/Yellow/Red Supergiants), we lack an accurate spectral classification, due to the demands that spectroscopy faces at these distances and for this number of stars. What we can do instead is to take advantage of machine learning algorithms (such as Support Vector Machines, Random Forests, Convolutional Neural Networks) to build an automated classifier based on a large multi-wavelength photometric catalog. We have compiled such a catalog based on optical (e.g. Pan-STARRS, OGLE) and IR (e.g. 2MASS, Spitzer) surveys, combined with astrometric information from the GAIA mission. We have also gathered spectroscopic samples of massive stars for a number of nearby galaxies (e.g. the Magellanic Clouds, M31, M33) and by using our algorithm we have achieved a success ratio of more than 80% for the training and test samples. By applying the fully trained algorithm to the available photometric datasets, we can uncover previously unclassified sources, which will become our prime candidates for spectroscopic follow-up aiming to confirm their nature and our approach.

Also Ming has presented his work in a couple of conferences:

1. As a poster presentation at the Supernova Remnants II, Chania, Greece, 3-8 June 2019,

“Evolved Massive Stars at Low-metallicity: A Source Catalog for the Small Magellanic Cloud”

Ming Yang, Alceste Z. Bonanos, Bi-Wei Jiang, Jian Gao, Panagiotis Gavras, Grigoris Maravelias, Yi Ren, Shu Wang, Meng-Yao Xue, Frank Tramper, Zoi T. Spetsieri, Ektoras Pouliasis, Stephan A. S. de Wit

We present a clean, magnitude-limited (IRAC1 or WISE1 ≤ 15.0 mag) multiwavelength source catalog for the SMC with 45,466 targets in total, with the purpose of building an anchor for future studies, especially for the massive star populations at low-metallicity. The catalog contains data in 50 different bands including 21 optical and 29 infrared bands, ranging from the ultraviolet to the far-infrared. Additionally, radial velocities and spectral classifications were collected from the literature, as well as infrared and optical variability statistics were retrieved from different datasets. The catalog was essentially built upon a 1′′ crossmatching and a 3′′ deblending between the SEIP source list and Gaia DR2 photometric data. Further constraints on the proper motions and parallaxes from Gaia DR2 allowed us to remove the foreground contamination. We estimated that about 99.5% of the targets in our catalog were most likely genuine members of the SMC. By using the evolutionary tracks and synthetic photometry from MIST and the theoretical J−Ks color cuts, we identified 1,405 RSG, 217 YSG and 1,369 BSG candidates in the SMC in five different CMDs. We ranked the candidates based on the intersection of different CMDs. A comparison between the models and observational data shows that the lower limit of initial masses for the RSGs population may be as low as 7 or even 6 M⊙, making RSGs a unique population connecting the evolved massive and intermediate stars, since stars with initial mass around 6 to 8 M⊙ are thought to go through a second dredge-up to become AGBs. We encourage the interested reader to further exploit the potential of our catalog.

2. As a talk at the ESO workshop “A synoptic view of the Magellanic Clouds: VMC, Gaia and beyond”, Garching near Munich, Germany, September 9-13, 2019

“Evolved Massive Stars and Red Supergiant Stars in the Magellanic Clouds”

Ming Yang, Alceste Z. Bonanos, Bi-Wei Jiang, Jian Gao, Panagiotis Gavras, Grigoris Maravelias, Yi Ren, Shu Wang, Meng-Yao Xue, Frank Tramper, Zoi T. Spetsieri, Ektoras Pouliasis, and Stephan de Wit

We present an ongoing investigation of infrared properties, variabilities, and mass loss rate (MLR) of evolved massive stars in the Magellanic Clouds, especially the red supergiant stars (RSGs). For the LMC, 744 RSGs compiled from the literature are identified and analysed by using the color-magnitude diagram (CMD), spectral energy distribution (SED) and mid-infrared (MIR) variability, based on 12 bands of near-infrared (NIR) to MIR co-added data from 2MASS, Spitzer and WISE, and ∼6.6 yr of MIR time-series data collected by the ALLWISE and NEOWISE-R projects. The results show that there is a relatively tight and positive correlation between the brightness, MIR variability, MLR, and the warm dust or continuum, where both the variability and the luminosity may be important for the MLR. The identified RSG sample has been compared with the theoretical evolutionary models and shown that the discrepancy between observation and evolutionary models can be mitigated by considering both variability and extinction. For the SMC, we present a relatively clean, magnitude-limited (IRAC1 or WISE1 ≤ 15.0 mag) multiwavelength source catalog with 45,466 targets in total, intending to build an anchor for the future studies, especially the massive stars at low-metallicity. It contains data in 50 different bands including 21 optical and 29 infrared bands, retrieved from SEIP, VMC, IRSF, AKARI, Heritage, Gaia, SkyMapper, NSC, Massey et al. (2002), and GALEX, ranging from the ultraviolet to the far-infrared. Additionally, radial velocities and spectral classifications are collected from the literature, as well as the infrared and optical variability information derived from WISE, SAGE-Var, VMC, IRSF, Gaia, NSC, and OGLE. The catalog is essentially built upon a 1” crossmatching and a 3” deblending between the Spitzer Enhanced Imaging Products (SEIP) source list and Gaia Data Release 2 (DR2) photometric data. Further constraints on the proper motions and parallaxes from Gaia DR2 allow us to remove the foreground contamination. We estimate that about 99.5% of the targets in our catalog are likely to be the genuine members of the SMC. By using the evolutionary tracks and synthetic photometry from MESA Isochrones & Stellar Tracks and the theoretical J−Ks color cuts, we identify 1,405 red supergiant, 217 yellow supergiant and 1,369 blue supergiant candidates in the SMC in five different CMDs. We rank the candidates based on the intersection of the different CMDs. A comparison between the models and observational data shows that, the lower limit of the RSGs population may reach to 7 or even 6M⊙, making RSGs an unique population connecting the evolved massive and intermediate stars, since stars with initial mass around 6 to 8M⊙ are thought to go through a second dredge-up to become asymptotic giant branch stars. We encourage the interested reader to further exploit the potential of our catalog, including, but not limited to, massive stars, supernova progenitors, star formation history and stellar population. Detailed analysis and comparison of RSGs in the LMC and SMC may be also presented depending on the progress of the investigation.

A new outburst of the yellow hypergiant star Rho Cas

Michaela Kraus, Indrek Kolka, Anna Aret, Dieter H. Nickeler, Grigoris Maravelias, Tõnis Eenmäe, Alex Lobel, Valentina G. Klochkova

Yellow hypergiants are evolved massive stars that were suggested to be in post-red supergiant stage. Post-red supergiants that evolve back to the blue, hot side of the Hertzsprung-Russell diagram can intersect a temperature domain in which their atmospheres become unstable against pulsations (the Yellow Void or Yellow Wall), and the stars can experience outbursts with short, but violent mass eruptions. The yellow hypergiant Rho Cas is famous for its historical and recent outbursts, during which the star develops a cool, optically thick wind with a very brief but high mass-loss rate, causing a sudden drop in the light curve. Here we report on a new outburst of Rho Cas which occurred in 2013, accompanied by a temperature decrease of ~3000 K and a brightness drop of 0.6 mag. During the outburst TiO bands appear, together with many low excitation metallic atmospheric lines characteristic for a later spectral type. With this new outburst, it appears that the time interval between individual events decreases, which might indicate that Rho Cas is preparing for a major eruption that could help the star to pass through the Yellow Void. We also analysed the emission features that appear during phases of maximum brightness and find that they vary synchronous with the emission in the prominent [CaII] lines. We conclude that the occasionally detected emission in the spectra of Rho Cas, as well as certain asymmetries seen in the absorption lines of low to medium-excitation potential, are circumstellar in nature, and we discuss the possible origin of this material.

arXiv.org: 1812.03065

From “A User’s Guide to CCD Reductions with IRAF” by Philip Massey (iraf’s photometry docs) we read at page 19 §3.8 that there are two useful options to set the reject parameter:

> reject parameter = crreject
when we know the gain and read noise of our camera

> reject parameter = avsigclip
where the “typical” signal is determined from the data, so no need (or lack of) gain and read noise

A previously published procedure for processing DSLR images with IRIS was somehow complicated. Though it may be more accurate since all the necessary steps for processing images are given, the procedure presented here seems to give the same results without making so much trouble!
Since we have a bright target and small total exposure times the dark/flat/bias frames are not so critical. So we just need to align the images and add them. Then we perform the photometry analysis.

For easiest use download the last version of IRIS 5.58 (especially for any Linux user because the new command convertraw is essential to load raw images).
Then you can install it very easily either to Windows or Linux (using wine).

At the final image that is loaded in IRIS you may not be able to see something. That does not mean that nothing exists! You can increase the visibility of the stars by moving the upper slider in the Threshold Window to the left until stars appear with enough contrast. This is just changing the display settings, not the image itself. In most cases the AUTO selection works fine. In the whole processing procedure you don’t really care to look these images as all these are intermediate steps to make the final image that we are looking for (step 6a).

Step 1 – Initialization:

1a. Open File>Settings and choose your working path (where you have COPIES of your images) and the file type (PIC for DSLR RAW) [IRIS will convert raw images to CFA images].
1b. Open Camera Settings (the button with the campera picture) and select only the right camera model you used and raw interpolation method linear [DO NOT tick the Apply button of White Balance]. The rest are not needed (leave the default values).

Step 2 – Loading:

2a. Click Digital photo > Decode RAW files > (opens a window where you drag and drop your files) and drag & drop your DSLR raw images (e.g. for Canon 300D: *.CR2 files).

The dialog lets you specify a sequence name. then you press “–>CFA” to decode the images. Do this for star field images (e.g. use name ‘img’) and you end up with a series of .pic files named img1,img2,… in the working directory configured in step 1a.

2aa. [alternative way – needed for linux users] Open the command line (button right left from the camera button) and type :

>convertraw input_sequence output_sequence number

input_sequence are the images from your camera and
output_sequence the images that IRIS created (in CFA format).

Step 3 – Alignment of Images:

3a. Go to tab Digital photo>sequence CFA conversion and select the sequence name that you gave at step 2 (‘img’). Give an output name (‘img-conv’) and number of your star images and select Color output files type .

3b. Go to Processing tab>Stellar registration and put the sequence name of 3a (‘img-conv’), give output sequence name (e.g. ‘img-reg’) and number. Select “one matching zone” (no need to define a specific region) and go on.

3c. Stacking of images: Processing tab>Add a sequence and give sequence name of 3b (img-reg) along with the number of images and method. There are 2 options: arithmetic (select also “normalization if overflow” if not selected) or median. They give more or less the same results
[Note: the median option removes extra features that may be visible in the images -you take a clearer image- but the S/N ratio should be smaller]

3d. Save the image!

Step 4 – Selecting green channel:

4a. Go to Digital Photo Tab > RGB Separation (so as to take only the green channel in which we are interested) and enter names for the color channel files, e.g.

final-r
final-g
final-b

You should be able to see see these images at your working directory.

Note: before performing the rgb separation take a look at the image (by selecting AUTO of Threshold Window). You can actually see the colours of the stars !! Then for scientifc reasons we select only the green channel and the images become less beautiful (but more interesting!).

Step 5 – Photometry:

5a. IRIS opens an image with a different way than some other programs. So load (either by File > Load or at command line by typing >load final-g ) the final-g image (from step 4a) and take a carefull look to figure out your field (it should not be that hard for eps aur!).

You can increase the visibility of the stars by moving the upper slider in the Threshold Window to the left until stars appear with eough contrast. This is just changing the display settings, not the image itself. In most cases the AUTO selection works fine.

5b. Go to Analysis>Aperture photometry. With our sample data, you can probably use the default values for aperture, but for your own data you may need to change these sizes so you may wish to review how to size an aperture . If you place the circles to any object you will take an output at the output box with values for Intensity and Magnitude (along with other parameters). The important one is the Magnitude (or Intensity which are equivalent).

[This section adapted from citizensky’s DSLR documentation & reduction team:
> The inner circle defines the area where the star has to fit in. The pixels inside this area must contain all (or at least almost all) light from the star. It should contain a bit of extra sky but never a second bright star.

> The outer ring between the middle and the outer circle defines the area that is considered to be “sky background”. As a rule of thumb, the radius of the outer ring should be about the same as the radius of the inner circle, but on the other hand it should not be so big that stars near any of the stars that you measure will be included in this ring.

> The area between the outer ring and the inner circle is just there to separate the two areas, pixels in this area are ignored. You will set one aperture to fit all stars (variable and comparison stars) that you will measure in an image.

As long as you are working with the same camera, lens, and focus setting, you need to make this decision only once and use the setting all over again for your measurements. When doing the measuremts, you use the aperture like a reticule to “capture” the light of the star you want to measure.]

5c. Now, locate all of the comparison stars you wish to use (see the table of comparison stars). If you have already activated the photometry tool (already done at step 5b) you should see the circles at the mouse point and a tick -activated- left to Aperture photometry option at the Analysis tab). Now, check your comparison star (i.e. eta Aur) and record the output in the spreadsheet or on paper to three decimal places. Repeat for as many check stars you want to use. Finally you select epsilon Aurigae to take an instrumental value of its magnitude.

5d. Repeat step 5c for all your sets of images.

5e. Use the Excel Data Reduction spreadsheet that is available to extract the final results.

Prepared for the CitizenSky.org ‘s DSLR tutorials (with the help of bikeman and Roger_Pieri). CitizenSky.org is dedicated to the eclipsing binary epsilon Aurigae, but the procedure is more general and can be applied to any star.

For easiest use download the last version of IRIS 5.58 (especially for any Linux user because the new command convertraw is essential to load raw images). Then you can install it very easily either to Windows or Linux (using wine).

[Note1: for astronomical images select the Linear interpolation mode for the RAW decoding toward a color images (less artifact around pointlike objects).

Note2: when you will be loading series of images the IRIS will pause for some time to read, trasform,calculate and save the images (depending on the number and the computer of course). Only the final output of each procedure will be visible but that does not mean that IRIS has rejected the rest (you specify a working directory and the program will look into this according to the input you give).

At the final image that is loaded in IRIS you may not be able to see something. That does not mean that nothing exists! You can increase the visibility of the stars by moving the upper slider in the Threshold Window to the left until stars appear with enough contrast. This is just changing the display settings, not the image itself. In most cases the AUTO selection works fine. In the whole processing procedure you don’t really care to look these images as all these are intermediate steps to make the final image that we are looking for (step 6a).

Step 1 – Initialization:

1a. Open File>Settings and choose your working path (where you have COPIES of your images), the script path (only if you have any scripts that you want to use, otherwise not necessary, the file type (PIC for DSLR RAW) [IRIS will convert raw images to CFA images].

[Note – different raw interpolation methods explained (linear, median, gradient):

For that the software must interpolate information collected by close pixels covered either with red filters, green filters and blue filters. The operation is so called demosaicing. The result is an image where each color plan is coded on 16 bits (either 48 bits of information by pixel). The bilinear method for interpolation is fastest, but it is passable concerning spatial resolution (restitution of the finest details). However this method remains often viable in many situation because the resolution is degraded at least same value at the time of the registration of a sequence of images. The median method preserves the resolution, but applies a filter which degrades the natural aspect of the stars for deep-sky imagery. This technique is not recommended for the stars images. The gradient method preserves well the resolution. The calculating time is long, but it is the method which gives the best results for the details point of view, even if in deep sky imagery the stars are not free from artifact.]

1b. Open Camera Settings (the button with the campera picture) and select only the right camera model you used and raw interpolation method linear [DO NOT tick the Apply button of White Balance]. The rest are not needed (leave the default values).

Step 2 – Loading:

2a. Click Digital photo > Decode RAW files > (opens a window where you drag and drop your files) and drag & drop your DSLR raw images (e.g. for Canon 300D: *.CR2 files).

The dialog lets you specify a sequence name. then you press “–>CFA” to decode the images.

You do this repeatedly for:

star field images (e.g. use name ‘img’)

dark frames ( suggested name ‘dark’)

flat field images ( suggested name ‘flat’)

bias frames ( suggested name ‘bias’)

After this you will have files named

img1,img2,…

dark1,dark2,…

flat1,flat2,….

bias1,bias2,…

in the working directory configured in step 1a.

2aa. [alternative way – needed for linux users] Open the command line (button right left from the camera button) and type :

>convertraw input_sequence output_sequence number

input_sequence are the images from your camera and

output_sequence the images that IRIS created (in CFA format).

Do it for all the images that you have (star-field images, dark, flat, bias).

Step 3 – Preprocessing > Bias, Dark, Flat:

3a. Go to Digital Photo Tab and select Make Offset (Bias frame). Type the appropriate sequence name (the one used in step 2, e.g. ‘bias’) and the number of bias images. Then select the command line button (the one on the left from camera button) and type:

>save master-bias

of the created image (you can use of course whatever name you want!)). You can also save it by File>Save , but the command line can be kept opened all the time as it will be easier to use.

3aa. IRIS needs a master bias frame (also called offset) to continue processing. So in case you don’t have bias images you can create a “fake” one. Open a star field image (or one from the sequence img) and type at command line:

>fill 0
>save master-bias

3b. Do the same for dark, you enter the appropriate generic name (only the sequence name), the offset image name you created previously (‘master-bias’), the number of dark frames and the method median (median or mean should both work well). Also, go and save the image as ‘master-dark’ (either type: save master-dark , or File>Save).

3c. Improve dark frame substraction: open the master dark frame (the image that you created at step 3b, which should be already opened) and at command line type:

>find_hot cosme 500

and check the output box (should open automatically). There is a value which should be 100<output<200.

3d. At the same Tab (Digital Photo) do the same for flat-fields and select Normalization value = 1000. [After creating the image you can check the output of the command STAT and see what is the max output value. This should be less that 32767. If not arrange accordingly the normalization value and check again. You just have to be under 32767].

Before saving the final flat image type at command line:

>grey_flat

so as to normalize the CFA-flat and then save the flat field to “master-flat” (either type: save master-flat , or File>Save — by now you should have understood why the command line is faster and easier !!)

3dd. In case you do not have flat images you can create a flat sequence as follows: first select 3 images from the star field images (prefferable the first, the median and the last one – as the camera is not moving the field rotates so by using these images you can remove the stars and take correction for system, lens+camera, of course this should not substitute entirely the processing for making flat images ! But in some cases…it can be really handy.). RENAME them as flat1 flat2 and flat3 (let’s call this sequence flat). In order to make the final master-flat image type at command line:

>sub2 flat master-bias i 0 number

(choose a region 100×30 pixels or more in the image – does not matter where exactly)

>opt2 i master-dark i number
>ngain2 i 1000 i number
>smedian i number
>grey_flat
>save master-flat

where flat is the sequence we created, master-bias (or offset or anything you named the saved image of step 3a), master-dark (step 3b) number the number of images (in our case 3). i is a sequence that helps us to build the master-flat. Also in this case you can check first with STAT command if the output after SMEDIAN is correct (according to what is written in step 3d) and then procceed with the grey_flat (see step3d) and saving the image.

3e. Go to Digital Photo>Preprocessing. Input the sequence name of you data images (img) along with the offset map (master-bias, step 3a), dark map (master-dark, step 3b), flat-field map (master-flat, step 3d), cosmetic file (step 3c). Give an output sequence name (e.g. ‘img-cal’) and the nunber of files. It is not necessary to tick dark optimize as it will take a little bit longer without real effect on the final result [of course you can use it if you want !]

Step 4 – Alignment of Images:

4a. Go to tab Digital photo>sequence CFA conversion and select the sequenace name that you gave at step 3e (‘img-cal’). Give an output name (‘img-cal-conv’) and number of your star images and select Color output files type .

4b. Go to Processing tab>Stellar registration and put the sequence name of 4a (‘img-cal-conv’), give output sequence name (e.g. ‘img-reg’) and number. Choose Global matching and Quadratic transformation

4c. Stacking of images: Processing tab>Add a sequence and give sequence name of 4b (img-reg) along with the number of images and method arithmetic (select also “normalization if overflow if not selected).

4d. Save the image!

Step 5 – Selecting green channel:

5a. Go to Digital Photo Tab > RGB Separation (so as to take only the green channel in which we are interested) and enter names for the color channel files, e.g.

final-r

final-g

final-b

You should be able to see see these images at your working directory.

Note: before performing the rgb separation take a look at the image (by selecting AUTO of Threshold Window). You can actually see the colours of the stars !! Then for scientifc reasons we select only the green channel and the images become less beautiful (but more interesting!).

Step 6 – Photometry:

6a. IRIS opens an image with a different way than some other programs. So load (either by File > Load or at command line by typing >load final-g ) the final-g image (from step 5a) and take a carefull look to figure out your field (it should not be that hard for eps aur!).

You can increase the visibility of the stars by moving the upper slider in the Threshold Window to the left until stars appear with enough contrast. This is just changing the display settings, not the image itself. In most cases the AUTO selection works fine.

6b. Go to Analysis>Aperture photometry. With our sample data, you can probably use the default values for aperture, but for your own data you may need to change these sizes so you may wish to review how to size an aperture . If you place the circles to any object you will take an output at the output box with values for Intensity and Magnitude (along with other parameters). The important one is the Magnitude (or Intensity which are equivalent).

[This section adapted from citizensky’s DSLR documentation & reduction team:

> The inner circle defines the area where the star has to fit in. The pixels inside this area must contain all (or at least almost all) light from the star. It should contain a bit of extra sky but never a second bright star.

> The outer ring between the middle and the outer circle defines the area that is considered to be “sky background”. As a rule of thumb, the radius of the outer ring should be about the same as the radius of the inner circle, but on the other hand it should not be so big that stars near any of the stars that you measure will be included in this ring.

> The area between the outer ring and the inner circle is just there to separate the two areas, pixels in this area are ignored. You will set one aperture to fit all stars (variable and comparison stars) that you will measure in an image.

As long as you are working with the same camera, lens, and focus setting, you need to make this decision only once and use the setting all over again for your measurements. When doing the measuremts, you use the aperture like a reticule to “capture” the light of the star you want to measure.]

6c. Now, locate all of the comparison stars you wish to use (see the table of comparison stars). If you have already activated the photometry tool (already done at step 6b) you should see the circles at the mouse point and a tick -activated- left to Aperture photometry option at the Analysis tab). Now, check your comparison star (i.e. eta Aur) and record the output in the spreadsheet or on paper to three decimal places. Repeat for as many check stars you want to use. Finally you select epsilon Aurigae to take an instrumental value of its magnitude.

6d. Repeat step 6c for all your sets of images. As long as you are using the same lens and DSLR, you don’t have to re-create a master-flat and master-bias for every observation session! You can save them for re-use and then skip the steps 3a and 3d.

6e. Use the Excel Data Reduction spreadsheet that is available to extract the final results.