Imaging the Great Dimming of Betelgeuse
During the Northern winter of 2019-2020, the nearby red supergiant star Betelgeuse experienced a historic drop of its brightness. By directly imaging the star, we found what happened to it. This is the story of a two year long adventure.
The story behind this Nature article actually started a year and a half before what we call the Great Dimming. In March 2018, I submitted an observing proposal at ESO to observe Betelgeuse, and another red supergiant (CE Tau) with the Very Large Telescope (VLT) and its interferometer (VLTI). This is my regular research: aiming to understand the mechanism(s) triggering the mass loss of cool evolved stars, and in this case red supergiants. We still do not know how these stars can lose up to 60% of their mass in the final stages of their life. This has important consequences for their final fate because this mass loss will set the nature of the supernova progenitor (red supergiant, or a blue or yellow giant if the star experiences a blue loop, see Meynet et al. 2015). It can also change the shape and peak value of the supernova light curve when it happens (Moriya et al. 2018), and set the final mass of the star, hence determine the nature of the compact remnant (neutron star or black hole).
The observations were approved in June 2018 in two parts:
- The VLTI/GRAVITY run was scheduled remotely in Designated Visitor Mode for January 2019. GRAVITY is one of the instruments recombining the light of four telescopes in the interferometric mode. It operates in the near-infrared (2-2.45 µm). Betelgeuse being the brightest star of the infrared sky (after the Sun), I applied for the small Auxiliary Telescopes (yet, 1.8m in diameter!) that can be moved into different configurations, which is crucial for interferometry. The goal of the GRAVITY observations was to image the convective pattern of the red supergiants.
- The VLT/SPHERE run was scheduled in visitor mode, meaning that I was going to travel to Chile. The observations would take place on the 30th and 31st of December 2018: I would spend New Year's eve observing my favorite star with the most powerful instrument/telescope combination in the world. Indeed, in the visible, SPHERE is reaching a better angular resolution than the Hubble Space Telescope, thanks to its extreme adaptive optics. Adaptive optics is a technique that compensates for atmospheric distorsions on the images, by using analysis of the light front to compute the corrective deformations (of a few nanometers) that are applied a thousand times per second on thousands of point across a deformable mirror, tens of centimetre in diameter. This equals transporting the 8.2m diameter VLT into space. With SPHERE I intended to look for dust in the circumstellar environment using polarimetry.
On December 26th 2018, I left Europe with Emily Cannon, the PhD student I am co-supervising. We arrived at Santiago in Chile on the morning of 27 December. In the early morning of December 28th, we took another plane to Antofagasta on the Pacific coast, the gate to the Atacama desert. After the usual 2h coach ride, we reached ESO's Cerro Paranal observatory, where many telescopes (not just the VLT/VLTI) are installed at 2600m above the sea level. It was my fourth time there, the happiness was still the same. I was even happier to share that with Emily, like my PhD co-supervisor Pierre Kervella did with me 7 years earlier. The first two days we prepared the observations: Emily was in charge of the backup program while I dealt with the main one. On December 30th, we went to the telescope. Bad luck: the Northern half of the sky remained cloudy. That's what the backup program on the Southern side was for. December 31st was the weirdest day. The families of the ESO staff members had joined them for New Year's eve. It was amazing to see and hear children running and laughing in the mineral and sometimes austere corridors of the VLT's residencia. At dawn, the traditional sunset moment before starting the night was replaced by a show. No fireworks of course, but the four lasers of UT4/Yepun were shot into the sky. Normally used to create artificial stars for the Adaptive Optics system, that evening they were there for the families to enjoy this incredible view. The night was better, I started the new year observing Betelgeuse. How could it be better? On January 1st 2019, on the way back to Antofagasta, I started to reduce the data through the pipeline on my laptop. The images were in the continuity with the 2015 observations (Kervella et al. 2016), it would make a nice follow-up paper. On January 3rd, we were back in Europe. During January 2019, I spent four mornings (between 3h and 8h) observing Betelgeuse with GRAVITY.
With other projects going on in 2019, I put these newly acquired data in standby for a while.
During the month of November 2019, we started to get reports that Betelgeuse was fainter than it should have been (according to predictions related to its pulsation cycle). On December 7th, the Astronomer Telegram #13341 announced The Fainting of the Nearby Red Supergiant Betelgeuse. On December 11th, I wrote to my collaborators Eric Lagadec (Lagrange, Nice, France) and Pierre Kervella (LESIA, Meudon, France) that, at that time, it could well have been a conjunction of the two pulsation cycles (periods of ~400 and 2000 days) like we witnessed in 2009.
The Great Dimming
The visual brightness of Betelgeuse continued to drop, and started to attract the media's and the public's attention. On December 19th, I decided to apply for observing time... to prove that nothing special was happening to the star! Instead of the regular call for proposal issued twice per year (in March and September for ESO), I had to use the Director Discretionary Time (DDT) channel, that allows to request time when something unexpected happens. My reasoning was that if the star got dimmer, it was because the brightness of its surface had apparently decreased. The best way to identify what was happening was then to resolve the stellar surface (photosphere): this means looking at the star as a disk using high angular resolution techniques (adaptive optics and optical interferometry) when most observations are still looking at stars as non-resolved points. Betelgeuse is one of few star that can actually be resolved.
My strategy was to use three of the VLT/VLTI instruments, the same ones I had the year before:
- VLT/SPHERE-ZIMPOL, the visible sub-unit of SPHERE, the only one that allows to resolve directly the surface of a handful of stars, and to look for dust in their circumstellar environment,
- VLT/SPHERE-IRDIS, the near infrared sub-unit of SPHERE, in its Sparse Aperture Masking (SAM) mode, to supplement the GRAVITY observations at short telescope separation,
- VLTI/GRAVITY, a near infrared interferometric combiner, equipped with both spatial and spectral resolution, in order to measure the diameter of the star (had it shrunk?) and assess the status of the convective pattern.
The proposal was sent to the co-investigators (CoIs) on December 20th, and to ESO after minor modifications on December 22nd (three days for the whole process when in regular calls it takes about a week!). This was the ideal time to start the Christmas holidays.
On the 27 December at 0h28, just as I was about to set my phone to 'do not disturb', I received an email from ESO asking me to send the OBS directly . It took me a few seconds to understand that my observations had been accepted, and that I was asked to bypass the normal procedure because all the chain of people supposed to validate it were on holidays! Instead of going to sleep, I took my laptop and prepared the OBs. I sent them at 1h38 in the morning and finally went to sleep, happy. At 6h18 CET that night while I was sleeping in Belgium, in Chile, mounted on UT3/Melipal, SPHERE was pointed towards Betelgeuse and then its calibrator star the blue giant Rigel. At 9h30 when I woke up, I saw the email confirming the successful execution of the observations. However, still because of the holiday seasons, I could not retrieve the data at 14h the same day and had to wait for December 29th when Paranal's shift coordinator Joe Anderson sent the data (many thanks for that!). I was on holidays at Rotterdam at that time, I reduced the data during the evening and to my great surprise saw that Betelgeuse was not as it should be: the Northern hemisphere was much dimmer than the Southern one. I had been wrong: something unusual was happening to the red supergiant!
It's only the day after on December 30th that I finally saw Betelgeuse with my own eyes. It was so faint compared to its usual brightness. Normally it rivals with the blue giant Rigel, however at that time it was similar to Bellatrix. How ironic can life be: being the only person knowing what was happening to Betelgeuse, I only saw the event with my eyes after having seen its stellar surface in my data. "Being the only person knowing what was happening to Betelgeuse..." or so I thought!
New observations, first analysis
At the beginning of the infamous year 2020, while Betelgeuse was still getting dimmer and dimmer, I finished the data reduction. My team (that was growing together with the attention on Betelgeuse) and I decided to target a fast publication of the images. Then convinced that something unusual was happening to the star, I decided to ask for more observations. I wrote a new proposal with a slight change in strategy:
- I chose the Target of Opportunity (TOO) mode, that allows to decide when the observations are executed. I wanted to get new images with SPHERE-ZIMPOL exactly at the minimum of light and afterwards.
- In addition to new images with VLT/SPHERE-ZIMPOL, I applied this time for observations with VLTI/MATISSE, the mid-infrared interferometric combiner. My goal was to characterize the thermal emission of a potential dust cloud (for example its composition). Being an interferometric instrument, MATISSE has a much better angular resolution than VISIR.
This new program was approved on January 21st 2020 and I immediately prepared the OBs. From there I started a daily monitoring of Betelgeuse's light curve provided by the American Association of Variable Star Observers.
It's around that time that I started the modelling effort, at that time only by using radiative transfer dust simulations based on the code RADMC3D (Dullemond et al. 2012). On January 27th, I was approached by ESO to publish a photo release (like a press release but showing only images without scientific interpretation). It was a risky move: showing the images before any publication might prevent publication of the result in a high impact journal, and give food for thought to other teams. We decided to go for it anyway, and go for a fast publication with minimal interpretation. On January 28th, the Dimming seemed to have reach a plateau so I triggered the first set of TOO observations, crossing my fingers that it was indeed the real minimum of the Dimming, and not an intermediate pause.
A secret revealed...
After many iterations with my research team and the great press relation team at ESO, the two pictures of Betelgeuse from January and December 2019 were released on February 14th, revealing to the world that it was my secret Valentine.
Scratching our heads
Meanwhile, I received new SPHERE, GRAVITY and MATISSE data. It was then clear that the minimum of light happened in early February (Astronomer Telegram #13512). It happened to be a historic minimum: not in 150 years had Betelgeuse been so dim! I worked on reducing, calibrating and analyzing these new data. On March 9th, the first refereed paper on the Great Dimming was published (Levesque & Massey 2020). It claimed that a drop in temperature alone could not explain the Dimming. We decided to pursue our publication strategy.
On March 11th, before any official recommendation from the authorities, the team in which I was working in Leuven decided to move to 100% teleworking. From this day, all the meetings happened through videoconference. On March 14th, Betelgeuse being brighter, I triggered the last set of TOO observations with SPHERE-ZIMPOL. These observations were executed on March 20th-21st, three days before the VLT closed due to the pandemic.
From there, I worked mainly on the data and their modelling. With the combination of VLTI/GRAVITY and VLT/SPHERE-IRDIS (SAM mode) data, I determined the size of the star in the near-infrared (where we are less screened by molecules). This was not as easy as just taking a ruler and measuring the size of the disk on an image. These instruments produce interferometric observables (squared visibility and closure phase) which are directly linked to the Fourier transform of the light intensity distribution of the star. The angular size is derived from model fitting after removing telluric and stellar line contributions (which make the star look larger) and smaller scale structure biases.
The size of the star (that had not changed significantly a year before and during the Dimming) served as input for the RADMC3D dust radiative transfer simulations. Basically, I put a star and dust cloud within a numerical grid. The star is then launching photons (at different wavelengths, according to its temperature) to set the dust equilibrium temperature. And then the code determines what the star and dust cloud look like from the Earth after light scattering. I ran thousands of simulations to determine the best set of parameters (clump position, size, dust composition, grain size) that would best reproduce the observations. I also tried more complex models using a spherical envelope to reproduce the circumstellar environment (Bad idea, it added too many parameters that could not be constrained. I changed for an extinction law accounting for both interstellar and circumstellar permanent dust extinction). Each single simulation took about one hour to run (to produce images and spectra). I used the KU Leuven Institute of Astronomy computer grid to run these simulations, still each set took several days (even with ~100 simulations running in parallel). While working on this, I realized that I had to better re-center the images between the different epochs and wavelength filters (there is no absolute pointing of SPHERE at these spatial scales).
During this part of the work, I encountered several ups and downs. With the pandemic, without being able to directly interact with my colleagues in Leuven, it was not easy to remain certain that these simulations would eventually converge to a result. Luckily through online meetings with my collaborators I received much suggestions, comments and advice that allowed me to stay on track.
In June 2020, an article was published by Dharmawardena et al. (2020) suggesting the dimming was caused by a localised cooling of the photosphere, this was contrary to previous assertion by Levesque & Massey 2020. That was an extremely interesting result, but also a difficult one that questioned my current work. After a discussion with my collaborators, we decided to test the hypothesis by creating cool spot models (filling the stellar disk with PHOENIX models at different temperatures). Again a large parameter space had to be explored through thousands of models.
After a computer crash in early July (luckily I had backups of everything), I converged towards the best models presented in the Nature article and whose images are shown below, together with the four epoch observations.
Early August 2020, I sent the abstract pre-submission to Nature. We got the green light to work on the real article a few days later. It was a difficult task. As always, formally writing a paper allows to uncover the biases in the method, and to refine the models. This article was no exception. Due to the length limits and targeted readership of Nature, the exercise was a bit different from standard scientific paper, but definitely an enriching experience. After few weeks, the time had come for iterations with the co-authors. Meanwhile, in October 2020, I prepared to move to my new PSL post-doctoral fellow position at the LESIA, in Paris Observatory (France). Of course, this did not happen as scheduled due to the pandemic. With the new rise of cases and the 2nd French lockdown, I moved to my family home in south of France. This is how, like many academics, I started a new position without directly meeting my new colleagues or going to my office.
The Nature article was submitted at the beginning of November 2020. The referee reports came in January 2021 and were positive and extremely constructive. The re-submission happened mid-February 2021, almost one year after the initial photo release. The green light from the referee occurred mid-March 2021, only editorial comments were left. Final submission happened end of March and final acceptance mid-April 2021. What a long journey!
What is the outcome of this adventure? We think the Great Dimming of Betelgeuse was caused by the presence of a dust cloud in the line of sight, whose formation was caused by the cooling of a patch of the stellar surface. Of course, this is what can be inferred from the data presented in the article: the images in the visible and the flux measurements in the visible and near-infrared. In the end, each observer is biased by the spectral range and the method he/she uses. Further observations, particularly with radio interferometry are ongoing/proposed to try to detect the gas component that should accompany the dust clump. The understanding of the Great Dimming will keep Betelgeuse enthusiasts busy for some time. This episode is a unique oppportunity to understand the dynamics and the mass loss mechanism of red supergiant stars, and to better constrain their spectacular fate as supernovae, and later on as neutron stars or black holes.
But this is not all. The V magnitude of Betelgeuse has been monitored continuously. The rise after February 2020 has been very fast, reaching the range of the usual bright magnitudes of the star in May, while according to the well known 420 days period (Stothers 2010), the light maximum was scheduled for September-October 2020. In July, while Betelgeuse was too close to the Sun to be observed from Earth, A. Dupree used the STEREO-A solar probe (trailing Earth's orbit and offering another vantage point, without the Sun in the line of sight) to obtain a new V-band magnitude measurement (Astronomer Telegram #13901). The star that was supposed to be on course towards its maximum brightness was already getting dimmer. And since then, its light curve has been eratic, with several seondary minima and maxima.
What is happening? I don't think anyone knows. What is certain is that we will continue to follow this very intriguing and famous star. Actually, my team and I obtained a new image with VLT/SPHERE-ZIMPOL on April 5th 2021. But I am not going to show it or discuss it here. You will have to wait for the next publication!
Meanwhile, I am organizing a Special Session at the European Astronomical Society (EAS) 2021 meeting, in order to bring together all the teams involved in the observations of this event. The idea is to discuss together our views of what happened, and trigger new collaborations. This is how science advances: through disagreements and friendly discussions.
I would like to deeply thank my amazing team of colleagues that worked with me throughout this great adventure, and also my friends that supported me during this difficult times.