≡ Menu

Proxima Centauri: Transits Amidst the Flares?

Discovered in 1915, Proxima Centauri has been a subject of considerable interest ever since, as you would expect of the star nearest to our own. But I had no idea research into planets around Proxima went all the way back to the 1930s. Nonetheless, a new paper from Emily Gilbert (University of Chicago) and colleagues mentions a 1938 attempt by Swedish astronomer Erik Holmberg to use astrometric methods to search for one or more Proxima planets. The abstract of the Holmberg paper (citation below) reads in part:

Many parallax stars show periodic displacements. These effects probably are to be explained as perturbations caused by invisible companions. Since the amplitudes of the orbital motion are very small, the masses of the companions will generally be very small, too. Thus Proxima Centauri probably has a companion, the mass of which is only some few times larger than the mass of Jupiter. A preliminary investigation gives the result that 25% of the total number of parallax stars may have invisible companions.

Holmberg (1908-2000) seems to be best known for his work on galaxy interactions, but the movement of nearby stars was clearly a lively interest. I think we can assume that the Proxima ‘detection’ was due to systematic factors; i.e., noise in the data. But because I was fascinated by this early flurry of exoplanet hunting, I dug around a bit in Michael Perryman’s The Exoplanet Handbook (Cambridge University Press, 2011) to learn that Holmberg came back in 1943 to use long-term time-series photographic plates in another astrometric hunt, this one at 70 Ophiuchi (he thought he detected a gas giant), while in the same year, the Danish astronomer Kaj Aage Gunnar Strand (1907-2000) found evidence for a 16-Jupiter mass planet around 61 Cygni. Neither of these worlds turned out to be any more real than Holmberg’s putative planet at Proxima Centauri.

So exoplanet hunting is replete with false positives. We’ve talked at some length in these pages about the work of Peter van de Kamp (1901-1995), a Dutch astronomer living in the US, on possible planets at Barnard’s Star. His detections were ultimately shown to have resulted from systematic errors in his equipment (though unless I am mistaken, he never accepted this conclusion). Van de Kamp’s work in the 1960s and later made the point that a small number of astronomers have been actively searching for exoplanets long before the detection of 51 Pegasi b, but the Holmberg paper, taking us back prior to World War II, came as a surprise I wanted to share.

Searching for Transits

On to today’s work on Proxima Centauri, which as we know has no gas giant of the sort Holmberg deduced, but does host at least two planets, among which is the fascinating Proxima Centauri b, the latter in the habitable zone of the star. Habitability, however, is problematic. Proxima is a flare star, so active that the atmosphere of a habitable zone planet may be threatened by the intense radiation, with obvious implications for surface life.

Proxima’s flares can dominate observation, as this passage from today’s paper makes clear:

We see 2-3+ large flares every day in the 2-minute cadence TESS light curve (Vida et al., 2019), and even with TESS 2-minute cadence, optical photometry, it can be hard to fully resolve flare morphology in order to search for transits. Davenport et al. (2016) even suggest that the visible-light light curve of Proxima Centauri may be so dominated by flares that the time series can be thought of as primarily a superposition of many flares.

Image: Proxima Centauri is a “flare star,” meaning that convection processes within the star’s body make it prone to random and dramatic changes in brightness. The convection processes not only trigger brilliant bursts of starlight but, combined with other factors, mean that Proxima Centauri is in for a very long life. Astronomers predict that this star will remain middle-aged — or a “main sequence” star in astronomical terms — for another four trillion years, some 300 times the age of the current Universe. These observations were taken using Hubble’s Wide Field and Planetary Camera 2 (WFPC2). Its two companions, Alpha Centauri A and B, lie out of frame. Credit: NASA/ESA.

Indeed, Proxima’s flares appear nearly continuous, occurring at a range of wavelengths. Flares can induce shifts in radial velocity measurements, making observations noisy and burying a potential planetary signal in a sea of misleading data. The same problem occurs for transit detection, where flare-induced variations in the lightcurve may mask an actual planetary signature. All this makes clear what fine work Guillem Anglada-Escudé and team performed at unpacking the radial velocity data that first revealed the existence of Proxima b in 2016.

Because flare activity may be masking transits at Proxima Centauri, Gilbert and team modeled the stellar activity in their planet search algorithm, refining the result. Previous flare detection algorithms have tried to identify the flares and remove them, revealing the more stable stellar signal beneath. The authors take a different approach, using their own algorithm to first identify flares, then modeling them using a template, subtracting them from the data and running a transit search on the result. The unique flare modeling they apply to Proxima, painstakingly presented in the paper’s section on methods, involves a multi-step process of filtering and fitting the flare data. The scientists injected transits into the light curves before modeling as a way of determining how sensitive their method was, with results that boosted the planet signal.

Image: This is Figure 3 from the paper. Caption: By subtracting a flare model from the light curve of Proxima Centauri, we are able to significantly increase the probability of recovering small planets. We are able to reliably recover planets down to around the radius of Mars across the period range searched, effectively ruling out any transit of Proxima Centauri b. Credit: Gilbert et al.

A transit at Proxima Centauri would be a huge boon to astronomers, allowing a precise radius and accurate determination of the composition of Proxima Centauri b (the same is true, of course, of Proxima c). Moreover, Proxima b’s short orbital period would mean frequent transits, and thus would elevate its status as a place to look for biosignatures in the atmosphere.

Alas, we have no transits. Gilbert’s team used TESS observations of Proxima Centauri to make this determination. The conclusion seems tight. From the paper:

We find no evidence for Proxima Centauri b in TESS data. This is not surprising because previous efforts using different telescopes have been similarly fruitless.

Using the known minimum mass of Proxima Centauri b (Msini = 1.27 M), we used the relationship from Chen and Kipping (2017) to derive an expected planet radius to be R = 1.08 ± .14 R. A 100% Iron planet would have an expected radius of 0.88 R (Zeng et al., 2019). Therefore, given our injection and recovery tests show that no planets larger than 0.4 R transit Proxima Centauri at periods between 10–12 days, we are confident that we would recover the signal from Proxima Centauri b if it were to transit.

The team’s flare modeling is a big part of the story here, improving the sensitivity to transiting planets from 0.6 Earth radii to 0.4, and allowing the team to put rigorous limits on the probability of transits. According to the authors, this kind of flare modeling is a technique that should be applicable to all active stars. This is good news for the continuing work of TESS and the future work of Plato (ESA’s PLAnetary Transits and Oscillations of stars mission). Our sensitivity to small planets transiting low-mass, nearby active stars receives a boost from these methods, even though the search for transits at Proxima Centauri comes up empty.

The paper is Gilbert et al., “No Transits of Proxima Centauri Planets in High-Cadence TESS Data,” accepted at Frontiers in Astronomy and Space Sciences (preprint). The Holmberg paper is Holmberg, E., “Invisible Companions of parallax stars revealed by means of modern trigonometric parallax observations,” Meddelanden fran Lunds Astronomiska Observatorium, Series II 92, 5–25.

tzf_img_post

{ 21 comments… add one }
  • Alex Tolley November 23, 2021, 13:42

    Does this work imply that the Keppler data might yield more planets around M dwarfs with active flares that might have been missed? If so, is there any sense of how many that could be?

  • Geoffrey Hillend November 23, 2021, 17:57

    I was so looking forward to the JWST looking at the spectra of Proxima Centauri b. Maybe there not any planets because of Proxima A being part of a triple star system and some kind of angular momentum being used up or gravitational tidal forces in the protoplanetary disk so there are not any planets?

  • Kamal Ali November 24, 2021, 4:25

    Can someone explain to me: I have seen articles stating that during a flare, the star can brighten by a factor of 10 or 15. That seems physically impossible to me. Does it brighten just in a very narrow wavelength band, or for a very short period of time? Would the brightening be noticeable to voyagers standing on the surface of, say, Proxima B ? Would the brightening be catastrophic in terms of temperature on the surface of a planet?

    • wdk November 25, 2021, 12:37

      You’ve got a good question that’s worth a sidebar. My comments are not definitive, but others can likely add or revise.

      Imagine the black body curve for the red dwarf planet with its characteristic temperature and its peak in the infrared or red vs. the middle of the visual for the sun – which is two or three times hotter ( 5800 K). Now, superimpose on that a flare event in the ultra violet with spill over onto other parts of the spectrum. The flare event itself could release more energy ove ran interval than the normal radiation of the star (e.g., Proxima).

      The flares come in all sizes. And for red dwarfs very frequently.
      The planets in the habitable zone are more closely placed to the star than with our sun. Their ringside seats make line of sight exposure
      more intense due to an inverse square of radial distance.

      A number of things could make the flare hazardous:
      The photo emissions could be directly aimed at the planet or else the charged particles released. The ultra violet contribution to the BB spectrum of the quiescent star is much lower, and the release is increased to levels larger than our own sun’s.

      Upper layers of atmospheres can be stripped by repeated flares. On the surface here on Earth, surface desert environments give us UV
      exposures serious enough to our health, if not necessarily to Little Green Men. I am sure that the intensity of a typical Proxima flare under these circumstances would make the Sahara or the Andes highlands a piker by comparison.

      Are there mitigating considerations? Perhaps. One that lies in the background is the question of prevalent distribution of flares. Possibly polar or out of the orbital plane – but just a hope for habitability. Others might be how other planets reach a state which we would assess as habitable. The Earth’s geological history is complicated and distinct. It could be a liability in another setting. Or circumstances of formation might in another star system might allow habitability reached by another route.

  • torque_xtr November 24, 2021, 5:44

    Unfortunately it seems that TESS does not have any bandpass filters, but M-dwarf flares and background photosphere have widely different spectra. In case of M-dwarfs, T_eff differs by as much as three times, quiescent photosphere at 3000K and flares at 10000K. Possibly could be thoroughly filtered out by building flare spectral distribution, observing flare activity in blue-to-ultraviolet bands and subtracting corresponding values from near-to-mid IR brightness where flare contribution is small and could be made even smaller.

  • Mike Serfas November 24, 2021, 8:39

    The paper mentions a peak time but I don’t see how long it usually is. But at a 2-minute cadence there seem to be some flares with one flux measurement more than twice as high as any other – quite a camera flash! Proxima b is 25 seconds away from the star… when it is “full”, almost 50 seconds there and back… is there any way to dig out reverse transit-like data by taking the 20-second cadence data and looking for late peaks at the tail end of flares that match with the phase of the planet? (If only you could figure out exactly when flares occur on the far side of the star…)

  • jonW November 24, 2021, 10:54

    Has the rotational plane of any star (and hence the likely plane of any orbiting planets) ever been determined by any means other than planetary transits?

    • Antonio November 24, 2021, 13:01

      The rotational plane of a star is not determined by planetary transits.

      • jonW November 25, 2021, 17:49

        Really? I thought that planets orbited in the same plane that the host star rotates in (I mean the plane such that the axis of rotation is perpendicular to it). So if you detected a transit, you know (to within some margin of error) what plane the planet is orbiting in (it’s the one that is edge-on from our point of view), and therefore (with a certain probablity) what plane the entire system is orbiting in, and therefore (roughly) what plane the star is rotating in.

        Which part of this have I misundertood? Thanks

        • Antonio November 26, 2021, 16:28

          It’s not always that way. There are cases were a planet inclination changes (due to the Kozai effect or others). Other times planets form that way in the very crowded environment of a stellar nursery. Binary stars also complicate things. Even in our Solar System there are TNOs with 90º inclination, probably caused by the influence of Planet 9 (thought to have a high inclination orbit too). Even a big TNO like Eris has 44º inclination.

        • Andrei November 26, 2021, 16:58

          While it generally is the case that a planet orbits in the same plane as the star rotates, any planet can have an orbit that takes it above or below the star from our point of view. And we got one example where only the innermost planet orbit in this way, while the others have been offset so much the orbit is in 90 degrees, meaning the planets pass over the poles of the star.

          • wdk November 28, 2021, 17:50

            The paper described below is well written and fascinating.
            However, I think there are some summary “observations” about this matter that might help. Encountered this subject sometime prior to exoplanet discovery and the Rossiter – McLaughlin effects were discussed with some reference to binary stars, but mostly to determine the orientation of a stellar axis.
            The overall effect was that the inclination of the rotation axis away from the perpendicular to the observer (e.g., us) would tend to broaden and depress an absorption line depth at the same time.

            Assuming the rotational axis was perpendicular, a reference absorption line would be relatively sharp, but the edge rotational velocities toward and away from the observer would cause some inherent broadening. Other phenomena can cause variations in line broadening as well such as the visible zone stellar surface pressure. But if you are looking at two stars of the same spectral type (e.g., G2), then differences in width would likely be due to differences in polar inclination, assuming similar rotation rates; tilt forward or backward causing an increase in width, but a diminishing of depth.

            Now so far so good, except that a star of a given spectral type could still have a variation in rotational velocity. There are statistical estimates for spectral types, but the exceptions would still be significant. So, means for discriminating rotational velocities vs inclination should enter into the analysis. A saving grace is that v sin i vs half width of an absorption line was found to be linear – extending from a base value such as given above for zero inclination. If a particular line was not sufficient to determine an axis inclination, then there were other candidate lines to select with differing zero inclination depths. Perhaps enough variation to separate v from i.

            “Observation and Analysis of Stellar Photospheres” by David Gray (1976) used the example of 4324 Angstroms ( 423,4 nanometers) with 0.1 Angstrom zero inclination width.
            At 80 degrees the width was about 10x greater. The depth or light drop decreased accordingly.

    • Mike Serfas November 25, 2021, 15:00

      https://arxiv.org/pdf/2110.02561.pdf – This paper describes some methods, and makes a refinement taking into account that a star does not rotate as a rigid object. Apparently more than 20% of giant planets are more than 30 degrees out of alignment with the star’s rotation? The paper seems nicely written and approachable, but there’s a lot to think about there.

      • jonW November 25, 2021, 17:54

        Thanks Mike this was super interesting! Agreed both about the accessibility and the food for thought.

  • kzb November 24, 2021, 12:27

    jonW: it is stated that Tau Ceti is seen pole-on from here, and there have not been any transits observed. So they must have a way of determining the orientation independently of planet transits.
    I guess Doppler shift, i.e there isn’t any for a pole-on star?

    • Paul Gilster November 24, 2021, 13:17

      I believe spectroscopic data as used in radial velocity analysis is the primary method, RV dropping to zero when the star is seen pole-on, as kzb suggests. Other methods seem to be considered in the literature, so if anyone has any further insights, please chime in.

      • Antonio November 24, 2021, 15:46

        Wikipedia also points to polarimetry.

        • Paul Gilster November 24, 2021, 20:15

          Thanks for that, Antonio.

  • wdk November 24, 2021, 15:08

    Judging from this history of Proxima Centauri planet searches
    via astrometric methods, participation did not necessarily mean paydirt, but likely a long career, judging by the passing age of the astronomers.

    Just to make sure I knew why or whether Proxima b was still considered a find, I noted that its origins are based on radial or doppler shifts in Proxima’s spectrum rather than transit or astrometric observation.
    And I’m sharing in case anyone else had the same concern.

    “Doppler” or radial velocity measurements, consequently, would place a minimum mass value on Proxima b, based on an orbital plane with zero deflection out of the line of sight. And considering how small a cross section a red dwarf is(e.g., 0.15 solar for Proxima), odds of alignment within transit observation are rather low too. If the odds are about 100 to 1 based on alignment, then it appears that we live in a 10 to 20 parsec radius region rich with planets.

    For doppler/radial velocity detection to work, the planet should at least have a low inclination angle for its orbital plane. By 45 degrees the mass estimate would go up by 1.4 from the minimum.

    o what I am rolling toward with this preamble is the question of when we observe Proxima amidst all this flare activity, are we observing Proxima from overhead of its own rotation axis or that of its magnetic pole? Or are we observing a phenomenon that is popping out of Proxima from all different latitudes more akin in distribution to terrestrial thunderstorms? There is a habitability and even a weathering consideration here for an orbiting planet.

    Where the axis is pointed from an observational point of view, I am not clear on this, but the period of rotation is about 85 days or about 1/3rd the rate for the sun.

  • FrankH November 24, 2021, 15:40

    I’m not sure it would be applicable to most stars, but some of the techniques in this old article could be used to extract the star’s inclination (but no guarantee that the planetary orbits will match). The citations list more recent work:

    Vogt, Steven S. ; Penrod, G. Donald ; Hatzes, Artie P.
    “Doppler images of rotating stars using maximum entropy image reconstruction”
    Astrophysical Journal v.321, p.496
    October, 1987

    https://ui.adsabs.harvard.edu/abs/1987ApJ…321..496V/abstract

Leave a Comment