You continue to bring extremely exciting planet candidates to our attention! When we find promising targets there are many steps that we have to go through in order to determine their planetary nature – sometimes the planets pass all our tests, and sometimes they turn out to be false positives. Read on for an awesome summary of how we investigated the possible planet around τ Ceti written by Benjamin Pope from New York University.
But first, none of this would be possible without all your amazing help, your classifications and your Talk discussions. I would like to say a special thanks to the 15 volunteers who classified this target: Anchel, LarryW, JobiMine, EEZuidema, lvwarren, ElisabethB, TaxiCab1729, LAIS_IONUT_ANTONEL, Jose-Adao, DanielRA37, baconsteven, bugge, gulpfumetti, adam211 and Vidar87, and to andrey373 who brought this candidate up on Talk. We couldn’t have done it without you all!
Notes on τ Ceti
by Benjamin Pope
Outside of our solar system, the nearest solar-like stars are only a few light-years away: the two bright components of the binary system α Centauri AB (orbited by a third component Proxima Centauri, a dim red dwarf). But to find a star like the sun with no companion, you have to look a little further away to τ Ceti (tau Ceti), a G8 dwarf (which means it is a little less massive and cooler than the G2 Sun) which is the second-closest star system visible to the naked eye at a distance of only about 12 light-years.
Such a close system is one of the first targets for bold proposals for interstellar travel and contact, and for science fiction: closest to my heart, in Ursula Le Guin’s The Dispossessed, τ Ceti is home to the twin habitable worlds of Urras and Anarres; the former capitalist, the latter the home for anarchist exiles. But to astronomers since then it has become increasingly exciting as a host for real exoplanets: from radial velocity observations (measuring the red and blue shift in the star’s spectrum as it is tugged back and forth by planets) it has been suspected since 2012 that it hosts a number of exoplanets, with orbital periods of a few weeks to a few years. If one of these transits, it would be a huge discovery – both by independently confirming the existence of these planets, but also because it would open up an unprecedented opportunity to study their atmospheres as they are illuminated by the star behind during transit. τ Ceti is ten times brighter than the next-brightest transiting planet host star and the extra light would be a significant boon to photon-starved spectrographs trying to detect chemistry (and life!) in its atmosphere.
When the data from the Transiting Exoplanet Survey Satellite (TESS) covering τ Ceti came out two weeks ago, I received an email from Prof. Suzanne Aigrain at Oxford, my former DPhil supervisor, saying that the Planet Hunters team had noticed evidence of a transit in the light curve of τ Ceti and asking if I could check this – without knowing when the transit they found occurred, so that I had to replicate the result blind! One of the difficulties is that τ Ceti is very bright, a third magnitude star in a telescope that saturates (overexposes – just like in other cameras) on stars three magnitudes (fifteen times) fainter. In my DPhil, I had worked with Dr Tim White (ANU) to develop the method of ‘halo photometry’ (the code halophot) to deal with this problem for similar data obtained by the previous mission Kepler, which we used to look at the Seven Sisters and the planet-hosting red giant Aldebaran. The way it works is by discarding the unusable ‘saturated’ pixels but looking at the broader distribution of light (the ‘halo’) around the star and teasing out a signal from these many good pixels. So I used this code to look at τ Ceti (and if you want to see how it’s used and the plots below were made, check out the Jupyter notebooks on the GitHub repo!). Running halophot, it produces a huge signal that looks just like a transit (light curve on the left – standard ‘PDC’ pipeline in orange, new halo light curve in blue, halo map on the right):
When I told Suzanne, she confirmed this was exactly when they thought the transit was. So we were on: time to check if it’s real! Suzanne’s postdoc Dr Oscar Barragán modelled the transit signal in the standard PDC data, assuming the transit was equatorial, the planet was in a circular orbit and using the stellar mass and radius to try to estimate a range of valid periods. The depth of the transit signal gives you a planet to star ratio of 0.0108835, translating to a planet radius of 0.94 Earth radii. So this would be by far the closest Earth-sized planet to be known to transit. Meanwhile from the transit duration of ~ 11 hours we get that the minimum orbital period should be ~ 230 days, corresponding to an orbital radius of ~0.7 AU (though somewhat worryingly, not to the known periods of any of the planets found by radial velocity – though it could just be because it has too low a mass to detect). At this distance from the star the planet’s equilibrium temperature would be ~225 K. At nearly fifty degrees C below freezing this is quite cold, but there is a lot of uncertainty about the effects of planetary atmospheres, and to me this is quite exciting – no cool transiting Earths are known around such a nearby star!
Oscar produced this great visualization of his model:
So with such high stakes we had to be very careful. In comparison to the PDC data, the transit in the halo light curve I made was much higher signal-to-noise, but seemed much deeper (a few percent rather than a fraction of a percent). This isn’t necessarily a killer, in that neither the absolute normalization of PDC nor halo light curves of saturated stars is completely accurate, but they are usually much closer than this. First thing to check: halophot doesn’t do anything obviously wrong, and generates a model (on the right of the figure above) that looks rather like the expected pattern of light from a star as seen by TESS. The light curve you get from this has a deep and clean transit, which is maybe a bit long and deep, but looks ok.
What is immediately suspicious, though, is that it occurs just before perigee: TESS has an eccentric orbit in a 2:1 resonance with the moon, which means that twice a month it approaches very close to the Earth briefly (perigee) and then swings back out again to spend most of its orbit far away. When it is at perigee it is subject to a lot of reflected light from the Earth – Earthshine! This is why there is a gap right in the middle of the light curve. So to me it immediately raised alarm bells that this signal happened just when the telescope was most vulnerable to contamination from background light.
To figure out what is going on with the Earthshine, we produce a ‘background’ light curve for τ Ceti using only pixels far away from where the star is contributing much light. Let’s plot this with a vertical line to note the midpoint of the ‘transit’ we found earlier:
Uh-oh! There is a ‘transit’ signal in the background light, a little later than the transit. This isn’t an absolute killer – τ Ceti is very bright, and it isn’t implausible that its light could have directly contaminated the background or done so via some electronic chip effect (‘cross-talk’). It is also not quite at the same time as our putative planet. But it is pretty alarming.
Let’s look at some less highly processed data. What we have been looking at so far has been a ‘target pixel file’ (TPF) produced with a frame every 2 minutes and a lot of sophisticated calibration. Insted let’s use the TESSCut tool to grab part of the Full Frame Image (FFI) data, which has 30 minute frames and a lot less processing but of a much wider field. If you use lightkurve.interact() and look at the individual pixel time series in the FFI, they all show this dip. But in the pipeline TPF this is different: above the mid-axis of the star, they go up during ‘transit’ and below they go down. You can actually see this at a global level if you use the slider and the right scalings: it seems that as a whole the background flux shifts upward on the detector for a few hours and then shifts back down. So something funky has happened to the spatially-varying background during processing.
So let’s look at another very bright star in the field: the giant star ζ Ceti. It shows the same background dip – but as Tim White pointed out, at a slightly different time! If you look at individual pixel time series from ‘above’ and ‘below’ the star midlines, above the line they go up and below they go down – features with the same midpoint and similar duration to the ‘signal’ at τ Ceti – they are something to do with TESS and not a planet around τ Ceti. The difference between the top and the bottom, Tim realized, can be ascribed to the fact that these pixel cutouts are very elongated in that axis, so if we have a spatially-varying background but subtract only a constant background, we will find this asymmetric pattern. This poor background subtraction may therefore have contaminated all the pixels and created the appearance of a transit where in reality there is none.
So to look at the spatial detail in the background, I downloaded all TPFs on the same camera as τ Ceti, extracted their background light curves, and made a video of their background flux over time. Each point below is coloured by the logarithm of the background flux, clipped at the top and bottom to bring out the features best. τ Ceti is a blue star bang in the middle and ζ Ceti orange to the top right, and the ‘transit’ occurs at day 1394.3 or about 11 seconds into the video.
As you can probably just make out, there is a lot of spatial structure there, mostly in the lower left. Just around the transit, there is a spur through the middle towards the top right that lights up a little, and then it switches back to the lower left, and then everything gets brighter overall towards perigee. When we contacted Dr Chelsea Huang (MIT) about this, she was able to dig into the huge full frame images and make a ‘difference image’, subtracting one from the next to look for where the background might be changing. In her image below, τ Ceti is highlighted with a red arrow:
The fuzzy blobs pick out bright stars (e.g. τ Ceti itself, or ζ Ceti top right). The vertical streaks are probably ‘straps’ on the back of the detector that reflect back some of the light that passes through. There is probably also some CCD smear like you get with cheap cameras at night, and it runs up and down every column with a sufficiently bright star in it (such as τ Ceti). But more importantly are the ripply concentric rings which are lens flare from the Earth just out of shot, and you can see τ Ceti lies smack bang in the middle of one of these rings. As the Earth seems to move and get brighter this ring runs over τ Ceti and causes this apparent transit effect.
Regrettably we have been ‘dispossessed’ of this potentially very exciting planet candidate. But this is how science is: something that is too good to be true very often turns out that way, but it doesn’t make it any less worth investigating. In digging through the data on τ Ceti we were forced to fix bugs in our code and grapple with unfamiliar systematics in TESS that we didn’t see in Kepler. Personally, I am concerned that systematics which affect so many pixels in common over such a wide field are going to pose a serious problem to the approach we had been taking to bright stars in K2, and playing with the τ Ceti data has been a valuable learning experience. While we haven’t yet detected her home planet’s transits, we should keep in mind the maxim of Le Guin’s Anarresti scientist Takver:
There was process: process was all. You could go in a promising direction or you could go wrong, but you did not set out with the expectation of ever stopping anywhere.
Kepler and TESS are both amazing space telescopes that have and will revolutionise our understanding of exoplanets. But let’s have a look at how these two telescopes differ?
Kepler was launched in March 2009 and used a 1.4-m primary mirror that observed a 12×12 degree patch of sky (for reference the Moon covers half a degree on the sky). The sensitivity of Kepler was significantly better than that of any other instruments at the time, thus enabling Kepler to find exoplanets as small as half the size of the Earth.
Conversely, TESS will survey the entire sky, looking at 400 times more stars than Kepler did throughout its lifetime. TESS will do this with four identical telescopes, which, combined, observe a 24-degree patch of sky at any one point. Each 27 days, TESS changes direction and looks for planets around a different set of stars in a new ‘sector’. The entire sky has been split into 26 overlapping sectors, and TESS will visit each one over the course of the next 2 years.
The two satellites also differ in their observing strategy and the types of stars that they focus on. Whilst Kepler observed one patch of sky for a long period of time, TESS will only spend a month looking at each sector. The long exposure times of Kepler allowed it to find the dimmer and more distant stars, whereas TESS will monitor the nearby, and brightest targets. In fact, the stars observed by TESS are 10 times closer and 100 times brighter than the Kepler target stars! Observing brighter and closer stars has the advantage that any planet candidates that we find will be easier to observe using ground based telescopes.
The main Kepler mission ended in 2013, when the telescope lost its ability to change orientation without the use of fuel. Luckily, engineers and astronomers quickly realised that the pressure from the Sun could be used to steer the telescope in order to keep it pointing at one patch of the sky. This new era of observations became known as the K2 mission.
K2 ran out of fuel in mid 2018, bringing the mission to a close. Luckily, by this point NASA’s new satellite TESS had already been launched. We now have brightness measurements of around 45,000 stars from the first three sectors, and we are already finding some promising planet candidates within the TESS data!
Will you help us find the planets hidden within the TESS data? Click here to give it a go!
Planet Hunters TESS is back with brand new data! The Sector 3 lightcurves have just been released and we are ready to find the planets hidden within them. This new data set consists of brightness measurements of 16 thousand bright stars that were observed by TESS between 22 September to 17 October, 2018.
You may notice some difference between this data release and the last one. During the Sector 3 observations TESS underwent some test in order to improve the data quality. This meant that the data collected during the first four and the last three days of the scheduled observations are not usable, leaving us with around 21 days, as opposed to the usual 28 days, of data. These tests are necessary as they give the amazing TESS system engineers and scientist the opportunity to learn more about how the satellite operates, allowing them to advance the system and improve the pointing stability. Due to this the TESS data will improve with every new sector. Each 21 day lightcurve has been split into three sections, providing you with higher resolution data and making it easier to spot even the smallest dips in brightness.
There are sure to be many planets hidden within this data, ranging from Earth-sized rocky planets to Jupiter-like gas giants orbiting around various different types of stars. What kind of planet will you find?
We hope you enjoy the new TESS data. Happy Planet Hunting!
Think you’ve found a great transit candidate? Can’t wait for us researchers to look into it? Here are a few things that you can do yourself to check whether your candidate could be a real planet. These are the first steps that we would do ourselves, so it’s a great help to us if you have the time or inclination to make a start yourself – and a great opportunity to learn a few cool things in the process. Note you can do as many or as few of the steps on this list as you like – it’s completely up to you!
1. Is it a TOI (Tess Object of Interest)?
TOI is the name used by the TESS team for good planet candidates that they have checked carefully and consider worthy of follow-up observations.
In order to check whether the candidate is a TOI you need to find the TIC number (you can view it by clicking the “i” icon below the subject image in Talk) and check if it appears on the TESS data alerts page: https://archive.stsci.edu/prepds/tess-data-alerts. TIC ID is the first column in the big table. If the candidate is on the TOI list, well done – you have found a candidate that the TESS team have identified as a planet candidate.
If the candidate you found is a TOI you’re doing really well. However, it’s already being looked into by the TESS team, so we won’t duplicate their efforts – we want to focus on objects that they haven’t already found. Before you leave the talk page for that subject though, please tell everyone else what you’ve found – you can say “This is Tess Object of Interest (TOI) XXX” where XXX is the number that appears in the 2nd column on the data alerts table.
2. Is it a TCE (Threshold Crossing Event)?
All of the TESS data are passed through the TESS transit search pipeline, which automatically flags any lightcurves that might contain a planet. TCEs are the raw flagged candidates of this pipeline (prior to any vetting done by the TESS team).
In order to check whether a candidate is a TCE you can download a CSV file, for each sector, where they are all listed:
Alternatively you can check if a given candidate is a TCE using EXOMAST (https://exo.mast.stsci.edu/). On EXOMAST, simply enter “TIC ” followed by the TIC number, and click ‘search’. If the candidate you are looking into is a TCE, you will be taken to a page containing some information about the host star and the potential planetary system.
If the candidate is not a TCE, you will see a notification below the search bar stating “No planet found”.
If you find a TCE, once again, you’re doing really well – it means that you’re as good at finding (some) transits as the pipeline that professional astronomers developed over a number of years!
Please flag such an object as a #TCE on the talk page (if possible including a link to the EXOMAST page for that TCE).
3. It’s a TCE but not TOI?
A candidate that is a TCE but not a TOI is an object that the TESS pipeline flagged, but the TESS team decided wasn’t a good enough planet candidate to be promoted to TOI status. Finding these is really great, not least because – in some cases – we might take a different view to the TESS team and consider them to be likely planet candidates. So if you find a TCE that isn’t a TOI, please let us know by including “@researchers” in your comment on talk. We will get notified automatically and – time permitting – we will look at it more closely.
When vetting the TCEs, the TESS team perform a long list of checks. These tests are designed to weed out instrumental false positive (the signal isn’t real) and astrophysical false positives (the signal is real but isn’t caused by a planet, but something else). The results of these tests are saved in a DV (data validation) report, which they have helpfully made publicly available – so we can use them to understand why the TCE didn’t become a TOI. This is a really quick way to look through candidates and to avoid repeating the hard work that the TESS team have already done. The DV reports are long and complex, and currently a little tricky to access for TCEs that aren’t TOIs, so we are not including instructions on downloading and using DV reports in this post (though we hope to do so at a later date).
Importantly, there are already a few TCE (and not TOI) candidates found by planethunters.org volunteers for which we have examined the DV reports and come to the conclusion that the candidates are promising. This mainly happens because the TESS pipeline requires at least two transits for a detection, so it only searches for transits that repeat with periods up to the duration of a TESS sector (~28 days). If there is only one real transit, it might be missed altogether (this is where you volunteers come in!) or it might be wrongly paired up with an artefact or noise feature somewhere else in the light curve. In that case, the diagnostics in the DV report, which are based on all the transits combined, might be misleading.
4. Create a cutout or movie of the TESS data
There is a fun tool at https://mast.stsci.edu/tesscut/ which allows you to extract a time-series of cutout images around a given target. You can use these to look at what is in the vicinity of the target, or even to make a movie! If the transit is deep enough, you might even see the star “blink” (this can be a fun thing to try out on variable stars or eclipsing binaries too).
Sometimes, what appears to look like a transit is actually due to some weird artefacts, affectionately dubbed “fireflies” or “fireworks” by the TESS team, that sweep through the field of view. These are probably due to scattered light from bright stars or moving objects inside the telescope and camera optics. If you notice that a promising candidate is actually due to such an artefact, please let everyone know on talk!
5. Want to play with the TESS data products yourself?
If you’re really keen and want to examine the TESS data in more detail, you can easily get your hands on them. Go to https://archive.stsci.edu/, enter “TIC” followed by the TIC number of the subject in the search box, and hit “search“. This should bring up a list of datasets stored by MAST (Mikulski Archive for Space Telescopes), including two that will have “TESS” in the “Project” column. The lightcurve is the one that lists the TIC number (rather than “TESS FFI”) under “Target name”.
You can download the data to your computer by clicking on the little floppy disk icon in the corresponding row. You can find more information on the format of these datasets in the TESS Science Data Products Description Document:
What to do with the data when you have it is a long story, far too long for this post… but again, we hope to provide a separate, dedicated article with some examples at a later date.
Thank you so much for all your amazing work! The next data release is just around the corner so hopefully everyone is ready to find some more planets. Until then, we have some preliminary results from the sector one data.
Over the past month the science team has been working hard on putting together a list of some of the most promising planet candidates. We find these by carefully looking at the lightcurves where many of you marked planets in the same location. With a careful eye we filter out lightcurves that show eclipsing binaries or that have transit-like events due to systematic effects. We can identify these by looking at features such as the shape and depth of the dips, as well as the time of the transit.
The candidates that withstand this initial filtering process have to go through a further screening before they can be promoted to be a high priority planet candidate. This screening involves looking at the variability of nearby stars, the depths of the alternating transits (if the lightcurve shows multiple transits), and stellar parameter of the host star.
So far, we have identified five high priority candidates, three of which are TCEs (you can see their lightcurves below). Even though these lightcurves have passed all of our tests up to this point, we cannot confirm that these transits are due to planets without further observations. As a next step we will, therefore, look to observe these targets with ground based telescopes in order to find out more about these fascinating systems.
We are very excited about these initial five candidates and look forward to finding many more as we finish looking through the sector one data. Stay tuned for more results!