Monthly Archives: November 2023

Finding six synchronised sub-Neptunes orbiting HD110067

We have a big press release today on the discovery of the HD110067 system, involving more than a dozen institutions across the US and Europe. That focuses on the science and the broader picture of these new planets. But I thought I would share a more personal viewpoint of the idiosyncratic (and somewhat unique way) of exactly how we found these planets.

The story begins in lockdown 2020. The Transiting Exoplanet Survey Satellite (TESS) had launched 18 months previously and I was at MIT working with the TESS science team. This space telescope scans a large fraction of sky for a month looking for the hints of transiting exoplanets – planets whose orbits pass between us and their stars, causing a brief dimming in their starlight which we can detect. The depth of the dip tells us the size of the planet and, when we have consecutive transits, we also get the orbital period of the planet.

One of my roles at MIT was to help run the candidate vetting process. Two separate automated pipelines churn through the data from all the stars TESS has observed and spit out planet-like signals. Our job is to look through the reports and determine which ones really were planetary (to become “TESS Obejcts of Interest”) and which ones weren’t. In late April 2020, the first hints of planets orbiting HD110067 arrived in the form of the PDC vetting report, and it immediately piqued my interest given the star was extremely bright. The data also showed a handful of clear U-shaped transit signals – good signs of planets – but the automatic pipeline itself didn’t do a great job linking up the various transits. One unfortunate problem with TESS’ observations is that it typically only looks at a patch of sky for 27 days before moving on, so even planets with 15 day orbits might not show two consecutive transits in each given sector. However, ESA’s CHEOPS satellite (the other half of my two-mission fellowship) is a targeted telescope, so we tell it where to point, and in this case I immediately wondered CHEOPS could help.

The photometry from TESS in 2020 – The purple transit marks a 9-day candidate, while the yellow triangles mark our guess for a 5.4 day signal (with a ? where one is missing). The other blue dips marked with “?” have unknown origins.

But the first thing to do was to model the data and see if I could figure out exactly what was going on… There seemed to be five clear transits, but from two or more planets. and so we needed to play a game of “match the dips” – which transits are from the same planet? Two relatively shallow dips 9 days apart seemed to link together well, as did two other deeper dips 5.4 days apart (although they should have produced a third transit, I thought this might have erased by noise). The fifth and deepest dip didn’t seem to link with anything else, so I thought this might just be a single (or mono-) transit. With this solution in hand, we got TESS PI George Ricker’s approval to speed up the TOI announcement process, and immediately gave the transit periods to CHEOPS for observations which… rejected them. It turns out we were about 10 days too late and HD110067 had just slipped out of the observability region of CHEOPS… Dammit!

Two years later

In the mean time, I worked on other projects involving TESS & CHEOPS, finding a bunch of other interesting planets. But by the time spring 2022 came around, I realised that TESS was about to observe HD110067 again! I was interested to see if our friend the monotransit would show her shadow again, as these planets with two transits separated by a large gap (which I have coined “duotransits”) were exactly the types of planets my CHEOPS programme was designed to follow. Luckily the guys at MIT had managed to create a rapid release pipeline for TESS data called TICA, and thanks to some other CHEOPS projects, we had even managed to convince them to release each TESS orbit extremely quickly. This was just the thing we needed, as HD110067 would set from CHEOPS’ view only a few weeks after TESS finished observing.

So I grabbed the TESS data from the new sector, and I have to say I was shocked! There were so many transits! I thought there could be three planets in this system, but it looks like I had been very wrong… After looking a bit more, I found our friend the planet at 9.11 days – so we had at least one planet with a confirmed orbit. But the 5.4 day planet was conclusively ruled out. And there were another four transits in these 14 days of TESS data which appeared new!

The first half of raw TESS data for HD110067 in March 2022. The 9 day planet is marked in yellow, but the origin of the other four transits are unknown…

Time to play a “Pair Matching” game…

Single transits are extremely tricky to deal with – the potential periods are effectively unlimited. But if you can find a second transit, then you drastically reduce the possible periods from infinite down to a few dozen of “aliases”. This is the game I had been playing with other TESS candidates, and it was the game I wanted to play with HD110067. But it’s a lot easier when you have only 2 transits as opposed to 12… How can we figure out which transits are from the same planets? Well, the best way was to play another “pair matching” game, just like for kids. Take the shape of a transit in 2020 and match it with the shape of a transit in 2022. Of course, I did this by fitting models to the transits and comparing the resulting parameters rather than eye-balling it, but the end result is the same – three of the 2020 transits seemed to match with three of the 2022 transits, while two more didn’t seem to match at all.

Playing “match the pair” with the transits. The right figure shows the transits & best-fit models. The left figure shows duration & depth for the unidentified dips. Red clouds are transits seen in 2020 & green clouds are transits seen in 2022 (both ignoring planet c – see next). Stars represent our guess for the transit shapes of 4 planets (2 pairs, 2 singles)

So, now we had three “duotransits”, we could compute those possible periods, and start using ESA’s CHaracterising ExOplanets Satellite (CHEOPS) to hunt for the true periods. Thanks to my position within the CHEOPS team, I could start checking possible periods immediately after TESS stopped observing. We also got the second orbit of TESS data, which didn’t bring any more unexpected transits (we had enough), but solidified our knowledge of the 9d planet. A little bit later we were also able to improve our lightcurve using algorithms developed by Andrew Vanderburg, and this revealed that one of the planets we thought was a duotransit (i.e. with only two transits) actually had another transit hiding in noise, turning it into a planet with a 13.5d orbit – two planets solved, four more to go!

After many observations with CHEOPS for these possible aliases, we finally got lucky in late April and caught a third transit of planet d, with an orbit of 20.5 days! Half of the planets now had orbits – we were sure three more planets were hiding in the system with periods beyond 20 days, but for the moment we weren’t sure where they might be. However, it was at the point that I got the CHEOPS data for this new planet and solved it’s period that a cog clicked into place – the periods for planets b, c and d all appeared to increase by 50% each time – a clear hint for “resonance”.

ESA/CHEOPS transit of planet HD110067d which confirmed a 20.5 day period.

Resonance is like a gravitational dance – the subtle influence of interior and exterior planets (or, in the case of Io, Europa & Ganymede, moons) lock their orbits into tight ratios and maintain this intricate balance, sometimes over billions of years. The planets may slightly oscillate around these precise integer period ratios (2:1, 3:2, & 4:3 being the strongest), but each time their periods drift the gravitational pull of a neighbour brings them back into lock-step. We know of a few key systems like this – TRAPPIST-1 & TOI-178 for example – but even they have certain planets which break from a perfectly “first order” resonance.

I also very quickly realised that the last remaining “duotransit”, which had more than 30 possible periods, was surely at the next position in the chain. And, sure enough, one of those resonances matched almost exactly (with a precision of <0.1%) a 3:2 orbital resonance with planet d. My colleague at the Observatory of Geneva, an expert in resonances, confirmed that the inner three planets were a stable resonance, and that planet e with a 31 day period was also stable and by far the most likely.

So, with the detection planet d, the dominoes started to fall and the next planet in the chain was solved. But the next two candidates didn’t have the luxury of producing two transits in the TESS data… So how to proceed. Luckily, Adrien came up with a plan to test every possible combination of resonant orbits for the outer two planets in order to test two things. The first was whether a planet at that orbit would have been spotted by either TESS or our subsequent CHEOPS observations, and the second was how far that planet was from a resonant equilibrium – effectively planets far from equilibrium are unstable over long timescales, and therefore far less likely to persist.

After trying 2:1, 3:2, 4:3, 5:4 & 6:5 orbits for each of the two outer planets (a total of 25 cases), we found that, remarkably, if the planets were resonant then there was a single likely solution – an orbit of 43 days for planet f and an orbit of 54 days for planet g. All other possible orbits were either ruled out by our data, or appeared far from the stable resonant equilibrium. At this point, we thought it was extremely likely that we had solved the whole system. But convincing the community (and a referee) was not going to be so easy… So we needed independent evidence for the planetary orbits.

Predicted orbital period ratios for planets f & g (with respect to the inner planet), alongside whether they are ruled out by past data or by unstable orbits (typically Laplace angles>70 degrees).

Ground-based telescopes pick up the baton

By this point, unfortunately, HD110067 was no longer visible to CHEOPS. But there was a chance that we might be able to spot the outer planets in data from ground-based telescopes via two methods. The first was to look at radial velocities, which our Spanish colleagues had been gathering using the HARPS-N and Carmenes telescopes. These spectroscopic measurements can precisely measure the tiny push-and-pull caused by a planet orbiting a star, so maybe we could find planets f & g in that data. However, HD110067 is a pretty active star, and the signal from stellar activity dwarfs that of the small planets in this case. However, thanks to some state-of-the-art activity modelling from Oscar Barrangan and Andrea Bonfanti, we were able to recover the signals from three of the six planets, including planet f at 43 days. This was a clear, independent sign that there was a planet at the orbit we predicted.

Radival velocity observations and models. The upper left plots show timeseries for RVs and FWHM (which traces stellar activity), the top right plots shows phase-folded RV signals for each of the six planets, while the lower six plots show the histograms for the amplitude of the planetary signal for each of the six planets.

The second independent sign came via a global campaign of ground-based telescopes at the time of a predicted transit of planet f, which I coordinated. Thanks to colleagues with access to the MuSCAT-2 & -3 telescopes, the Saint-Ex telescope, the Tierras observatory, the NGTS telescope array of 10 telescopes, and a DDT on the LCO network. All these multiple observatories spread across the globe sent their data back to me and, although no single observatory found a clear-cut transit signal, careful analysis of the entire dataset showed a clear hint of a transit at the expected time! So it looked like the resonant orbit of planet f was a certainty.

Detrended ground-based photometry from each of the telescopes used in the global campaign.

But planet g still only had a single TESS transit, and no hint in RVs… So we were resigned to publish the paper like that: effectively with five and a half planets. Until, at a CHEOPS meeting, Rafa Luque, who we had tasked to lead the paper, showed us a plot that Josh Twicken had made, which had re-analysed some “missing” TESS data from 2020, right when our orbits predicted transits of both planets f & g… And there they were – two transits perfectly matching those which we had seen in 2022, but which had been masked due to poor-quality. This was a wonderful confirmation of Adrien’s predictions, with the two planets being once again less than 0.1% away from perfect resonance orbits.

The full TESS & CHEOPS photometry, with the additional “missing” TESS data which revealed transits of f & g in green (and highlighted in yellow)

With that additional TESS data, we had a confirmed system of six resonant transiting planets. The system is by far the brightest resonant system of transiting planets on the sky. The architecture of the system is also unique – it is only system with six planets all in first-order resonances (the strongest), and their orbits are the closest to perfect resonance of any such system. HD110067 is also now the brightest star to host a transiting multi-planet system with four or more planets, and it could even be the brightest we will ever find. The six planets are all sub-Neptunes with radii between 1.9 and 2.9 times that of the Earth. We also know they are low-density, likely with large amounts of hydrogen-helium gas (or possibly steam). This makes them ideal candidates for JWST spectroscopy, although the star is if anything too bright for JWST making many instruments saturate!

But I am most looking forward to the next few years – CHEOPS will continue to monitor the system looking for transit timing variations – the hints of gravitational tugs between planets which can help tell us the masses. I also lead a successful monitoring proposal on the star with CARMENES and HARPS-N which should independently improve our precision on the planetary masses. There is also a very real possibility that the resonance chain is not finished at six planets. And, with our current data, we do not really know what might be lurking beyond planet g on orbits of 75 or even 150 days, although we know they should transit as the system is extremely edge-on. And if that is the case then each additional planet will be cooler and closer to earth-like temperatures… Only time will tell for this brightly shining new star of exoplanetary science.

I created a detailed animation of the system, as well as annotations describing the discovery.
I also created a new mug for the office, featuring the six HD110067 planets.