It is a beautiful, sunny, but cool day in the little village of Servoz in the French Alps: surrounded by breathtaking snow-capped mountains – among them the legendary Mont Blanc – I am sitting on a tiny railway station waiting for the little red mountain train that will carry me out of the valley. With still an hour to go before the train I am hoping that getting out will be easier than getting in: this was the first time a lecture I was about to give had to be postponed due to an avalanche.
I had spent five amazing days here, three of which at the Les Houches School for Physics, where Nicolas Iro, Francois Forget, and Valerio Lucarini organized an outstanding school on planetary circulation. This was definitely a school to remember – great lectures, lots of discussions, great food and skiing opportunities allowed by the long mid-day breaks, and fun evening discussions, all set in a picturesque Alpine setting. This was also an important meeting for me personally because we announced here the publication of our new study which, I believe, is an important step in exoplanet characterization with the Hubble Space Telescope. It may even be important for Hubble’s successor, the James Webb Space Telescope. Our study offers a solution to the infamous “ramp effect”. This vaguely understood effect has been plaguing all Hubble Space Telescope observations — among them some of the most beautiful data on exoplanet atmospheres — ever since the instrument was installed in the memorable 2009 Hubble Servicing Mission 4.
Hubble was never designed to do exoplanets: it really was built to be a cosmology machine, mankind’s most advanced telescope (for science) peering into the depths of the universe and — due to the fact that the speed of light is finite — also to the depth of time. Its sharp vision has revealed galaxies back to about one billion years after the Big Bang (or about 12.5 billion years ago). Yet, somewhat paradoxically, that sharp vision is not enough to explore planets even around the closest stars. Planets orbit their bright host stars so closely that even Hubble’s resolution and image quality are not enough to distinguish the light of the planets from the light of their host stars. But that is not the end of the story for Hubble: as a testament to human ingenuity, different teams of astronomers realized that for some systems observing brightness changes in time can help separate the light of the planet from that of the star. Soon after different techniques popped up that used this idea: from changes in the star’s brightness we are now able to deduce when a planet passes in front of its host star (planetary transit) and from precise measurements of the level of dimming in different colors we can start figuring out what the atmosphere is made of (if it contains water vapor, for example, the planet will appear larger at the wavelengths where light can be absorbed by water than at wavelengths where it can’t). For other systems, observations showed that very close in and very hot Jupiter-like planets – as they orbit their host stars – will add varying levels of extra light to the pixel where the star’s light is collected. With a very hot dayside, these planets are very bright and depending on how much of their bright dayside we see (in different parts of their orbits) the extra light will vary. This allowed astronomers to create the first crude maps of hot jupiters, a spectacular result from the last decade. This is now an unexpected but booming field of discovery and results from planets received visibility and generated excitement on par with the exciting cosmology results. And this was just a beginning – exoplanet programs more ambitious than ever are being executed on Hubble.
But there was one problem that continued to annoy astronomers, continued to limit HST’s accuracy for these measurements. It led to both some unreliable results and to forcing astronomers to discard large amounts of precious Hubble time. The problem became known as “the hook” or the “ramp effect”. Although HST exoplanet programs rely on extracting tiny changes in time, Hubble – simply put – will see all stars change regardless of whether there are planets around them. Measuring a typical star’s brightness with Hubble – which should yield a precisely flat line – will instead result in a funny hook-shaped light curve: a shape that will be different for each star and even for each Hubble orbit observations of the same star! This effect is typically 1.5% – it is small for most other studies with Hubble, but huge for exoplanets where we are after much tinier effects. How can we then use Hubble to map clouds on other planets if even stars appear to change?
The Les Houches school had many participants who have been developing impressive models for how exotic exoplanet atmospheres should behave – and compared their model predictions to mostly data from Hubble, most of which was affected by – to different levels – by the infamous ramps. Therefore, in my lecture I decided to highlight both the problem and the solution we found for it.
To explain how this works, let me tell you more about tourists in the Alps. After the school ended on Friday I stayed for the weekend in Les Houches and, following Alain Leger’s advice, I used Saturday to go up to one of the highest peaks – Aiguille du Midi (3,842m), which offered incredible views of the Chamonix valley. I took a thrilling cable car ride up to the top – a whopping 2,807-meter ascent in just about forty minutes!
On top of the peak is a crazy tower – it seems small from below (see the photo with the moonrise that I took from the village of Les Houches) but standing on top of it is a majestic and humbling experience. At the top of the tower is a viewing terrace with one of the most beautiful and panoramic views I have ever seen. Standing on the terrace on top of a ten-story-high needle-like tower carefully balancing on a 100 meter-high cliff, buffeted by strong, icy winds, and blinded by the bright, untamed sun of the high altitudes, I can all but wonder about the raw power of nature. Gigantic mountains and majestic peaks all around – among them Mont Blanc (4,810m), Dome du Gouter (4,304m), Mont Maudit, Aiguille de Verte (4,121m). I went up on a good day but the low-level of oxygen (only 45% of that at sea level), the high wind, the sun, and the brightness of the snow offered a glimpse of what it may be like to try to climb one of these stairs (although the only climb I did was the stairs to cafeteria, one of the highest in Europe).
Dozens of skiers traveled with me on the cable car from Chamonix to the top of the mountain: most equipped not only with skis but with ice axes, ropes, and climbing harness. I suspected that their goal was not the cafeteria but to ski down from the top, which must be an incredible (and seemingly dangerous) experience.
From the peak I could watch these skiers; holding on to chains and inching on top of an impossibly steep cliff to reach the slightly gentler slope that does not end in a thousand meter free fall. Then they started, one by one, their descent – following a gradient in gravitational potential energy.
Interestingly and unknown to the skiers, they follow a similar pattern – for a very similar reason — to what we have seen in the HST data.
Imagine now that you want to figure out how many people are brought up to the mountain top by the cable cars by counting the skiers that arrive back to Chamonix. You could determine that, say 50%, of the people on average will ski (while the rest of us enjoy the view and a glass of French wine); so you would know that for every person you count at the bottom there was – on average – another one that went up. That is easy. However, if you were to count the people arriving to the bottom via ski – right after the arrival of the first cable car – you would see first only a few arriving, then more and more, until the number of skiers arriving each minute reaches an approximately constant value. If you plot the number of skiers arriving in 5-minute chunks of time, you may get a curve that is similar to what HST measures when it observes a star of constant brightness. You may wonder why do you see fewer skiers first, then more skier a bit later, even though the cable cars run precisely on schedule and they are always packed to full capacity.
The solution for both skiers and for HST has to do with what happens to them after departure and before arrival. Experienced skiers can ski all the way down; they start one by one and arrive roughly the same amount of time later than they started. But some skiers – perhaps the less experienced or less lucky – will fall, often hitting obstacles covered in snow. Recovering from a fall could be easy or – if the skier hit a bad obstacle – could take a longer time. With many skiers going down, some of those hidden obstacles will be visible as skiers will be trying to stand up and recover from the fall: as long as a skier is stuck at an obstacle, other skiers will easily see and avoid those.
If you start observing the arrival of the first skiers in the morning, those that arrive first are the ones who did not fall – then you start seeing skiers arrived who started early, but fell once. Just by counting the skiers’ arrival rate you may think that cable car is not bringing up skier at a constant rate – but if you look carefully and you understand the pattern, you can figure out how many obstacles are there and what is the chance that a skier hits an obstacle.
In our paper – led by University of Arizona graduate student Yifan Zhou – we proposed a model in which HST’s detector is like a slope with hidden obstacles (traps). When the detector is illuminated — say, by a bright star — electric charges will be freed that will travel in the detector (helped by an electrical potential difference) until they are detected. However, if a charge hits an obstacle, which are most likely imperfections in the detector, they can get trapped and it will take time until they can find their way out of the traps, leading to their delayed arrival. Yifan has done a great job in translating this idea into a set of equations and then went on to show that this works perfectly for many different HST datasets! It is an exciting result that was accepted to the Astronomical Journal (https://adsabs.harvard.edu/abs/2017arXiv170301301Z) and which we are already using to revisit some of the most interesting HST exoplanet observations.
It seems to work so well that, who knows, one day we may even use it to figure out how many people ski down from the Aiguille du Midi peak.
Eventually, the little red train did come and is taking me to the next adventure – the UK Exoplanet Community meeting. Scotland, here I come!
