On Exoplanets and Elements: A Chat with Dr Natalie Hinkel
We at Nakshatra recently had the opportunity to get in touch with Dr Natalie Hinkel, Senior Research Scientist at the Southwest Research Institute, San Antonio and Co-Investigator for the Nexus for Exoplanet System Science (NExSS) research network at Arizona State University. Specialising in the study of elemental compositions of Sun-like stars and how they affect the planets formed around them, she was a treasure trove for information regarding planetary physics. We hope you have as enlightening an experience as we did, with her responses to our many questions.
We are honored to get this opportunity! Seeing you enthusiastically guiding budding scientists, and by how well you answer their burning questions, we were excited to ask a few ourselves!
Q: How did you make the AI understand the spectrometer data and teach it to analyze the specific things it looked for? Were there any uncanny things the AI did, that you found impressive or puzzled you of why it is doing that?
A: This is a great question because it cuts to the heart of the data. We did not have the algorithm (not quite an AI yet, unfortunately) look at data from the spectrometer directly. Instead it looked at data that had already been preprocessed or “reduced.” Namely, we had already converted the bumps, dips, and wiggles within the spectra to be the total amount (or abundance) of each particular element. By giving the algorithm data we had already processed, i.e. data from the Hypatia Catalog (www.hypatiacatalog.com), we knew in advance that the element abundances were correct, we could look a large variety of different elements, and we didn’t have to have the algorithm processes the data each time it looked at the data (which was hundreds of thousands of times!).
When we first started working with the algorithm, we wanted to see how far we could push it in one direction or another. So sometimes we gave it only a few hundred stars, sometimes we gave it thousands, sometimes we gave it some “weird” elements, etc. That was kind of fun to test the limits and see what came out at the other end.
Overall, the biggest puzzle (that we have yet to solve) is that the algorithm found that sodium (Na) is in important indicator for the presence of a giant planet. While Na is important on the Earth (for example, salt), we weren’t sure why it would be important in a big gaseous planet like Jupiter. So, that’s something I’ve been trying to ask others scientists to see if they might understand it.
Q: Why is it that the presence of only certain elements within a star system strongly impact how the planets are created and changed? Why do other elements play a lesser role?
A: Fundamentally, the answer to this question depends on how common are certain elements in the universe (https://en.wikipedia.org/wiki/Abundance_of_the_chemical_elements ) and how do they interact with each other chemically. When you combine a lot of iron, oxygen, silicon, and magnesium in the universe (in our Solar System, in exoplanetary systems), with the fact that they like to bond with one another, then you get the building blocks of planets. Certain elements aren’t as likely to bond with these major elements, meaning that they aren’t as important or influential in creating planets.
Q: Is there an understandable reasoning behind the trends (“high Fe/H ratio implies giant exoplanets”) observed?
A: Right now, this result is mostly empirical — meaning that we’ve observed it over and over again. When this happens, we try to make sure that our theories match the observations. So we currently think that, if a star has a lot of iron, than this means that there is a lot of heavy material — so much so that the material left over from star formation is enough to build a giant planet.
Q: The validation technique you’ve incorporated attempts to use ‘Golden Sets’ on the Hypatia Catalog. The ‘Golden Set’ number to skip stars in each iteration is 10, as used in the implementation. In what way does that value affects the validation of the algorithm?
A: This is a very good question. There is a total number of 290 stars in the training set (i.e. stars with known exoplanets). When it comes to the sort of statistics implemented in the algorithm, this is not a large training sample. Therefore, during each iteration, I didn’t want to skip too many stars…because we needed them! After a bit of testing, I determined that 10 was a good number to skip that wouldn’t strongly influence the training sample. However, 10 is not a huge number to help me determine how well the algorithm is able to predict that these known planet hosts are, in fact, planet hosting stars! Hah. So, I had the algorithm run hundreds of times, choosing a new set of 10 each time, so I could get a bigger sample.
Q: Do you think the false positives are that big of a problem? What would you do to improve this scouting process?
A: Honestly, it’s really hard to say how big of a problem false positives might be. Of course, I can run the statistics and trust the algorithm, but that doesn’t mean much compared to actually going out and observing those stars with a very high planet prediction probability. So, that’s exactly what I’m planning on doing, for example using the Discovery Channel Telescope in Arizona or the McDonald Telescope in Texas.
Q: What is the magnitude of change that your work brings? How different is it from what a layperson might think about your research?
A: For other scientists, my work means shifting the way that we look for planets. Most of the techniques used to find exoplanets are based on the physical properties of the planets (compared to the star): for example the planet transiting the host star or the planet’s mass causing a wobble in the star. However, the technique I’ve introduced in this paper means looking at the chemistry, trying to see how the planet influenced the elements within the star. That’s a big change, especially since the chemistry of the planet plays a big part in making a planet habitable…or not (for example, Venus is very similar to the Earth physically, but the chemical variations are huge!).
As far as what a layperson might think — well, that’s hard to know. I would hope that the average person would know about how we find planets, especially since I think the transiting method is fairly applicable after a number of recent total solar eclipses. So, if they understand transits, than I would hope that finding planets via the chemistry of the star alone would be really cool. I mean, I think it’s really exciting! Hah.
Q: Is there any special thing that you want people to understand about your work, that nobody gets/asks?
A: Aww, thanks for asking!
I think there is a kind of weird perception that scientists are hyper-focused geeks who don’t do anything else other than study science. But when it comes down to it, we are really working on the edge of human understanding.
There are a lot of things that we are exploring for the first time; we are asking questions that we might not fully understand. As a result, there is a lot of trial and error, there are a lot of failures before we reach the successes. So what people end up seeing is really the polished tip of all of our work: the ideas that we thought of, tried, experimented with, and eventually finalized. All of this means is that, at the end of the day, we are usually very invested and emotionally involved with the projects that we work on — we’ve spent many months (if not years!) trying to make it work! It also means that we don’t always have the answers to everything, because we are still learning how everything is connected — how stars and planets interact with each other, what it means for a planet to be habitable. And while sometimes that’s frustrating, it’s also really exciting. :)
For more info on Dr Hinkel’s work, outreach initiatives, podcast, TEDx talks and much more, do visit her website, www.nataliehinkel.com!
Dr Hinkel and team’s work was previously covered by Nakshatra in ‘Astronomy Rewind: June’, here.
We at Nakshatra gleefully thank Dr Natalie Hinkel for her patience and enthusiasm in giving the interview, and wish her spirited team success in all future research adventures. Hoping to see the comet trails of your work brighten the skies of scientific history!