The start of the school year brings about the end of an internship for twelve exceptional interns, including six Trottier fellows, that were working with us over the summer. The goals they have achieved are truly astounding. Learn more about the research projects and the experience of Mariya Krasteva, Emily Pass, Caroline Piaulet, Maryum Sayeed, Jessica Speedie, Thomas Vandal, Ben Leblanc, Juliette Goeffrion and Evelyn MacDonald in the interviews below.
Trottier intern from Concordia University, working with Prof. René Doyon and researcher Étienne Artigau at the Université de Montréal, and Claire Moutou at the Canada-France-Hawaii Telescope over the summer of 2018
The project was mostly to develop tools, but this required extensive research on the instruments, the catalogues, and the correction methods used. The most rewarding part is that these tools will be very useful and help countless researchers navigate the new data. The TESS mission’s first data release is expected for January 2019, and the SPIRou first data release for October 2018. This is a very busy time in the field of exoplanet research, and it is very important to develop a sufficient number of visualisation tools to handle all the information.
I think the biggest challenge was to design a web application without any prior knowledge about web development. Having to create the entire search algorithm itself, then create the user interface, and make the whole thing interactive was not an easy task for half an internship. I also think this internship challenged my ability to adapt very quickly to a new project, new country, and new supervisor midway through the project. I am very happy that this internship challenged my abilities, that I gained a whole new set of skills, and that I managed to finish both projects on time.
I think my favourite part was the people I met throughout the journey. The iREx interns were from all over Canada with a large range of different interests and ambitions. Working with René Doyon and Étienne Artigau gave me a very positive experience in the field of instrumentation. Visiting the CFHT and Mont-Mégantic gave me an entirely different perspective on astronomy. Being immersed in a mostly French community at the CFHT opened my eyes to the huge collaboration behind the development of instruments. The people at the CFHT were from all kinds of backgrounds and were all extremely welcoming. I loved learning about the field and my own ambitions during my stay there, not to mention the friendships I developed and the amazing time I had working with Claire Moutou!
Trottier intern from Waterloo University, working with Prof. Nicolas Cowan at McGill University over the summer of 2018
I was studying the temperatures of hot Jupiter planets using Gaussian processes, a type of machine learning technique.
It is amazing that we have the technology to study the atmospheres of planets that are hundreds of lightyears away! When the planet passes behind its star, we observe a dip in light that corresponds to the thermal emission of the planet. By observing this dip at many wavelengths, we can see the emission at different layers of the planet’s atmosphere, as the atmosphere is more opaque at some wavelengths than others. I was using this information to study the planet’s temperature, but you can use it to learn about some of the planet’s other properties too.
I developed my Gaussian process method, tested it on simulated data and found it performed very well. My most important result was determining new temperature estimates for twelve real exoplanets.
I had programmed before, but this was my first time working with machine learning. Machine learning is the future of programming, and it was really exciting getting to develop skills in such a cutting-edge, in-demand field.
People typically use Gaussian processes on data with thousands of points, but I only had three! It took a lot of time and creativity to adapt the method to work with so little data. I ended up having to “train” my Gaussian process on other data sets — for example using the spectrum of water since hot Jupiter atmospheres are water-dominated in the infrared.
I met so many amazing people during my internship, both at iREx and the McGill Space Institute. By attending the weekly paper discussions, seminars, and workshops, I really felt like I was part of the astronomy community.
Trottier intern from Université de Montréal, working with Prof. Björn Benneke at the Université de Montréal over the summer of 2018
My research advisor, Prof. Björn Benneke, developed a modelisation and retrieval framework using a Markov Chain Monte Carlo implementation, which allows us to constrain the parameters of exoplanet atmospheres from transmission (when the planet passes in front of its star), secondary eclipse (as the exoplanet passes behind its host star) and directly observed spectra (when the planet is hot and far enough from the star to be resolved separately). Until now, parametrised forms were used to describe and constrain the vertical temperature profile of the planet (its temperature at different altitudes), allowing little flexibility. My main project consisted of implementing a new way of describing this temperature profile from spectral datasets using a parametrisation-free retrieval method. This will be especially helpful when higher-resolution, higher-quality spectra will be available following the launch of the James Webb Space Telescope (JWST).
The temperature profile in the atmosphere of a planet is crucial to constrain its habitability for two main reasons. First, we need to know the temperature and pressure at the surface of the planet to determine whether or not the covalent bonds required for the formation of complex organic molecules could form. Second, the temperature of a planet at each “layer” in the atmosphere is also the starting point of the calculation of its chemical composition, which may or may not make it hospitable for the formation and development of life.
I demonstrated that, in the case of the giant planet HD 209458b, the quality of the data provided by the instrument NIRISS aboard JWST will be enough to precisely constrain its temperature profile, as well as its composition. Furthermore, for more challenging targets with a lower signal-to-noise ratio, the new method I implemented to constrain the temperature profile will enable us to retrieve more information than the methods previously used in SCARLET from the same dataset.
This summer, I learned a lot about the various observational methods used to obtain exoplanet spectra. On the modeling side, I learned about the different processes that come into play in the atmospheres that give rise to the features in the emerging observed spectra. I also had the opportunity to drastically improve my skills in object-oriented programming in Python.
The idea of this project was based on a similar retrieval method for the temperature structure presented in Line et al. (2015), which applies to spectra from well-isolated brown dwards, yielding higher-quality data than what can be obtained for exoplanets. Thus, when I tried to apply the same method as theirs on exoplanet spectra, I was confronted with a degeneracy that exists between the temperature-pressure profile and the atmospheric composition, which also existed in their case but became predominant when using lower-quality spectral datasets from extrasolar planets. This was a challenge which we overcame in the end by parametrising the “layers” in the atmosphere using optical depth rather than pressure, allowing us to break this degeneracy.
Over the summer, I really appreciated the bonds that we created and the complicity arising between the interns and with other graduate students. This made my experience even more enjoyable, as we all knew that we could get help from other people when we were struggling with our respective projects. I also loved the feeling of contributing to the understanding of things that are utterly beyond us, and I am still amazed by this aspect of the research project.
Trottier intern from UBC, working with Prof. Jason Rowe at Bishop’s University over the summer of 2018
The topic of my internship was to analyze K2 photometry of Uranus to investigate seismic activity, weather patterns, and analyse reflected light from the Sun on the planet. Since most of the exoplanets Kepler & K2 have discovered resemble Uranus or Neptune, it’s important for us to understand the behaviour and composition of these planets.
All parts of the project were interesting: learning how to extract the photometry, reduce, clean up, and analyse the data using various tools and methods. Every step was new and exciting for me, while also being useful to the project.
While the data was too noisy to accomplish all our goals, we were still able to use reflected light from the Sun on Uranus to investigate solar oscillations. Our most important result was that we were able to determine the stellar parameters of our Sun using this reflected light to a very small uncertainty.
I learned how to reduce and clean up noisy data, the importance of trying different methods to accomplish tasks, and how to write code efficiently. I also learned about the various tools we used throughout the analysis, the different aspects of astroseismology, and the composition and importance of ice giants.
My biggest challenge during the internship was to write codes that run efficiently, as well as to understand and apply the different methods used to clean up the data. Because I didn’t have previous research experience applicable to this project, it was challenging to understand the importance of the steps we were taking. However, once the data clean-up was complete, it was easy to see how our models were used.
I liked everything about my internship! I was able to use the knowledge I acquired in my courses — such as Fourier Transforms, convolution, etc. — to the project directly. It was also really fun to follow through with the project from beginning to end; I extracted the data, reduced it and analysed it which produced significant results. I also really liked the iREx community; everyone was really supportive and excited about doing research in astrophysics. Most of all, I liked doing exciting and cutting-edge research in an astronomy field that I’m really interested in!
Trottier intern from McMaster, working with Prof. David Lafrenière at the Université de Montréal over the summer of 2018
The topic was finding the mass of brown dwarfs which are stars on the cusp of being planets. The mass of some of these objects is currently not known to high precision, and it’s their mass which determines their identity. We had two ideas for ways in which we could improve the mass estimates: (1) by detecting the presence of deuterium in the emission spectrum and thereby constraining the mass to below the minimum deuterium burning mass of approximately 13 Jupiter masses, or (2) by constraining the surface gravity first and subsequently solving for the mass. Both methods require high resolution spectroscopy, and only recently has the necessary instrumentation become available for this. The Canada-France-Hawaii Telescope’s new spectrograph, SPIRou, will be able to provide the necessary high resolution emission spectra for these objects, and so we applied these two methods to simulated SPIRou observations to predict the methods’ success.
The results from the simulated observations will be used to inform SPIRou proposals to obtain real observations. The two methods that we used are creative ways of using the high resolution of SPIRou to our advantage. My work on improving the mass estimates of brown dwarfs and giant planets contributes to a field whose progress on this front has been stalled for 10-15 years and needs a breakthrough.
There are two important results. Firstly, according to the models that were available to us this summer, SPIRou would not be able to see deuterium in the atmosphere of an object at a temperature of 900K — but we would need more models to say something about objects at higher temperatures. Secondly, we found that there is a way to determine a brown dwarf’s surface gravity by comparing the small-scale features of its emission spectrum (i.e. absorption lines) to reference models, instead of the large-scale features yielded by the current mass estimates. More work is needed to confirm, but preliminary results are promising.
My time at iREx gave me a genuine taste of how research is done. I learned just how difficult research is — how you’re always encountering problems, and how things always take longer than you think they will. An important lesson that is not often considered is how to deal with a negative result, or the lack of the positive result that you were looking for. I learned to have the same respect for a plot that says “no” as for a plot that says “yes”, because regardless of what the answer is, it’s the answer itself that you’re looking for.
The biggest challenge was dealing with the unknown. Research is the process of answering questions that no one has answered before. For this reason, there is no “solutions manual” against which you can compare your answer, like there is in school. Because you don’t know a priori what the result should be, and because some procedures can be long and complicated, it’s impossible not to worry that you’ve made a mistake somewhere. I struggled to be comfortable with this, like I am comfortable with the contents of a textbook, but I think that’s how science works. A result becomes accepted as it is reproduced over time.
I really enjoyed being a part of the research community. Starting my day at the office by checking the arXiv over morning coffee — now I could get used to that! I also enjoyed the freedom which comes with doing research. You’re free to be creative and try something to see if it works. As you get more experience, your sense of judgement becomes sharper, and your ideas become more intricate as they incorporate the successful aspects of what you tried in the past.
The topic of my internship was using Gaussian processes to remove stellar noise from the radial velocities (RV) of a star named Beta Pictoris, and putting constraints on the mass of its planet, Beta Pictoris b. Beta Pictoris is a young, active star that is well known for its circumstellar disc, which is a good laboratory to study planet formation. In 2010, the direct imaging of a giant planet (Beta Pictoris b) was confirmed around Beta Pictoris. A Gaussian process is a non-parametric method that uses a covariance function (kernel function) to measure the similarity between data points and to return a predictive model. We tested several covariance functions to find which one was best suited to our problem. The Gaussian process was trained (covariance function parameters optimization) on a light curve before being used on the RV data.
Beta Pictoris is a rare case for which a planet was directly imaged around the star and for which there is RV coverage of the star. Also, most of the orbital parameters of the planet (Beta Pictoris b) are well known from observations. However, one of its fundamental parameters — the mass — still relies on indirect estimations from evolutionary models (based on luminosity). Obtaining the mass of Beta Pictoris b from RV measurements would allow us to test evolutionary models by comparing them to a model-independent measurement. This project was also an opportunity to learn more about the efficiency of Gaussian processes when working with high-activity stars.
We found out that the choice of covariance function is very important in a project like this. Some covariance functions that we tested would completely remove the planetary signal from the data, making it impossible to find the mass of the planet. We thus had to focus on the kernels that would not completely remove the signal of the planet. Even with these covariance functions, we were unable to directly constrain the mass of Beta Pictoris b. However, performing multiple tests allowed us to conclude that the planetary signal is probably in the data and that longer light curve and RV coverage might be necessary for active stars like Beta Pictoris.
I learned a lot about Gaussian processes since it is the main method I used throughout this project. I also improved my coding abilities in Python since I spent most of the summer analysing data with this programming language. I also acquired a lot of knowledge on exoplanets, especially on Doppler spectroscopy (radial velocity method).
My biggest challenge was to go through a lot of literature at the beginning of the project to correctly understand the context, the objectives, and the methods related to the project. It was especially challenging to get a solid grasp of Gaussian processes, but was helpful for the data analysis portion of the project.
What I liked the most about this internship is that it introduced me to astronomy research. I have been interested in exoplanets since high school, so I enjoyed working on this topic over the summer. I also liked the work environment at iREx and the great support of my supervisors.
Trottier intern from McGill, working with Prof. Nicolas Cowan at McGill University over the summer of 2018
This summer, I worked on something called exo-cartography. More specifically, I helped develop a tool allowed multi-rotational longitudinal mapping of exoplanet surfaces. I was interested in using this tool to look at data of Earth taken with EPIC, seeing if we could create cloud-corrected albedo maps of Earth using it, and how this could be translated to cloud-corrected maps of exoplanets. The idea behind this project is that clouds can skew albedo data coming from exoplanets, and we want to know how to correct the data for this to learn more about exoplanet surfaces.
Exo-cartography in general is really interesting because you can get a rough idea of what very distant exoplanets might look like. My project was interesting because it was looking at how you could use Earth to learn more about other planets.
I discovered that creating cloud-corrected albedo maps of Earth is not a trivial exercise! I’m still working on getting good results from my longitudinal mapping tool.
I learned a lot about coding! Specifically, I honed my Python skills and learned about creating more efficient code and using algorithms I previously didn’t know.
The biggest challenge was gaining the necessary background understanding in exo-cartography to get started on my project, as well as all the math and coding involved in it.
I loved always being surrounded by other people studying astronomy and astrophysics, and learning more by attending the events organized by iREx and MSI.
Summer intern from Bishop’s University working with Prof. Jason Rowe at Bishop’s over the summer of 2018
The topic of my internship was to analyse photometric Uranus data collected by K2. The main aspect I worked on was data reduction. Since K2 is not built to look at objects as close or bright as Uranus, there was a lot of it to do. The saturation of the image and the movement of Uranus across the detector were the largest data artifacts that had to be removed. Following data reduction, we hope to examine the weather patterns of Uranus and to recreate the Sun using the light reflected off Uranus.
The most interesting thing about my project was seeing the data get less and less noisy as the summer went on. Every time another section of data reduction was completed, the difference in the lightcurve was noticeable. It made going through all the reduction steps much more fun since we could see our work having a direct impact on the data set.
Since we are still in the data reduction stage, we do not have many final results. However, the biggest data reduction steps were successfully obtaining the size and shape of the aperture that we needed to use to calculate all the light from one image correctly, followed by finding the best method to remove the photometric variations that show up when Uranus moves across the detector.
I learned a lot about the photometric process of retrieving data and producing a light curve. I also learned a tonne about Python, and I believe I can do almost anything in Python after four months straight of exclusively using it. Finally, I learned about the sheer number of data reduction techniques that exist, and that the biggest problem is trying to find the one which works the best and fastest for your specific goals.
The biggest challenge during my internship was creating my first light curve. We started from the raw K2 data and I could not see my progress until I had created that first light curve, so I was never really sure if I was going in the right direction. There were many times where I would go through all the steps only to realise that it did not work at all. Once I got that first light curve, I was able to take the data reduction step by step and check my progress along the way.
The best thing about my internship was the opportunities. I was given the opportunity to meet a lot of people when I went to the fourth Emerging Researches in Exoplanet Science conference (ERES IV) at Penn State in Pennsylvania and when I went to the Observatoire du Mont-Mégantic throughout the summer.
Trottier intern from McGill, working with Prof. Nicolas Cowan at McGill University over the summer of 2018
My internship was on using infrared limb transmission spectra from the Atmospheric Chemistry Experiment (ACE) to create an empirical transit spectrum of Earth. I used this spectrum to determine what an Earth-like planet would look like in transit, with various host star types, as seen by JWST. This will help us recognize an Earth-like atmospheric composition based on the molecular features in an exoplanet’s transit spectrum.
The ACE data that I used have a high spectral resolution and are also organised by altitude, latitude, and season. These data therefore provide more detailed information that can be obtained from a transit spectrum in which only an overall flux decrease from a system is measured. Although my main goal was to convert these data into a transit spectrum, I was able to identify where in the atmosphere photons are blocked by molecules, and how this changes as spectral resolution is decreased.
In transparent wavelength regions of Earth’s spectrum, most of the sunlight passing near the surface is transmitted. However, the atmosphere still blocks some of the incident light, and therefore contributes significantly to the transit depth. This means that in a transit spectrum, we will receive photons from lower altitudes than the transit depth suggests when the atmosphere appears transparent.
This summer, I learned to communicate with others about my work. I wrote my first paper, so I learned many things related to that, particularly presentation of results and scientific writing. I also learned how to present a poster and give a talk, and I greatly improved my Python skills.
The most difficult thing for me was acquiring a deeper understanding of my own research through a literature review. It was challenging for me to read papers and understand them in the context of what I had already learned. I had to learn to work independently, building upon the work of others to obtain and interpret my own results.
The best part of the whole experience was feeling like I was valued for the work I was doing. As I progressed in my project, I learned to explain what I was doing to professors and other students and answer questions about my work. I spent the summer feeling like a new, but not entirely lost, member of the academic community. I am grateful for the inclusivity of iREx and the McGill Space Institute, which encouraged summer interns to participate in discussions and ask questions.