In which I observe an exoplanet transit, and explain how I fit a model to my observations. For a brief intro to, and some artsy pictures of, the Maryland Space Grant Observatory I've been starting to use, visit this previous post.
What if I told you I observed a planet, orbiting a star hundreds of light years away, from the middle of Baltimore? It’s true, I battled back the stadium lights and light pollution and got a glimpse of another world, and in this post I hope to show you how.
what’s an exoplanet transit?
Planets can orbit their host stars at a variety of distances and orientations, and astronomers can learn a lot of information about a given planet by studying its orbit. When a chance alignment brings the orbit of a planet between its star and our line of sight to that star from Earth, the planet blocks a small percentage of the light that travels from that star all the way to us.
You’re probably familiar with other forms of eclipses, like a solar eclipse where the Earth’s moon blocks the light from our sun. By using telescopes and cameras to measure how bright a star is over time, astronomers can measure how much of a star’s average brightness is blocked by the planet during its eclipse, and use that information to model how large that planet might be, and what its orbit might look like.
My colleagues at JHU do this ~professionally~ and use very complex instruments on powerful telescopes in order to make groundbreaking measurements of the exoplanets that populate our galaxy. For my day job, I use different techniques to study exoplanets, but on one of my nights off, I wanted to see if it was possible to use the transit technique to observe exoplanets using a ~relatively~ small telescope situated in the middle of Baltimore.
I started by using a tool called Tapir [Jensen 2013] which computes observable transit events from known exoplanets based on the location of your observatory. Tapir tipped me off that a gas giant planet would be transiting a bright star that evening. Gas giants are big, and so they block a larger percentage of their star’s light, making their transits easier to observe. We had our target, HD 189733 b. That’s a star in the Henry Draper catalogue, HD 189733, orbited by a planet astronomers have labeled “b.”
My friend Natalie and I reserved the roof and set up a monochromatic camera for astrophotography on the Morris-Offit telescope at the Maryland Space Grant Consortium Observatory. We trained the telescope on the position of HD 189733, and set our camera to taking short exposures. We then used a program called AstroImageJ to extract the brightness of our target star when compared to the other stars in our images. Plotting this “relative brightness” over time reveals a signal that resembles the expected transit curve in the above figure – we caught the transit!
If you squint, you could imagine drawing a line through all the data points and coming away with a “dip.” The only issue, is that because we used a smaller telescope in the middle of a city near bright lights and through the turbulent atmosphere, our data is very noisy. You can also see some gaps in the data where we had to fiddle with our instrument setup, and stopped taking images, or took fewer images per minute.
trust the process
We still have a lot of work to do, if we want to understand our planet a bit more deeply.
Natalie and I used Juliet, a package of python code written by Nestor Espinoza, in order to fit a model of a planetary eclipse to our model. The package will take a model for how a star’s brightness should change as a planet passes in front of it, and fit that model to our data.
The cool thing about Juliet, is that it allows us to fit a model to the stochastic processes which might add complex sources of noise to our observations. You can see that there’s a lot of variation in our observations, and some of it doesn’t seem entirely random. We might expect benign sources of noise (the random thermal jitter of our camera, for instance) to behave randomly, adding some scatter to our datapoints, but we observe weird trends that can’t be explained by “white” noise. These weird stochastic processes are most likely either variations due to the Earth’s atmosphere, or activity (like starspots) on the surface of HD 189733. Both are interesting to study in their own right, but we don’t really care about them when we want to study HD 189733 b – in fact they get in the way!
By fitting a Gaussian Process, a smoothly varying random function defined by the joint distribution of normally distributed random variables (whatever that means), we can make sure our eclipse model isn’t biased by these extraneous sources of noise. We set the expectation that our data should look like the basic transit dip, plus some weird wiggly random variations, and test models that are combinations of various transit dips and weird wiggles. The figure below is an example “gaussian process” I generated that is fit to two data points. Juliet fits a gaussian process to the many hundreds of data-points in our data.
Juliet runs really smoothly out of the box, so basically the only trouble is making sure your data is in the correct format for it to be read-in and fit. The result is a cleanly detrended lightcurve (below), that shows a smooth transit event!
The size of the planet we measure from our best fit model is 1.1 times the radius of Jupiter, but it orbits its host star at a distance of just 0.03 astronomical units (the Earth is 1 au from the Sun, and Jupiter is ~5 au). This planet, HD 189733 b, is what astronomers call a “hot Jupiter.” My PhD co-advisor, Dr. David Sing, actually published a paper based on observations of this exoplanet taken by the Hubble Space Telescope. His observations, combined with many others, have contributed to our understanding that the atmosphere of this planet is a blue color, comprised of molten glass, which whirlwinds in fierce storms across the planet’s surface. NASA made a cool horror movie style poster about it (below)
That’s about all I’ve got, but I hope you enjoyed this brief overview of transit photometry (the observation of the intensity of light from a source). I had a lot of fun learning this stuff on my own earlier this semester. Maybe more experiments with small telescopes to come in the future? Until then, clear skies!