The Road to Characterizing PH1: Transits and Initial Modeling
Today we have a blog post from Dirk Terrell. Dirk is an astrophysicist and the manager of the Astronomy and Computer Systems section at the Southwest Research Institute in Boulder, CO. His hobbies include coaching football, racing motorcycles and space art. He is a Fellow and former President of the International Association of Astronomical Artists.
The discovery of PH1, a planet orbiting an eclipsing binary with another pair of stars about 1000 AU away, was exciting for everyone involved. As someone who has worked with amateur astronomers for over twenty years, I was particularly happy to see that the initial discovery of the transits was made by two amateur scientists, Kian Jek and Robert Gagliano. My involvement came shortly thereafter when Meg Schwamb was visiting my institution , Southwest Research Institute in Boulder, Colorado, last spring to give a talk on another topic. She was meeting with Hal Levision, a dynamicist here at SwRI, and was describing the rather complicated Kepler data on the system. That’s when Hal gave me a call.
I get calls from Hal all the time, almost always on some computer-related topic. But this time he said “Do you have time to talk science?” My area of expertise is eclipsing binary stars, and here at SwRI I am surrounded mostly by scientists who do research on topics in planetary science, so I don’t often get phone calls from people here to talk about science. Curious, I walked down the hall to Hal’s office.
When I walked in, he introduced me to Meg and she began to describe this eclipsing binary system that had been observed by the Kepler satellite. She pulled up the light curve on the Planet Hunters web site. Having been doing the analysis of eclipsing binary light curves for nearly a quarter century, I immediately began to develop an idea of what the system looked like. It had an eccentric orbit and the stars were reasonably small compared to their separation, so they would not be very distorted by each other’s gravity. It was also clear that there were regular variations in the light curve outside the eclipses of the stars, probably pulsations or spots on one of the stars. It was a very busy looking light curve! Then she showed me the dips that Kian and Robert had noticed. With all of the activity in the light curve, it wasn’t quite as obvious as it is in simpler systems that we had transits by a planet. She asked me if I could help clear things up enough to warrant turning very expensive big telescopes like Keck on the system to get complementary data. I told her I’d be happy to look at it.
When working with data, I like to start as closely as possible to what the instruments produce and then do what’s needed to get the data into the form needed for an analysis. That way I know exactly what’s been done. Kepler data, like those from any instrument, have various effects that have nothing to do with the object you are observing. For example, the spacecraft has to make periodic maneuvers to keep its solar panels properly aligned. These rolls show up clearly in the data because the stars being observed are in different places on the imaging detectors before and after the rolls. Kepler, as you probably know, measures the brightness of stars very, very precisely with CCD arrays, the same device inside your digital camera. The pixels that make up the CCDs are not all uniformly sensitive, so when a star is moved from one set of pixels to another, the response to the same brightness level from the star will be a little different. But even a relatively small 1% difference in sensitivity will show up as a big jump with detectors as precise as those on Kepler. To give you an idea of what these raw data look like, here is a plot of the raw Kepler data for PH1 from quarter four:
You can see five deep eclipses of the hotter star, the higher frequency pulsations or spot modulation, as well as large instrumental effects and gaps. And, as we later showed, there is indeed a transit by the planet in there too, but it certainly doesn’t jump out at you because of everything else that is going on. So, my job was to clean up the light curve so that we could isolate the potential transit signatures. While we will later be very much interested in the stellar eclipses and the spots/pulsations, at this stage of the game we just needed to get rid of them. My approach was pretty simple: break the light curve up into parts to which I could fit a combination of linear and periodic terms, and then subtract those fits to get the residuals (i.e. what was left). Since spot/pulsation modulations happened at a much higher frequency (period of about 2.6 days respectively) than the suspected transits (period of about 138 days), removing the high frequency terms would leave the transits and the stellar eclipses (20 day period) intact and the linear term would remove most of the instrumental effects. Then I could fit the stellar eclipses with the Wilson-Devinney light curve program and remove the eclipse signatures leaving only the transits. This approach worked well enough to answer our big question at the time: are there transits? Lurking in the residuals were things like this: