The Road to Characterizing PH1: Visual-band Imaging
Today we have a guest post from Bill Keel. Bill is a member of the science team for Galaxy Zoo, and is more accustomed to dealing with stars by the billion than one at a time. He is a University of Alabama astronomer, weekend trombonist, and occasional photographer, being gradually trained by two cats with names out of Tolkien. Both his Twitter stream and his posts on the Galaxy Zoo forum can be found under the name NGC3314, and his other professional exploits may be found at http://astronomy.ua.edu/keel.
Kepler is sometimes most effective when properly backed up by other instruments, since its design was tightly optimized for precision in measuring bright stars at the expense of other things (such as angular resolution). Here’s a case showing how interpretation of Kepler results on planetary transits can be assisted by fairly routine ground-based measurements.In late June, I got an email request from Meg Schwamb:
“We’ve found a planet with ~130 days orbit going around a eclipsing binary. The eclipsing binary has a 20 day orbit so the planet is circumbinary and there’s a third star in the binary+planet system orbiting out at ~1000 AU with a period of 10E4-10E5 years. We’ve been following up the system with Keck observations.” [We didn’t yet know at the time that this third star would itself turn out to be a binary star].
The region around this star from our perspective is very busy (like the whole Kepler field), and the Kepler measurement includes light from additional faint stars. One, in particular, appears about 3 arcseconds away from the star of interest, well within the 6-acsecond radius of a Kepler measurement. Knowing its brightness would help narrow down the planet’s properties, making sure we have the right starting points in brightness for the Kepler target star my itself.
My institution is a partner in the SARA consortium, which operates telescopes in Arizona and Chile remotely. As a result, I have fairly regular nights scheduled, and indeed there were a couple of nights I could use at our northern telescope, a 0.9m instrument on Kitt Peak, Arizona, in July (just before shutdown for the monsoon season). After a couple of tries when the weather didn’t quite cooperate, including one night that was clear but the air to turbulent for this project, I got an hour’s worth of images on the evening of July 17. The image quality (seeing, in astronomical jargon) was 1.5-1.8 arcseconds, meaning that these values give the diameter across which a stars image drops to half its peak intensity due to atmospheric turbulence. That makes separating stars 3 arcseconds apart tractable. The timing worked out well during the night – the field was within 15 degrees of the zenith, minimizing atmospheric and tracking problems.
Trying to get precision measurements of bright and faint stars simultaneously takes some care – good data on the faint star isn’t much help if the bright star is hopelessly saturated in the data. So instead of one long exposure, I took 60 1-minute observations, using a red filter to roughly match the midpoint of the very broad spectral band used by Kepler. For further analysis, that gave both the grand average of all 60, and I also used averages of subsets of 10 to help estimate certain sources of error in the processing.
Even though the fainter interfering star was clearly separated from the bright one in these images, there was enough spillover to need correction.I tried several procedures or this – the most successful took as a reference point a similarly bright star with no companion in that direction, subtracting variously scaled versions of its image to eliminate as much of the bright star’s light as possible (the subtracted images looked a little odd in the middle – much later I realized that might come from the very close companion star seen in other data).
To make sure we understood how our brightness measurements relate to the Kepler data, I checked published magnitudes for Kepler stars in this neighborhood. This gave me some bad moments until I realized that the published values were often based on short exposures with a telescope no bigger than I was using – bright stars are OK, faint stars become quickly much less accurate. Phew. Now I know this, so that if it comes up again, I’m ready.
The result? That fainter star has magnitude R=18.73, making it only 1.02% as bright as the Kepler target with planet. Other contaminating stars are still fainter, down to 0.03% of the target star’s red-light intensity.