HAT-p-7b and the Grail quest – With Jon Jenkins
I went to bed the evening of May 13th exhausted from the long, intense campaign of commissioning the Kepler spacecraft. The long march started about a week after launch when we began to receive data from the photometer and needed to process it to verify that it was behaving as we expected and to prepare all the data products needed for nominal science operations. These included taking very special data sets to characterize the 2D bias frame of the CCDs (the image you get with no light falling on the detectors), the noise characteristics, the sky to pixel mapping, the science data compression tables, and the detailed shape of the stellar images (the Point Spread Functions) across the focal plane. We had been calculating the PSFs and getting our first science target tables together while the Combined Differential Photometric Precision (CDPP) data set was being collected during the last ten days of Commissioning. This was the first science-like data to be collected. So we had a target table in place with 52,496 targets and were compressing the 30-minute samples for each pixel of interest and storing these on board the Solid State Recorder. (During nominal science operations we collect pixel data for ~145,000 stars.) On Monday May 11 we turned the spacecraft to point the High Gain Antenna to Earth and downlinked the CDPP data set, all ten days of it, to the Deep Space Network, who transferred it through our Ground System* to the Science Operations Center at NASA Ames Research Center where we process the pixels, extract the photometric light curves and search for transiting planets. Nominal science operations commenced on May 12 and we could turn our attention to processing the CDPP data.
Just before I fell asleep Wednesday evening I received email from our operators that the data were fully processed and ready to examine. There wasn’t much chatter about the light curves, and hearing no alarms, I fell asleep. I suppose everyone else was just as tired as I was and weren’t prepared to stay up half the night examining the results (as we’d done when the first light images came down). I did wake up very early and went straight in to work in the SOC, managing to beat our earliest commuter who usually arrives at 6:30 am. I did procrastinate a bit — taking the time to wash the coffee pot and brew a full pot before sitting down at the workstation to begin pulling the light curves off the server. I have to admit that it was a bit scary to confront the first light curves, harbingers of habitable planets, if they were only good enough.
My worries were unfounded: True, the light curves were not perfect. The cosmic ray cleaning algorithm needed some tweaking – it was letting obvious impulsive outliers through, and there was evidence of instrumental effects such as pointing offset-induced photometric variations. However, the light curves were simply stunning! Eclipsing binaries and other variable stars looked as if they had been rendered from analytical models. The SOC staff started collecting within the next few hours and we looked at thousands of the light curves, paying attention to those with strong coherent behavior (the light curves or flux time series could be easily sorted to identify the stars with the strongest periodic nature and the strongest variations in time). An astrophysicist colleague, Doug Caldwell, our Instrument Scientist, brought out his textbook on variable stars and we kept asking him “what kind of variable star is this one (on screen)?”. He riffled through the book and in many cases said he didn’t know, that none of the light curves in the book looked anything like the ones on screen.
After a bit, we looked up the target id numbers for the three known transiting planets in the Kepler field of view (FOV). These include TrES-2, HAT-P-7b and HAT-P-11b. All three exhibited their transits in glorious detail. This was quite a relief! If we couldn’t simply see the transits for giant planet transits, Kepler’s prospects for finding 100x weaker signals would be in danger. The light curve for HAT-P-7b was different from those of TrES-2 and HAT-P-11b. Even before I folded the light curve I could see evidence for a bow-shape rise and fall in the flux level from the end of one transit to the start of the next. Additionally, there appeared to be a small dip in the light curve centered between transits. We’d hoped to find evidence for reflected light from close in giant planets, but we anticipated this would take many months of observations given the small amplitudes expected. What could this feature be?
Elisa Quintana, one of our scientific programmers with a Ph.D. in astrophysics, found and printed out the discovery paper for this planet and the information available on the Extrasolar Planets Encyclopaedia. HAT-P-7b is about 40% larger than Jupiter in a 2.2 day orbit that holds it merely 0.038 AU from its star, which is 86% larger than Sol and burns at 6350 K. The planet is approximately 2650 K so it’s glowing like a super heated electric stove coil. When the planet is in or near transit it’s facing its cool night side towards us, but as it parades around to the back side of its orbit we see more and more of its star-heated face. This thermal emission is strong enough to increase the flux that we measure by ~122 parts per million (ppm). When the planet disappears behind the star at the furthest point in its orbit from our perspective, this thermal emission winks out, taking a small “bite” out of the light curve.
I folded the light curve and displayed it on the screen. The thermal emission “bow” feature and the “bite” caused by the occultation of the planet by the star were unequivocal. The entire team was awestruck: we were looking at raw light curves from the first run of the pipeline and were able to see the signature of a planet in the data that is comparable to the Holy Grail Kepler seeks: an Earth-sized planet transiting a Sun-like star produces a drop in brightness that is about 100 ppm, similar to the 130 ppm drop in brightness of the secondary occultation for HAT-P-7b.
By this time our Deputy Principal Investigator, David Koch, and our Target Scientist, Natalie Batalha, had shown up and we trolled through the light curves until noon, breaking for a quick snack of leftover pizza from commissioning in the fridge, and sitting down again for a few more hours. I called our Science Principal Investigator, Bill Borucki, to let him know that I thought we already had our first Science paper topic. He arrived and I gave him a tour of the light curves of the variable stars. Some of these have such high frequency oscillations that they look like human voice recordings. I joked with him that we should play a few through the sound card to see if we could hear “hello, Earthlings!”. A civilization that can modulate the brightness of their star to transmit such messages would be awesome, indeed. Then I showed him the light curves for TrES-2, HAT-P-11b and finally, HAT-P-7b. Confronted by the thermal emission and occultation features, Bill, as usual, attempted to find some flaw in our interpretation of the data. How could we be sure? One by one, all his objections fell by the wayside and he was left with a gentle smile on his face and a gleam in his eye.
Bill knows full well that regardless of how well our instrument works, discovering Earth-sized planets won’t be easy. We can’t directly confirm true Earth analogs. Their radial velocity signatures are too faint to observe. We can reject most false positives from background eclipsing binaries by using the Kepler data itself, but there will likely be a handful of triple systems with a bright, hot star in a wide orbit with an eclipsing binary composed of much cooler, dimmer stars. The eclipses will be diluted by the glare of the hotter star so that their depths are similar to those of planetary transits. These will be tough nuts to crack, but we do plan on making observations of these systems with the Spitzer infrared telescope, which will allow us to reject this type of masquerader.
Kepler has been a dream becoming reality since 1984 (and I joined the team and the dream in 1995). We’ve scaled many peaks along the way, but with the CDPP data, and the light curve for HAT-P-7b in particular, we’re finally in sight of the summit we seek to conquer. It will take the full three and one half years of the mission to reach this summit and to begin detecting and vetting and announcing discoveries of Earth-like planets transiting Sun-like stars. This will not be a cake walk: we are in the nascent phases of getting to know the personalities and quirks of the Kepler Photometer. Nobody has seen stellar light curves at this level of detail. Most of our targets are dimmer than HAT-P-7b at a Kepler magnitude of 10.5. Their light curves will be intrinsically noisier and we won’t be able to “see” at a casual glance individual transits in the data — we’ll have to rely on our automated transit search tools to comb through the data. I expect we’ll re-write 80% of the SOC science pipeline software in the course of the mission.
Nevertheless, we can take hope and respite in this discovery even as we continue the steep and winding trail to our ultimate goal, and we can look forward to many unexpected and pleasant surprises along the way.
* The Kepler Ground System includes the Mission Operations Center at LASP in Boulder CO, the Flight Planning Center at Ball Aerospace in Boulder CO, the Data Management Center at STScI in Baltimore MD, and the SOC at NASA Ames Research Center.