NASA Kepler Misison

I worked on the Kepler Mission at NASA Ames Research Center in Mountain View and with the SETI Institute, from 01/2011 - 08/2012.

The Mission

The Kepler Mission is NASA's 10th mission in the Discovery program. Kepler addresses the fundamental question how frequent habitable planets - like Earth - are in our galaxy. The mission objective is to find earth-like planets around distant stars, and to determine the frequency of planets in the habitable zone. For that purpose, the Kepler Spacecraft (see right) in earth-trailing orbit continuously observes about 150,000 stars in the Cygnus region along the Orion arm. A large array of CCD-cameras, totalling about 100 Million pixels, onboard the Kepler Spacecraft takes images of the stars in the field of view, and the data is then downlinked to Earth in monthly intervals.

In addition to the main mission objective of planet detection, the Kepler Mission also produces a wealth of unequaled other astrophysical information. The scientific data Kepler collects is made available to the public, and allows for new studies in cosmology, astroseismology and other related disciplines.

Kepler uses Transit Detection to detect planets: when a planet passes between the star it is orbiting and the camera aboard the Kepler Spacecraft, it obscures a fraction of that star. Each star covers only 10-20 pixels on the CCD camera, so it is not possible to directly see any planet or its shadow. Instead, transit detection tries to measure the resulting dips in apparent brightness of the star. The magnitude of the dip is proportional to the size of the planet relative to the star, which means that minuscule brightness fluctuations of 1/10,000 have to be detected (the earth being about 1/100 the diameter of the sun). When considering the range for the habitable zone, and the stellar variability many stars exhibit, the required precision surmounts to 20 ppm (parts per million).




The Kepler Spacecraft

(from http://kepler.nasa.gov)

The Challenge

This precision requirement demands a completely unprecedented photometric precision. It is the equivalent of trying to detect a tiny flea passing across a car headlight from many miles away. To make things worse, the light curves of the stars are not quiet and flat lines, but they are often seemingly chaotic signals polluted with noise and complicated systematic errors from electronic effects (drift of the CCD sensitivity, interference with the electronics aboard the spacecraft, etc) as well as astrophysical noise like the light of other stars in proximity to the target star.

And this was where my job at Kepler started. At NASA Ames Research Center, we were developing the signal processing algorithms to process Kepler lightcurves and detect the minuscule planet transit signals. My particular responsibility was the development and implementation of algorithms to identify and correct the systematic errors in the lightcurves, while preserving the underlying astrophysical signals. This challenging endeavor requires sophisticated statistical analysis methods and state of the machine learning techniques