A modern HR diagram. The features here encode a wealth of information about stellar physics. We are currently working toward building a similar diagram for exoplanets. Credit: Kenneth Lang


You can learn a lot by counting.

About 100 years ago, Ejnar Hertzsprung and Henry Norris Russell tabulated the colors and luminosities of a few dozen stars. They noticed that stars clustered in certain regions of this diagram and not in others. The features in the Hertzsprung-Russell Diagram are fundamental tests for models of stellar structure and evolution. Today, we understand much of the physics that accounts for these features. Can we do the same for planets?

Since its launch in 2009, NASA's Kepler mission has played a key role in characterizing the population of exoplanets by detecting over 4000 planets. In Petigura, Howard, & Marcy (2013) we measured the prevalence planets as small as Earth around Sun-like stars. We found that:

  1. Nearly every star (74%) has at least one planet, just within 1 AU.

  2. Small planets are common. Earth-size planets outnumber Jovians by 16 to 1.

  3. About one in five stars like the Sun host a planet similar to Earth in size and temperature. This implies that there are about 40 billion Earth-like* planets in the Milky Way. (Disclaimer: the range of sizes and temperatures that qualify as Earth-like are uncertain and under debate.)

This work received the 2014 Cozzarelli Prize from the National Academy of Sciences.

Recently, my group completed the California-Kepler Survery (CKS), a 50-night Keck/HIRES survey. Our goal was to improve the precision of Kepler host star properties and thus better understand the properties of their planets. In Fulton, Petigura, et al. (2017) we observed a gap dividing planets between the size of Earth and Neptune into two populations, which are likely high-density rocky "super-Earths" and  low-density "sub-Neptunes" with thick H/He envelopes.

Distribution of planets in the period radius plane . Points: detected planets, shading: planet occurrence after correcting for survey incompleteness. From Petigura et al. (2018).