Close Friends of Hot Jupiters: The WASP-47 system

Guest post by Juliette Becker and Andrew Vanderburg

Ever since a mechanical failure caused the end of the original Kepler mission in 2013, the Kepler spacecraft has been conducting a survey of new stars, searching for planets across the ecliptic plane in its new K2 mission. The K2 dataset is a goldmine of fascinating science results. One such result is the recent discovery of two new planets in the WASP-47 system.

Until a few months ago, everyone knew that hot Jupiter planets don’t have “friends”, or nearby small planets in close orbits to the host star. These other planets had been searched for extensively, through radial velocity measurements, analysis of the transit times of the hot Jupiters, and even through transits by Kepler during its original mission. All of these searches turned up nothing.

This all changed one day last July, when Hans Martin Schwengeler, a citizen scientist who enjoys poring over Kepler and K2 data searching for new transiting planets by eye, came across the telltale signatures of two extra transiting planets in the hot Jupiter system WASP-47. WASP 47b was, by all indications, a perfectly normal hot Jupiter -- in the discovery paper, Coel Hellier wrote “With an orbital period of 4.16 days, a mass of 1.14 Jupiter masses, and a radius of 1.15 Jupiter radii, WASP-47b is an entirely typical hot Jupiter”. The discovery of additional transiting planets dramatically changed the narrative.

When Hans came across the planets, he posted them to the Planet Hunters forum, where he and other citizen scientists discuss their findings. Andrew Vanderburg came across the post suggesting that a known hot Jupiter had planetary companions. Using his K2 data reduction pipeline, he analyzed the light curve and confirmed Hans’s discovery - there were additional planets in the system, a super-Earth at a 0.8 day period and a Neptune at a 9 day period!

The WASP-47 transit signals.

Andrew emailed me, and at first I hardly believed that the light curve was real. How could a hot Jupiter have close-in planetary companions? I knew people had been looking for this type of companion for years via both photometry and transit timing variations, but the lack of discoveries indicated that they might not exist. I performed some numerical stability simulations (because it seemed at first like this system could not be dynamically stable!) and sure enough, the N-body simulations showed that the system was likely stable on timescales of 10 million years.

At that point, we formed a team with Hans, Andrew, MIT Professor Saul Rappaport, University of Michigan Professor Fred Adams (my advisor!), and me. Once this team was formed, we devoted ourselves to understanding as much about the systems as we could. Some work by Saul and Andrew confirmed that the planets were all orbiting the same star, Andrew fit the light curve to get the planet properties, and I ran more stability simulations. Soon enough, Fred suggested that I look at what transit timing variations (or TTVs, which happen when transits come late or early because of the gravity of other planets in the system) we would theoretically expect to see from the system - and I found that for the outer two planets, the TTVs should be observable.

I then measured the TTVs from the light curve, and sure enough - there was something there. After some discussion, we realized we could measure the masses of the planets from those TTVs! Though I had never done dynamical fits before, I wrote the code to utilize Kat Deck’s TTVFAST code in a Markov Chain Monte Carlo fit. With some advice from Kat and help from Fred, I eventually got the fits working and we were able to measure or put limits on the masses of each planet.

Transit timing variations. As the planets gravitationally
tug on one another, they arrive "early" and "late" to their
expected transit times

In a little less than two weeks, we had put together a paper deriving planet properties from the light curve, mass limits from the TTVs, and showing that you CAN detect companions to hot Jupiters using TTVs!

This result is exciting because it is the very first time a hot Jupiter has been found to have such close-in other planets. Before this discovery, it was unclear if hot Jupiter could have nearby friends, as they might destabilize the friends’ orbits during migration. This discovery opens up new questions about how these systems form - it is possible that there is more than one migration mechanism for hot Jupiters.

The system has become even more exciting, too. Marion Neveu-Van Malle has recently published a paper announcing yet another planet in the system, a Jupiter-mass planet at a 571-day period, which engenders even more questions about how this system formed. Additionally, there has been RV followup resulting in another constraint on the masses (led by Fei Dei, a grad student working with Josh Winn at MIT), Roberto Sanchis-Ojeda of U.C. Berkeley also recently published a paper where he and his coauthors used the Rossiter-Mclaughlin effect to find that the stellar obliquity of WASP-47b is close to 0. All this work has begun to paint a fascinating picture of this system - the first hot Jupiter with hot friends.

Our paper on WASP-47 and its new companions, which was published October 12th, 2015, in ApJ Letters and is available at, was a collaboration between myself (Juliette Becker, a graduate student at the University of Michigan), graduate student Andrew Vanderburg (Harvard CfA), Professor Fred Adams (the University of Michigan), Professor Saul Rappaport (MIT), and Hans Schwengeler (citizen scientist).