The Science Case for SPADES

The diversity of exoplanet properties uncovered in the last 15 years presents tremendous challenges to our understanding of planet formation and evolution. There have been many new surprises, from planets orbiting pulsars (e.g. Wolszczan & Frail 1992, Nature, 355, 145), to the “hot Jupiters” with densities too low to exist (e.g. Mandushev et al. 2007, ApJ, 667, L195), to planets with highly eccentric orbits (e.g. O’Toole et al. 2009, MNRAS, 392, 641). One of the most surprising discoveries has been the significant number of planets and substellar companions in binary star systems: approximately 17% of exoplanetary systems discovered to date contain at least two stars (Mugrauer & Neuhäuser 2009, A&A, 494, 373). Most of these systems contain planets orbiting a Main Sequence star with a companion at large distance – 250 to 6000 AU – however three systems have been discovered where the binary separation is small: less than 40 AU (Queloz et al. 2000, A&A, 354, 99; Hatzes et al. 2003, ApJ, 599, 1383). These are sometimes referred to as “S-type” (“satellite-type”) planetary systems. Perhaps the most remarkable discoveries though, have been those of systems where the planets orbit both stars in the binary system (e.g. Lee et al. 2009, AJ, 137, 3181); these are the “P-type” (“planet-type”) or circumbinary planets.

Formation and Stability

The detection of circumbinary planets orbiting close stellar binaries provides us with an important way to test and constrain planet formation models. While the detection of planets around single stars has led to a paradigm shift in our understanding of planet formation, there are still several different models that can produce effectively the same planetary system configurations (e.g. Boss 2007, ApJ, 599, 577; Ida & Lin 2004, ApJ, 604, 388). By investigating how stellar multiplicity affects planet formation we can ask questions such as: does binary star formation promote or hinder planet formation?

What is expected is that in a binary system the planet-forming nebula will be disrupted by the dynamics of the two stars. The perturbations of the central binary star will lead to the substantially different evolution of a protoplanetary disk compared with that of a single star (Artymowicz & Lubow 1994, ApJ, 421, 651). In a series of papers, Pierens & Nelson (2008, A&A, 483, 633; 2008, A&A, 478, 939; 2007, A&A 472, 993) investigated the formation, evolution and migration of planets in circumbinary discs. In binary systems where the initial separation of the two stars is 1 AU, they found that Saturn-mass circumbinary planets are likely to be quite common, while Jupiter-mass planets will be less common, and exist at larger distances from the binary. There appear to be no formation studies of systems where the binary orbital periods are on time-scales of days or tens of days. Holman & Wiegert (1999, AJ, 117, 621) studied the long-term stability of planets in binary systems over a range of mass ratios and binary eccentricities. They found that orbits beyond 3.7 times the binary separation should be stable.

Currently known systems

Up to now, all circumbinary planets or brown dwarfs have been discovered orbiting systems containing either a white dwarf or hot subdwarf. The secondary star is an M star in a short-period orbit in each case (e.g. HS0705+6700, Qian et al. 2009, ApJ, 695, L163). There is a strong selection effect at work here, however: each of these systems have been discovered by teams studying (pre-)cataclysmic variables rather than searching for circumbinary planets (at least initially). All systems have been discovered using eclipse timing: the times of each primary eclipse minimum are compared in an O–C analysis, and periodic signals that cannot be explained in any other way, such as magnetic effects or angular momentum loss (Applegate 1992, ApJ, 385, 621), must be due to third bodies.

It is unclear when these circumbinary planets formed: did they form as “first generation” planets, at the same time as the progenitor star; or did they form after a vast amount of mass was lost from the system through a symbiotic or common envelope phase? Perets (2010, ApJ submitted, astro-ph/1001.0581) has suggested that there may enough material in, e.g., an accretion disc around a symbiotic star, to form planets, and has used this idea to explain the known systems. There is some observational evidence that protoplanetary discs can form around these systems (e.g. Ireland et al. 2007, ApJ, 662, 651). Detection of – or stringent limits on – a planetary system around a close Main Sequence binary would be an important step towards answering these questions.

Circumbinary discs

Can circumbinary planets form around binary systems where both stars are on or near the Main Sequence? Circumbinary discs have been detected around young spectroscopic binaries such as CoKu Tauri/4 (Ireland & Kraus 2008, ApJ, 678, L59), HH 30 (Guilloteau et al. 2008, A&A, 478, L31) and GG Tau. In the latter system, the disc has been resolved and an inner disc cavity has been observed, which is due to the tidal torques exerted by the central binary (Dutrey et al. 1994, A&A, 286, 149). Brightness asymmetries have also been seen in subsequent HST imaging, which suggests the disc is warped by tidal effects (e.g. Krist et al. 2005, AJ, 130, 2778).

It therefore appears that the necessary factors (discs and stability) are there for circumbinary planets to form and survive around binary systems containing Main Sequence stars. Detecting such planets is now the final piece in the puzzle.

Previous studies

There have been a limited number of attempts to detect such systems, mainly because standard planet search methods are difficult to implement. The Doppler velocity method used by planet search teams such the Anglo-Australian Planet Search (e.g. Vogt et al. 2010, ApJ, 708, 1366; O’Toole et al. 2009, ApJ, 697, 1263) cannot easily be applied to double-lined spectroscopic binaries. Konacki et al. (2009, ApJ, 704, 513) initiated a radial velocity search for circumbinary planets around Main Sequence binaries using a novel combination of the standard method and the two-dimensional cross-correlation technique TODCOR introduced by Zucker & Mazeh (1994, ApJ, 420, 806). To date, they have managed to achieve precisions down to ~2m/s for the individual components, but significantly higher (> 10 times) for the systems overall. They detected no candidate exoplanets based on the examination of 10 double-lined systems with periods longer than ~5 days. It is uncertain however, whether this technique will be precise enough to detect the tiny variations in the systemic velocities caused by circumbinary planets.

The eclipse timing method does not suffer from these problems. If the target systems are properly selected, it therefore provides the best chance to detect planets around the large number of eclipsing binaries with periods less than 5 days. For this reason, we have started the SPADES project.

The SPADES project

The Search for Planets Around Detached Eclipsing Systems (SPADES) project is designed to look for substellar companions orbiting around eclipsing binaries containing Main Sequence or Sub-giant stars with spectral types from A to M. We will use the eclipse timing method discussed above and involve multiple telescopes at different sites. The eclipsing systems have been selected from the General Catalogue of Variable Stars (GCVS – Kazarovets et al. 2009, Astronomy Reports, 53, 1013), have orbital periods between 1 and 5 days, and are Algol-type binaries where the stars are detached; i.e. there is no mass transfer between them. By selecting targets in this way, we are ensuring that the eclipse minima are sharp and can be measured with high precision, and the eclipse durations are short enough that a whole eclipse can easily observed in one night. Other periodic and quasi-periodic variations – due to magnetic variations and angular momentum loss, for example – will also be minimised. To achieve adequate phase coverage for our O–C analyses, we are collaborating with several experienced amateur astronomers through the Variable Stars South network (, and plan to work with undergraduate student programs where appropriate. Variable Stars South has participants across the world. (Note that the conditions of joining the SPADES network are access to a telescope with a minimum aperture of about 30cm, a CCD camera, and an accurate method of calibrating and recording observing times.)

The project will last for at least five years, as, unlike Doppler velocity planet searches, eclipse-timing measurements become more sensitive to lower masses at longer periods. This can be seen in Figure 1, where we show the discovery space for circumbinary planets using the eclipse timing method. The timing precision thresholds assume a binary system mass of 1.5M€, equivalent to a late G-type plus K-type Main Sequence binary; they are calculated using equation 5 of Sybilski et al. (2010, MNRAS, 405, 657). The green crosses in Figure 1 show the currently known planets detected using the timing method. The two crosses at the bottom of the figure near 100 days are the pulsar planets around PSR 1257+12 (Wolszczan & Frail 1992), while the other systems are orbiting (pre-)white dwarfs, as discussed above. Also shown in the figure are sensitivity levels for four different Doppler velocity precisions for! comparison. Finally, the data points for currently known exoplanets are taken from the Exoplanets Encyclopaedia (


Figure 1: Discovery space for circumbinary planets around a binary system with a total mass of 1.5M!. The dashed lines marked ET represent various timing accuracies, while the magenta dotted lines marked RV represent various Doppler velocity accuracies for a star with one solar mass. Also shown are all known Doppler-detected planets (black diamonds), transiting planets (red squares), and timing-detected planets (green crosses), along with Earth, Jupiter and Saturn (blue asterisks). The cyan lines are marked at 1, 4 and 10 years.

The aim of SPADES is to achieve an eclipse timing precision of approximately 1 second, which would allow us to detect a Jupiter-like planet in a Jupiter-like orbit, however the eclipses of all of our targets may not be sharp enough to achieve this, so we have also plotted lower precisions. Even with timing precisions of 5 seconds, we will still detect companions with only a few Jupiter masses in four years. Note that Jupiter-mass planets are more likely to be formed and stable at these kinds of periods (Pierens & Nelson 2008). We have already achieved this precision for the extremely difficult case of the semidetached eclipsing system CU Hya (Richards 2010, {jd_file file==4} ). In brief, this star is fainter than all of our targets (CU Hya is not a SPADES target), which means longer exposure times are required. It is a semi-detached system, meaning the eclipses are not as sharp as a fully detached system; it is therefore much more difficult to measure times of minima precisely. Despite this, the mean measurement error for times of minima is only 3.8 seconds! With the better-defined eclipse minima of brighter, fully detached eclipsing binaries, we are confident of reaching our target precision of 1 second for the majority of systems.

There are online databases such as the O–C Gateway ( that contain more than 50 years of archival data for some of our targets, however much of it – but not all – are low quality and based on visual or photographic observations. The majority of eclipsing binary systems in the southern sky are in fact poorly studied, with little known about their spectral types, component masses and other properties. Since in many cases even the ephemerides are poorly constrained, we are currently using data from the SuperWASP archive ( to measure them; an example light curve is shown in Figure 2. We will provide the light elements to the community in the form of a catalogue. Without the total mass of the binary system, we cannot determine the masses of any low-mass companions we find with SPADES. Therefore a time resolved spectroscopic study of these systems forms a vital part of the SPADES project.


Figure 2: SuperWASP light curve of the eclipsing binary RU Gruis, phased with a period of 1.89 days and the time of minimum light. This star shows a clear size difference between its two components, and will probably be a single-lined spectroscopic binary. SuperWASP has observed a large number of SPADES targets over the last 4-5 years. We are using John Southworth!s JKTEBOP code to measure the light elements of each of these systems. These updated ephemerides will be useful for both the SPADES project and the eclipsing binary community.