Getting HD 99429 ready for its screen test
Nine extrasolar planets are known to transit their parent stars, but all of these planets have periods shorter than 5 days. They are frying beneath the brilliance of their parent stars. It would be nice to find a transiting planet with a longer period. Preferably, this would be a giant planet with towering thunderstorms and warm, drenching rains, and orbited by a habitable Earth-sized moon that we could detect with HST photometry.
Advanced observers can make the discovery of a transiting room-temperature Jupiter a reality by participating in the systemic team‘s distributed observing project: Transitsearch.org. A brief blurb on the transitsearch.org home page describes the basic strategy:
Transitsearch.org is a cooperative observational effort designed to allow experienced amateur astronomers and small college observatories to discover transiting extrasolar planets. In order to utilize the advantages of a network of small telescopes most effectively, our strategy is to observe known planet-bearing stars at the dates and times when transits are expected to occur.
At present, the majority of confirmed extrasolar planets have been discovered using the Doppler radial velocity technique (see the tutorials at www.oklo.org). The Doppler method, however, cannot determine the inclination of a planetary orbit to the line of sight from Earth. Therefore, each planet discovered by the Doppler method has an a-priori probability of transiting, which depends mainly on the orbital period of the planet. Short-period planets have relatively high transit probabilities, whereas long-period planets have low transit probabilities.
Transitsearch.org hasn’t found a new transiting planet. But if we can maintain the enthusiasm of the collaboration, then eventually it will. Every planet that is detected by the radial velocity technique has a finite a-priori probability of transiting. Hence we need to work systematically down through the list. Chances of success (among planets that have not yet been fully checked) range from a sporty 12.9% for HD 118203 “b” down to a depressingly low 0.1% for 55 Cancri “d” (for which our best-guess next opportunity to observe a transit center occurs, curiously enough, a week after the start of the next long count).
We’ve found that transitsearch.org works best in a “campaign mode”, in which a critical mass of observers, distributed across the world, are marshalled to observe a single candidate star through a full transit window with as much redundancy as possible. One of my responsibilities is to pick good candidate stars, and then to get the campaigns rolling. In a series of posts, I’ll trace through how this candidate vetting and observing process works.
First, I go to the transitsearch.org candidates site, and start scrolling.
Whenever a new planetary system is announced (for example see this interesting chronology of extrasolar planet detections maintained by Dominique Naef), I take the planet’s published orbital elements and then enter them into the transitsearch.org orbit database. I also enter in basic information about the star (usually this is included in the discovery paper announcing the planet) including the estimated stellar radius, the temperature, and the mass. I also check to see whether the radial velocities that led to the detection are actually tabulated in the paper.
I’m always very happy whenever the full list of radial velocities on which a new planet detection is based is published, and over the past several years, this has increasingly become the de-facto procedure among all of the radial velocity survey teams. A list (through 2003) of which planet-bearing stars have radial velocities available, and which ones do not has been posted at exoplanets.org. Occasionally however, the radial velocities underlying published planetary fits are still published only in graphical form, and the actual tabulated radial velocity values are not released to the public domain.
The unavailability of the radial velocity data sets for some of the planet-bearing stars has led to something of a gray market industry in which the radial velocity plots of the parent stars of interesting multiple-planet systems such as HD 82943 and HD 202206 are digitized, and the radial velocities are reconstructed from the graphs. For an example of this technique, see this preprint on astro-ph.
Now it’s easy for me, Mr. Armchair Theorist, to advocate that the planet hunters publish the radial velocities supporting the planets they discover in tabular form. I can tune in to the alternate point of view, however. The radial velocity observers put a huge amount of effort into obtaining their data, and into refining their observational pipelines. Each velocity point in the actual data sets linked to the systemic console represents a night at the telescope. I have spent just enough nights actually at the telescope actually observing (about 20) to know how much work they signify. Furthermore, from a strictly economic perspective, radial velocities aren’t free. For example, according to a 2003 analysis by the National Optical Astronomical Observatory, the amortized cost of one night of observing on the 10-meter Keck telescope is $47,400. For stars in the V=6-8 magnitude range, it’s reasonable to obtain about 120 radial velocities measurements during one night at Keck (assuming that the template spectra have already been obtained). That means that each radial velocity obtained at Keck costs about $400. (Note that observing with Keck is at least an order of magnitude less expensive than observing with a facility such as HST). It makes some sense to let the observers keep control of their data-sets to prevent poaching, and to prevent opportunists from using tools such as the systemic console to find and publish improved, self-consistent fits to tricky systems such as 55 Cancri.
In any case, as I scroll down the target list in search of a transit candidate for an observational campaign, I look for a number of attributes. First, the prospective transiting planet needs to be adequately “in play”. That is, the star needs to have a window of phase coverage that has not been adequately checked for transits. In general, all of the planets with periods less than 10 days have been fairly well picked over. This is because the a-priori geometric transit probability increases dramatically as the planetary period decreases, and because it is much easier to definitively rule out or confirm transits for a short-period planet than for a longer-period planet. Exciting exceptions do, however, arise from time to time. For example, I realized one afternoon last summer that HD 188753A (P=3.35 d) had apparently not been checked for transits. Also, the red dwarf, GL 581, with a 5.3 day period has not yet been checked (and is hence currently being worked up for a transitsearch campaign).
For planets that are still in play, I look for stars that are visible during the time of year that the campaign will take place. I gauge this by looking at the Right Ascension Column. In general, for a January-March timeframe, I am looking for stars with right ascensions in the neighborhood between 4 hours and 14 hours. A good resource for evaluating the visibility of a particular star at a particular time, at a particular place is the Hourly Airmass Table by Brian Casey, which uses John Thorstensen’s Skycalc algorithm. Transitsearch.org also has a transit assignment algorithm, written by Aaron Wolf, in which you enter in your preferred observing time, and your location on Earth, and a rank-ordered list of candidate stars is returned. Stars with declinations above +20 degrees are tagged for Northern Hemisphere campaigns, whereas stars with declinations below -20 degrees are assigned to Southern Hemisphere campaigns. Everybody can participate for stars near the celestial equator.
I also keep an eye on the a-priori transit probability, and the expected transit depth. Stars with a less than 1% chance are currently not being given much attention, although this may change soon, as more observers begin to aquire high-quality data. Star-planet pairs with predicted transit depths greater than 0.5% are also preferred, although some transitsearch participants are now reliably pushing down to the 0.3% photometric depth level. In any event, for the purposes of this post, HD 99492 caught my attention. Here’s a Goddard Skyview image of the star:
HD 99492 is a K dwarf main sequence star, with roughly 80% the mass of the Sun. It’s a member of a wide binary pair, and is separated from its slightly brighter companion HD 99491 by about half an arc minute. The planet orbiting HD 99492 has a 17.04 day period, and a very low mass — just a bit more than 1/10th that of Jupiter. Our theoretical models predict that the planet will have a radius of about 0.7 Jupiter radii, leading to a readily detectable expected transit depth of 0.87%. The a-priori transit probability is a respectable 3.4%.
The presence of the binary companion may be a bit of a snag. If the two stars are not resolved on a CCD image, or if their individual images overlap, then the central transit depth will be reduced to ~0.3-0.4% of the combined light of the stars. If, however, an observer has a larger-aperture telescope and can cleanly separate the light from stars, then they will serve as an excellent comparison pair for differential photometry. (Check the comments field of this post to see observer updates on this issue as they come in.)
In order to make up-to-date transit predictions, it is best to fit the radial velocities. I used the console to reproduce the fit published in the discovery paper of Marcy et al (2005), and then entered the resulting parameters into my transit prediction code. This code generates a distribution of possible transit times using the so-called bootstrap monte-carlo method for analysis of uncertainties. A simple histogram of the first 100 predictions indicates that the (March 2006) transit window is several days wide, with the most probable time of central transit falling at about JD 2453803. A central transit is predicted to last about 4 hours, so clearly a worldwide campaign will be required.