Aug 022014

By Peter McCullough, Associate Astronomer at STScI

“The third time’s the charm” applies to this story.

Some planets pass in front of the stars they orbit. We call that a “transit.” One of the brightest stars with a transiting planet is HD 189733. That makes it attractive for scientific study, because the more photons one collects, the more precise a measurement can be made. The planet that orbits it is also a large one, a gas giant planet like Jupiter, similar in size and mass, but much hotter because the planet that orbits HD 189733 does so with a period of only 2 days! I am eight thousand “years” old in HD 189733b “years”!

That star also happens to have a lot of prominent star spots; from the rotational modulation of its light curve, with an amplitude of ~2% peak to valley, HD 189733 is “spottier” than 95% of all stars of similar spectral type. We know the statistics of stars in general from the Kepler mission.

Since it was first discovered to have a transiting planet approximately a decade ago, the planet HD 189733b has been the subject of scores of observational campaigns. The depth of the transit tells us the square of the ratio of the planet’s radius to the star’s radius. However, the radius of the planet’s silhouette in front of the star depends on wavelength because the grazing rays of starlight passing by the planet pass through its upper atmosphere and hence are affected differently according to the composition of the planet’s atmosphere. Hence, by taking spectra of the star-planet system during transits, we can measure the composition of the planet’s atmosphere. In this case we were searching for water, and our “divining rod” was the WFC3 instrument on Hubble. We were not sure we’d detect it – prior attempts had failed: with the older NICMOS instrument the results were plagued by systematic effects, and even two recent attempts with the new WFC3 instrument had failed. The first attempt failed because the bright star saturated the detector, and second attempt failed because of a miscommunication about a software update that caused the planet to be observed when the Earth itself was between Hubble and the star – blocking out the scene at the critical hour when the planet was transiting the star. We hoped that this third attempt “would be the charm”!

It was. We obtained excellent WFC3 data on June 5, of 2013 and found a clear signature of water vapor at 1.4 microns in the spectrum taken during transit that was not there before or after the transit. In a recent press release, dated July 24, 2014, my colleague N. Madhusudhan describes this measurement, and two others, as the first measurements of a chemical compound in exoplanets. To be sure, although we and others have detected water vapor before in exoplanets, those detections were tentative enough that the water concentration could not be quantified much better than to say that it was greater than zero.

For planet HD 189733b, there were some that expected we would not see the signature of water vapor. While they expected that gaseous water ought to be present, they imagined that it would be obscured from our view by a haze layer, much like the buildings of Los Angeles or Beijing can be obscured by smog. This expectation was based on prior Hubble observations made with the instruments STIS and ACS.

The working hypothesis, developed over the past few years by Frederic Pont and his colleagues, to explain the ensemble of data from the Hubble and Spitzer space telescopes has been that a haze layer in the upper atmosphere of the planet Rayleigh-scatters light, creating a circular silhouette of the planet that is largest in the ultraviolet and smoothly decreases ever so slightly with wavelength into the near-infrared until it bottoms out and is nearly flat in the thermal infrared. Figure 1 comes from our ApJ paper (McCullough et al. 2014, 791, 55) . One way to shoe-horn the new detection of water vapor into what had become the consensus model is to presume that the water vapor occurs at higher altitudes than the haze layer. Although that is still plausible, in investigating the various models I became less and less confident that the consensus was correct. Another model may apply equally well or better.



Figure 1. Transmission spectrum of the exoplanet HD 189733b. Observations: The upper WFC3 spectrum from our analysis (black open circles) is as-observed; the mean transit depth for the WFC3 data is approximately at the same level as the ∼1 μm end of the ACS data (blue open circles) reported by Pont et al. (2008), which had been corrected for an assumed unocculted star spot level of 1%. The lower WFC3 spectrum (black filled circles) has been shifted down by 300 ppm to better match the ACS data (blue filled circles) corrected for an unocculted star spot level of 1.7% by Pont et al. (2013). Models: One model (upper, red line) combines two effects: 1) unocculted star spots with temperature T(spot) = 3700 K and spot fractional area δ = 0.056, and 2) a clear planetary atmosphere of solar composition, a mixing ratio for water of 5×10-4, and zero alkali metal lines (Na and K) for a gas giant planet with physical parameters commensurate with HD 189733b. The other model (lower, orange line) is solely the contribution of the unocculted star spots of the first model. Both models have been smoothed with a Gaussian of FWHM=0.089 μm for clarity.

The core of my idea was expressed also by Pont and his colleagues before our WFC3 observations. They had noted that uncertainty about spots on the face of the star could create uncertainty in the slope in the spectrum from the UV through the visible to the near-infrared, nominally attributed to Rayleigh-scattering in the planet’s atmosphere. Recall that the measurement, namely the transit depth, depends on the square of the ratio of the planet’s radius to the radius of the star. Obviously, either the planet or the star can affect the ratio. If the star has spots off the transit chord, i.e. they are on the star but never occulted by the planet, those spots can mimic the “Rayleigh-scattering slope.” Here’s why. The spots are cooler and redder than the stellar photosphere. Such a spot will cause the transit depth to be deeper than it would be otherwise, which we might misinterpret as the planet’s radius being larger.  And because cooler spots are also redder than the photosphere, the spot is relatively darker in the blue than the red. Thus, we might interpret the observed fact that the transit is deeper in the blue than the red, as either (A) the planet being larger in the blue (i.e. Rayleigh scattering) or (B) the star having unocculted spots that have not been fully accounted for. Pont’s team was aware of all that, but their best estimate of the number and temperature of spots seemed too small to them to account for all of the spectrum’s slope, which they therefore attributed to model A. When we examined the same data, and added the new WFC3 data showing a water-vapor feature in the near infrared, we re-interpreted the combination of old and new data with model B.

The overall lesson from this story is that even for one of the best objects to study, namely one of the closest transiting exoplanets to Earth, and a large gas-giant planet at that, the interpretation of the observations is still unsettled. That is unsatisfying, challenging, and fun, all at the same time. So it is with science. Hopefully a more thorough interpretation of existing data and/or even better data obtained with future instruments or telescopes such as JWST will solve this puzzle.