An outline of the paper: A Spitzer Transmission Spectrum for the Exoplanet GJ 436b, Evidence for Stellar Variability, and Constraints on Dayside Flux Variations by Knutson et al.

My last post described the comprehensive study that Eric Agol and collaborators had made of the Spitzer/IRAC data set that exists for the transiting hot Jupiter HD 189733 b. In this paper by Heather Knutson and collaborators (including many from the HD189733b paper) a similarly thorough treatment is applied to the Spitzer/IRAC data available for the transiting hot Neptune GJ 436 b. Specifically, the authors analyse a total of 8 primary transit measurements made in the 3.6, 4.5 and 8.0 micron IRAC channels, as well as 11 secondary eclipse measurements in the 8.0 micron channel. The secondary eclipses allow some interesting constraints to be placed on the brightness variability and dayside temperature of GJ 436 b, but for the rest of this post I’ll be focusing on the primary transit measurements only (Figure 1).

Figure 1. Values that Knutson et al derive for the depths of the primary transits. Different symbols indicate the IRAC channel that was used to make the measurement, with stars corresponding to 3.6 microns, triangles corresponding to 4.5 microns, and circles corresponding to 8.0 microns. Adapted from Knutson et al (2011).

And this is where things start to get a little messy. For starters, the two 3.6 micron transit depths differ by 4.7σ! The two 4.5 micron transits aren’t all that much better, differing by 2.9σ, and the observed scatter in the 8.0 micron transits is certainly not negligible. Basically, transit depths measured for a given wavelength seem to change depending on when you’re looking. Could this be because the atmosphere actually undergoes major changes between observations? Could, for instance, the planet be getting puffed up or contracting between observations? The authors consider this latter possibility, but using a quick back-of-the-envelope calculation are able to show it’s highly unlikely, given that the amount of energy required would be a factor of $10^5$ larger than the total amount of stellar radiation incident upon the atmosphere over the entire intervening period.

Another explanation could be that a large layer of opaque haze or cloud developed at the limb of the planet between observations, causing the apparent size of the disc to increase. Witnessing time-variable atmospheric phenomena such as this is an exciting prospect for the future, but the limited data we currently have available is certainly not enough to claim the detection of such an event. Accordingly, Knutson et al note that although this intriguing scenario cannot be totally ruled out, there are more mundane solutions to the problem that should be favoured for the time being.

In particular, the authors argue that increased stellar activity, in the form of star spots and faculae, are somehow implicated in the puzzling transit depth variations. For instance, the passage of the planet across dark spotty regions would tend to reduce the measured transit depth, whereas if the planet crossed a region of bright faculae, the observed transit would be deepened. Active regions that the planet does not pass over could also influence the measured transit depth, with unocculted spots biasing the measured transit depth to deeper values, and unocculted faculae again acting in the opposite direction.

There are a number of reasons to suspect that such effects are at play for this particular data set. To start with, the disagreement between transit depths measured at different times gets worse as you move to shorter wavelengths, which is qualitatively what you would expect if stellar activity is to blame. This is because the brightness contrast between active regions and the ambient photosphere gets stronger as you go to shorter wavelengths, thus amplifying the shift in the measured transit depth.

Furthermore, long-term photometric monitoring of GJ 436 from the ground seems to show that the overall brightness varies with a period of around 57 days and an amplitude of a few millimag. It just so happens that the three deepest transit depths measured in 2009 occurred close to the minimum of one of these brightness cycles, which is presumably when the star is most active and there is a relatively large number of dark star spots covering the stellar surface. However, the brightness variations of the star are too small for unocculted spots to explain the larger depths that are measured for these particular transits, so the other obvious option is for occulted faculae to be responsible the deeper transits. This seems a little counterintuitive to me at first: it would seem to imply that the planet is by-and-large passing across brighter regions of the stellar disc during these transits, despite the star being near the low-point of its brightness cycle. But there’s no reason to suspect that there wouldn’t be faculae present, even if there are more dark star spots overall. Indeed, the authors argue that the in-transit points exhibit a higher degree of scatter than the out-of-transit points for these observations, which is what you would expect if the planet was passing across regions of uneven brightness.

Now, given these difficulties, is it possible to put together a transmission spectrum for GJ436b? Of course, it is, but the extent to which we can trust it is the real question. Figure 2 shows transit depth measurements across a range of wavelengths, including the Spitzer points of the current study in the near infrared. For each of the IRAC channels, the transits that Knutson et al believe are most contaminated by stellar activity are indicated by the open red circles. When these points are ignored, they find that an atmosphere depleted in CH4 relative to H2O and CO by a factor of $\sim 10^{-3}$ best explains the remaining data, with the model fit shown by the green line. This is consistent with the findings from the dayside emission spectrum presented by Stevenson et al. (2010).

Figure 2. Red circles show the transit depths from Knutson et al. (2011) in the 3.0, 4.5 and 8.0 micron Spitzer/IRAC channels, along with published transit depths at other wavelengths. The open red circles indicate the transits that Knutson et al consider less reliable. A methane-poor model, shown by the green line, provides the best match to the data when these points are excluded from the modelling. The grey line shows a methane-poor model with an opaque cloud deck at a pressure of 50mbar, which primarily has the effect of improving the match with the data between 1-2 microns. Meanwhile, when the deeper 3.6 and 8.0 micron transits are included, the data is more consistent with a methane-rich model, shown by the blue line, which is what Beaulieu et al found. Green, blue and grey circles show the models integrated over the each of the corresponding passbands. Adapted from Knutson et al. (2011).

However, it differs significantly from the picture painted by Beaulieu et al. (2011), who, using the same data set as Knutson et al, proposed that the transmission spectrum showed evidence of enhanced methane absorption. The reason for this disagreement is that Beaulieu et al included the deeper 3.6 and 8.0 micron transit measurements that Knutson et al discarded, arguing that those points were less likely to be affected by residual instrumental systematic effects which they considered to have a more pronounced effect than stellar activity. When these points are included in the analysis, the best fit atmospheric model becomes something like the blue line shown in Figure 2, where the broad absorption feature between 3.6-8.0 microns is caused by methane.

Regardless of who, if anyone, is correct, these two studies (Knutson et al. and Beaulieu et al.) highlight what a tricky business it is to use a few multi-colour transit depths to constrain chemical species in the atmospheres of faraway planets such as GJ 436 b. In the face of confounding factors such as stellar activity and instrumental systematics, the conclusions reached can be highly sensitive to the most basic assumptions made about the scarce data that is available, such as which points are reliable and which are not.

Alonso et al, 2008, 487, L5

Ballard et al, 2010, ApJ, 716, 1047

Cáceres et al, 2009, A&A, 507, 481

Pont et al, 2009, MNRAS, 393, L6

Stevenson et al, 2010, Nature, 464, 1161

Feature image: NASA/ESA

Share.