We recently published a paper to the Monthly Notices of the Royal Astronomical Society (Irwin et al., 2020. MNRAS, 493, 106 - 125. doi: https://doi.org/10.1093/mnras/staa238) presenting a reanalysis of phase curve observations of WASP-43b using a novel 2.5-D retrieval model. The ArXiv version of this paper is here .
In this paper we attempt to fit the temperature profile and abundance at all longititudes, altitides and latitudes simultaneously, making the assumption that properties vary with latitude as (cos(latitude))^n, where n is an assumed coefficient. We call this approach a 2.5-D retrieval.
Using a model based on a GCM we simulate what a planet such as WASP-43b will look like, here in the Spitzer/IRAC channel centred at 4.5 microns. It can be seen that the temperatures vary greatly across the planet. We then perform a full disc ingration across our synthetic planet at all 15 HST/WFC3 and 2 Spitzer/IRAC channels at the fifteen different phases reported by Stevenson et al. (2017) to produce phase curves that approximately match observations. We then add on noise and and fit the synthetic phase curves with our NEMESIS radiative transfer and retrieval model. Having fitted the synthetic observations, we then recomputed how the planet would actually look at 4.5 microns.
Here are the results using a simple 1-D model. Here we assume that at each phase the temperatures and abundances are the same at all points on the disc. While this is easy to model, it does not recognise that in fact temperatures vary across the disc. Hence in this sort of retrieval the dayside temperatures are cooler than they should really be and the nightside temperatures are warmer.
However, when we use our 2.5-D model, with n=0.25, we simulate this appearance. It can be seen here that the hot parts of the planet are more tightly constrained to the dayside. In addition the the magnitude of the temperature variations is closer to the simulation. The north-south variation in our model is contained in the n-coefficient.
If instead we set n=1.0 we get the following simultion. This simulation looks less realistic and actually doesn't fit the synthetic observations as well.
It can be seen that the 2.5-D model approach more realistically simulates the appearance of our model planet and recovers more consistent temperatures. We then applied our model to the observed phase curves of Stevenson et al. (2017).
Here are the results of the simulated appearance of WASP-43b using our 1-D, 2.5-D (n=0.25) and 2.5-D (n=1.0) models:
It is clear that the nightside of the real WASP-43b is much colder than the GCM-based simulation. Such a stark temperature difference is difficult to explain with dynamical models, but one explanation is that the nightside may be covered with a cloud that lowers the radiation to space. To model this we repeated our retrievals, but set the nightside temperatures to be very cold (500K).>
Here are the results using a 2.5-D model, with n=0.25, nightside set to 500K:
Here are the results using a 2.5-D model, with n=1.0, nightside set to 500K:
Our 2.5-D model, with n=0.25 and a dark nightside provides the best fit to the observations.