Composing a sample of stars

We built a complete sample of stars for which we could have detected a planetary transit. To reduce the number of targets, some assumptions were made. We cannot find planets around stars fainter than a critical V magnitude. The Poisson uncer- tainty in each brightness measurement would mask even the largest transit depths, and fainter stars are more difficult to follow up with radial velocity. We excluded the bright stars to avoid including saturated objects for which the lightcurve char- acteristics are atypical. The spectral type was restricted to only include FGK stars as these are the bulk of targets in the WASP archive. Likely giants were not studied as they would be rejected at the eyeballing stage.

V magnitude The faint V magnitude limit was set by considering the known

transiting planet sample. The target planets were only found around stars with V magnitudes brighter than 12.8 so we restricted our input catalogue to this range. This is advantageous also since there are exponentially more faint stars than bright stars, and we would potentially have wasted a lot of effort analysing stars for which the photon noise would be considerably higher. The saturation point for WASP is variable and sensitive to sky background and focus levels. We used the known planets sample again to restrict the magnitude range to stars fainter than 9.3.

Spectral type We also made a restriction on the J-H colour of the stars which is a

proxy for spectral type. The WASP project is sensitive to predominantly FGK stars. Later type stars tend to be too faint in the WASP bandpass. Earlier type stars are observed in large numbers but planets around these stars have shallower transits, and they are not well suited to radial velocity follow up as they are fast rotators. We used the J-H colour to infer spectral type, and used the known planets to restrict the range. No WASP planets have been found around stars with J-H colours outside the range of 0.17 ≤(J-H)≤0.49 which maps to the effective temperature range of

-0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 J-H -9 -4 1 6 11 R P M J 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Figure 3.1: Dwarf probability used in the classification of stars for the WASP re- duced stellar input catalogue.

5000 ≤ Tef f ≤ 6500, covering late K stars to early F stars (see Section 3.1.2 for an explanation of how the effective temperature was derived from J-H colour.) We excluded WASP-33 b from calculating this range as it orbits an A5 star [Collier Cameron et al., 2010].

Dwarf probability Another cut was made on the dwarf probability, calculated

empirically from proper motion information. Giant stars are a common contaminant for exoplanet surveys as they have similar photometric colours and brightnesses than main sequence stars, but are larger leading to shallower transits. They can be identified by their lack of proper motion as they are significantly more distant than dwarfs of the same magnitude so exhibit substantially smaller proper motions. We followed the method of Collier Cameron et al. [2007a] which is performed in the WASP analysis pipeline.

The classification was modified from Gould & Morgan [2003] but using the J-H colour index, and proper motions from the USNO-B1.0 catalogue [Monet et al., 2003]. A subsample of 2000 FGK stars with high quality high resolution spectra for which the surface gravity had been determined to better than±0.1 dex were used to calibrate the classification system, with dwarfs exhibiting logg≥4.0 and giants logg ≤ 3.0. The stars were cross-matched in the USNO-B1.0 catalogue and their J-H colour. The number density of dwarfs and giants were independently placed in reduced proper motion HJ versus J-H space. The fraction of dwarfs to total objects in each bin was calculated to give the dwarf probability, shown in Fig. 3.1. The known WASP exoplanets all have a dwarf probability of>98% suggesting that

Table 3.1: Parameter cuts used to restrict the WASP stellar sample. Parameter Range

V magnitude 12.8 - 9.3 J-H 0.16 - 0.49 Dwarf probability >98% Number of data points >1000

known planets are only detectable around very likely dwarfs. We excluded stars with a dwarf probability lower than this value.

Additionally to the cuts outlined above a cut was performed on the number of data points in the lightcurve. Objects with fewer than 1000 points are rejected by the selection cuts, and so we do not simulate these poorly sampled systems. These cuts are summarised in Table 3.1 and reduce the input stellar catalogue from around 37 million to 326620, a much smaller sample. Figure 3.2 shows the distributions of the parameters of the first 40 known planets used to reduce the stellar sample, along with the full stellar sample in the WASP catalogue. The J-H distribution peaks at around 0.25 indicating that the typical target stars are similar in effective temperature, and therefore spectral type. The underlying sample distribution shows that we are selecting the bluer stars preferentially and are not including some of the redder stars but these are likely to be much fainter in the WASP bandpass so will be rejected most likely through the V magnitude cut. Alternatively the red stars are distant giants and are rejected by the dwarf probability cut. The V magnitude distributions match well down to V = 12 but the full sample continues to around V = 13 when it starts to decrease, suggesting that the known planets are preferentially found around the brighter stars, as expected due to increased difficulty in detecting the transits, and radial velocity follow up constraints. The dwarf probability of the selected stars has a strong peak near 1.0. The full sample shows a strong peak with a low but flat distribution between 0 and 1, with a slight increase at 0 showing that the number of objects for which the probability is neither exactly 0 or exactly 1 is low. We are deliberately rejecting giant stars for our sample, and the lower cut of 98% rejects uncertain objects.