Now that we have collected 1/2 of the observations with HARPS, LCOGT.net, and ASH2, let us share a glimpse of how the HARPS Doppler data looks. As seen in the article ‘The signal’, we expect some variability of a ‘few’ meters per second, but not larger than 4–5 m/s, otherwise the UVES survey would have spotted it. Unfortunately, we cannot disclose the real data to avoid biasing the future revision of the manuscript. Instead, we present you with a few examples of Doppler measurements similar to those we are obtaining on Proxima Centauri. These are simulated datasets using the same HARPS observation dates as in the Pale Red Dot Campaign, and they reproduce the three most likely outcomes of these first thirty measurements. But there is a twist! One of the six data sets actually corresponds to the ‘real’ observations… can you guess which one it is?
Case 1 – Radial velocity variability dominated by random noise
The left panels in Figure 1 show two examples of typical datasets that only contain random noise. The vertical small lines on each point are called error bars, and illustrate how uncertain each measurement is (~1 m/s). Note that depending on weather conditions some measurements have larger error bars. The central panel shows a graphic called a periodogram. Periodograms tells us which are the most significant possible periods in the data and allows us to quantify whether or not a signal is strong enough to be detected. In this example we set the detection threshold at a false alarm probability of 0.1%. That is, peaks over the blue dashed line would correspond to signals with false alarm probabilities smaller than 0.1%. Neither of these two datasets reveal a significant signal.
Case 2 – Hints of a signal, but corrupted by activity
Here we show two datasets that contain a possible Doppler signal, but these have been corrupted by non-periodic stellar activity. As in case 1, neither of the sets is sufficient to confirm a signal. Accumulation of data over the campaign should boost true planet candidates above the detection threshold, while pushing down the significance of the spurious ones. In cases like these, we would try to model stellar activity using photometry and other spectroscopic measurements to see if part of the variability could be explained by stellar noise. As an example of techniques used to achieve this, see the interview to Prof. Suzanne Aigrain.
Case 3 – A signal is well detected despite stellar activity
In this case we have two simulated data-sets with bona fide planet signals that clearly dominate over the noise (Figure 3). This would be the best case scenario for the Pale Red Dot campaign. Still we would need to investigate the photometry and other activity indices for activity related variability.
Thanks for following!