This week we look at some of the progress on our analysis of variability. This involves what we call variability fingerprints (see week 214 post). These are plots made from the light curves that allow us to find out the probability that a star varies by a given amount in brightness after a given amount of time. There are some further details on these plots and their reliability in the post from week 275.

We have now taken all these plots for a set of variable young stars in HOYS and run them through a clustering algorithm. Specifically for this example we have used k-means clustering, and principle component analysis for dimension reduction (check out the Wikipedia links for details! 🙂 ). On top of this we tried to see how the fingerprints for the young stars compare to a simple sine-wave like light curve. Such a light curve obviously does not look like one of the young stars, so one expects these to not coincide with the real data. But you better check if this is true 😉 .

In the plot above in blue are the positions from the clustering for the real young star light curves. There is no real physical meaning of the axis shown. But we can see that the young stars spread out in the parameter space. We are now investigating if any of the object properties (such as age, mass, disk mass) are related to the position in the diagram. We then repeated this clustering multiple times by adding an additional fingerprint into the group that was obtained by a sine-wave of 0.5mag amplitude and 1, 2, and 4yr periods, observed with randomly selected HOYS like observing times (cadences).

The mean position in the plot for each period and the typical scatter of these positions are shown as coloured points and shaded areas in the respective colours indicated in the legend. Indeed these non-realistic (as young star light curves) do not overlap with the real data. Though some individual objects are close to them. Some further tests are now to be done to quantify this in more detail, but this is a promising start. We ultimately aim to find simple to describe models of the small scale structure in the disks that fall within the real data to determine how the disks are structured on very small scales, that cannot be directly observed directly or interferometry.