Plots as Art

In the daily course of research I will often generate tens to hundreds of plots, depending on the project and how automated the task is. Naturally most of these hit the "cutting room floor", or are simply sanity checks in real time, never meant to be seen by others.

Sometimes these figures are too cool to just discard, and it occurred to me that they might even qualify as "art". So I present to you: three artistic looking figures that were too cool to throw out, and a short explanation of what each was. Each figure is a real byproduct of research, but I have taken some liberties with the color schemes.

1. Wavelet Tree

While looking at time series data from the Kepler mission, I started to mess about with the continuous wavelet transform. This was intended to search for changing starspot periods over time. The result was not terribly exciting, since the starspot in question is seemingly fixed on the star's surface (detailed results still in the works). The x-axis in this figure is time, the y-axis is frequency. The wavelet shows the power from each frequency in the light curve as a function of time. The real information I wanted is in the top 1/2 of the figure, and needed to be expanded with a logarithmic axis.

UPDATE: here is the actual science-grade plot for the wavelet, which is arguably just as artistic!

2. City Crashing

I've been working hard on a paper researching an eclipsing binary system for about a year. Part of the analysis required me to run some simple simulations using Python. Unfortunately my PySkills were pretty weak at the time, and it kept crashing!  I wanted to know if the code crashed after a certain number of iterations, so I had it spit out the number of each iteration before it crashed to a file. I ran the script a couple dozen times, and had a file that contained a list of increasing integers that (sort of) looked like:

1,2,3,4,5,6,7
1,2,3,4,5
1,2,3,4,5,6,7,8,9,10,11...

and so on. If the code was crashing at the same point, it would run to the same number each time. Instead it varied quite a bit, from only 5 to almost 50 iterations. The problem turned out not to be in my Python code, but rather a more subtle and easily fixed issue of file I/O with multithreaded processes. The x-axis here is just the entry number in the output file, and the y-axis is the simulation iteration for the respective run.

3. Velocity Owl

Another artistic byproduct of the eclipsing binary project was "The Owl", probably my favorite decorative plot yet (a version hangs in my office). I wrote my own version of a 2 dimensional cross correlation code (aka TODCOR in Astronomy). This method takes the spectrum of a double-lined eclipsing binary and compares two template spectra. Both templates are scanned back and forth in radial velocity, until the best fit velocities for both objects are found. In this visualization of my TODCOR code, the x-axis is the velocity separation between the two templates, and the y-axis is the velocity shift of the two templates to the system as a whole. It's the funny definition of the x-axis that leads to the "mirrored" look of the figure, otherwise it would look more like a "figure 8" at a 45deg angle.

The colors represent the quality of the fit, but using the ridiculous "hardcandy" color scheme. The two "eyes" are the regions of best fit velocities. Since the templates are quite similar, these represent the two solutions where v1 and v2 are swapped. The left "eye" is the better fit, and thus provides the velocities of both stellar components.