Methods

The development of our visualization tool involved a comprehensive approach, integrating both physical and computational materials. An essential goal was to create visualizations that not only facilitate scientific research for experts but also effectively communicate scientific concepts to the broader public. To achieve both of these objectives, the methods employed can be categorized into two main approaches: illustrative and computational. By combining the strengths of both approaches, we successfully established an intuitive and coherent visual framework that effectively merges the power of illustrative representations with the precision of computational visualization.

Illustrative Methods

Prior to computational work, initial illustrations were pivotal in creating the stellar evolution visualization tool. They depicted evolution stages and events while offering a flexible foundation for updates. Crafted with varying pencil hardness, archival ink pens, and color pencils in a muted palette, each detail was meticulously attended to. Spheres used compasses, and curves employed a flexible curve ruler. High-quality scans used 250 GSM white Stonehenge paper.

Adobe Illustrator converted scans to vector images, ensuring resolution quality and customization. Standardized dimensions guaranteed smooth computation for the final tool. Incorporating the essence of traditional visualizations like Van Den Heuvel diagrams, we aimed to develop a more comprehensive and detailed visual tool that is fit for POSYDON data. The layout facilitates a portrayal of binary star evolution, ensuring that researchers can easily grasp the relationships and transitions between different evolutionary stages in the evolutionary narrative. Stars are color-coded by temperature in the binary system, and their sizes are relative. We designed a minimalist and representative motif for accreting matter, accommodating the in-spiral of such matter for stages such as Roche lobe overflow (RLOF). A 3-dimensional perspective was applied to the Roche lobe, aiding in its dimensional understanding, and enabling future visualization coding. However, the spheres are kept nearly 2 dimensional to maintain the balance of informative but minimalist motifs present throughout the visualization. Simplified potential diagrams (Tauris & van den Heuvel 2023), inspired by those in The Physics of Binary Star Evolution, were chosen for clarity, as the equipotential surfaces present in each event stage acted as a visual link between the RLOF event with its physical effect in the gravitational field of each star. The layout of the visualization adds visual dimensionality to the algorithmic diagram, Fig. 1 (Fragos et al. 2023). The Illustration of this tool, however, just represents a singular end result. BinaryStory must be able to be auto-generated with respect to any population run of the POSYDON code. Therefore, it adopts a computational contribution that allows the outcome of the visualization to be based on the POSYDON output data.

Computational Methods

After creating the visualizations, the next step was developing an interactive version using the POSYDON code. Leveraging Python packages such as Matplotlib, PIL, Cv2, we created an integrated Van Den Heuvel diagram gen- erator. First, a BPS model run, with specific keyword arguments, and a large population yielded a large Pandas DataFrame (DF). The function getstar returns one star from the population and displays its important information, such as state and event names, step names, mass, state, separation, etc., for the following visualization functions. Having established a DF, MiniVizView, with relevant stellar evolution information, we assign a function named getimage which receives the DF and extracts pertinent information for constructing a Van Den Heuvel diagram. Each chartobject acts as an image in the visualization and requires an axis and an image as input.

The function vdh is the integral function that created the Van Den Heuvel Diagrams. It iterates over columns, searches for corresponding visualizations in a database based on a specific file path that is stitched together by the output of the getimage function, and places connected chart objects in chronological order. During this process, it uses PIL to turn the image into a numpy array of pixels with RGB values.

A list of effective temperatures of the stars at each stage is passed to a function that uses a database of corresponding Hexadecimal (hex) and RGB codes based on spectral type and temperature (Harre & Heller 2021). This is then used to color each star in the diagram. The visualizations, thus, hold the temperature and color information of the stars in the binary system. The function utilizes connectionpatches as the connecting pathways between each stage/event. The function accounts for null results and provides researchers with appropriate messages to notify if a visualization doesn’t exist or isn’t found. At each time step, points are plotted on a vertical line adjacent to the chart, introducing a temporal dimension to the visualization. The vdh function facilitates the creation of a well-organized timeline, documenting the stellar evolution of each binary system based solely on the output data from POSYDON.