One of the most important aspects of the tracking system are the extremely high resolution wafer-thin silicon disks that offer a wide range of energies that particles can be detected at. The production of these disks is a very intensive and expensive process, and the budget will dictate what kind of precision can be offered through these trackers. These disks are constructed with many small modules that can be oriented and designed for maximal efficiency and reliability. The most preferable choice would be a disk constructed of 1 by 1 modules, allowing for the most area to be covered and therefore the most precision attained. Of course, this is also the most expensive option, and is likely not possible. Other configurations involve modules that are 1 by 3 or 2 by 6, which all offer varying degrees of expense and reliability. In order to attain the most funding possible, I have to create a compelling case for a certain type of module that can balance cost and reliability. This is where machine learning comes in, allowing me to test my hypotheses for different design ideas. By the end of my time at MARC, I've developed a model that is approaching an optimal design for these silicon trackers that will eventually be incorporated into the final design plans for the ALICE 3 upgrade.

Final Product

At the end of my time at MARC, I was able to create a model that is approaching the ability to accurately create sensor layouts. My project had a very steep learning curve that unfortunately left me with insufficient time to create a useable model, but I don't intend to leave my research here. I have plans to continue work over the summer on my model, further refining its design.

To read more about my model in its current state, you can read my paper: Identifying the optimal layout of the ALICE ITS3 using machine learning