Exploring the application of neural networks in predictions of nuclear binding energies

In this project the locations of the proton and neutron drip-lines are predicted using neural networks and theoretical data obtained by applying the HFBTHO program. For each of the neural networks a comparison is made between neural network predictions and experimental data in the region experimental data exists. By comparing the effectiveness of the networks at reproducing experimental results with the effectiveness of the HFBTHO program it is found that extensive improvements can be made these results. This indicates that the application of machine learning exists as a potential method for making corrections to theoretical modes. Whether the final predictions are sufficiently trustworthy to reach a conclusion is difficult to determine, however this seems to be a potential path for future development into obtaining data.What are the heaviest and lightest possible nuclei for each element? And why is it important to answer this question? An atomic element in nature consists of a nucleus (made of neutrons and protons) which is surrounded by electrons. An element is identified by the number of its protons. For example hydrogen has one proton in the nucleus, helium has 2 and so on. On the other hand, it is the number of neutrons in the nucleus that determines the isotope of the element, where isotope is a term referring to a specific neutron and proton number configuration. Many isotopes have the right balance of protons to neutrons and thus their nuclei are stable. However, many other isotopes have too many or too few neutrons per proton, with the result that these isotopes are unstable. Nevertheless, an important question which arises is: “how unstable can you get”? Differently stated, “how heavy could the nucleus of an element possibly become”? These questions are important as their answers would help when planning experiments by indicating where those experiments might be impossible. Further experimentation into the relatively unique properties of nuclei close to the limit of stability could then reveal critical information for guiding and improving theoretical models which could in turn be applied throughout nuclear physics from nuclear reactors to nuclear security, or nuclear medicine. Additionally experimentation and increased understanding of elements in this region is also crucial for a proper understanding of how matter around us came to be. The vast majority of all elements heavier than Iron are produced from ether supernovae referring to the explosion of a star due to runaway nuclear reactions or kilonovae where two neutron stars violently merge. In both cases a ridiculous quantity of neutrons are expelled into the surroundings, slower neutrons are hit by faster ones and neutron clumps form. Through nuclear processes neutrons become protons and along with the continuous barrage of new neutrons the heavy elements such as Uranium and Gold that exist in nature are produced. In order to properly understand the processes the elements undergo and thereby better understand the origins of heavy elements we need knowledge concerning how elements filled with mostly neutrons behave, one of these behaviours is the neutron drip-line which is the technical name for the before mentioned limit of nuclear stability. Investigating the neutron drip-line means investigating the properties of nuclei far from where we have experimental results. This means we will be relying on theory which is a problem as experiments produce far more trustworthy results and we don't know if the theoretical models are accounting for everything. My solution to this is to see if we can't teach a computer to improve the results. There are multiple options for applying machine learning to produce results and my work applies a form of deep learning called a neural network. This name comes from the fact that it uses a bunch of nodes connected together sort of like how neurons are connected in a brain. The neural network is trained to make the results obtained from theory closer to the more trustworthy experimental values. The final goal being to make new predictions for the location of the limit of nuclear stability