Understanding the physics of stellar environments is critical to unraveling how the Universe evolves as a whole.  The exotic environs of core-collapse supernovae (CCSN, the explosive death of a supermassive star) are candidate sites for the creation of the heavy elements, which seed life on Earth - and whose origin is unknown.  In seeking a means to access these distant environs, we focus on the information contained in Earth-based measurements of neutrinos.  Neutrinos are elementary particles generated in CCSN in copious numbers.  Typically they interact weakly with other matter particles, but a CCSN is sufficiently dense that neutrinos profoundly impact this "matter profile" within the cloud of expanding gas.  We ask: what signature may Earth-based neutrino observations contain about the CCSN cloud with which they interacted prior to arriving here?  This cloud of matter seeds interstellar space with the products of nucleosynthesis, and neutrino "flavor" - a term that defines the way neutrinos interact with the cloud as they stream outward through it - profoundly influences what those seeds will be.

In nuclear astrophysics, the traditional approach to tackling such problems is forward integration.  While powerful, this formulation renders aspects of the physics difficult to access.  Inference is an alternative methodology, related to machine learning.  In the geosciences and neurobiology, inference has illuminated problems akin to those noted to hinder progress within astrophysics. The potential for inference to illuminate these problems is high, and thus this project explores inference to bear upon them.  Innovations cultivated within one scientific arena can be transformative when applied to disjoint fields.

One difficulty with forward integration is its required knowledge of "initial conditions": the initial state of a system.  In practice, initial conditions are poorly constrained.  Thus that a priori assumption may artificially hide important physics. This is particularly true for the case of direction-changing backscattering, which neutrinos can undergo.  Neutrino flavor in large part sets the neutron-to-proton ratio as well as energy and entropy deposition, thereby shaping nucleosynthesis, and backscattering can further affect the process.  But backscattering creates a two-point boundary-value problem - where "initial conditions" are not defined.  On the other hand, inference does not require known initial conditions.  This project investigates the capabilities of statistical data assimilation (SDA), a Bayesian inference methodology invented for numerical weather prediction, well-suited for boundary-value problems, and expected to outperform integration in efficiency.  The underlying ambition of this project is to hone SDA for application to interpreting a real galactic CCSN signal.  They occur every 50-100 years in our galaxy, the last in 1987.  So we are due. 

Broader Impacts are threefold.  First, the research involves the training of undergraduates from a primarily-undergraduate institution, with student demographics and socio-economic backgrounds not well represented in STEM fields.  Second, we extend our work with SDA to neurobiology.  This will inform theory in that field, and those advances feed back to the astrophysics work.  In addition, the neurobiology community is our chief means to stay abreast of new developments in SDA methodology.  Third, we harness tools from storytelling and comedy to teach communication skills to young scientists.

Outcomes:

1. Astrophysics: In a simple model of back-scattering, SDA outperforms forward integration in that it does not require known initial conditions.  Thus inference can solve a problem that is hidden from integration techniques, and one that represents an important feature of flavor evolution in CCSN (https://doi.org/10.1103/PhysRevD.105.083012)
2. Astrophysics: In a small-scale CCSN model, we can use simulations of Earth-based neutrino flavor measurements to infer whether oscillations in neutrino flavor  occurred within the CCSN envelope.  If they occur, such oscillations could affect nucleosynthesis.  The implication for a real neutrino detection from a CCSN is that it may show whether such behavior occurred (https://doi.org/10.1103/PhysRevD.105.103003).
3. Astrophysics: SDA is employed for the first time using real neutrino measurements from the Sun, to obtain independent estimates of solar properties (https://doi.org/10.1103/PhysRevD.107.023013,    https://doi.org/10.1103/PhysRevD.110.043011)
4. Broader Impacts: We can use measurable quantities to infer properties of biological neuronal networks, as well as features of a population model of COVID-19 (https://doi.org/10.1007/s11538-023-01176-x)
5. Broader Impacts: Since fall 2022, PI Armstrong has led workshops in comedy and storytelling to teach communication skills to young scientists, for two demographics.  One is the undergraduate population at New York Tech, with an annual performance at the Symposium for University Research and Creative Expression.  The other is the astrophysics community across New York City: NY Tech, CUNY, Columbia, NYU, and the Flatiron Institute, with performances at the American Museum of Natural History.

Last Modified: 09/10/2024
Modified by: Eve Armstrong