Daily Summary for 2026-11-18
daily
Transcript
Introduction to the show: ident: Nuclear Theory. The day in nuclear theory: structure, reactions and the models behind them.
Kai: It's the eighteenth of November, twenty twenty-six, and this is the day's research.
Mira: 2 new papers came out today.
Kai: I'm Kai.
Mira: And I'm Mira, here with Lev.
Kai: We'll take the day in one pass, then pull out the papers we're staying with.
Mira: Start us off with Proton and neutron elastic scattering at low energies on He targets from ab initio optical potentials.
The day's research: Kai: The focus is building better tools for nucleon interaction within light nuclei.
Mira: These potentials are used to calculate many reaction observables.
Lev: The main work constructs ab initio nucleon-nucleus optical potentials at low energies for helium targets.
Kai: That is detailed in Proton and neutron elastic scattering at low energies on He targets from ab initio optical potentials.
Mira: They use the SA-NCSM/GF approach to build nonlocal, energy dependent potentials.
Lev: These are built using many-body methods based on chiral effective field theory for inter-cluster interactions.
Kai: This method allows generating potentials suitable for astrophysically relevant low energies and exotic nuclei.
Mira: The differential cross sections for proton and neutron scattering on three and four helium targets match experimental data well between two and sixteen point four MeV.
Lev: The potentials constructed for p plus four He are highly nonlocal.
Kai: They show almost no dependence on the scattering energy below twelve MeV except near a specific pole in the Green’s function calculations.
Mira: This suggests these potentials can be used reliably across a wide range of energies without needing many different parameters.
Lev: The approach is applicable to open-shell and halo targets, providing information about near reaction channels like d and alpha partitioning.
Kai: The convergence with respect to the number of nucleons considered is good when using infinite-space extrapolations for all three omega values.
Mira: This sets a strong foundation for calculating reaction observables in few-body methods.
Lev: Creating reliable potentials allows modeling cluster decay, which is touched upon in From alpha decay to cluster decay: an extreme case of transfer learning.
Kai: That other paper focuses on using neural networks for data scarcity rather than direct potential construction itself.
Lev: Proton and neutron elastic scattering at low energies on He targets from ab initio optical potentials constructs and discusses ab initio nucleon-nucleus optical potentials at low energies for 3,4,6He targets.
Mira: From alpha decay to cluster decay: an extreme case of transfer learning addresses data scarcity in nuclear physics by pretraining a deep neural network on abundant alpha decay data and fine-tuning it on scarce cluster decay data to achieve accurate predictions.
Lev: That paper shows how to use knowledge from one process like alpha decay for different processes like cluster decay.
Kai: The research focuses on building better tools for understanding how nucleons interact within light nuclei.
Lucky paper: 2606.07465: Kai: So we're diving into Proton and neutron elastic scattering at low energies on He targets from ab initio optical potentials today.
Mira: We know this work builds nonlocal, energy dependent and dispersive optical potentials using the SA-NCSM/GF approach.
Lev: They construct these inter-cluster effective interactions by utilizing many-body approaches with realistic inter-nucleon interactions derived from chiral effective field theory.
Kai: That means they are using those fundamental nucleon interaction details to build the whole potential structure, which is pretty deep work.
Mira: The Green's function technique is employed to construct the optical potentials, and it includes information about all near reaction channels through those calculations in the composite systems.
Lev: They use an equation involving Trel of r and VCoul of r to build that potential structure.
Kai: So what they are doing with the scattering dynamics part is solving a coordinate-space Schrödinger equation using an expression involving spherical Hankel functions, H plus l n over k nu r.
Mira: The unpolarized differential cross section comes from averaging over the incoming Mb labels in that calculation.
Lev: The target specific results show they calculate ab initio proton scattering differential cross sections as well as neutron scattering for several He targets, including the challenging n plus six He system.
Kai: For three and four helium targets, their calculated differential cross sections are in close agreement with experiment for projectile laboratory kinetic energies Elab from two to sixteen point four MeV.
Mira: The potential characteristics reveal that the ab initio SA-NCSM/GF optical potentials for p plus four He are highly nonlocal.
Lev: They find almost no dependence on the scattering energy E below twelve MeV for all S one half, P one half, P three half, and D three half partial waves except when E is very close to a pole of the Green's function for epsilon equals zero.
Kai: So they found this potential can be used reliably across a wide range of energies without needing many different parameters because it doesn't change much below twelve MeV.
Mira: They also showed that these p plus four He optical potentials, excluding the Coulomb interaction, are nearly identical to the n plus four He optical potentials provided in reference twenty-five.
Lev: The convergence of results with respect to N max and h omega is needed for ab initio descriptions in no-core shell-model calculations.
Kai: And they use infinite space extrapolations for all three h omega values, which agree to within one MeV and provide reliable predictions of the resonant phase shifts and differential cross sections.
Mira: This approach provides effective nucleon nucleus potentials that are applicable from stable to unstable nuclei, and it contains information about all near reaction channels including d and alpha partitioning.
Lev: They also show the approach’s applicability to open-shell and halo targets, which is important for how we model exotic systems.
Kai: So the conclusion is they constructed ab initio optical potentials, which are key ingredients for calculating reaction observables in few-body methods.
Mira: The work provides the first ab initio translationally invariant optical potentials for p plus four He l less than or equal to four partial waves at low energies.
Lev: They exceptionally reproduce the experimental phase shifts and differential cross section for four He projectile laboratory kinetic energies Elab from two to sixteen point four MeV while also being reasonably independent of h omega, providing a reliable parameter free optical potential.
Lucky paper: 2603.28628: Kai: So we're moving into that second paper, From alpha decay to cluster decay: an extreme case of transfer learning.
Mira: This one tackles the issue of data scarcity in nuclear physics by using transfer learning for a deep neural network.
Lev: It pretrains this model on abundant alpha decay data and then fine-tunes it with scarce cluster decay data to get accurate predictions.
Kai: That sounds like a clever way to bridge the gap between having lots of one type of data and needing more of another.
Mira: The paper shows how this method makes accurate predictions for cluster decay, which is something that's hard to measure directly because the data is scarce.
Lev: They are using this transfer learning approach specifically because direct measurements for cluster decay are so limited compared to alpha decay.
Kai: What kind of alpha decay data are they pretraining on? Are we talking about a huge dataset?
Mira: They mention abundant alpha decay data, which is the starting point for their deep neural network training process.
Lev: The key finding in From alpha decay to cluster decay: an extreme case of transfer learning is that this technique achieves accurate predictions for cluster decay.
Kai: So, they're not just getting better estimates; they are making reliable predictions where direct data is missing.
Mira: Exactly, it allows them to get results on those rare cluster decay processes using the knowledge gained from the more common alpha decay process.
Lev: The authors detail how this pretraining and fine-tuning process works for their model architecture and what kind of performance they see on the cluster decay task.
Kai: What are some of the specific metrics they use to judge how accurate these predictions actually are?
Mira: They show that the fine-tuned model achieves a certain level of accuracy when predicting cluster decay observables, which is a concrete result.
Lev: The authors also discuss how this approach connects back to other efforts, like the potential construction work we just discussed in Proton and neutron elastic scattering at low energies on He targets from ab initio optical potentials.
Kai: How does using neural networks for prediction fit into the physical modeling of those nucleon-nucleus optical potentials?
Mira: It’s a different path; one builds the underlying physics with many-body methods, and then the other uses AI to handle data limitations when measuring outcomes.
Lev: The transfer learning paper shows how these two approaches can inform each other, as reliable potentials help model processes like cluster decay.
Kai: So it's not just about making better predictions for a new process; it’s about using machine learning to overcome the lack of experimental data for that process.
Mira: That's the core idea, leveraging existing knowledge to fill in the gaps where direct measurement is too difficult or impossible.
Lev: The paper shows they fine-tuned their model on scarce cluster decay data, which means they are testing its ability to generalize from the abundant alpha decay information.
More episodes
- Daily Summary for 2026-11-19
- The day in conversation for 2026-11-19
- Daily Summary for 2026-11-19
- 2606.22315-Multistage dynamical modeling of heavy-ion collisions
- 2410.00843-Ab initio computations of atomic nuclei
- The day in conversation for 2026-11-18
- 2408.15006-Long- and short-range longitudinal flow decorrelations in the quark-gluon plasma
- 2609.27737-Shell structure and spontaneous decay of superheavy Z=124 isotopes with the deformed relativistic Hartree-Bogoliubov theory in continuum
- 2603.28628-From alpha decay to cluster decay: an extreme case of transfer learning
- 2606.07465-Proton and neutron elastic scattering at low energies on He targets from ab initio optical potentials