The nuclear many-body problem 🤯

A nucleus:

AXNZ

N= Neutron number
Z= Proton number
A=N+Z

We would like to get a precise structure of the ground state and low lying energy states for a set of nuclei. This will allow us to get the form factors for neutrino interactions when the momentum transfer is low. In particular, we are interested in:

If I get into the nuclear deexcitation model we may need to look at others, but that would be way into the future.

There are mainly three (?) ways that this problem is approached.

I have found many of these approaches only do the computations for even-even nuclei or just for magic number (2, 8, 20, 28... p or n) nuclei. Even if they do do it for general nuclei, they just give results for certain properties like the different excitation energies, which is not enough. We need the actual wavefunctions to extract the form factors.

In any case, for our purposes we should use some form of HF-BCS/Bogoliubov with QRPA or an ab-initio method. The rest simply can't deal with the discrete states correctly.

Tensor networks

As far as I can tell, it would be more interesting to use tensor networks for the ab initio method, as

  1. Mean field theory seems less restricted by computational resources, and more by theoretical uncertainties in the models.
  2. Writing the many-body state in terms of a tensor network seems quite natural (see the references) and can reduce the computations by orders of magnitude (you know, MPS and that stuff).
  3. This doesn't mean we won't need to deal with HF or its variants: it seems that an initial single-particle basis is needed in some cases.

Some references I found

  1. Low-rank representations of nuclear interactions. Talks about matrix decomposition in momentum space and "advanced tensor formats" for nuclear interactions (Presentation).
  2. This one is pretty crazy but take a look (Presentation).

This is where the fun begins 🔥
From here on is what I think we should use. But the previous two references are still worth checking out for context.

  1. The nuclear many-body problem for large, many-shell nuclei: An exact solution using tensor networks. Here they use the density-matrix renormalization group (DMRG) to solve the pairing Hamiltonian exactly in the language of matrix-product states (MPS). They use a 1D chain. This one is pretty interesting. Q: is the pairing force they use in this paper only between proton-proton and neutron-neutron?.

    "DMRG approach does not neglect any diagrams and does not rely on particle-hole excitation cutoffs as, e.g., in the configuration interaction approach".

  2. Symmetry reduction of tensor networks in many-body theory. Paper 3 is ended by saying that including "shell model arbitrary terms" to increase the accuracy of the calculations is unfeasible, but that one avenue to do so is to exploit the rotational SU(2) symmetry of the problem. This paper deals with this.
    I'm still not sure whether we can use SU(2) symmetry in nuclei with open shells (like ours), though. Maybe axial symmetry?

  3. A problem cited in Paper 3 with this type of symmetry reduction is that, while it shrinks the Matrix Product Operator (MPO) bond dimension, it increases the size of the local basis (D). They suggest the need to optimise the choice of the local basis, something covered in this paper. Didn't follow it very well but the final results seem amazing. They also use DMRG and MPS! Also this paper (still haven't read it in detail), which is much more recent.

  4. Who knows? I think we have more than enough to start with this.