Exterior complex scaling enables physics-informed neural networks for quantum scattering

summary

The gist

The gist The exterior complex scaling enables physics-informed neural networks for nuclear reactions.

In short

The Exterior Complex Scaling (ECS) method allows physics-informed neural networks (PINNs) to solve complex quantum scattering problems for nuclear reactions. ECS transforms outgoing waves into exponentially decaying functions, making them suitable for neural networks. This enables PINNs to accurately predict phase shifts in nucleon-nucleus and heavy-ion scattering by minimizing the differential equation residual.

Key concepts

Exterior Complex Scaling (ECS): ECS is a mathematical transformation that rotates spatial coordinates into the complex plane beyond a certain radius. This rotation changes outgoing waves, which normally oscillate, into exponentially decaying waves. This modification allows the problem to be solved using neural networks because the resulting functions become square-integrable and decay at large distances.
Physics-Informed Neural Networks (PINNs): PINNs are machine learning models trained to solve differential equations. In this context, a PINN is used to learn the wave function of nuclear scattering by minimizing the residual of the Schrödinger equation. The network learns the solution while simultaneously ensuring it adheres to the physical laws described by that equation.
Differential Equation Residual: The differential equation residual is a measure of how well a proposed function (the neural network's output) satisfies the governing physics, which is the radial Schrödinger equation for nuclear scattering. The PINN minimizes this residual to find the wave function that accurately describes the physical system.
Partial Waves: In quantum scattering, partial waves are distinct solutions corresponding to different orbital angular momenta ($ ext{l}$). These waves describe how two colliding nuclei scatter based on their specific rotational states. The method validates its accuracy across 21 spin-orbit channels in nucleon-nucleus scattering and 41 partial waves in heavy-ion scattering.

Terminology used across episodes

Reinforcement Learning (RL) — explained in 25 episodes
Retrieval-Augmented Generation (RAG) — explained in 16 episodes
Interpretability — explained in 15 episodes
Sample Complexity — explained in 15 episodes
Federated Learning — explained in 14 episodes
Generalization — explained in 13 episodes

This episode discusses

The paper

Exterior complex scaling enables physics-informed neural networks for quantum scattering · Read on arXiv

School of Physics Science and Engineering, Tongji University · Southern Center for Nuclear-Science Theory (SCNT), Institute of Modern Physics, Chinese Academy of Sciences

Physics-informed neural networks (PINNs) have emerged as a powerful tool for solving differential equations, yet their application to nuclear scattering has been hindered by the oscillatory, non-decaying nature of scattering wave functions. In this work, I demonstrate that exterior complex scaling (ECS) transforms scattering boundary conditions into exponentially decaying waves suitable for neural network solutions, enabling PINNs to solve nuclear reaction problems for the first time. I develop a driven-equation formulation where the source term is confined to the real axis, avoiding the need to analytically continue nuclear potentials into the complex plane. The method is validated on nucleon-nucleus scattering (n+ 40 Ca at E lab=20 MeV) with 21 partial waves, achieving phase shift accuracy of Δδ 0.1 for the strongly absorbed channels (at most 4) and Δδ at most 0.60 for all channels up to = 10, when compared to conventional solvers. I further demonstrate the approach on heavy-ion scattering (6 Li+ 208 Pb at 40 MeV) with 41 partial waves and strong Coulomb effects, where an auto-adaptive anchor warm-down for weak-source channels yields a mean S-matrix accuracy of ΔS about 3 times 10-3 across the full angular momentum range, including the absorption-to-transparency transition region. This work establishes the foundation for extending PINNs to inverse problems where end-to-end differentiability enables direct fitting of optical potential parameters, coupled-channel reactions, and few-body scattering where traditional grid methods face exponential scaling.

Transcript

Introduction to the show: ident: AI Radio. Generated commentary on the latest Artificial Intelligence papers.

Tom: Today's paper: "Exterior complex scaling enables physics-informed neural networks for quantum scattering".

Jane: The gist The exterior complex scaling enables physics-informed neural networks for nuclear reactions.

Tom: First, who's behind it and why it matters.

Paper summary: Tom: So, we're talking about this paper called "Exterior complex scaling enables physics-informed neural networks for quantum scattering." Essentially, they’re tackling a big problem in nuclear physics where standard methods struggle because the wave functions are oscillatory and don't decay nicely.

Jane: It claims that by using exterior complex scaling, which transforms those messy boundary conditions into waves that decay exponentially, you can finally use physics-informed neural networks to solve these kinds of reaction problems for the first time.

Lu: What’s really interesting here is how they set up the mathematical problem; they developed a driven-equation formulation where the source term stays on the real axis, which avoids having to analytically continue those nuclear potentials into complex space.

Meng: That sounds like a huge step because traditional methods often require you to deal with those tricky complex continuations of potentials, which can be really messy computationally.

Lalam: From what I’m seeing in the model's understanding, the core idea is that this exterior complex scaling acts as a translator, taking the problem from oscillatory behavior into something squareintegrable that a neural network can actually handle.

Conclusion: Tom: So, looking at "Exterior complex scaling enables physics-informed neural networks for quantum scattering," the main idea is that this new technique lets us use these powerful AI tools to solve nuclear reactions in a way we couldn't before because of how the waves behave.

Jane: They achieved this by showing that rotating the spatial coordinates into the complex plane turns those outgoing waves, which are normally wiggly, into exponentially damped functions. This decay means you can truncate the computational domain and still get a good answer for where it’s going at large distances.

Lu: The authors validated this on nucleon-nucleus scattering, specifically with n+40Ca at an energy of twenty MeV, testing it across twenty-one partial waves to get phase shift accuracy of less than zero point one degrees for the strongly absorbed channels where is four or less <ref:2602.04553#pg1>.

Meng: That level of accuracy on those channels is actually pretty impressive when you think about how sensitive these potentials are to small changes in the physics input.

Lalam: It means that if we can apply this concept broadly, it could help us fit optical potential parameters directly to experimental scattering data, which is a really practical application for inverse problems.

More episodes

← Home