Axial Symmetric Navier Stokes Equations and the Beltrami /anti Beltrami spectrum in view of Physics Informed Neural Networks

This paper establishes the theoretical framework for solving axial symmetric Navier-Stokes equations in a cylindrical topology by constructing a complete basis of harmonic 1-forms comprising Beltrami, anti-Beltrami, and closed components, thereby reducing the problem to a hierarchy of quadratic relations suitable for future optimization via Physics-Informed Neural Networks.

The Gist

Imagine you are trying to predict how water swirls inside a long, straight pipe. This is the classic problem of Fluid Dynamics, governed by the famous Navier-Stokes equations. These equations are notoriously difficult; they are the "Mount Everest" of math problems. Usually, scientists use supercomputers to chop the pipe into tiny pixels and simulate the water step-by-step. This works, but it's like trying to understand a symphony by listening to one note at a time—you get the data, but you miss the underlying melody or the "why" behind the chaos.

This paper, written by Professor Pietro Fré, proposes a different way to tackle this mountain. Instead of brute-force computing, he wants to build a mathematical Lego set that perfectly fits the shape of the pipe, and then use a special kind of "smart brain" (a Neural Network) to snap the pieces together to find the perfect solution.

Here is the breakdown of his idea using simple analogies:

1. The Pipe and the Symmetry

Imagine a long, cylindrical pipe. If you rotate the pipe around its center, it looks exactly the same. This is called Axial Symmetry.

• The Old Way: Scientists usually look at the pipe as a 3D box (like a cube), which is messy and complex.
• The New Way: Fré says, "Let's just look at the pipe as a cylinder." This simplifies the problem massively. Instead of tracking water in $x, y, z$ directions, we only need to track it moving out/in (radius) and up/down (length).

2. The "DNA" of the Flow: Beltrami Fields

The paper focuses on special types of water flows called Beltrami flows.

• The Analogy: Imagine a corkscrew or a helix. In a Beltrami flow, the water spins around its own axis while moving forward, like a giant, invisible tornado inside the pipe.
• The Magic: These flows are mathematically "perfect." They are the building blocks of turbulence. Fré discovered that inside this pipe, there are exactly six types of these building blocks for every energy level:
  1. Left-spinning (Beltrami)
  2. Right-spinning (Anti-Beltrami)
  3. Non-spinning (Closed forms - just flowing straight or in loops without twisting)
  • (And their mirror images).

Think of these six types as the primary colors of fluid motion. Any complex swirl in the pipe can be created by mixing these six "colors" in different amounts.

3. The "Diamond" Interaction

The hardest part of the Navier-Stokes equations is the non-linear part. This means that when two swirls meet, they don't just add up; they interact in a crazy, complex way to create new swirls.

• The Metaphor: Imagine you have two Lego bricks. If you just put them side-by-side, you have two bricks. But in fluid dynamics, if you smash two swirls together, they might snap together to form a completely new shape (a third brick).
• Fré calls this interaction the "Diamond Product." He has spent the paper calculating exactly how these six "primary color" bricks smash together to create new bricks. He created a massive "instruction manual" (tables in the appendix) that says: "If you smash a Left-Spinner with a Right-Spinner, you get a Non-Spinner."

4. The "Smart Brain" (Physics Informed Neural Network)

This is where the paper gets futuristic. Fré isn't just listing the bricks; he is proposing a way to find the exact recipe for a stable flow.

• The Problem: We know the bricks (the basis functions) and we know the rules for smashing them (the Diamond Product). But we don't know how much of each brick to use to make a perfect, steady flow.
• The Solution: He proposes using a Neural Network (a type of AI).
  • Instead of teaching the AI to guess, we give it the "Lego rules" (the math of the pipe and the Diamond Product).
  • The AI's job is to adjust the "amount" of each of the six bricks until the resulting flow satisfies the laws of physics perfectly.
  • It's like a chef trying to find the perfect recipe. The AI tastes the soup (checks the math), realizes it's too salty (the physics are off), and adjusts the ingredients (the coefficients) until the soup is perfect.

5. Why This Matters

Usually, when we simulate fluids, we get a number: "The water moves at 5 meters per second here."
Fré's approach aims for something deeper: Understanding the structure.

• By breaking the problem down into these specific "Beltrami" bricks, the AI might reveal hidden patterns.
• It might show us why turbulence starts.
• It might help us design better pipes, blood vessels, or even understand how the atmosphere moves, not just by calculating numbers, but by understanding the "geometry" of the flow.

Summary

Professor Fré has built a mathematical toolbox specifically for water flowing in a pipe. He has identified the six fundamental "notes" (Beltrami flows) that make up the music of the fluid. He has written down the rules for how these notes interact (the Diamond Product). Now, he plans to use an AI chef to mix these notes together to cook up the perfect, stable flow, hoping to uncover the secret recipe of turbulence that has baffled mathematicians for a century.

It's a shift from "calculating the chaos" to "understanding the harmony" hidden inside the chaos.

Technical Summary

Here is a detailed technical summary of the paper "Axial Symmetric Navier Stokes equations and the Beltrami/anti Beltrami spectrum in view of Physics Informed Neural Networks" by Pietro Fré.

1. Problem Statement

The paper addresses the challenge of finding exact or high-precision approximate solutions to the Navier-Stokes (NS) equations for incompressible, viscous fluids. Specifically, it focuses on axially symmetric flows confined within a cylindrical tube with periodic boundary conditions in the longitudinal direction.

• Context: Traditional Computational Fluid Dynamics (CFD) relies on discretizing differential operators (finite differences/volumes), which often obscures underlying analytical structures and symmetries.
• Goal: The author proposes a novel approach using Physics Informed Neural Networks (PINNs). However, instead of training a network to approximate the solution directly on a grid, the goal is to reduce the NS equations to an algebraic problem involving the coefficients of a specific functional basis. This allows for the discovery of hidden analytical rules and symmetries.
• Specific Focus: The study investigates the role of Beltrami flows (where velocity is parallel to vorticity, $\nabla \times \mathbf{u} = \lambda \mathbf{u}$) and anti-Beltrami flows in the onset of turbulence, extending previous work done on 3-tori ($T^3$) to a cylindrical topology.

2. Methodology

A. Geometric Reformulation

The paper reformulates the Navier-Stokes equations using differential geometry (exterior calculus):

• The velocity field $\mathbf{u}$ is treated as a 1-form $\Omega$.
• The NS equation is rewritten as:
  $$-d(p + \frac{1}{2}\|\mathbf{u}\|^2) = \partial_t \Omega + i_{\mathbf{u}} d\Omega - \nu \Delta_g \Omega - \mathbf{f}$$
• For steady flows with constant Bernoulli function, the equation reduces to a balance between the non-linear term ($i_{\mathbf{u}} d\Omega$) and the viscous term ($\nu \Delta_g \Omega$).

B. Construction of the Functional Basis

The core mathematical contribution is the construction of a complete orthogonal basis for the space of axially symmetric square-integrable vector fields ($L^2_{axial}$) on a compact tube.

• Topology: The domain is modeled as a cylinder with identified bases (topologically a portion of a torus), imposing periodicity in the $z$-direction and vanishing radial velocity at the boundary $r=1$.
• Eigenfunctions: The basis is constructed from harmonic 1-forms (eigenstates of the Laplace-Beltrami operator).
• Decomposition: For every energy shell defined by spectral indices $(n, k)$ (where $n$ relates to radial Bessel zeros and $k$ to longitudinal wavenumbers), the 6-dimensional eigenspace decomposes into:
  1. Two Beltrami modes (Left-handed rotation, eigenvalue $+\varpi$).
  2. Two Anti-Beltrami modes (Right-handed rotation, eigenvalue $-\varpi$).
  3. Two Closed (Irrotational) modes (Zero eigenvalue, gradient flows).
• Basis Functions: The radial dependence is expressed using Bessel functions ($J_0, J_1$) scaled by their zeros ($j_{1,n}$) to satisfy boundary conditions. The longitudinal dependence uses trigonometric functions ($\sin, \cos$).

C. The "Diamond Product"

The non-linear convective term in the NS equation ($\mathbf{u} \cdot \nabla \mathbf{u}$) is mapped to an algebraic operation called the diamond product ($\diamond$).

• This product represents the antisymmetric vector product generalized to the manifold.
• The product of two basis elements is expanded back into the same basis:
  $$U_{s_1} \diamond U_{s_2} = \sum_s C(s_1, s_2 | s) U_s$$
• The coefficients $C(s_1, s_2 | s)$ are derived analytically and depend on integrals of products of Bessel functions.

D. Reduction to Algebraic Recurrence

By expanding the velocity field $\mathbf{u} = \sum c_s U_s$ into this basis, the differential NS equation transforms into a hierarchy of quadratic algebraic equations for the coefficients $c_s$:
$$c[s] = \frac{1}{\nu \varpi(s)^2} \sum_{s_1, s_2} \varpi(s_2) c[s_1] c[s_2] C(s_1, s_2 | s)$$
This system is proposed to be solved via a recursive optimization algorithm (a specialized PINN) that minimizes a loss function based on the residual of these quadratic relations.

3. Key Contributions

1. New Functional Basis: The paper provides a complete, explicit basis of axially symmetric harmonic 1-forms (Beltrami, anti-Beltrami, and closed) on a cylindrical tube, utilizing a specific rescaling of Bessel functions to handle the boundary conditions rigorously.
2. Tripartite Beltrami Index: Unlike previous work on 3-tori where the Beltrami index was binary (chiral), the author demonstrates that in a cylindrical geometry, the spectral index is tripartite, distinguishing between:
   • Left-rotatory (Beltrami)
   • Right-rotatory (Anti-Beltrami)
   • Irrotational (Closed forms)
     This clarifies the physical mechanism where Beltrami and anti-Beltrami modes can coalesce into irrotational flows via non-linear interaction.
3. Algebraic Formulation of NS: The paper successfully reduces the complex PDEs of fluid dynamics to a set of quadratic recurrence relations on expansion coefficients. This shifts the problem from numerical PDE solving to an algebraic optimization problem.
4. Diamond Product Decomposition: The author explicitly calculates the structure constants (coefficients) for the diamond product of the basis elements, providing the necessary "interaction rules" for the proposed neural network algorithm.
5. Zero-Mode Analysis: A detailed analysis of zero-modes (where $k=0$ or $n=0$) is provided, identifying which modes are physically admissible under the boundary conditions (specifically, only certain helical modes survive).

4. Results

• Analytical Derivation: The paper derives the explicit forms of the 6 basis vectors for each spectral shell $(n,k)$ and the transformation matrix relating the general harmonic coefficients to the Beltrami/anti-Beltrami/closed components.
• Convergence of Products: Numerical tests in Appendix B demonstrate that the product of two basis functions (e.g., $G^{[n_1]}_0 \times G^{[n_2]}_1$) can be re-expanded into the same basis with impressive precision using only a finite number of terms (specifically, truncating at $n \approx n_1 + n_2$). This validates the feasibility of the algebraic approach.
• Streamline Visualization: The paper visualizes streamlines for single modes and superpositions, showing that while single modes form toroidal surfaces, superpositions create complex, deformed structures, illustrating the richness of the solution space even within the axially symmetric constraint.

5. Significance and Future Work

• Scientific Insight: The approach aims to move beyond "black box" numerical solutions to uncover the analytic structure of NS solutions. By analyzing the coefficients found by the optimization algorithm, the author hopes to deduce the rules governing exact solutions.
• PINN Strategy: This paper lays the theoretical groundwork for a future publication [10] where a Physics Informed Neural Network will be constructed. Unlike standard PINNs that differentiate the network output, this method uses the network to optimize the coefficients of a pre-defined, symmetry-adapted basis.
• Symmetry Breaking: The framework is designed to be extensible. Once axially symmetric solutions are found, the symmetry can be broken step-by-step by introducing perturbations organized into irreducible representations of the broken symmetry group ($SO(2)$), potentially leading to fully 3D turbulent solutions.
• Physical Relevance: The identification of the tripartite nature of the flow spectrum offers a new perspective on how energy transfers between chiral (rotational) and irrotational modes in confined flows, which is relevant for applications in pipelines, blood vessels, and industrial reactors.

In summary, this paper is a foundational theoretical work that bridges differential geometry, group theory, and fluid dynamics to create a new algebraic framework for solving the Navier-Stokes equations, specifically tailored for implementation with advanced neural network optimization techniques.