🔬 physics

A Local Gauge-Covariant Formulation of Classical Dynamics

This paper proposes a local gauge-covariant framework for classical dynamics where physical laws emerge from the asynchronous relaxation of local incompatibilities between state variables and transport geometry, recovering familiar continuum equations without assuming predefined evolution laws or an external time parameter.

Original authors: Gunjan Auti, Hirofumi Daiguji, Gouhei Tanaka

Published 2026-04-17

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Gunjan Auti, Hirofumi Daiguji, Gouhei Tanaka

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Big Idea: Fixing a Broken Map While Driving

Imagine you are driving a car through a foggy, unfamiliar city. In traditional physics, the map is perfect, the roads are fixed, and you have a master clock that tells everyone exactly what time it is. You just follow the rules: "If you are at this spot, turn left." The map never changes, and the rules are written in stone before you even start driving.

This paper proposes a different way of thinking.

Imagine instead that you don't have a map, and you don't have a master clock. You only have a car (the state) and a vague sense of where your neighbors are (the transport geometry). You can only talk to the people in the car next to you.

The core idea of this paper is: Dynamics (movement and change) happens because things don't quite match up.

The Problem: "Incompatibility"

Let's say you and your neighbor are both holding a piece of paper with a drawing on it.

You draw a red circle.
Your neighbor draws a blue square.

In a perfect world, if you both agree on how to rotate and translate your papers (the transport relation), your drawings should match perfectly when you compare them. But right now, they don't. There is a mismatch or incompatibility.

In this paper, the authors say: This mismatch is the engine of the universe.

How the System Works (The "Fix-It" Process)

The paper suggests that nature doesn't follow a pre-written script. Instead, it follows a simple, local rule: "Fix the mismatch."

Here is the step-by-step process, using our car analogy:

The Glitch (Incompatibility): You look at your neighbor's drawing. It doesn't line up with yours. There is a "glitch" in the connection between you two.
The Reaction (Relaxation): To fix this glitch, two things happen simultaneously:
- You change your drawing (State): You might erase the red circle and draw a blue square to match them.
- You change the road (Geometry): Maybe the road between you was too bumpy or twisted. You smooth out the road so that when you look at their drawing, it naturally aligns with yours.
The Result: You both adjust until the mismatch disappears.

The Magic: The paper shows that if you do this "fix the mismatch" thing everywhere, all at once, and locally (only talking to your neighbors), the complex laws of physics (like how water flows or how electricity moves) emerge automatically. You didn't need to program "water flow" into the system; it just happened because everyone was trying to make their local drawings match.

Key Concepts Explained Simply

1. No Master Clock (Asynchrony)

In normal physics, we assume time flows like a river, ticking the same for everyone.

This Paper: Imagine a room full of people fixing a giant puzzle. No one has a watch. Person A fixes a piece, then Person B fixes a piece, then Person C. They don't wait for a signal. They just fix the piece next to them whenever they feel like it.
The Result: Even though everyone is working at their own speed (asynchronously), if you zoom out and look at the whole room, it looks like a smooth, continuous movie. The "global time" we feel is just an illusion created by watching many tiny, local fixes happen very fast.

2. The Map is Alive (Dynamic Geometry)

Usually, we think space is a fixed stage where the play happens.

This Paper: The stage itself is part of the play. If the actors (the state) can't agree on where they are, the stage (the geometry/roads) changes shape to help them agree.
Analogy: Imagine a dance floor made of rubber. If the dancers are out of sync, the floor stretches and twists to help them get back in rhythm. The floor isn't just a background; it's a participant.

3. Learning Without a Teacher (Adaptive Plasticity)

In machine learning, we usually teach a computer by giving it a "correct answer" (a teacher).

This Paper: The system teaches itself. It only knows when it is "wrong" because its neighbor says, "Hey, that doesn't match!"
The Rule: If there is a mismatch, the system changes its internal connections (roads) and its state (drawings) to reduce the mismatch. It's like a self-correcting network that builds its own rules as it goes.

Why Does This Matter?

The authors show that when you run this "fix the mismatch" simulation, it naturally creates the famous equations of physics:

Diffusion: How ink spreads in water.
Fluid Flow: How water moves in a pipe (Navier-Stokes).
Electromagnetism: How electric and magnetic fields interact (Maxwell's equations).

The Takeaway:
The universe might not be a machine running on a pre-set program. Instead, it might be a giant, self-organizing network where everything is constantly trying to get along with its neighbors.

Classical Physics: "Here are the rules. Follow them."
This Paper: "Here is a mismatch. Fix it."

By simply trying to resolve local disagreements between neighbors, the complex, beautiful laws of the universe emerge on their own. The "laws" aren't the starting point; they are the result of the system finally getting its act together.

1. Problem Statement

Traditional classical dynamics relies on closed evolution equations defined on fixed geometric backgrounds with a global time parameter. In standard formulations (e.g., Lagrangian mechanics or standard continuum equations):

The spatial structure (metric, connection) is prescribed a priori.
A global clock synchronizes the evolution of all parts of the system.
The rules for comparing neighboring states (transport relations) are fixed and independent of the state itself.

This approach struggles with far-from-equilibrium and multiscale systems (e.g., active matter, under-resolved flows, electrochemical systems) where:

The effective transport relations are not known in advance.
Interactions depend on local history or unresolved processes.
Information propagates at finite speeds, making instantaneous global coordination physically unrealistic.

The authors ask: Can dynamical laws emerge from local relational constraints without assuming a predefined action principle, fixed geometry, or global time?

2. Methodology

The authors propose a constructive, axiomatic framework where dynamics arises from the restoration of local gauge consistency. The methodology proceeds through the following steps:

A. Fundamental Axioms

The framework is built on eight axioms that reject predefined evolution laws:

State is Primitive: Systems are described by local densities of conserved quantities ( $U$ ).
Locality: Interactions occur only between immediate neighbors.
Manifold vs. Geometry: A smooth manifold exists, but the transport geometry (connection) is not fixed; it is a dynamical variable.
Local Frame Freedom (Gauge Covariance): Local states are defined only up to an arbitrary internal frame ( $g_i \in G$ ). Physical descriptions must be invariant under independent local frame transformations.
Coercivity: Incompatibility has a non-negative cost.
Finite-Rate Adaptation: Transport geometry evolves over a finite timescale; it does not adjust instantaneously.
Local Linearizability: Transport operators can be locally represented in the Lie algebra (tangent space).
Normability: State spaces admit norms to measure mismatch.

B. Core Concepts

Transport Operators ( $T_{ij}$ ): Maps that transport the state of a neighbor $U_j$ into the local frame of node $i$ . These are treated as dynamical variables that co-evolve with the state.
Local Incompatibility ( $r_{ij}$ ): Defined as the covariant difference between a state and its transported neighbor:
$r_{ij} \propto T_{ij}U_j - U_i$
If $r_{ij} = 0$ , the descriptions are compatible. If not, a "mismatch" exists.
Residual ( $R_i$ ): The accumulation of interfacial incompatibilities at a node, representing the local drive for change.
Global Incompatibility Measure ( $L$ ): A scalar functional quantifying the total system mismatch. Under the axioms, the simplest admissible form is a quadratic measure:
$L[U, W] = \frac{1}{2}\sum_i \langle R_i, R_i \rangle + \frac{\lambda}{2}\sum_{(i,j)} \langle W_{ij}, W_{ij} \rangle$
where $W_{ij}$ represents the transport geometry (via $T_{ij} \approx e^{W_{ij}}$ ) and $\lambda$ is a structural stiffness parameter.

C. Dynamics as Relaxation

Instead of postulating an evolution equation $\partial_t U = F(U)$ , the authors define dynamics as the gradient descent (steepest descent) of the incompatibility measure $L$ :

State Relaxation: The state $U_i$ evolves to minimize local residuals.
Structure Relaxation: The transport geometry $W_{ij}$ $W_{ij}$ evolves to minimize the mismatch between neighbors.
- The update rule for $W_{ij}$ resembles a Hebbian learning rule (correlation of error and activity) but is derived purely from geometric consistency, not supervised learning.
- Far from compatibility, the system "learns" transport pathways. Near compatibility, the structure decays to a trivial state.

3. Key Contributions

Dynamical Transport Geometry: The paper treats the rules of comparison (transport relations) as independent dynamical variables that co-evolve with the physical state, rather than fixed background structures.
Emergence of Time: Global time is not a primitive assumption. It emerges only as a coarse-grained effective limit of asynchronous local relaxation events.
Derivation of Classical Laws: The authors demonstrate that familiar continuum equations (Diffusion, Incompressible Navier-Stokes, Ampère-Maxwell) are limiting cases of this local relaxation process when transport relations stabilize.
Gauge-Covariant Mismatch: They prove that locality and gauge covariance strictly restrict the form of admissible incompatibility measures to a simple, globally additive, quadratic form.
Adaptive Plasticity: The framework provides a mechanism for "adaptive plasticity" where the system's internal connectivity and constraints restructure themselves in response to persistent local mismatch, without external rewards or loss functions.

4. Results

Mathematical Derivation: The authors derived explicit update rules for both state ( $U$ $U$ ) and structure ( $W$ $W$ ) based on the variation of $L$ $L$ .
- $\Delta W_{ij} \propto -(\text{Residual}_i \times \text{Activity}_j + \text{Decay})$
- $\Delta U_i \propto (\text{Local Restoration}) - (\text{Neighbor Back-propagation})$
Limiting Cases: In the Supplementary Information (referenced in the text), they show that:
- Diffusion arises when transport geometry is fixed and mismatch relaxes.
- Navier-Stokes emerges when the incompressibility constraint is enforced via the relaxation of velocity field incompatibility.
- Electromagnetism (Ampère-Maxwell) arises from the relaxation of gauge-incompatibility in a vector field.
Geometric Interpretation: Nonzero incompatibility is interpreted as curvature (holonomy) on the interaction graph. The dynamics drive the system toward a flat (integrable) geometry.

5. Significance

Foundational Shift: This work challenges the standard paradigm of physics where laws are postulated. Instead, it suggests that laws emerge from the requirement of local consistency in a relational universe.
Non-Equilibrium Physics: It offers a robust framework for modeling systems where the "rules of interaction" are unknown or evolving, such as active matter, biological development, and complex fluids.
Machine Learning & AI: The framework provides a "bottom-up" approach to world modeling. Instead of training a global predictor, an agent could learn a gauge-covariant relational geometry by minimizing local incompatibility. This aligns with recent interests in "World Models" and offers a physical basis for adaptive plasticity and open-ended learning without predefined objectives.
Unification: It unifies concepts from gauge theory, non-equilibrium thermodynamics, and geometric deep learning into a single constructive framework where time, geometry, and dynamics are co-emergent.

In summary, the paper posits that dynamics is the process of restoring local compatibility. By treating geometry as a dynamical variable subject to finite-rate adaptation, the authors derive classical physical laws as effective descriptions of a system striving for local gauge consistency.