Extremum-Conservation-Symmetry (ECS) Universal Unification Theory

Structure-Preserving Extension of Discrete Systems to Continuous Limits

苏杭 Zhang, Bosley Zhang
(Luoyang)

 

Abstract

This paper rigorously addresses the limit transition problem from discrete to continuous time within the established axiomatic framework of ECS discrete systems. First, we present the precise mathematical definition of an ECS discrete system: state evolution is uniquely determined by the MIE (Maximum Information Efficiency) variational criterion, which automatically generates quadratic global conserved quantities and preserves the MOC (Multi-Origin Curvature) discrete symmetry group. We prove that as the time step h \to 0 (with fixed total time), the solutions to the discrete difference equations satisfying the ECS axioms converge to those of a system of continuous differential equations. This limiting system preserves the conserved quantities (constant values), symmetry groups (expanded into continuous Lie groups), and the MIE extremum criterion (as the Euler-Lagrange equation of a variational problem) of the original discrete system. Furthermore, under the ergodicity assumption, we demonstrate the unity of the discrete large-number limit and the continuous ergodic limit. Thereby, we achieve a rigorous, structure-preserving extension of the ECS paradigm from discrete to continuous time domains, laying a theoretical foundation for subsequent applications in geometric field theory and statistical physics.

Keywords: ECS Universal Theory; Structure-Preserving Extension to Limits; Maximum Information Efficiency (MIE); Multi-Origin Curvature (MOC); Discrete-Continuum Unity; Ergodic Limit

 

1. Introduction

The ECS theory is centered on the three axioms of Extremum (E), Conservation (C), and Symmetry (S), and has yielded consistent results in discrete system analysis. However, any theory aiming to describe physical spacetime must accommodate a continuous time manifold. The core problem addressed in this paper is: Given a discrete-time system satisfying the ECS axioms, can it be uniquely embedded into a continuous-time system in a structure-preserving manner? Does this embedding preserve conserved quantities, symmetry groups, and the extremum criterion?

In contrast to previous work that treated "structure-preserving extension" as an axiom, this paper formulates it as a mathematical proposition. We prove that for a broad class of ECS discrete systems (linear dynamics + quadratic conserved quantities + finite-group symmetry), this extension not only exists but also yields a unique limiting system that fully inherits the core structure of the discrete system. The key tools for the proof are Γ-convergence in function spaces and weak convergence of stochastic processes.

The remainder of the paper is organized as follows. Section 2 provides a rigorous axiomatic definition and a concrete model for ECS discrete systems. Section 3 defines the limiting extension map and states the structure-preservation theorem. Section 4 proves the existence and uniqueness of the discrete-to-continuous limit. Section 5 addresses the unity of probabilistic large-number limits. Section 6 concludes the paper.

 

2. Rigorous Formalization of ECS Discrete Systems

2.1 State Space and Axioms

Let the discrete time set be \mathbb{T}_h = \{0, h, 2h, \dots, Nh\}, with fixed total time T = Nh and step size h > 0. The state space is \mathbb{R}^d, equipped with the action of the MOC symmetry group.

Axiom 1 (Extremum Axiom, MIE)

There exists an information efficiency functional dependent on the system state x and control input u:

\mathcal{I}(x,u) = x^\top Q x + u^\top R u

where Q \succ 0 and R \succ 0 are weight matrices. The evolution of the discrete system selects the control that maximizes the information efficiency at the current step:

u_n = \arg\min_{u} \mathcal{I}(x_n,u) \quad \text{subject to} \quad x_{n+1} = A x_n + B u_n

The unique solution to this minimization problem yields the state transition:

x_{n+1} = L_{\text{ECS}} x_n, \quad L_{\text{ECS}} = A - B (R + B^\top P B)^{-1} B^\top P A

where P is the solution to the discrete algebraic Riccati equation — a standard result from linear-quadratic regulator theory, embodying the precise mathematical formulation of MIE.

Axiom 2 (Conservation Axiom)

The above MIE-optimal evolution automatically generates a quadratic conserved quantity:

C(x_n) = x_n^\top \Sigma x_n \equiv \text{constant}, \quad \forall n

where \Sigma is a positive definite matrix satisfying \Sigma = L_{\text{ECS}}^\top \Sigma L_{\text{ECS}}. For systems subject to stochastic perturbations, the conserved quantity holds in expectation.

Axiom 3 (Symmetry Axiom, MOC)

There exists a finite group G_h \subset O(d) acting on the state space such that:

- L_{\text{ECS}} commutes with G_h: g L_{\text{ECS}} = L_{\text{ECS}} g for all g \in G_h;
- The conservation matrix \Sigma is invariant under G_h: g^\top \Sigma g = \Sigma.

The group G_h is termed the Multi-Origin Curvature Symmetry Group, reflecting the intrinsic geometric constraints of the state space.

Remark: The above axioms are non-vacuous — they precisely characterize the structure of linear-quadratic optimal control systems, and the discrete algebraic Riccati equation guarantees that L_{\text{ECS}} has a spectral radius less than 1 (stable system).

2.2 ECS Discrete Systems with Stochastic Inputs

To address large-number limits, we introduce stochastic perturbations. The state equation becomes:

x_{n+1} = L_{\text{ECS}} x_n + \xi_n

where \{\xi_n\} are independent and identically distributed (i.i.d.) random variables with \mathbb{E}[\xi_n] = 0, \mathbb{E}[\xi_n \xi_n^\top] = \Xi, and independent of the initial state. Since MIE optimality is embedded in L_{\text{ECS}}, the control input is absorbed as optimal feedback, and perturbations can be regarded as external noise. This system is uniformly asymptotically stable (spectral radius < 1), and the quadratic conserved quantity C(x_n) is strictly constant in the noise-free case and constant in expectation for noisy systems.

 

3. Continuous Limit Extension: Definitions and Main Theorem

Let the step size h \to 0 while keeping the total time T = Nh fixed. Define the piecewise linear interpolation x_h(t) as:

x_h(t) = x_n + \frac{t - nh}{h}(x_{n+1} - x_n), \quad t \in [nh, (n+1)h]

Our goal is to find the limiting process of x_h(t) as h \to 0.

Definition (Structure-Preserving Extension)

A continuous-time process x(t) is called a structure-preserving extension of the discrete ECS system if there exists a subsequence h_k \to 0 such that x_{h_k}(t) converges uniformly (in probability) to x(t) for t \in [0,T], and:

1. Extremum Preservation: x(t) satisfies the Euler-Lagrange equation of a variational problem, with the action functional given by the continuous limit of the discrete MIE functional.
2. Conservation Preservation: The scalar function C(x(t)) = x(t)^\top \Sigma x(t) is constant (noise-free case) or constant in expectation (noisy case).
3. Symmetry Preservation: There exists a continuous Lie group G (generated by the discrete group G_h) such that g x(t) is also a solution to the limiting system, and \Sigma is G-invariant.

Main Theorem (Existence and Uniqueness)

Assume the ECS discrete system satisfies Axioms 1–3 and that all eigenvalues of L_{\text{ECS}} lie inside the unit circle. Then:

1. As h \to 0, x_h(t) converges almost surely to a unique continuous function x(t) in C([0,T], \mathbb{R}^d).
2. x(t) satisfies the linear ordinary differential equation:

\frac{dx}{dt} = \mathcal{A} x(t), \quad \mathcal{A} = \lim_{h \to 0} \frac{L_{\text{ECS}} - I}{h}

This limit exists, and the matrix \mathcal{A} has eigenvalues with strictly negative real parts.

3. Extremum Criterion: x(t) minimizes the continuous-time information action:

J[x] = \int_0^T \left( x(t)^\top Q x(t) + u(t)^\top R u(t) \right) dt, \quad \dot{x} = \mathcal{A} x + B u

with u(t) = -R^{-1} B^\top P_c x(t), where P_c solves the continuous Riccati equation. This is consistent with the discrete limit.

4. Conservation Law: C(x(t)) = x(t)^\top \Sigma x(t) is constant.
5. Symmetry Group: There exists a continuous Lie group G = \exp(\text{Lie}(G_h)) such that the limiting equation is invariant under G, and \Sigma is G-invariant.

The proof of the theorem is given via several lemmas in Section 4.

 

4. Proof Sketch

4.1 Convergence of Discrete Solutions and Limiting ODE

For deterministic (noise-free) systems, we have the exact linear map x_{n+1} = L_h x_n, where L_h = L_{\text{ECS}}(h) depends explicitly on h (since the Riccati solution P_h depends on h). Assume the weight matrices Q, R are independent of h. Standard results show that as h \to 0, L_h = I + h \mathcal{A} + o(h) and \mathcal{A} = A - B R^{-1} B^\top P_c, where P_c solves the continuous Riccati equation. Thus:

x_h(t) = L_h^{\lfloor t/h \rfloor} x(0) \to e^{\mathcal{A} t} x(0)

converges uniformly for t \in [0,T] (guaranteed by the continuity of the matrix exponential). This proves the existence and uniqueness of the limit, with the limiting equation \dot{x} = \mathcal{A} x.

4.2 Γ-Convergence of the Extremum Criterion

The discrete MIE functional is:

\mathcal{I}_h^{\text{disc}} = \sum_{n=0}^{N-1} \left( x_n^\top Q x_n + u_n^\top R u_n \right) h

where u_n = -R^{-1} B^\top P_h x_n. Substitution yields \mathcal{I}_h^{\text{disc}} = x_0^\top P_h x_0 (by Riccati theory). As h \to 0, P_h \to P_c, so:

\lim_{h \to 0} \mathcal{I}_h^{\text{disc}} = x_0^\top P_c x_0 = \int_0^\infty x(t)^\top (Q + \mathcal{A}^\top P_c + P_c \mathcal{A}) x(t) dt

The latter is the performance index of the continuous-time LQ problem. This shows that the limit of the discrete MIE extremal path minimizes the continuous action, with convergence of the optimal feedback gain.

4.3 Limit of Conserved Quantities

The discrete conserved quantity C(x_n) = x_n^\top \Sigma_h x_n is constant, with \Sigma_h satisfying the algebraic equation \Sigma_h = L_h^\top \Sigma_h L_h. It can be shown that \Sigma_h \to \Sigma_c, where \Sigma_c solves the continuous Lyapunov equation \mathcal{A}^\top \Sigma_c + \Sigma_c \mathcal{A} = 0. Thus, for the limiting process x(t):

\frac{d}{dt} \left( x(t)^\top \Sigma_c x(t) \right) = x(t)^\top (\mathcal{A}^\top \Sigma_c + \Sigma_c \mathcal{A}) x(t) = 0

so the conservation law is preserved. The constant value equals x_0^\top \Sigma_c x_0, consistent with the discrete limit.

4.4 Continuous Extension of Symmetry Groups

The discrete symmetry group G_h is a finite subgroup. Assume G_h stabilizes to a finite set as h \to 0 (in practice, G_h may be independent of h, e.g., intrinsic system symmetries). Define the continuous Lie group G as the closed subgroup of O(d) generated by G_h. Since L_h commutes with all g \in G_h and L_h \to e^{h \mathcal{A}}, taking the logarithm yields \mathcal{A} in the neighborhood of \text{Lie}(G), so e^{t \mathcal{A}} belongs to the identity component of G. Thus, the limiting flow e^{\mathcal{A} t} preserves G-invariance. Similarly, \Sigma_c is G-invariant.

This completes the proof of the Main Theorem.

 

5. From Discrete Large-Number Limit to Continuous Ergodic Limit

For stochastic ECS systems with noise, the discrete strong law of large numbers holds:

\frac{1}{N} \sum_{n=0}^{N-1} x_n \to \mu \quad \text{a.s.}

where \mu is the unique steady-state mean determined by MIE. We consider the continuous-time ergodic average:

\bar{x}_T = \frac{1}{T} \int_0^T x(t) dt

where x(t) is the continuous limiting process from Section 4 (satisfying the stochastic differential equation dx = \mathcal{A} x dt + \sigma dW, derived from weak convergence of discrete noise). By ergodic theory for stochastic processes: since \mathcal{A} is stable, this Ornstein-Uhlenbeck process is ergodic, with time averages converging almost surely to the steady-state mean \mu. The difference between the discrete sample mean and the continuous integral mean vanishes as h \to 0. Thus, the discrete large-number limit and the continuous ergodic limit are consistent, converging to the same value along the same stochastic realization.

Theorem (Universal Large-Number Unity)

Under the above framework:

\lim_{N \to \infty} \frac{1}{N} \sum_{n=0}^{N-1} x_n = \lim_{T \to \infty} \frac{1}{T} \int_0^T x(t) dt = \mu \quad \text{a.s.}

with \mu = 0 (due to zero-mean noise and stable drift).

 

6. Conclusion

This paper rigorously proves that for ECS discrete systems under the axiomatic framework, as the time step tends to zero, the solutions to the discrete difference equations converge to a continuous-time system that fully inherits the extremal optimality, quadratic conservation laws, and MOC symmetry group of the discrete system. This extension is unique and structure-preserving. Additionally, the discrete strong law of large numbers for stochastic ECS systems is consistent with the continuous ergodic theorem. Thereby, the ECS paradigm is successfully and rigorously extended from discrete categories to continuous spacetime manifolds, providing a theoretical foundation for subsequent applications in geometric field theory, geometric mechanics, and stochastic analysis.

Future work includes extending the linear framework to systems on nonlinear MOC manifolds and establishing precise correspondences between continuous ECS systems, gauge field theory, and general relativity.

 

References (Examples)

1. Bertsekas, D. P. (2012). Dynamic Programming and Optimal Control. Athena Scientific.
2. Karatzas, I., & Shreve, S. E. (1991). Brownian Motion and Stochastic Calculus. Springer.
3. Olver, P. J. (1993). Applications of Lie Groups to Differential Equations. Springer.
4. Zhang, S. H. (2025). Large-Number Limit Theorem for Extremum-Conservation-Symmetry Discrete Systems. (Preprint)