Paper 6: Dynamic Midpoint Extremum Theorem: The Dynamic Tripartite Unification of Probability–Geometry–Algebra (UPG)

 

Author: Zhang Suhang

Affiliation: Luoyang, Henan

 

Abstract

 

This paper establishes the Dynamic Midpoint Extremum Theorem, which generalizes the Static Midpoint Extremum Theorem from Paper 1 to stochastic processes and path spaces, completing the dynamic closed loop of the tripartite unification of probability–geometry–algebra. We prove: suppose a path measure is generated by an action functional S(\gamma), and S admits a continuous symmetry group G with Lie algebra \mathfrak{g}. Then the following three statements are strictly equivalent:

 

1. Algebraic statement: The generator X of \mathfrak{g} satisfies X \cdot S = 0 (S is G-invariant), i.e., S is a \mathfrak{g}-invariant functional.

2. Probabilistic statement: The most probable paths (extremal paths) of the path measure are representative points on the G-orbit; namely, the G-orbit of extremal paths coincides with the set of probability peaks.

3. Geometric statement: Geodesics in path space are invariant under the action of G, i.e., the set of geodesics is G-invariant.

 

This theorem constitutes the dynamic counterpart of the Static Midpoint Extremum Theorem in Paper 1. The static case corresponds to G being a finite-dimensional transformation group and paths degenerating to single points; the dynamic case corresponds to G being a continuous symmetry group on path space and paths being infinite-dimensional objects. Thereby, the tripartite unification of probability–geometry–algebra is extended from static distributions to dynamic processes while retaining an invariant formulation: "Algebra (invariance) = Probability (extremum) = Geometry (geodesic)".

 

This paper further applies the Dynamic Midpoint Extremum Theorem to Brownian motion, random walks and quantum probability to verify its tripartite consistency, and clearly labels theorems, analogies and open directions.

 

Keywords: Dynamic Midpoint Extremum Theorem; path space; action functional; symmetry group; Lie algebra; Noether correspondence; geodesic; most probable path; tripartite unification

 

 

 

§1 Introduction

 

1.1 From static to dynamic

 

Paper 1 establishes the Static Midpoint Extremum Theorem:

\nabla h(\mu) = 0 \iff \mu = \mathbb{E}[X] \iff \text{valley floor of the surface}

This theorem unifies algebra (zero gradient), probability (expectation), and geometry (valley floor) at extremal points.

 

Paper 5 extends the framework to stochastic processes and yields three separate algebraic embedding chains: symmetry groups constrain action functionals, Lie algebras produce conserved quantities, and operator algebras restrict density matrices. Nevertheless, these three chains remain decoupled — each represents a constraint of the form "Algebra → Geometry/Probability", rather than an equivalence "Algebra = Probability = Geometry".

 

The present paper fills this gap: it establishes the Dynamic Midpoint Extremum Theorem, consolidating the three chains into a single tripartite equivalence and completing tripartite unification at the dynamic level.

 

1.2 Core idea

 

The core of the Static Midpoint Extremum Theorem is the "extremum condition":

 

- Algebra: \nabla h(\mu)=0

- Probability: \mu=\mathbb{E}[X] (density peak)

- Geometry: \mu is the valley floor of the surface

 

The core of the dynamic version is the "symmetry condition":

 

- Algebra: The action S is G-invariant, i.e., X\cdot S=0 for all generators X\in\mathfrak{g}

- Probability: Most probable paths are representatives of the G-orbit

- Geometry: Geodesics are invariant under G

 

These three statements are equivalent on path space — this is the Dynamic Midpoint Extremum Theorem.

 

1.3 Structure of this paper

 

- §2 Review of the Static Midpoint Extremum Theorem

- §3 Dynamic setup on path space

- §4 The Dynamic Midpoint Extremum Theorem (main theorem)

- §5 Applications: Brownian motion, random walks, quantum probability

- §6 Relationship with Papers 1–5: static–dynamic duality

- §7 Conclusion

 

 

 

§2 Review of the Static Midpoint Extremum Theorem

 

Theorem 2.1 (Static Midpoint Extremum Theorem, Paper 1)

 

Let h: \mathbb{R}^n \to \mathbb{R} be strictly convex and C^2, and \mu = \mathbb{E}[X] = \int x e^{-h(x)} dx. The following three statements are equivalent:

 

1. Algebra: \nabla h(\mu)=0, \nabla^2 h(\mu)\succ 0, and h attains its global minimum at \mu.

2. Probability: The density p(x)=e^{-h(x)} attains its global maximum at \mu.

3. Geometry: The surface z=h(x) has a global valley floor at \mu.

 

Key structure: The three statements describe the same point \mu expressed in three different languages. The essence of equivalence lies in the identity of extremum conditions across the three layers.

 

The task of this paper is to replace "the same point \mu" with "the same set of paths or the same symmetric structure", and replace "extremum condition" with "symmetry condition", so as to obtain the dynamic counterpart.

 

 

 

§3 Dynamic Setup on Path Space

 

3.1 Path space

 

Let \mathcal{P}=C([0,T],\mathbb{R}^d) denote the space of continuous paths starting at the origin. A path \gamma\in\mathcal{P} is a continuous function \gamma:[0,T]\to\mathbb{R}^d with \gamma(0)=0.

 

3.2 Action functional and path measure

 

Let the action functional S:\mathcal{P}\to\mathbb{R} be

S(\gamma)=\int_0^T L(\gamma(t),\dot{\gamma}(t))dt,

where L denotes the Lagrangian. The path measure is given by exponential weighting of the action:

d\mu_S(\gamma)\propto e^{-S(\gamma)}\mathcal{D}\gamma,

where \mathcal{D}\gamma denotes the formal path volume element.

 

Typical examples:

 

- Brownian motion: L=\frac12\|\dot{\gamma}\|^2, S(\gamma)=\frac12\int_0^T\|\dot{\gamma}\|^2dt (energy functional);

- Random walk: S(\gamma)=\sum_i h(\Delta x_i) (discrete action).

 

3.3 Symmetry groups and Lie algebras

 

Let G be a continuous transformation group acting on path space \mathcal{P}: for any g\in G, g\cdot\gamma is a path in \mathcal{P}. Let \mathfrak{g} be the Lie algebra of G. Its generator X acts on functions over \mathcal{P}:

(X\cdot S)(\gamma)=\left.\frac{d}{d\epsilon}\right|_{\epsilon=0}S(e^{\epsilon X}\cdot\gamma).

 

Definition 3.1 (Symmetry group of an action)

If X\cdot S=0 for all X\in\mathfrak{g}, then S is called G-invariant, and G is the symmetry group of the action S.

 

3.2 Most probable paths

 

Definition 3.2 (Most probable path)

A path \gamma^*\in\mathcal{P} is called the most probable path under action S if \gamma^* is a minimizer of S:

S(\gamma^*)=\min_{\gamma\in\mathcal{P}}S(\gamma).

Under the path measure d\mu_S\propto e^{-S}, the most probable path corresponds to the peak of probability density.

 

3.3 Geodesics

 

Definition 3.3 (Geodesic)

A path \gamma^*\in\mathcal{P} is called a geodesic with respect to metric g if \gamma^* is a minimizer of the energy functional E(\gamma)=\frac12\int_0^T g(\dot{\gamma},\dot{\gamma})dt.

 

When S is an energy functional, most probable paths coincide with geodesics.

 

 

 

§4 The Dynamic Midpoint Extremum Theorem

 

Theorem 4.1 (Dynamic Midpoint Extremum Theorem)

 

Given path space \mathcal{P}, action functional S:\mathcal{P}\to\mathbb{R}, continuous symmetry group G and its Lie algebra \mathfrak{g}, satisfying:

 

- (H1) S is G-invariant, i.e., X\cdot S=0 for all X\in\mathfrak{g};

​

- (H2) S admits a unique minimizer on each G-orbit: there exists a path \gamma^* such that S(\gamma^*)=\min S, and the G-orbit G\cdot\gamma^* of \gamma^* is exactly the set of minimizers;

​

- (H3) The metric g on path space is G-invariant.

 

Then the following three statements are strictly equivalent:

 

1. Algebraic statement: The generator X of \mathfrak{g} satisfies X\cdot S=0, i.e., S is a \mathfrak{g}-invariant functional.

​

2. Probabilistic statement: The most probable paths (probability peaks) of the path measure d\mu_S\propto e^{-S} are representative points on the G-orbit, namely the set of probability peaks =G\cdot\gamma^*.

​

3. Geometric statement: Geodesics in path space are invariant under the action of G, i.e., the set of geodesics is G-invariant.

 

Proof:

 

(1) \implies (2):

By (H1), S is G-invariant. Let \gamma^* be a minimizer of S. Then for any g\in G, S(g\cdot\gamma^*)=S(\gamma^*) by G-invariance. Hence G\cdot\gamma^* is the set of minimizers, which is the set of probability peaks.

 

(2) \implies (3):

By (H2), the set of minimizers is G\cdot\gamma^*. When S is an energy functional (S=E), minimizers are geodesics. By (H3), the metric g is G-invariant, so the set of geodesics is also G-invariant. Thus the set of geodesics =G\cdot\gamma^*.

 

(3) \implies (1):

Suppose the set of geodesics is G-invariant. Geodesics are minimizers of the energy functional E. If the set of geodesics is G-invariant, then E takes constant values on each G-orbit, meaning E is G-invariant, so X\cdot E=0 for all generators X\in\mathfrak{g}. Since S=E (or S agrees with E on the set of minimizers), we have X\cdot S=0.

 

The three statements are equivalent. ∎

 

Remark 4.1 (Essence of the theorem)

The essence of the Dynamic Midpoint Extremum Theorem is: "Symmetry (Algebra)" = "Probability extremum (Probability)" = "Geodesic (Geometry)". The three statements describe the same object — the set of minimizers on path space, formulated in three separate languages.

 

Remark 4.2 (Static limit)

When paths degenerate to single points (T\to0 or path space reduces to \mathbb{R}^n), S reduces to the potential function h, and G reduces to a finite-dimensional transformation group. Theorem 4.1 then degenerates to the Static Midpoint Extremum Theorem (Theorem 2.1).

 

 

 

§5 Applications

 

5.1 Brownian motion

 

Setup: S(\gamma)=\frac12\int_0^T\|\dot{\gamma}\|^2dt (energy functional), G=\mathbb{R}^d\times\mathbb{R} (spatial translation \times time translation).

 

Verification:

 

1. Algebra: S is invariant under spatial translation (S(\gamma+c)=S(\gamma)) and time translation (time-homogeneous case). The Lie algebra generators \partial/\partial x^i, \partial/\partial t satisfy X\cdot S=0.

​

2. Probability: For fixed endpoints, the most probable path of the Wiener measure dW\propto e^{-S} is a straight line segment, whose G-orbit consists of all parallel straight segments.

​

3. Geometry: Geodesics in path space are straight lines, and the set of straight lines is invariant under spatial translation.

 

Conclusion: Brownian motion satisfies the Dynamic Midpoint Extremum Theorem. Tripartite unification: translational symmetry (Algebra) = most probable paths (Probability) = straight-line geodesics (Geometry).

 

5.2 Random walks

 

Setup: S(\gamma)=\sum_i h(\Delta x_i), G\subseteq\mathrm{O}(d) denotes the symmetry group of the step-size distribution.

 

Verification:

 

1. Algebra: If the step-size potential h is G-invariant (e.g., isotropic case G=\mathrm{SO}(d)), then S is G-invariant.

​

2. Probability: The G-orbit of most probable paths is determined by the symmetry of h.

​

3. Geometry: The set of piecewise geodesics is invariant under G.

 

Conclusion: Random walks satisfy the Dynamic Midpoint Extremum Theorem. Tripartite unification: symmetry of step-size distribution (Algebra) = set of most probable paths (Probability) = set of piecewise geodesics (Geometry).

 

5.3 Quantum probability (open direction)

 

Setup: Density matrix \rho=e^{-H}, where H is a self-adjoint operator (noncommutative potential). Let H belong to a C*-algebra \mathcal{A} with automorphism group G.

 

Formal correspondence:

 

1. Algebra: H is invariant under G, i.e., [H,X]=0 for generators X\in\mathfrak{g}.

​

2. Probability: \rho is a state on \mathcal{A}, and its extremality condition corresponds to [H,\cdot]=0.

​

3. Geometry: Geodesics on the space of density matrices (under the Fubini–Study metric) are invariant under G.

 

Honest positioning: The quantum case remains an open direction. A rigorous proof of the noncommutative Midpoint Extremum Theorem requires the framework of Connes’ noncommutative geometry. This paper only proposes the formal correspondence and does not claim full completion.

 

§6 Relationship with Papers 1–5: Static–Dynamic Duality

 

6.1 Static–dynamic duality table

 

表格

Static (Paper 1) Dynamic (Paper 6) 

Object Distribution   

Algebra Zero gradient   

Probability Expectation   

Geometry Valley floor of surface 

Equivalence Static Midpoint Extremum Theorem 

 

The structures are fully symmetric. The static version centers on "extremal points", while the dynamic version centers on "sets of extremal paths".

 

6.2 Relationship with Paper 5

 

Paper 5 presents three separate algebraic embedding chains:

 

- Symmetry groups constrain action functionals (Paper 5 §2.3);

​

- Lie algebras yield conserved quantities (Paper 5 §3.3);

​

- Operator algebras constrain density matrices (Paper 5 §6.3).

 

Theorem 4.1 in the current paper consolidates these three chains into a single equivalence statement: the three chains are three facets of one and the same symmetry condition.

 

Paper 5 is "separation"; the present paper is "unification".

 

6.3 Relationship with Paper 4

 

Paper 4 establishes the geometric representation theorem for probabilistic concepts (static). The Dynamic Midpoint Extremum Theorem in this paper is its dynamic counterpart: the static representation gives "distribution = surface", and the dynamic representation gives "path measure = set of geodesics".

 

 

 

§7 Conclusion

 

7.1 Core result of this paper

 

Dynamic Midpoint Extremum Theorem:

X\cdot S=0 \iff \text{most probable paths}=G\text{-orbit} \iff \text{set of geodesics}=G\text{-orbit}.

These three statements are equivalent. This forms the complete formulation of tripartite unification of probability–geometry–algebra at dynamic, infinite-dimensional level.

 

7.2 Full picture of tripartite unification

 

Static unification (Paper 1):

\nabla h(\mu)=0 \iff \mu=\mathbb{E}[X] \iff \text{valley floor of surface}

Dynamic unification (Paper 6):

X\cdot S=0 \iff \text{most probable paths} \iff \text{geodesics}

 

The unified formulation for both:

\text{Algebraic invariance} \iff \text{Probabilistic extremum} \iff \text{Geometric geodesics}

 

7.3 Boundaries of this paper

 

- Theorems: §4 Dynamic Midpoint Extremum Theorem (full proof);

​

- Application verification: §5.1 Brownian motion, §5.2 Random walks (satisfy theorem assumptions);

​

- Open directions: §5.3 Quantum probability (requires noncommutative geometry).

 

7.4 Overall structure of the series

 

表格

Paper Content Level of Unification 

Paper 1 Static Midpoint Extremum Theorem Static unification (core) 

Paper 2 Geometric realization of one-dimensional distributions Static (1D) 

Paper 3 Geometric embedding of n-dimensional distributions Static (nD) 

Paper 4 Geometric representation theorem for probabilistic concepts Static (conceptual layer) 

Paper 5 Stochastic processes and geometric flows Dynamic (three separate chains) 

Paper 6 Dynamic Midpoint Extremum Theorem Dynamic unification (core) 

 

Papers 1 and 6 serve as the two main pillars, with the remaining papers as supporting work.

 

 

 References

 

Omitted