Loop Closure

rigorous

Overview

This derivation answers the question: what does it take for an observer to persist?

Having a conserved quantity (Axiom 1) and a self-maintaining identity (Axiom 2) is not enough. An observer must actively sustain itself — its current state, processed through its own dynamics, must reproduce a valid observer state. This is self-reference: the observer is a process that instantiates itself.

The argument. Self-reference under finite resources forces loop structure:

• An observer’s internal state must be self-generating: the current configuration determines the next configuration. This is a deterministic flow on the state space.
• The observer has finite resources — a compact state space with finite coherence. A flow that drifts without returning will eventually cross the observer’s boundary and dissolve.
• Approximate return is not enough. Each imperfect cycle accumulates drift, giving the observer a finite lifetime proportional to how tightly the cycle closes (the drift bound).
• Only exact return — the state returning precisely to its starting configuration — gives indefinite persistence. Exact return of a continuous flow is periodicity, and periodicity is $U(1)$ symmetry.

The result. Every persistent observer has a natural frequency (how fast its loop cycles), a coherence cost (the minimum resources needed to complete one cycle), and a conserved charge paired with its cyclic symmetry. The loop is not an additional constraint imposed on observers — it is what self-reference looks like for a finite system.

Why this matters. Periodicity gives rise to frequency, frequency gives rise to energy (via the Planck-Einstein relation derived later), and the requirement of a minimum coherence cost per cycle is what ultimately produces Planck’s constant. The wave-like behavior of quantum mechanics traces back to this requirement: to exist is to oscillate.

An honest caveat. The step from “indefinite persistence” to “exact periodicity” relies on ruling out quasi-periodic flows (where the orbit is dense in a higher-dimensional torus rather than closing as a circle). The drift bound motivates this step but does not rigorously exclude quasi-periodicity; the additional physical input is that the observer’s dynamics generates a single frequency, not a continuum. See the Remark following Theorem 3.2 for a precise discussion.

Statement

Axiom 3 (Loop Closure). Every observer’s internal dynamics is self-sustaining and periodic: the state returns to its initial configuration after a finite period. A self-referencing system with finite resources that does not close its loop eventually dissolves. Loop closure is what makes an observer persistent in the sense of Axiom 2.

State Space Structure (Formerly Structural Postulate S1 — Now a Theorem)

Theorem 0.2 (State space is a finite-dimensional compact smooth manifold). The state space $\Sigma$ of any observer $\mathcal{O} = (\Sigma, I, \mathcal{B})$ is a finite-dimensional, compact, smooth manifold — without assuming this as a postulate. The manifold structure follows from the constructive mechanism by which observer state spaces are built.

Proof. The proof proceeds by induction on the bootstrap level, using the fact that the framework constructs observer state spaces from $U(1)$ orbits rather than placing observers into a pre-existing manifold.

(i) Base case. The minimal observer has state space $S^1$ — a compact, connected, 1-dimensional smooth manifold. This follows from Axiom 3 ($U(1)$ loop closure gives a faithful $U(1)$ action, Corollary 4.3 below) and O1 of Axiom 2 (compactness and connectedness). $S^1$ is a Lie group, hence automatically smooth. No postulate is needed: the manifold structure is intrinsic to the $U(1)$ Lie group.

(ii) Inductive step (interaction). Suppose observers $\mathcal{O}_1$ and $\mathcal{O}_2$ have state spaces $\Sigma_1$ and $\Sigma_2$ that are compact smooth manifolds (inductive hypothesis). When they interact via a Type III interaction, a new relational invariant $I_{12}$ is generated (Relational Invariants, Theorem 3.2). The composite observer’s state space is the level set (Bootstrap Mechanism, Theorem 1.1):

$\Sigma_{12} = \{(\sigma_1, \sigma_2) \in \Sigma_1 \times \Sigma_2 : I_{12}(\sigma_1, \sigma_2) = c\}$

The product $\Sigma_1 \times \Sigma_2$ is a compact smooth manifold (products of compact smooth manifolds are compact smooth manifolds). The relational invariant $I_{12}$ is a smooth function on this product (its smoothness is inherited from the $U(1)$ Lie group structure via the symplectic construction of Relational Invariants, Theorem 0.1). At any regular value $c$, the level set is a smooth submanifold by the regular value theorem. It is compact (closed subset of a compact space) and connected (the $U(1)$ action on the level set is transitive on connected components).

(iii) Finite dimensionality. Each Type III interaction adds at most one $U(1)$ degree of freedom (the orbit of the new relational invariant’s symmetry). The observer network has finite local complexity (Aperiodic Order, Proposition 2.0 — the network is a Delone set with finitely many distinct local neighborhoods). Each observer therefore has finitely many interaction partners, hence finitely many relational invariants, hence finitely many $U(1)$ factors. $\dim \Sigma < \infty$.

(iv) Summary. Every observer’s state space is constructed by a finite sequence of operations — products and level sets — starting from $S^1$ (a Lie group manifold). Each operation preserves compactness, smoothness, and finite dimensionality. The manifold structure is a consequence of the construction, not an assumption about the state space. $\square$

Remark (Structural Postulate S1 — now a theorem). S1 originally postulated that $\Sigma$ is a finite-dimensional topological manifold. This has been progressively tightened: compactness was derived from O1, free orbits from Lemma 0.0, smoothness from Theorem 0.0. The final step — deriving the manifold structure itself — comes from recognizing that the framework constructs state spaces from $U(1)$ orbits via the bootstrap mechanism, and the construction operations (products, level sets at regular values) preserve manifold structure at each step. Finite dimensionality follows from the finite local complexity of the observer network. S1 is now fully derived: no component remains postulated.

Lemma 0.0 (Free orbits from Axiom 3). The $U(1)$ action on $\Sigma$ is free: no point of $\Sigma$ is fixed under all $\phi_t$.

Proof. Case $\dim \Sigma = 1$ (minimal observer): $\Sigma \cong S^1$. A non-trivial continuous $U(1)$ action on $S^1$ has no fixed points (a rotation with a fixed point is the identity, contradicting non-triviality from Axiom 3). So the action is free.

Case $\dim \Sigma \geq 2$: Suppose $\sigma^* \in \Sigma$ is a fixed point: $\phi_t(\sigma^*) = \sigma^*$ for all $t$. The orbit through $\sigma^*$ is the single point $\{\sigma^*\}$, which has zero length. The coherence cost of this orbit is zero: $S(\sigma^*) = 0$. But by Proposition 7.2 (below), every point in an observer’s state space must have positive coherence cost — zero-cost states carry no coherence and cannot contribute to the observer’s conserved invariant $I$. A state that does not participate in the dynamics is not part of the observer (it violates the self-referential character of loop closure: to persist is to cycle). Therefore no fixed point exists, and the action is free. $\square$

Remark (Axiom coherence — why smooth structure is structurally necessary). Beyond being the unique consistent setting, smooth structure is required for the three axioms to communicate. The framework’s core mechanism is the Noether pair (Theorem 5.1 below): the $U(1)$ symmetry from Axiom 3 is linked to the conserved invariant $I$ from Axiom 2, which is related to the coherence measure $\mathcal{C}$ from Axiom 1, through the moment map. This identification requires a Lie algebra (to define the generating vector field), a smooth action (for the moment map to exist), and a smooth manifold (for the tangent bundle and differential forms).

On a non-smooth space, all three ingredients fail. The three axioms remain individually satisfiable — a conserved measure, a continuous invariant, and a periodic homeomorphism can coexist on a non-smooth space — but the Noether mechanism that connects them is severed. Coherence conservation becomes an accidental coincidence rather than a structural consequence of the $U(1)$ symmetry. The axiom system requires smooth structure not as mathematical convenience but for internal coherence: without it, the axioms cannot communicate.

Remark (Selection argument for S1). Three independent downstream requirements converge on requiring at least $C^1$ differentiability, from which $C^\infty$ follows:

1. Noether’s theorem (Step 5, Observer Definition Theorem 5.1) requires a Lie group acting smoothly on a smooth manifold.
2. Coherence cost (Step 7) requires the velocity $\dot{\phi}_t$ along the observer loop, presupposing $C^1$ differentiability.
3. Fisher information metric (Fisher Metric) requires $\mathcal{C}$ to be $C^2$ on $\Sigma$.

Whitney’s theorem (1936) upgrades $C^1$ to the unique compatible $C^\infty$ structure. The following theorem shows this smooth structure is not assumed but derived from S1 and Axiom 3.

Theorem 0.0 (Smooth structure). $\Sigma$ admits a smooth manifold structure, unique up to equivariant diffeomorphism, such that the $U(1)$ action from Axiom 3 is smooth.

Proof. By Theorem 0.2, $\Sigma$ is a compact topological manifold of dimension $n \geq 1$ with a free continuous $U(1)$ action (Lemma 0.0).

Step 1 (Smooth structure exists). Since the $U(1)$ action is free, $\Sigma$ is a principal $U(1)$-bundle over the orbit space $B = \Sigma/U(1)$, which is a compact topological manifold of dimension $n - 1$ Bredon, 1972. By the equivariant smoothing theorem for compact Lie group actions on topological manifolds Illman, 1983; Orlik, 1972, there exists a smooth structure on $\Sigma$ making the $U(1)$ action a smooth Lie group action.

Step 2 (Uniqueness). For $n \leq 3$: every compact topological manifold of dimension $\leq 3$ admits a unique smooth structure (dimension 1: elementary; dimension 2: Radó, 1925; dimension 3: Moise, 1952). The smooth $U(1)$ action is then the unique smoothing of the continuous action. For $n \geq 4$: by the uniqueness part of the equivariant smoothing theorem Illman, 1983, any two smooth structures making the compact Lie group action smooth are equivariantly diffeomorphic, so the equivariant smooth structure is unique. $\square$

Remark. For the minimal observer ($\Sigma = S^1$, $\dim \Sigma = 1$), Theorem 0.0 is trivial: $S^1$ has a unique smooth structure and every continuous $U(1)$ action on $S^1$ is automatically smooth. The theorem’s content becomes substantive for higher-dimensional state spaces. With the promotion of S1 to Theorem 0.2, all inputs to Theorem 0.0 are now derived from the axioms.

Theorem 0.1 (Invariant Riemannian metric). The state space $\Sigma$ carries a $G_\mathcal{O}$-invariant Riemannian metric $g$ with $\phi_t^* g = g$ for all $t$.

Proof. Step (a): Existence of a Riemannian metric. Since $\Sigma$ is a smooth manifold (Theorem 0.0) and is compact (O1 from Observer Definition), it admits a Riemannian metric $g_0$. (Standard result: every smooth manifold admits a Riemannian metric, constructed via a partition of unity subordinate to an atlas. Compactness ensures a finite partition suffices.)

Step (b): Weyl averaging. The dynamics $\phi_t$ generates a $U(1)$ action on $\Sigma$ (Corollary 4.3), and $U(1)$ is compact. Define the averaged metric:

$g(\sigma)(v, w) = \int_{U(1)} (\phi_\theta^* g_0)(\sigma)(v, w) \, d\mu(\theta)$

where $\mu$ is the normalized Haar measure on $U(1)$ and $\phi_\theta = \phi_{\theta T_\mathcal{O}/2\pi}$.

• $g$ is Riemannian: For any non-zero $v \in T_\sigma\Sigma$, $(\phi_\theta^* g_0)(v, v) = g_0(d\phi_\theta(v), d\phi_\theta(v)) > 0$ since $g_0$ is positive-definite and $d\phi_\theta$ is a linear isomorphism. The integral of positive functions with respect to a probability measure is positive. So $g(v, v) > 0$ for all $v \neq 0$.
• $g$ is $G_\mathcal{O}$-invariant: For any $\phi_s$ in the $U(1)$ action, $(\phi_s^* g)(v, w) = \int_{U(1)} ((\phi_\theta \circ \phi_s)^* g_0)(v, w) \, d\mu(\theta) = g(v, w)$ by translation-invariance of Haar measure on $U(1)$. $\square$

Remark (Canonical choice). Theorem 0.1 establishes existence using an arbitrary initial metric. The Fisher Information Metric provides a canonical choice: the unique (up to scale) Riemannian metric on $\Sigma$ invariant under sufficient statistics (Čencov’s theorem). Since the Fisher metric is already $U(1)$-invariant when $\mathcal{C}$ is constant along orbits (which holds by invariant preservation), it provides a distinguished $G_\mathcal{O}$-invariant metric without averaging.

Remark (Former Postulates S1 and S2 — both now theorems). S1 originally assumed smooth manifold structure + compactness + free orbits + finite dimensionality. Each component has been progressively derived: compactness from O1 (Axiom 2), free orbits from Lemma 0.0, smoothness from Theorem 0.0, and now manifold structure + finite dimensionality from Theorem 0.2 (inductive construction from $U(1)$ Lie group orbits via the bootstrap mechanism). S2 originally assumed a $G_\mathcal{O}$-invariant Riemannian metric; this is Theorem 0.1, derived from Theorem 0.0 + compactness + Weyl averaging. No component of S1 or S2 remains postulated. The state space’s entire structure follows from the axioms plus the constructive mechanism of the bootstrap.

Formalization

Step 1: Self-Sustaining Dynamics

Definition 1.1 (Internal dynamics). Let $\mathcal{O} = (\Sigma, I, \mathcal{B})$ be an observer (Axiom 2). The internal dynamics of $\mathcal{O}$ is a smooth group homomorphism:

$\phi: (\mathbb{R}, +) \to (G_\mathcal{O}, \circ), \quad t \mapsto \phi_t$

satisfying:

• Homomorphism: $\phi_{t_1 + t_2} = \phi_{t_1} \circ \phi_{t_2}$ and $\phi_0 = \text{id}_\Sigma$
• Smoothness: The map $(t, \sigma) \mapsto \phi_t(\sigma)$ is smooth as a map $\mathbb{R} \times \Sigma \to \Sigma$
• Invariant preservation: $I(\phi_t(\sigma)) = I(\sigma)$ for all $t \in \mathbb{R}$, $\sigma \in \Sigma$ (follows from $\phi_t \in G_\mathcal{O}$)

Remark (Self-reference). The requirement $\phi_t \in G_\mathcal{O}$ encodes self-reference: the observer’s dynamics preserves the very invariant that defines the observer. The state at time $t$ is produced by the observer’s own symmetry group acting on the state at time $0$. The observer is, in this precise sense, a process that instantiates itself.

Step 2: Approximate Closure and Dissolution

Before stating the axiom’s content, we establish why self-sustaining dynamics must eventually close into a loop.

Definition 2.1 (Boundary diameter). The boundary diameter of observer $\mathcal{O}$ is $D_\mathcal{B} = \text{diam}(\Sigma) < \infty$ (finite by compactness, using the geodesic distance $d$ induced by the Riemannian metric $g$ from Theorem 0.1).

Definition 2.2 ($\epsilon$-approximate closure). An observer satisfies $\epsilon$-approximate closure with period $T$ if:

$\sup_{\sigma \in \Sigma} d(\phi_T(\sigma), \sigma) < \epsilon$

Proposition 2.3 (Drift bound). An $\epsilon$-approximately closed observer’s state drifts by at most $N\epsilon$ from the initial state after $N$ cycles:

$d(\phi_{NT}(\sigma), \sigma) \leq N\epsilon$

The state remains within $\Sigma$ as long as $N\epsilon < D_\mathcal{B}$, giving $N_{\max} \leq \lfloor D_\mathcal{B}/\epsilon \rfloor$.

Proof. By induction. For $N = 1$: $d(\phi_T(\sigma), \sigma) < \epsilon$ by definition. For the inductive step, define $\sigma_k = \phi_{kT}(\sigma)$. Then:

$d(\sigma_{k+1}, \sigma_k) = d(\phi_T(\sigma_k), \sigma_k) < \epsilon$

since $\epsilon$-approximate closure holds uniformly over $\Sigma$ (and each $\sigma_k \in \Sigma$). By the triangle inequality:

$d(\sigma_N, \sigma_0) \leq \sum_{k=0}^{N-1} d(\sigma_{k+1}, \sigma_k) < N\epsilon$

Once $N\epsilon \geq D_\mathcal{B}$, the drift may carry the state to $\partial\mathcal{O}$ (the boundary of the self/non-self partition from Axiom 2). At $\partial\mathcal{O}$, the state is exposed to the full transformation group $\text{Aut}(\mathcal{H})$, not just $G_\mathcal{O}$. Transformations in $\text{Aut}(\mathcal{H}) \setminus G_\mathcal{O}$ do not preserve $I$, so the observer dissolves. $\square$

Definition 2.4 (Observer lifetime). The observer lifetime is $\tau_\mathcal{O} = N_{\max} \cdot T$ where $N_{\max} = \lfloor D_\mathcal{B}/\epsilon \rfloor$.

Proposition 2.5 (Exact closure gives persistence). Exact closure ($\epsilon = 0$) gives $N_{\max} = \infty$ and $\tau_\mathcal{O} = \infty$.

Proof. With $\epsilon = 0$, $\phi_T = \text{id}_\Sigma$ exactly, so $\phi_{NT} = \text{id}_\Sigma$ for all $N \in \mathbb{N}$. The drift is identically zero. $\square$

Step 3: Persistence Forces Periodicity

Theorem 3.1 (Persistence requires exact closure). If an observer persists indefinitely ($\tau_\mathcal{O} = \infty$), then its dynamics satisfies exact closure: there exists $T > 0$ with $\phi_T = \text{id}_\Sigma$.

Proof. Suppose $\phi$ has no exact closure: $\phi_T \neq \text{id}_\Sigma$ for all $T > 0$. Then for every $T > 0$, there exists $\sigma_T \in \Sigma$ with $\phi_T(\sigma_T) \neq \sigma_T$, so $\epsilon(T) = \sup_\sigma d(\phi_T(\sigma), \sigma) > 0$. By Proposition 2.3, the observer’s lifetime at period $T$ satisfies $\tau(T) = \lfloor D_\mathcal{B}/\epsilon(T) \rfloor \cdot T < \infty$.

The observer’s effective lifetime is $\tau_\mathcal{O} = \sup_T \tau(T)$. For $\tau_\mathcal{O} = \infty$, we would need $\epsilon(T_n) \to 0$ sufficiently fast along some sequence $T_n$ — that is, the dynamics would need to approximate the identity arbitrarily well. By compactness of $\Sigma$, if $\phi_{T_n} \to \text{id}_\Sigma$ uniformly, then any limit point $T^*$ of a bounded subsequence gives $\phi_{T^*} = \text{id}_\Sigma$ (exact closure). If $T_n \to \infty$ with no bounded subsequence, the dynamics approximates the identity only at diverging times — the observer’s “renewal period” grows without bound, meaning it takes arbitrarily long to approximately reproduce its state. This is not self-sustaining persistence in the physical sense required by the axiom: a system whose self-reproduction time diverges is not maintaining itself.

Therefore, indefinite persistence requires $\phi_T = \text{id}_\Sigma$ for some finite $T > 0$. $\square$

Remark (Quasi-periodic flows). The argument above rules out non-recurrent flows but leaves open the case of quasi-periodic flows — dynamics where the orbit is dense in a higher-dimensional torus $T^k$ ($k > 1$) rather than closing as a circle. For such flows, $\epsilon(T_n) \to 0$ along a sequence with $T_n \to \infty$, and the effective lifetime can be infinite without exact periodicity.

The framework excludes quasi-periodicity on physical grounds: a quasi-periodic observer has $k > 1$ independent incommensurate frequencies, producing a continuous power spectrum rather than a discrete one. Quantum mechanics requires discrete energy levels ($E = n\hbar\omega$), which arise only from a single fundamental frequency — i.e., from exact periodicity generating $U(1)$, not $T^k$. The requirement of a discrete spectrum is equivalent to the requirement that the dynamics generates exactly $U(1)$.

Axiom 3 (Precise statement). The homomorphism $\phi$ is non-trivial ($\phi \neq \text{const}$), and there exists $T > 0$ such that $\phi_T = \text{id}_\Sigma$.

Proposition 3.2 (Minimal period). Under Axiom 3, the minimal period $T_\mathcal{O} = \inf\{T > 0 : \phi_T = \text{id}_\Sigma\}$ exists, is positive, and satisfies $\phi_{T_\mathcal{O}} = \text{id}_\Sigma$.

Proof. Define $P = \{T > 0 : \phi_T = \text{id}_\Sigma\}$. By Axiom 3, $P \neq \emptyset$.

$P$ is closed: Let $T_n \in P$ with $T_n \to T^*$. For any $\sigma \in \Sigma$, $\phi_{T_n}(\sigma) = \sigma$ for all $n$. By smoothness of $\phi$, $\phi_{T^*}(\sigma) = \lim_{n} \phi_{T_n}(\sigma) = \sigma$. Hence $T^* \in P \cup \{0\}$.

$P$ is discrete: Suppose $P$ is not discrete. Then there exists a sequence $T_n \in P$ with $T_n \to 0^+$. For any $\sigma \in \Sigma$, $\phi_{T_n}(\sigma) = \sigma$, so the orbit $t \mapsto \phi_t(\sigma)$ returns to $\sigma$ at arbitrarily small positive times. By smoothness, the generating vector field $X(\sigma) = \frac{d}{dt}\big|_0 \phi_t(\sigma)$ must vanish at every $\sigma$ (since the orbit is at $\sigma$ at times converging to 0). Hence $\phi_t = \text{id}_\Sigma$ for all $t$ — the dynamics is trivial. But triviality contradicts the non-triviality requirement of Axiom 3 ($\phi \neq \text{const}$).

Therefore $P$ is closed and discrete in $(0, \infty)$, so $T_\mathcal{O} = \inf P$ exists in $P$ and $T_\mathcal{O} > 0$. $\square$

Remark (Discrete case). If $\Sigma$ is finite, the dynamics is a bijection $\phi: \Sigma \to \Sigma$ with $\phi^n = \text{id}$ for some $n \geq 2$. The minimal period is the order of $\phi$ in $\text{Sym}(\Sigma)$.

Step 4: The Observer Loop as Geometric Object

Definition 4.1 (Observer loop). The observer loop of $\mathcal{O}$ based at $\sigma_0 \in \Sigma$ is the image of the orbit map:

$\gamma_\mathcal{O}(\sigma_0) = \{\phi_t(\sigma_0) : t \in [0, T_\mathcal{O})\} \subset \Sigma$

Proposition 4.2 (Loop is a smooth embedding). For each $\sigma_0 \in \Sigma$, the orbit map $\alpha: S^1 \to \Sigma$ defined by $\alpha(e^{2\pi i t/T_\mathcal{O}}) = \phi_t(\sigma_0)$ is a smooth embedding. The observer loop $\gamma_\mathcal{O}(\sigma_0)$ is a compact, connected, one-dimensional submanifold of $\Sigma$.

Proof. Well-defined: If $e^{2\pi i t_1/T_\mathcal{O}} = e^{2\pi i t_2/T_\mathcal{O}}$, then $t_1 - t_2 = k T_\mathcal{O}$ for some $k \in \mathbb{Z}$, so $\phi_{t_1}(\sigma_0) = \phi_{t_2 + kT_\mathcal{O}}(\sigma_0) = \phi_{t_2}(\sigma_0)$ by periodicity. Smooth: Composition of smooth maps $S^1 \hookrightarrow \mathbb{R}/T_\mathcal{O}\mathbb{Z} \to \Sigma$. Immersion: The derivative $d\alpha$ has image spanned by $X(\phi_t(\sigma_0)) \neq 0$ (non-vanishing since $T_\mathcal{O}$ is the minimal period; if $X(\sigma^*) = 0$ at some $\sigma^* = \phi_{t_0}(\sigma_0)$, then $\phi_t(\sigma^*) = \sigma^*$ for all $t$, giving $\phi_{t_0}$ as a period element with $\phi_{t_0}(\sigma_0) = \sigma^*$ fixed, contradicting minimality). Injective: If $\phi_{t_1}(\sigma_0) = \phi_{t_2}(\sigma_0)$ with $0 \leq t_1 < t_2 < T_\mathcal{O}$, then $\phi_{t_2-t_1}(\sigma_0) = \sigma_0$ with $0 < t_2 - t_1 < T_\mathcal{O}$, contradicting minimality of $T_\mathcal{O}$. An injective immersion from a compact manifold is an embedding. $\square$

Corollary 4.3 ($U(1)$ action). The periodicity defines a faithful smooth $U(1)$ action on $\Sigma$:

$\rho: U(1) \to \text{Diff}(\Sigma), \quad e^{i\theta} \mapsto \phi_{\theta T_\mathcal{O}/2\pi}$

Proof. Homomorphism: $\rho(e^{i\theta_1}) \circ \rho(e^{i\theta_2}) = \phi_{\theta_1 T/2\pi} \circ \phi_{\theta_2 T/2\pi} = \phi_{(\theta_1+\theta_2)T/2\pi} = \rho(e^{i(\theta_1+\theta_2)})$ by the group property of $\phi$. Faithful: If $\rho(e^{i\theta}) = \text{id}_\Sigma$, then $\phi_{\theta T/2\pi} = \text{id}_\Sigma$, so $\theta T/2\pi \in P$. By minimality of $T_\mathcal{O} = T$, $\theta T/2\pi = kT$ for some $k \in \mathbb{Z}$, giving $\theta = 2\pi k$, i.e., $e^{i\theta} = 1$. $\square$

Proposition 4.4 (Orbit decomposition). For any $\sigma_0, \sigma_1 \in \Sigma$, the observer loops $\gamma_\mathcal{O}(\sigma_0)$ and $\gamma_\mathcal{O}(\sigma_1)$ are either identical or disjoint. The decomposition $\Sigma = \bigsqcup_{\text{orbits}} \gamma_\mathcal{O}(\sigma)$ is the orbit decomposition under the $U(1)$ action.

Proof. Standard orbit decomposition for group actions. If $\sigma_1 = \phi_s(\sigma_0)$ for some $s$, then $\gamma_\mathcal{O}(\sigma_1) = \{\phi_t(\phi_s(\sigma_0)) : t \in [0, T)\} = \{\phi_{t+s}(\sigma_0) : t \in [0, T)\} = \gamma_\mathcal{O}(\sigma_0)$ by periodicity. If $\sigma_1 \notin \gamma_\mathcal{O}(\sigma_0)$, the orbits are disjoint (orbits of a group action partition the space). $\square$

Step 5: The Noether Pair

Theorem 5.1 (Loop closure ↔ Noether pair). Assume $\Sigma$ carries a symplectic structure $\omega$ preserved by $\phi_t$ (i.e., $\phi_t^*\omega = \omega$). Then:

(a) The $U(1)$ action of Corollary 4.3 has an associated moment map $\mu: \Sigma \to \mathbb{R}$, which is a conserved quantity: $\mu \circ \phi_t = \mu$ for all $t$.

(b) Conversely, given a conserved quantity $\mu: \Sigma \to \mathbb{R}$ generating a Hamiltonian flow that is periodic with period $T$, the flow defines a $U(1)$ action satisfying Axiom 3.

Proof. (a) The Lie algebra $\mathfrak{u}(1) \cong \mathbb{R}$ has generator $\xi$, whose fundamental vector field is $X(\sigma) = \frac{d}{dt}\big|_0 \phi_t(\sigma)$. Since $\phi_t^*\omega = \omega$, we have $\mathcal{L}_X \omega = 0$. By Cartan’s formula, $d(\iota_X \omega) + \iota_X(d\omega) = 0$, and $d\omega = 0$ (symplectic), so $d(\iota_X \omega) = 0$. When $H^1(\Sigma; \mathbb{R}) = 0$ (or more generally when $[\iota_X \omega] = 0$ in de Rham cohomology), there exists $\mu: \Sigma \to \mathbb{R}$ with $d\mu = \iota_X \omega$. Conservation: $\frac{d}{dt}\mu(\phi_t(\sigma)) = d\mu_{\phi_t(\sigma)}(X) = \omega(X, X) = 0$ (antisymmetry).

(b) Given periodic Hamiltonian flow with period $T$, define $\rho(e^{i\theta}) = \phi_{\theta T/2\pi}$. This is a $U(1)$ action by the argument of Corollary 4.3, and periodicity is Axiom 3. $\square$

Remark (Assumptions). Theorem 5.1 requires: (i) a symplectic structure on $\Sigma$, (ii) the flow preserves it, and (iii) the moment map exists (automatic if $H^1(\Sigma) = 0$). These are additional structural assumptions beyond S1 (and its consequence, Theorem 0.1), physically motivated by the fact that state spaces of classical mechanical systems and projective Hilbert spaces carry natural symplectic structures.

Proposition 5.2 (Relationship between Axioms 2 and 3). Axiom 3 is not logically independent of Axiom 2 — it specifies the dynamical realization. Precisely:

• Axiom 2 requires the existence of an invariant $I$, but permits $G_\mathcal{O}$ to be any non-trivial subgroup of $\text{Aut}(\mathcal{H})|_\Sigma$ — including purely discrete groups.
• Axiom 3 strengthens Axiom 2 by requiring $G_\mathcal{O}$ to contain a $U(1)$ subgroup acting as internal dynamics.

Proof. Axiom 2 requires $G_\mathcal{O} = \{g \in \text{Aut}(\mathcal{H})|_\Sigma : I \circ g = I\}$ to be a non-trivial subgroup. This is satisfied by discrete groups (e.g., $\mathbb{Z}_n$), which have no continuous one-parameter subgroups. Axiom 3 requires a smooth homomorphism $\phi: \mathbb{R} \to G_\mathcal{O}$ with period $T$, whose image is isomorphic to $U(1) \cong \mathbb{R}/T\mathbb{Z}$. This is a strictly stronger requirement. $\square$

Remark (Why a separate axiom). Despite this logical dependence, stating loop closure separately is clarifying: it emphasizes that the observer must be dynamically persistent, not merely a static configuration with a conserved quantity. It distinguishes continuous cyclic dynamics from the discrete symmetries that Axiom 2 alone would permit.

Step 6: Stability — Lyapunov Formulation

Definition 6.1 (Lyapunov stability). Using the geodesic distance $d$ induced by the Riemannian metric $g$ (Theorem 0.1), the observer loop $\gamma_\mathcal{O}$ is Lyapunov stable if for every $\epsilon > 0$, there exists $\delta > 0$ such that:

$d(\sigma, \gamma_\mathcal{O}) < \delta \implies d(\phi_t(\sigma), \gamma_\mathcal{O}) < \epsilon \quad \forall t \geq 0$

where $d(\sigma, \gamma_\mathcal{O}) = \inf_{\sigma' \in \gamma_\mathcal{O}} d(\sigma, \sigma')$.

Definition 6.2 (Asymptotic stability). The loop is asymptotically stable if it is Lyapunov stable and additionally $\lim_{t \to \infty} d(\phi_t(\sigma), \gamma_\mathcal{O}) = 0$ for all $\sigma$ with $d(\sigma, \gamma_\mathcal{O}) < \delta$.

Remark (Lyapunov vs. asymptotic). Asymptotic stability requires dissipation (contraction of phase space volume), which is inconsistent with coherence conservation in an isolated system (Axiom 1 implies phase space volume preservation via Liouville’s theorem). The physical requirement is Lyapunov stability (neutral stability), consistent with Hamiltonian dynamics. Dissipative stability occurs only for open subsystems exchanging coherence with their environment.

Step 7: The Coherence Cost of a Loop

Definition 7.1 (Coherence cost / action). Using the Riemannian metric $g$ (Theorem 0.1) and the coherence measure $\mathcal{C}$ from Axiom 1, the coherence cost of an observer loop is:

$S_\mathcal{O} = \int_0^{T_\mathcal{O}} \mathcal{C}(\phi_t(\sigma_0)) \sqrt{g(\dot{\phi}_t, \dot{\phi}_t)} \, dt$

where $\dot{\phi}_t = \frac{d}{dt}\phi_t(\sigma_0)$ is the velocity along the loop.

Remark (Structural dependence). This integral requires both the metric $g$ from (Theorem 0.1) and the coherence measure $\mathcal{C}$ from Axiom 1 restricted to $\Sigma$. Their compatibility (that $\mathcal{C}$ is smooth on $(\Sigma, g)$) is an assumption of the framework.

Proposition 7.2 (Positive minimum). The coherence cost satisfies $S_\mathcal{O} > 0$ for every observer. There exists a positive lower bound $S_{\min} > 0$.

Proof. Since $\gamma_\mathcal{O}$ is a smooth embedding $S^1 \hookrightarrow \Sigma$ (Proposition 4.2), its arc length $L = \int_0^{T_\mathcal{O}} |\dot{\phi}_t| \, dt > 0$ (the loop is not a point, by minimality of $T_\mathcal{O}$ and the non-degeneracy argument in Proposition 3.2). The coherence measure satisfies $\mathcal{C}(\Sigma) > 0$ by axiom C2 (normalization: $\mathcal{C}(\mathcal{H}) = C_0 > 0$) combined with the fact that $\Sigma \subseteq \mathcal{H}$ is a non-empty observer state space carrying non-zero coherence (an observer with $\mathcal{C}(\Sigma) = 0$ would have zero coherence cost and could not sustain a loop — contradicting the physical content of Axiom 3). Since $\mathcal{C}$ is continuous and $\gamma_\mathcal{O}$ is compact, $\inf_{\sigma \in \gamma_\mathcal{O}} \mathcal{C}(\sigma) > 0$. Therefore:

$S_\mathcal{O} \geq \inf_{\sigma \in \gamma_\mathcal{O}} \mathcal{C}(\sigma) \cdot L > 0$

The infimum $S_{\min} = \inf_\mathcal{O} S_\mathcal{O}$ over all observers satisfying Axioms 1–3 is positive because any observer has both $L > 0$ and $\mathcal{C} > 0$. The observer achieving this infimum is the minimal observer, developed in Minimal Observer Structure. $\square$

Remark (Identification with $\hbar$). The framework identifies $S_{\min} = \hbar$. This identification is not part of the loop closure axiom; it is derived in Action and Planck’s Constant from dimensional analysis and the minimal observer structure.

Step 8: Frequency, Energy, and Mass

Proposition 8.1 (Natural frequency). An observer loop with period $T_\mathcal{O}$ has natural frequency $\omega_\mathcal{O} = 2\pi/T_\mathcal{O}$.

Proof. The $U(1)$ action (Corollary 4.3) parameterizes the loop by phase $\theta \in [0, 2\pi)$. The physical period $T_\mathcal{O}$ corresponds to $\Delta\theta = 2\pi$, giving $\omega_\mathcal{O} = 2\pi/T_\mathcal{O}$. $\square$

Proposition 8.2 (Action-energy relation). If $\mathcal{C}(\phi_t(\sigma_0))$ is constant along the loop and the loop is traversed at constant speed, then $S_\mathcal{O} = \mathcal{C}(\Sigma) \cdot v \cdot T_\mathcal{O}$, where $v$ is the constant speed and $\mathcal{C}(\Sigma)$ the constant coherence value.

Proof. Substituting constants into Definition 7.1: $S_\mathcal{O} = \int_0^{T_\mathcal{O}} \mathcal{C}(\Sigma) \cdot v \, dt = \mathcal{C}(\Sigma) \cdot v \cdot T_\mathcal{O}$. $\square$

Remark (Constancy justification). The constancy of $\mathcal{C}$ along the loop holds when $\mathcal{C}$ depends only on the invariant $I$ (which is conserved along the flow). Constant speed is guaranteed by $G_\mathcal{O}$-invariance of the metric (Theorem 0.1): since $\phi_t$ is an isometry, $|\dot{\phi}_{t+s}(\sigma_0)|_g = |\dot{\phi}_s(\sigma_0)|_g$ for all $t$, so the speed is constant for a $U(1)$ orbit.

Corollary 8.3 (Planck-Einstein relation). Defining $E_\mathcal{O} = S_\mathcal{O}/T_\mathcal{O}$ as the energy (action per period), and using $S_{\min} = \hbar$ (cf. Action and Planck’s Constant):

$E_\mathcal{O} = \frac{S_\mathcal{O}}{T_\mathcal{O}} \geq \frac{S_{\min}}{T_\mathcal{O}} = \hbar \omega_\mathcal{O}$

For the minimal observer ($S_\mathcal{O} = S_{\min} = \hbar$), equality holds: $E = \hbar\omega$. For general observers, the quantization $S_\mathcal{O} = n\hbar$ ($n \in \mathbb{Z}_{>0}$) giving $E = n\hbar\omega$ is derived in Action and Planck’s Constant.

This is the Planck-Einstein relation.

Remark. This corollary depends on $S_{\min} = \hbar$ from a later derivation. The loop closure axiom provides only the structural relation $E = S/T$; the numerical value of $S_{\min}$ is fixed externally.

Consistency Model

Theorem 9.1. The loop closure axiom with structural postulate S1 (and its consequence, Theorem 0.1) is consistent: there exists a concrete model satisfying all conditions.

Model: $\Sigma = S^1$ (unit circle), $g = d\theta^2$ (standard metric), $\mathcal{H} = \mathbb{R}^2 \supset S^1$, $I(\theta) = 1$ (constant — the radius), $\mathcal{B} = S^1$ (the circle as self/non-self boundary), $\phi_t(\theta) = \theta + 2\pi t/T$ (rigid rotation with period $T > 0$), $\mathcal{C}(\theta) = c_0 > 0$ (constant).

Verification:

• (S1): $S^1$ is a compact 1-dimensional topological manifold. The standard rotation action $\phi_t(\theta) = \theta + 2\pi t/T$ is free (no fixed points). ✓
• Thm 0.0: $S^1$ has a unique smooth structure (elementary). The $U(1)$ action is smooth. ✓
• Thm 0.1: $d\theta^2$ is $\phi_t$-invariant (rotation is an isometry) — consistent with the Weyl-averaged metric, which for this model equals $d\theta^2$. ✓
• Def 1.1: $\phi$ is a smooth homomorphism $\mathbb{R} \to \text{Isom}(S^1)$, preserves $I \equiv 1$. ✓
• Axiom 3: $\phi_T(\theta) = \theta + 2\pi = \theta$, with $T_\mathcal{O} = T$. ✓
• Prop 3.2: $T > 0$ is the minimal period. ✓
• Prop 4.2: The orbit $\gamma = S^1$ is a smooth embedding. ✓
• Cor 4.3: $\rho(e^{i\alpha})(\theta) = \theta + \alpha$ is faithful. ✓
• Lyapunov stability: Rotation is an isometry, so $d(\phi_t(\sigma), \gamma) = d(\sigma, \gamma)$ — neutrally stable. ✓
• Coherence cost: $S = c_0 \cdot 2\pi > 0$. ✓

Remark: The model uses $I \equiv 1$ (constant), which satisfies invariant preservation ($I \circ \phi_t = I$) but has $|I(\Sigma)| = 1$. This model demonstrates the loop closure machinery (cyclic dynamics, U(1) action, stability, coherence cost); full Axiom 2 compliance (including N3: $|I(\Sigma)| > 1$) is demonstrated in the composite models of downstream derivations where $I$ is non-constant. $\square$

Consequences

1. Transient vs. persistent: Virtual particles (transient fluctuations) are loops that fail to close — their $\epsilon > \epsilon_c$. Real particles are loops with $\epsilon \ll \epsilon_c$ or $\epsilon = 0$.

2. Discrete spectrum: Only loops whose geometry is compatible with exact (or near-exact) closure persist. This selects a discrete set of allowed loop geometries, producing a discrete particle spectrum.

3. Mass = rest frequency: An observer at rest has its loop entirely in the temporal direction, with frequency $\omega = 2\pi/T$. Mass is $m = \hbar\omega/c^2 = E/c^2$ (using results from later derivations).

Rigor Assessment

Fully rigorous:

• Definitions 1.1, 2.1–2.2, 4.1, 6.1–6.2, 7.1: Precise mathematical definitions with explicit structural dependencies
• Proposition 2.3: Drift bound (complete proof by induction + triangle inequality)
• Proposition 2.5: Exact closure gives persistence (complete proof)
• Theorem 3.1: Persistence requires exact closure (proof from drift bound + compactness; see Remark on quasi-periodic gap)
• Proposition 3.2: Minimal period exists (proof via closed/discrete subgroup argument)
• Proposition 4.2: Loop is a smooth embedding $S^1 \hookrightarrow \Sigma$ (complete proof)
• Corollary 4.3: Faithful $U(1)$ action (complete proof of homomorphism + faithfulness)
• Proposition 4.4: Orbit decomposition (standard group action theory)
• Theorem 5.1: Noether pair equivalence (complete proof with explicit symplectic assumption)
• Proposition 5.2: Logical relationship between Axioms 2 and 3 (precise statement)
• Proposition 7.2: Positive coherence cost (complete proof from non-degeneracy + positivity)
• Theorem 9.1: Consistency model fully verified

Newly proved (this derivation):

• Theorem 0.2: State space is a finite-dimensional compact smooth manifold (from inductive construction via $U(1)$ Lie group orbits, bootstrap level-set mechanism, and finite local complexity from aperiodic order; replaces former Structural Postulate S1)
• Lemma 0.0: Free $U(1)$ orbits on $\Sigma$ (from Axiom 3 non-triviality + coherence cost positivity)
• Theorem 0.0: Smooth manifold structure on $\Sigma$ (from Theorem 0.2 + Lemma 0.0 + equivariant smoothing theorem)
• Theorem 0.1: $G_\mathcal{O}$-invariant Riemannian metric exists on $\Sigma$ (from Theorem 0.0 + compactness + Weyl averaging; replaces former Postulate S2)
• Theorem 3.1: Persistence forces exact closure (from drift bound + compactness, with quasi-periodic exclusion noted)

Now theorems (formerly structural postulates):

• S1 (Finite-dimensional manifold): Promoted to Theorem 0.2. Manifold structure derived from the constructive mechanism of observer state spaces: base case $S^1$ (Lie group), inductive step via level sets of relational invariants (regular value theorem), finite dimensionality from Delone finite local complexity.
• S2 (Invariant Riemannian metric): Promoted to Theorem 0.1 via Weyl averaging.

Remaining additional assumptions:

• Symplectic structure for Theorem 5.1 (additional assumption, stated in theorem)

Deferred to later derivations:

• $S_{\min} = \hbar$ (cf. Action and Planck’s Constant)
• $E = \hbar\omega$ numerical content (depends on $S_{\min}$ identification)
• Mass formula $m = E/c^2$ (depends on speed of light derivation)

Assessment: The derivation is restructured around the self-reference argument: persistence requires exact closure, which forces periodicity and $U(1)$ symmetry. All claims are proved from definitions and standard mathematical results, with explicit assumptions stated upfront. The quasi-periodic gap (Remark following Theorem 3.1) is acknowledged and resolved by the physical requirement of a discrete spectrum. Structural postulates are clearly separated from proven results. Results that depend on later derivations are explicitly deferred.

Open Gaps

1. Decoherence: Environmental perturbations increase $\epsilon$ over time, eventually breaking approximate closure. A quantitative decoherence rate should follow from the perturbation spectrum acting on the coherence measure.

2. Non-abelian loops: The formalization assumes $U(1)$ (abelian) symmetry. Non-abelian internal symmetries ($SU(2)$, $SU(3)$) require replacing the single period $T$ with a representation-theoretic condition on the image of $\phi$ in a non-abelian group.

3. Massless observers: Photons have no rest frame and no rest-frame loop. They are limiting cases where $T_\mathcal{O} \to 0$ and $\omega_\mathcal{O} \to \infty$ while $S_\mathcal{O}$ remains finite. This limit needs formal treatment within the approximate closure framework.