Introduction

Epimechanics is a framework that models everything — physical systems, biological organisms, economies, minds — as self-representing causal structures. It rests on seven foundational definitions. This document shows that when those definitions are quotiented by three declared equivalences, the resulting dependency graph converges to a four-node fixed point whose spectral radius is $\varphi = (1+\sqrt{5})/2$ , the golden ratio. The first two equivalences formalize co-primitive pairings; the third ( $E \cong R$ ) is a well-motivated postulate. The golden ratio is a structural consequence of this definitional architecture — real, but specific to the graph it produces.

The Seven Definitions

Epimechanics defines seven primitive concepts, organized into three groups:

Group	Definitions	Role
Co-primitive pair 1	Causation (C), Entity (E)	Ontological ground: what exists and what connects
Bridge	State (S)	The inaccessible referent that representations approximate
Co-primitive pair 2	Computation (Co), Representation (R)	Epistemological ground: how approximations are produced
Derived pair	Information (I), Prediction (P)	Evaluative: how good the approximation is, and what it implies

Each definition in one sentence:

Causation: the relation by which one entity's state constrains another's.
Entity: anything with causal presence; equivalently, anything representable.
State: what determines an entity's causal dispositions; the posited referent that representations approximate.
Computation: producing a representation — discriminating, partitioning, compressing.
Representation: the product of computation; an approximation of state.
Information: predictive information — the degree to which a representation's distinctions reduce uncertainty about future states, above background noise.
Prediction: reduction of uncertainty about future states given a representation's information content.

These definitions form a directed graph with 7 nodes and 17 edges (entry $M_{ij} = 1$ when definition $i$ references definition $j$ ). The adjacency matrix is:

$A_7 = \begin{pmatrix} & C & E & S & \text{Co} & R & I & P \\ C & 0 & 1 & 1 & 0 & 0 & 0 & 0 \\ E & 1 & 0 & 0 & 0 & 1 & 0 & 0 \\ S & 1 & 1 & 0 & 0 & 1 & 0 & 0 \\ \text{Co} & 0 & 0 & 1 & 0 & 1 & 0 & 0 \\ R & 0 & 0 & 1 & 1 & 0 & 0 & 0 \\ I & 0 & 0 & 1 & 0 & 1 & 0 & 1 \\ P & 0 & 1 & 1 & 0 & 0 & 1 & 0 \end{pmatrix}$

The graph is strongly connected with diameter 5. Every definition is reachable from every other. The spectral radius of $A_7$ is $\approx 2.247$ .

The Three Equivalences

Three equivalences are declared among these definitions:

1. Causation $\cong$ Entity (co-definitional). Causation is defined as what connects entities; entities are defined as what causation connects. Neither is prior. They are two descriptions of a single primitive.

2. Computation $\cong$ Representation (co-definitional). Computation is the process of producing a representation; a representation is the product of computation. Again, neither is prior.

3. Entity $\cong$ Representation (well-motivated postulate). If something has causal presence, it can — at least in principle — be represented, because detecting its causal effects produces a distinction. If something can be represented, it has causal presence, because representation is a causal interaction.

$E \cong R$ is a well-motivated postulate — hard to deny for entities we can interact with, fruitful in its consequences, but not logically necessary. Dark matter has causal presence detectable only through bulk effects: we can name our ignorance but cannot construct a structural representation. Whether "barely representable" is sufficient for the equivalence is a substantive question. The golden ratio follows from this postulate. If the postulate is wrong, the spectral radius stays at $2.247$ .

The Quotient Cascade

Quotienting by $X \cong Y$ means: merge $X$ and $Y$ into a single node $[XY]$ , taking the union of their edges.

Step 1: $C \cong E$

Merge C and E into $[\text{CE}]$ . The graph reduces from 7 to 6 nodes: $\{[\text{CE}], S, \text{Co}, R, I, P\}$ .

$A_6 = \begin{pmatrix} & [\text{CE}] & S & \text{Co} & R & I & P \\ [\text{CE}] & 1 & 1 & 0 & 1 & 0 & 0 \\ S & 1 & 0 & 0 & 1 & 0 & 0 \\ \text{Co} & 0 & 1 & 0 & 1 & 0 & 0 \\ R & 0 & 1 & 1 & 0 & 0 & 0 \\ I & 0 & 1 & 0 & 1 & 0 & 1 \\ P & 1 & 1 & 0 & 0 & 1 & 0 \end{pmatrix}$

The self-loop at $[\text{CE}]$ arises because C $\to$ E and E $\to$ C both exist. Spectral radius: $\approx 2.247$ .

Step 2: $\text{Co} \cong R$

Merge Co and R into $[\text{CoR}]$ . The graph reduces to 5 nodes: $\{[\text{CE}], S, [\text{CoR}], I, P\}$ .

$A_5 = \begin{pmatrix} & [\text{CoR}] & [\text{CE}] & S & I & P \\ [\text{CoR}] & 1 & 0 & 1 & 0 & 0 \\ [\text{CE}] & 1 & 1 & 1 & 0 & 0 \\ S & 1 & 1 & 0 & 0 & 0 \\ I & 1 & 0 & 1 & 0 & 1 \\ P & 0 & 1 & 1 & 1 & 0 \end{pmatrix}$

The self-loop at $[\text{CoR}]$ arises because Co $\to$ R and R $\to$ Co both exist. Spectral radius: $\approx 2.247$ .

Step 3: $E \cong R$

Since E lives in $[\text{CE}]$ and R lives in $[\text{CoR}]$ , this equivalence forces $[\text{CE}] \cong [\text{CoR}]$ . Merge them into a single node $\Phi$ . The graph reduces to 4 nodes: $\{\Phi, S, I, P\}$ .

No further quotienting is possible. The cascade has reached a fixed point.

The Fixed-Point Graph

The converged adjacency matrix is:

$A_\star = \begin{pmatrix} & \Phi & S & I & P \\ \Phi & 1 & 1 & 0 & 0 \\ S & 1 & 0 & 0 & 0 \\ I & 1 & 1 & 0 & 1 \\ P & 1 & 1 & 1 & 0 \end{pmatrix}$

The four nodes and their interpretations:

$\Phi$ — the fused causal-representational primitive, unifying causation, entity, computation, and representation. It is self-referential ( $\Phi \to \Phi$ ) because representing requires entities which require causation which requires entities.
$S$ — state. The inaccessible referent that $\Phi$ approximates. $S \neq \Phi$ because representation is not reality.
$I$ — information. How well the representation approximates the state.
$P$ — prediction. Using information to constrain expectations about future states.

Spectral Analysis

The characteristic polynomial of $A_\star$ is:

$\det(A_\star - \lambda \mathbf{1}) = \lambda^4 - \lambda^3 - 2\lambda^2 + \lambda + 1 = 0$

This factors cleanly:

$(\lambda^2 - \lambda - 1)(\lambda^2 - 1) = 0$

The four eigenvalues are:

Eigenvalue	Approximate value	Source
$\varphi = \frac{1+\sqrt{5}}{2}$	$1.618$	$\Phi \leftrightarrow S$ subblock
$1$	$1.000$	$I \leftrightarrow P$ interaction
$-1$	$-1.000$	$I \leftrightarrow P$ interaction
$-1/\varphi = \frac{1-\sqrt{5}}{2}$	$-0.618$	$\Phi \leftrightarrow S$ subblock

Spectral radius: $\rho(A_\star) = \varphi \approx 1.618$ .

Determinant: $\det(A_\star) = \varphi \cdot 1 \cdot (-1) \cdot (-1/\varphi) = 1$ . The mapping preserves volume.

Trace: $\text{tr}(A_\star) = 1$ . Only $\Phi$ is self-referential among the four nodes.

Why the golden ratio?

The $\Phi \leftrightarrow S$ subblock is:

$\begin{pmatrix} 1 & 1 \\ 1 & 0 \end{pmatrix}$

This is the Fibonacci matrix (Graham, Knuth, Patashnik, Concrete Mathematics, 1994). Its characteristic polynomial is $\lambda^2 - \lambda - 1 = 0$ , whose positive root is $\varphi$ . The golden ratio appears whenever the next state depends on both the current state and the previous one: self-referential growth with memory depth 2.

The spectral radius of a static adjacency matrix governs walk counts (length- $k$ walks grow as $\varphi^k$ ), not dynamical growth. A dynamical interpretation requires treating the definitional system as iterated self-application, which is open work.

For background on spectral methods: Cvetkovic, Rowlinson, and Simic, An Introduction to the Theory of Graph Spectra (2010). On self-referential structures in category theory: Lawvere's fixed-point theorem (1969); Yanofsky, "A Universal Approach to Self-Referential Paradoxes" (2003).

The $E \cong R$ Equivalence as Spectral Taming

The three equivalences do not contribute equally to the convergence.

Stage	Nodes	Spectral radius
Original graph	7	$2.247$
After $C \cong E$	6	$2.247$
After $\text{Co} \cong R$	5	$2.247$
After $E \cong R$	4	$\varphi \approx 1.618$

The first two equivalences leave the spectral radius unchanged. The third — $E \cong R$ — drops it to $\varphi$ , a $28\%$ reduction. This taming is a structural consequence of this definitional architecture.

Robustness

The result depends on the specific 17 edges of $A_7$ . Flipping a single edge generically changes the eigenvalues.

The golden ratio survives if the $\Phi \leftrightarrow S$ subblock $\begin{pmatrix} 1 & 1 \\ 1 & 0 \end{pmatrix}$ is preserved — which requires that $\Phi$ has a self-loop (a self-referential primitive) and $\Phi \leftrightarrow S$ is bidirectional (representation approximates state, state constrains representation).

These two structural features — self-referential primitive plus bidirectional bridge — are arguably the core architecture, not an encoding choice. But this argument should be formalized before claiming universality.

Interpretation

The converged graph tells a four-part story:

The growing mode ( $|\lambda| = \varphi$ ): The $\Phi \leftrightarrow S$ cycle dominates walk counts ( $\sim \varphi^k$ for length- $k$ walks). Whether this governs dynamical growth requires a dynamical formulation, which remains open.
The persistent mode ( $|\lambda| = 1$ ): The $I \leftrightarrow P$ pair neither grows nor decays, forming a stable evaluative scaffold. Connects to self-multiplication, where entities are eigenvectors with $|\lambda| = 1$ .
The decaying mode ( $|\lambda| = 1/\varphi \approx 0.618$ ): Non-self-referential structure is transient in walk-count terms.
Volume preservation ( $\det = 1$ ): The definitional system as a whole neither creates nor destroys information. Growth in one mode is exactly compensated by decay in another.

[!sidenote] Parallel to second quantization. The co-primitive structure resembles QFT's algebraic structure (entities as states, causation as operators). The spectral structure suggests an algebraic constraint on $\Phi \leftrightarrow S$ involving $\varphi$ , but the precise form is open.

Open Questions

Is there an algebraic constraint? The Fibonacci matrix governs the $\Phi \leftrightarrow S$ subblock. Is there a deeper algebraic relation (analogous to $[a, a^\dagger] = 1$ ) from which this matrix can be derived rather than read off?
Does the golden ratio constrain dynamics? The spectral radius $\varphi$ governs walk counts on the definitional graph. Does it also constrain the dynamics of specific systems modeled within Epimechanics — for instance, bounding auto-causal density growth rates? This requires a dynamical formulation.
Is the fixed point unique? The quotient cascade reaches a 4-node graph that admits no further quotienting. Is this the unique fixed point for any sequence of valid equivalences on the original 7-node graph, or are there alternative equivalence declarations that lead to different fixed points with different spectral properties?
Is the golden ratio robust to reasonable variations in the dependency encoding? The Robustness section identifies the $\Phi \leftrightarrow S$ subblock as critical, but a systematic sensitivity analysis is needed.

The Golden Ratio in Epimechanics' Definitional Structure

Contribute to Epiphysics