The Pirangi cashew tree — cajueiro — is a single organism that covers more than eight thousand square metres of ground. It grew from one seed. The seed does not contain the forest. It contains the instruction that, given the right medium, inflates into the forest. This is not metaphor. It is the exact mechanism by which the dm³ framework operates: a compressed geometric instruction delivered to a receptive medium, which unfolds through a ascent whose stable endpoint is the target structure. These theorems formalise that mechanism.
How does a single photon build a protein? How does a three-base codon specify a fold with three hundred degrees of freedom? How does a seed become a forest? The naive answer — the instruction must contain the full description of the output — is wrong, and provably so.
Consider the degrees of freedom gap. A structured light beam (optical vortex, topological mode) carries perhaps ten independent parameters: frequency, wavevector direction, polarisation, orbital angular momentum quantum number ℓ, intensity. A single protein of one hundred residues has approximately three hundred backbone dihedral angles specifying its fold, plus one hundred side-chain identity choices from twenty amino acid types. The ratio is at least one order of magnitude: d_O / d_I > 10.
This is not a gap to be crossed — it is a necessary feature of efficient generative systems. The compression of the instruction is what makes it transmissible. The expansion is what happens in the medium. Nature solved this four billion years ago with the ribosome. The dm³ framework formalises the principle.
The Pirangi cashew (Anacardium occidentale, Natal, Brazil) is the world's largest individual fruit tree: one organism, one root system, covering 8,500 m² of ground, producing 80,000 fruits per year. Its seed — the cashew nut — contains roughly 10 kB of heritable information (the operative portion of its genome). The tree at maturity encodes approximately 10¹⁷ atoms arranged in a specific topology. The compression ratio between instruction and output is approximately 10¹³. The medium (soil, water, sunlight, atmosphere) provides the materials. The seed provides the selection rule: which attractor, of all possible arrangements of those atoms, to converge to.
In the contact bundle J¹(ℝ³) with contact form α = dz − p·dq, a photon propagating along trajectory γ defines a Legendre submanifold Λ_γ ⊂ J¹(ℝ³) via the lift (q(t), p(t) = ℏk(t), z(t) = ∫ℏω dt). The mode numbers (ℓ, σ) specify the helical phase structure of the wavefront — equivalently, the curvature of the Legendre submanifold in the fiber direction.
At the moment of absorption at q₀, the photon delivers the contact element (q₀, ℏk₀, (ℓ+σ)ℏ) to the absorber. The absorber's internal dynamics are perturbed by this element. Two modes with ℓ₁ ≠ ℓ₂ deliver different angular momentum to the fiber coordinate — they activate different normal modes of the absorber (selection rules for rotational transitions are exactly ΔJ = ℓ ± σ). Therefore they are distinct instructions. The photon is a section of the contact bundle that specifies a preferred direction in the full contact geometry of the absorber. □
By contradiction. Suppose no inflation mechanism exists. Then the map f: I → O is direct. By the data-processing inequality, mutual information satisfies I(I; O) ≤ H(I) ≤ d_I · log₂(K) for some maximum K values per degree of freedom.
If d_O > 10 · d_I, the minimum description complexity of O — the minimum program length to specify O — exceeds H(I) by more than one order of magnitude (for comparable K). Therefore H(I) < H(O)/10, and the map f cannot determine O uniquely from I. But O is a definite state (a specific protein fold, a specific organism, a specific crystal). Its degrees of freedom must be determined by something. That something must reside in the medium — as stable attractors that are consistent with I and complete the specification. This is precisely the inflation mechanism M. M exists necessarily. □
Remark. The ribosome is M for protein synthesis (I = mRNA codon sequence, O = folded protein, d_O/d_I ≈ 50). The rhizome is M for the cajueiro (I = seed genome, O = tree topology, d_O/d_I ≈ 10¹³). Rhodopsin is M for visual transduction (I = one photon, O = neural spike train, d_O/d_I ≈ 10⁶).
Let V: M → ℝ be the Lyapunov function for the dm³ flow. On B(A), V is strictly decreasing along trajectories. Let x₀ be the initial state and let V(∂B(A)) = V_c be the critical value at the basin boundary.
The instruction I must supply a perturbation δx such that V(x₀ + δx) < V_c. This is a condition on δx alone — it requires pushing x₀ across the level set {V = V_c}. The minimum norm of such δx is dist(x₀, ∂B(A)), which depends only on the geometry of ∂B(A), not on the internal structure of A.
In the dm³ framework, ε₀ = 1/3 bounds the stability radius: the basin B(A) contains a ball of radius ε₀ around A. Therefore, if x₀ is within ε₀ of A, no instruction is needed — the system already converges. If x₀ is outside, the minimum instruction complexity is determined by the single real parameter dist(x₀, ∂B(A)). This is O(1) in the number of real parameters, and is O(ε₀⁻¹) = O(3) at the basin boundary. It is independent of d_O. □
Corollary. The genetic code does not need to contain the protein fold. It contains enough information to select the attractor (the native fold) of the folding energy landscape. The landscape does the rest. The minimum codon-level information needed per amino acid is log₂(20) ≈ 4.3 bits — far less than the bits needed to specify all dihedral angles.
Convergence of the η-weighted sum. Since η ≈ 1.8393 > 1, we have |η⁻¹| ≈ 0.5437 < 1. The geometric series Σ_{k=0}^∞ η⁻ᵏ converges by the ratio test, with sum 1/(1−η⁻¹) = η/(η−1).
Numerically: η/(η−1) = 1.8393/0.8393 ≈ 2.191.
Self-similarity. The subseries starting at level k is Σ_{j=k}^∞ η⁻ʲ = η⁻ᵏ · Σ_{j=0}^∞ η⁻ʲ = η⁻ᵏ · η/(η−1). The sum of the ascent from level k onward is exactly η⁻ᵏ times the full ascent-to-threshold sum. The structure at every level is the same series scaled by η⁻ᵏ. This is the algebraic expression of self-similarity: the fruit contains a seed which generates, when planted, an ascent to the threshold of the same type scaled by η⁻¹.
Convergence to τ. Let ηₙ denote the n-bonacci constant (η₂ = φ, η₃ = η, η₄ = Δ, η₅ = Σ, η₆ = Ω, η_∞ = τ = 2). Then ηₙ/(ηₙ−1) → 2/(2−1) = 2 = τ as n → ∞. The ascent-to-threshold sum converges to the convergence threshold. The forest is bounded. Its limit is τ. □
Corollary (Cajueiro Ratio). The ratio of the total ascent-to-threshold output to the seed instruction is η/(η−1) ≈ 2.191. This is the Cajueiro ratio: the factor by which the medium inflates the seed. At the limit of the n-bonacci ladder, the Cajueiro ratio is exactly 2 = τ.
The architecture is not additive construction. The seed does not add itself to more seed. It activates a medium — soil, cytoplasm, photon field — which has pre-existing attractor structure. The instruction's job is solely to select which attractor. Once selected, the medium assembles the output autonomously.
The dm³ operator chain G = U∘F∘K∘C is the formalisation of this architecture. C (Contact/Compression) is the seed — the compressed instruction. K (Konvergence) is the cascade activation. F (Fold/Catastrophe) is the transition through the basin boundary. U (Unfolding) is the expansion into the full output state. The output at τ = 2 is the fold-free attractor: the tree.
| dm³ Operator | Cajueiro Analogue | Physical Instance | Compression/Expansion |
|---|---|---|---|
| C — Compression | Seed encoding | Genome, codon sequence, photon mode | Compress: d_I ≪ d_O |
| K — Konvergence | Germination trigger | Ribosome, rhodopsin, photosynthesis | Activate medium |
| F — Fold | Basin crossing | Protein folding transition state | Threshold crossing at ε₀ |
| U — Unfolding | Growth — ascent to the threshold | Gene expression, morphogenesis | Expand: attractor fills d_O |
| G = U∘F∘K∘C | Seed → forest | Full generative transition | Cajueiro ratio η/(η−1) → τ |
All four theorems mechanised in AXLE (Algebraic eXpression Language for Evaluation). Repository: github.com/TOTOGT/AXLE. Zero sorry.
Rhodopsin is the vertebrate photoreceptor. A single photon (d_I = 1 mode) triggers isomerisation of 11-cis-retinal to all-trans-retinal in under 200 femtoseconds. This conformational change (d_O ~ 3 dihedral angles) activates a G-protein cascade that amplifies to approximately 10⁶ cGMP hydrolysis events per second per activated rhodopsin. One photon → one neural spike. The Cajueiro ratio here is ~10⁶. The photon is the seed. The cascade is the forest. Theorem C2 applies: the gap (10⁵ in DOF) is spanned by the medium inflation mechanism (the GPCR signalling cascade).
The ribosome reads three nucleotides (one codon, d_I = 6 bits of information per amino acid position) and selects one of twenty amino acid types to attach to the growing chain. The native fold of a 300-residue protein has approximately 600 backbone dihedral angles plus 300 side-chain identity choices — far more information than the codon provides. The medium inflation mechanism is the folding energy landscape: an attractor that is pre-existing in the physics of the amino acid sequence. Theorem C3 applies: the instruction (codon sequence) selects the attractor basin. The basin does the rest.
A photon at 680nm (Photosystem II) or 700nm (Photosystem I) triggers an electron transfer chain that ultimately reduces CO₂ to carbohydrate. One photon deposits ~1.8 eV. The synthesis of one glucose molecule requires ~2880 kJ/mol — approximately 18 eV total (8-10 photons, one per step). The photon is a partial instruction that, combined with the enzyme machinery, selects and drives the specific metabolic cascade. d_I per photon ≈ 1. d_O of one glucose molecule ≈ 24 degrees of freedom (bonds, stereocenters). The inflation mechanism is the enzyme complex. Theorem C1 applies: the photon at 680nm and at 700nm are distinct geometric commands — they activate different reaction centres.
One seed. 8,500 m² of ground. 80,000 fruits per year. The seed contains the instruction — the genome — that selects the tree-shaped attractor from all possible arrangements of carbon, hydrogen, oxygen, nitrogen, and mineral atoms available in the soil. The soil is the medium. The cascade is the organism. The Cajueiro ratio for this system: d_O/d_I ≈ 10¹³. The forest is bounded. Its limit — the adult organism at full expression — is the attractor. It does not grow forever. It converges to τ.
The four theorems in this chapter span the series by volume:
| Theorem | Volume | Primary Connection |
|---|---|---|
| C1 — Geometric Command | Book 2 · TOGT | Generative transitions via photon contact elements |
| C2 — Medium Inflation | Book 3 · Mini-Beast | Biological cascade: autophagy, protein assembly, triple-alpha |
| C3 — Attractor Selection | Book 3 · Mini-Beast | Energy landscape, folding, ε₀ = 1/3 stability radius |
| C7 — Seed Compression | Book 5 (series-wide) | The η-weighted sum as the unified recursion across all volumes |
The Naming Gate theorem (C6 — perceptual access requires nomenclature) is housed in chPsi-quantum-mind.html as an epistemological theorem connecting consciousness, language, and the dm³ framework. The S³/H4 theorems (C4 — Generative Equidistance, C5 — Chord-Arc Shortcut) are housed in chE-gtct.html as the Galilean coordinate transformation theorems.