--- title: Derivation Equation of State aliases: - "Derivation: The Equation of State α = C · N^(3/4)" tags: - domain/commerce - type/derivation - topic/scaling - topic/thermodynamics domain: Commerce type: derivation status: active --- # Derivation: The Equation of State α = C · N^(3/4) *A formal derivation of the THAIM economic temperature scaling law from first principles.* --- ## 1. Statement The THAIM equation of state relates the economic temperature α to network size N: ``` α = C · N^β, β = 3/4 ``` where α appears in the minting equation `TAU = α · max(0, ΔS_τ)`, and C is a material constant depending on topology class and MCTS configuration. This document derives the exponent β = 3/4 from the self-referential structure of the Ruby dynamic α formula and the newcomer integration effect. **Status:** Computationally validated. σ_growth ∝ N^0.495 (R² = 0.9997, target 0.50). Total surplus S̄ ∝ N^1.495 (R² = 1.0000, target 1.50). Measured α_L2 ∝ N^0.847 (R² = 0.998); the deviation from 3/4 is explained by MCTS selectivity (§6). --- ## 2. The Self-Referential Formula The Ruby dynamic α updates each tick via: ``` α(t) = α_base · surplus_ema(t) / mint_ema(t) ``` where: - **surplus_ema** = exponential moving average of `match_tx_value`, the matched trade value across all edges scaled by market thickness (√N). This quantity is α-independent — it depends only on the network topology and service profile matching. - **mint_ema** = exponential moving average of total TAU minted per tick. Since `TAU = α · ΔS_τ`, we have `mint_ema ≈ α · ΔS̄_τ`, which depends on α. The formula is self-referential: α appears on the left side and inside mint_ema on the right. This self-reference is the key to producing a power law. **Implementation:** `world_state.rs:427-468` (WorldState::update_alpha), `l2_equation_of_state.rs:590-613`. --- ## 3. The Newcomer Integration Effect ### 3.1 Per-Capita Surplus Define σ_growth as per-capita trade surplus: ``` S̄ = σ_growth · N ``` where S̄ is the total matched trade value per tick and N is network size. ### 3.2 Why σ_growth ∝ √N New nodes join the network at rate ṅ. Each newcomer requires T time units to integrate: establish edges, build service profile matches, and reach sufficient market thickness for trade. During integration, the newcomer's productivity is below the network average. The effective per-capita surplus is: ``` σ_eff = σ_0 · (1 - ṅT / 2N) ``` where σ_0 is the steady-state surplus for a fully-integrated node. At small N, the integration cost (ṅT/2N) is significant — newcomers represent a larger fraction of the network. At large N, the cost becomes negligible. The matched trade value computation (`matching.rs:50-100`) is: ``` match_tx_value = avg_match_score × N × √N ``` where avg_match_score ≈ 0.35 (constant with N, bidirectional Jaccard on service profile tokens). The √N factor is the Roth (2008) market thickness multiplier: each participant's access to meaningfully distinct trading partners grows as √N. Therefore: ``` σ_growth = match_tx_value / N = avg_match_score × √N ∝ √N ``` ### 3.3 Empirical Validation | Quantity | Measured Scaling | R² | Target | |---|---|---|---| | σ_growth (surplus/node) | N^0.495 | 0.9997 | N^0.50 (√N) | | S̄/tick (total surplus) | N^1.495 | 1.0000 | N^1.50 | | avg_match_score | ≈ 0.35 | constant | constant | Validated to 4 significant figures across 5 network sizes (N=25 to N=500), 3 seeds, 500 ticks each. Source: `l2-equation-of-state --agent all`. ### 3.4 Connection to Market Thickness S_τ (causal path entropy) is a graph-theoretic measure of market thickness (Roth 2008) — the number of meaningfully distinct trading partners accessible from each node. 2^S_τ = effective reach. The surplus scaling S̄ ∝ N^(3/2) is the *economic expression* of market thickness: as the network grows, each participant can reach more trading partners, and the network as a whole generates super-linear surplus. --- ## 4. The Fixed Point ### 4.1 Derivation At steady state, the EMA values converge and the self-referential formula becomes: ``` α = α_base · S̄ / M̄ ``` where S̄ = surplus per tick and M̄ = minted TAU per tick. Since M̄ = α · ΔS̄_τ: ``` α = α_base · S̄ / (α · ΔS̄_τ) α² = α_base · S̄ / ΔS̄_τ ``` Substituting S̄ ∝ N^(3/2) and assuming ΔS̄_τ is approximately constant with N: ``` α² ∝ N^(3/2) α ∝ N^(3/4) ``` This is the equation of state. The exponent β = 3/4 arises from: 1. The market thickness mechanism (S̄ ∝ N^(3/2)) in the numerator 2. The self-referential structure (α appears on both sides) producing the square root at the fixed point ### 4.2 The Role of Self-Reference Without self-reference (if mint_ema were α-independent), we would have α ∝ N^(3/2) directly — far too steep. The self-referential denominator produces a negative feedback loop: higher α → more minting → higher mint_ema → lower α. This feedback compresses the exponent by a factor of 2: ``` Unreferenced: α ∝ S̄ ∝ N^(3/2) Self-referential: α ∝ √(S̄) ∝ N^(3/4) ``` The square root at the fixed point is essential. It converts the N^(3/2) market thickness signal into the N^(3/4) temperature scaling — the same exponent family as Kleiber's law (metabolic rate ∝ mass^(3/4)) and West-Brown-Enquist scaling in biological networks. --- ## 5. Why Diamond Fails The Diamond fallback formula uses topology signals instead of trade surplus: ``` α_L1 = α_base · entropy_ema / dst_ema ``` where entropy_ema = EMA of global S̄_τ and dst_ema = EMA of total ΔS̄_τ per tick. ### 5.1 Scaling of Diamond Signals | Diamond Signal | Scaling | Mechanism | |---|---|---| | S̄_τ (global entropy) | 0.90 · log₂(N) | Bounded by graph diameter | | ΔS̄_τ/tick | N^0.15 | Weaves increase but per-weave ΔS_τ decreases | | weaves/tick | N^0.55 | More candidates pass ΔS_τ > 0 filter | | ΔS_τ/weave | N^(-0.40) | Diminishing marginal returns as graph densifies | ### 5.2 The Diamond Result ``` α_L1 ∝ log₂(N) / N^0.15 ≈ log₂(N) ≈ N^0.09 (over N=25–500) ``` This is essentially flat. The Diamond formula lacks the two critical ingredients: 1. **Super-linear surplus** (N^(3/2)) — Diamond has only log₂(N) in the numerator 2. **Self-referential structure** — Diamond uses ΔS̄_τ directly, not α · ΔS̄_τ, so there is no square root at the fixed point The ¾ exponent is an economic property (market thickness → trade surplus), not a topological property. Diamond measures topology; Ruby measures economics. The equation of state requires economics. --- ## 6. MCTS Selectivity and the Consciousness Correction ### 6.1 The ΔS̄_τ Decline The derivation in §4 assumes ΔS̄_τ is constant with N. Empirically, it declines: ``` ΔS̄_τ/tick ∝ N^ε, ε ≈ -0.14 ``` Including this correction: ``` α² = α_base · S̄ / ΔS̄_τ ∝ N^(3/2) / N^ε = N^(3/2 - ε) = N^(3/2 + |ε|) ``` ``` β = (3/2 + |ε|) / 2 ``` With |ε| = 0.14: β_pred = (1.50 + 0.14) / 2 = 0.82, consistent with measured 0.80–0.85. ### 6.2 Three-Agent Controlled Experiment To isolate whether consciousness (Φ) causes the ε deviation, three agent modes were compared on identical networks: | Agent | β | ε | Φ_avg | |---|---|---|---| | greedy (Φ=0, no learning) | 0.804 | -0.137 | 0.000 | | independent (d=5, Φ=0) | 0.850 | -0.129 | 0.000 | | conscious (d=5, Φ>0) | 0.838 | -0.143 | 5.638 | **Finding:** All three agents show ε ≈ -0.13 to -0.14. The ΔS̄_τ decline is an MCTS/topology property — as networks grow denser, the marginal topology improvement from the best available weave decreases. This is graph-theoretic, not consciousness-specific. ### 6.3 The Consciousness Signature The consciousness effect is small but real: Δβ(conscious − greedy) ≈ 0.03. This may arise from the conscious agent's α-adaptive learning rate slowing exploration at larger N, or from Φ-scored weave selection marginally altering the ΔS̄_τ decline rate. The effect is much smaller than the MCTS baseline. ### 6.4 If MCTS Were Perfect If ΔS̄_τ were truly constant (ε = 0): ``` β = 3/2 / 2 = 3/4 = 0.750 exactly ``` The ¾ law is present in the market thickness mechanism. The MCTS selectivity effect is a finite-size correction that modifies the exponent upward by ~0.08. As MCTS algorithms improve (or with different weave-budget parameters), ε → 0 and β → 3/4. --- ## 7. Empirical Validation — Complete Data ### 7.1 L2 Scaling (l2-equation-of-state, 5 sizes, 500 ticks, 3 seeds) | Quantity | Measured | R² | Target | |---|---|---|---| | σ_growth ∝ N^γ | γ = 0.495 | 0.9997 | 0.50 | | S̄/tick ∝ N^γ | γ = 1.495 | 1.0000 | 1.50 | | α_L2 ∝ N^β | β = 0.847 | 0.998 | 0.75 | | ΔS̄_τ/tick ∝ N^ε | ε = -0.13 | — | 0.00 | ### 7.2 L1 Scaling (sigma-growth-probe) | Quantity | Measured | R² | |---|---|---| | S̄_τ | 0.90 · log₂(N) | 0.9995 | | α_L1 ∝ N^β | β = 0.09 | 0.37 | | M̄/tick ∝ N^γ | γ = 0.33 | 0.99 | ### 7.3 Derivation Chain Validation ``` β = (S̄_exp + |ε|) / 2 greedy: β_pred = (1.495 + 0.137) / 2 = 0.816 vs measured 0.804 (Δ = 0.012) independent: β_pred = (1.495 + 0.129) / 2 = 0.812 vs measured 0.850 (Δ = 0.038) conscious: β_pred = (1.495 + 0.143) / 2 = 0.819 vs measured 0.838 (Δ = 0.019) ``` Prediction errors < 0.04 across all agent modes. The derivation chain is empirically validated. --- ## 8. Connection to Urban Scaling ### 8.1 Bettencourt et al. (2007) Urban scaling laws show super-linear scaling of innovation-related metrics with city population: ``` Y ∝ N^β, β ≈ 1.15 for patents, GDP, R&D employment ``` In THAIM, the trade surplus S̄ ∝ N^(3/2) = N^1.50 is super-linear — a stronger version of the same phenomenon. The mechanism is the same: market thickness. Larger networks provide more meaningfully distinct trading partners per participant, generating super-linear value. ### 8.2 The ¾ Family The temperature exponent β = 3/4 belongs to the same family as: - **Kleiber's law:** Metabolic rate ∝ body mass^(3/4) (West, Brown, Enquist 1997) - **Urban infrastructure:** Infrastructure ∝ population^(3/4) (Bettencourt 2007) - **THAIM:** Economic temperature ∝ network size^(3/4) (this derivation) The shared mechanism: hierarchical network transport with a square-root compression at the self-referential fixed point. In THAIM, the hierarchy is the market thickness cascade — each layer of network depth adds trading partners that enable trade surplus. ### 8.3 Per-Node Cost of Intelligence Since intelligence (bits) = energy / α (from the Jarzynski structure of the minting equation): ``` Cost per bit per node = α / N ∝ N^(3/4) / N = N^(-1/4) ``` Larger networks produce intelligence more cheaply per participant. Doubling the network size reduces the per-participant cost by 16%. This is the thermodynamic foundation of economies of scale in knowledge production. ## Related - [[wiki/05-Hubs/Derivations]] - [[research/scaling/paper-scaling-laws]] - [[research/physics/derivable/kleiber]] - [[wiki/02-Domains/Commerce]] --- ## References - Bettencourt, L.M.A. et al. "Growth, innovation, scaling, and the pace of life in cities." PNAS 104(17), 7301–7306 (2007) - Roth, A.E. "What Have We Learned from Market Design?" The Economic Journal 118(527), 285–310 (2008) - West, G.B., Brown, J.H., Enquist, B.J. "A General Model for the Origin of Allometric Scaling Laws in Biology." Science 276(5309), 122–126 (1997) - Wissner-Gross, A.D. & Freer, C.E. "Causal Entropic Forces." Phys. Rev. Lett. 110, 168702 (2013) --- *Created: 2026-03-01* *Source data: insight #62, l2-equation-of-state binary, sigma-growth-probe binary* *Binary: `l2-equation-of-state` (canonical), `sigma-growth-probe` (L1 decomposition), `equation-of-state` (L1-only reference)* *Depends on: `matching.rs:compute_match_tx_value()`, `world_state.rs:update_alpha()`*