Contribution
A bound tying the precision of a steady-state output to its dissipation: Q ≡ σ t ε² ≥ 2k_BT, where σ is the entropy production rate, t the observation time and ε² = Var X/⟨X⟩² the squared relative uncertainty of a current X (eqs. 4, 10). The bound is time-independent, because both ε² and σt scale with t. To reach 1% precision needs at least 20,000 k_BT of dissipation (p. 2; 2/10⁻⁴, checked). The letter gave the relation its name; its general proof came later and is not in this paper.
Key insight
Precision is bought with dissipation, at a fixed exchange rate. A nonequilibrium current is a biased count of transitions. Near equilibrium its mean grows with the affinity and its variance does not shrink, so making its relative noise small requires driving it hard or for long, and both cost entropy production in proportion.
Assumptions
- A finite Markov jump network in a nonequilibrium steady state.
- Thermodynamic consistency: every transition has its reverse (“if kᵢⱼ ≠ 0 then kⱼᵢ ≠ 0”, p. 2) and local detailed balance holds, ln(kᵢⱼ/kⱼᵢ) = Σ_β dᵢⱼ^β A_β + Eᵢ − Eⱼ (eq. 5). Irreversible transitions put the affinity at infinity, where the bound is trivially met.
- The observable is a current, a count of transitions weighted by an antisymmetric increment dᵢⱼ. Occupation-time observables are excluded explicitly: their uncertainty “is finite even in equilibrium” (p. 1).
- Steady state, long times: J and D are the stationary drift and diffusion coefficient of X (eqs. 6–7).
Key results
- Biased walk (eq. 4). Q = A coth(A/2k_BT) ≥ 2k_BT, increasing in A, so the cheapest precision is near equilibrium, where however the time to produce output diverges (p. 2). Checked.
- Linear response, any network (eqs. 11–13). With the Einstein relation D_α = L_αα, Q_α = 2[1 + Σ G_βγ A_β A_γ / (Σ L_αβ L_αγ A_β A_γ)] with G_βγ = L_αα L_βγ − L_αβ L_αγ positive semidefinite (supplement §I, by Sylvester’s bordered-determinant identity). So Q_α ≥ 2, with equality when only A_α is nonzero.
- Unicyclic networks, any affinity (eq. 20). For N states and cycle affinity A, Q ≥ (A/N) coth(A/2N) ≥ 2, the minimum at uniform rates. The supplement shows uniform rates are a stationary point of the relevant ratio and says “we are not able to provide a rigorous proof of this uniqueness” (§III). So the unicyclic bound is proved up to that gap. Checked at N = 3 (A = 2, 10) and N = 5 (A = 4): Q agrees with the formula to six digits.
- Fano factor (eq. 22). F ≥ (1/N) coth(A/2N), sharper than the known F ≥ 1/N when the affinity is known; a diagnostic for the number of states in an enzyme cycle.
- Multicyclic networks beyond linear response. Numerical only: one two-substrate enzyme and six further networks, rates random and minimised; “we conjecture that Q_α ≥ 2 also for general multicyclic networks beyond linear response” (p. 4). My random check of the complete four-state graph found min Q = 2.08 over 2,998 draws.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | Q ≥ 2k_BT within linear response for any network | strong | eq. 13 with supplement §I |
| C2 | Q ≥ (A/N) coth(A/2N) ≥ 2 for unicyclic networks at any affinity | strong, with a uniqueness gap the authors name | supplement §III; checked at uniform rates |
| C3 | Q ≥ 2 for multicyclic networks far from equilibrium | moderate (conjecture) | numerics on seven networks; my random check |
| C4 | Minimal cost for given precision is reached near equilibrium | strong | eqs. 4, 20 |
Concepts
- Q, the dissipation–precision product — σtε², dimensionless with k_B = T = 1 (eq. 10).
- Generalized distance dᵢⱼ^α — how much X_α changes on transition i → j, antisymmetric.
- Koza’s formulas — J and D from the coefficients of the tilted generator’s characteristic polynomial (eqs. 14–17).
Connections
- Uses Seifert’s framework (LIT-523) for σ = Σ J_β A_β (eq. 9).
- The review LIT-1892 (§VIII.B.1) states the macroscopic form: for a system prepared in one attractor and observed for less than its lifetime, the bound involves only that attractor’s own dissipation.
- Later proofs of the general case (Gingrich et al. 2016, Horowitz & Gingrich 2020) are not held and were not read.
Bearing on the record
- CASE-189 (maintenance) and A13’s use of it. The exchange said the relation ties “resource cost and precision of maintenance”. It ties the precision of a current to dissipation. A maintained state is an occupation, and the paper explicitly sets occupation observables aside. What it does bound is the precision of the throughput that maintenance runs on, for instance a production flux counted over time. So it can price a reliable production rate in a construction like CASE-189’s, once that construction is made thermodynamically consistent (every production step reversible). It does not price how long the maintained state lasts; that is the escape-rate question of THEORY-1264.
- THEORY-671 (maximum entropy production). Nothing here selects a steady state; the bound constrains fluctuations at a given state.
- Produces THEORY-1268.
Limitations
- The general multicyclic case is conjectural in this paper.
- Steady states only; finite time and driven protocols are not treated.
- The supplement’s algebra is long and I followed, not re-derived, it.