Source
England (2012; J. Chem. Phys. 139:121923, 2013), LIT-515, read in full in NOTE-419 from arXiv v1, the only arXiv version. The published text was not seen. NOTE-419 re-derived Eq. (6) and recomputed every numerical estimate.
What was actually shown
The general bound (Eq. 6). Take stochastic dynamics in a heat bath with detailed balance, π(j→i)/π(i→j) = exp(−βΔQ_{i→j}) (Eq. 2), and “by stipulation” no external drive. Let I and II be ensembles fixed by an observer’s classification of microstates. Then β⟨ΔQ⟩ + ln π(II→I) + ΔS_int ≥ 0. It follows from an exact average (Eq. 5) by Jensen’s inequality. Since π ≤ 1 it implies ⟨ΔS_tot⟩ ≥ 0, and the paper calls it “simply a precise statement of the Second Law”. NOTE-419 re-derived it and finds it correct given the definitions.
The replication bound (Eqs. 7–8). The paper argues that the likeliest way back from two cells to one is a daughter hydrolysing completely, which gives ln π(II→I) ≤ 2 ln p_hyd (Eq. 7). With Poisson statistics for bond hydrolysis, β⟨Q⟩ ≥ 2n_pep ln[(n_pep τ_hyd)/τ_div] − ΔS_int (Eq. 8). A copy that lasts longer, or is made faster, must shed more heat.
The estimates.
- E. coli. With n_pep = 1.6×10⁹, τ_div = 20 min and τ_hyd = 600 years, the irreversibility term is about 75 n_pep (NOTE-419 recomputes 75.6). The measured heat is 220 n_pep (Rothbaum & Stone 1961), a ratio of about 2.9: “less than three times as large as the absolute physical lower bound”.
- RNA and DNA. For a ligase-type RNA replicator with a 1-hour doubling and a 4-year half-life the bound is printed as 7 kcal/mol, against a reaction enthalpy near 10. NOTE-419 recomputes 6.2–6.5 kcal/mol. The same speed in DNA would need about 16 kcal/mol (15.6–16.2 recomputed), more than the reaction releases.
What could have come out otherwise. The measured heat of E. coli could have fallen below the bound. That would have refuted either the bound or the argued reverse probability. It came out 2.9 times above. The RNA bound could have exceeded the RNA reaction’s enthalpy; it did not.
What this does not say
- That copying information has an intrinsic thermodynamic cost. The heat priced is the irreversibility of making a durable copy in finite time. The term ln(τ_hyd/τ_div) falls away as τ_div grows, as a reversible copying limit requires (NOTE-419).
- That the E. coli figure is a measurement of efficiency. The reverse probability in Eq. (7) comes from an argument about which path back dominates, not a bound proved over all paths. The internal-entropy term, −ΔS_int ≤ 10 n_pep, is set “arbitrarily” and “generous[ly]”. So the factor of three could reflect adaptability, as the author suggests, or the looseness of the reverse probability.
- That it holds for driven systems. The text says the same form holds for time-symmetric driving, by Crooks’s relation, but does not derive it here.
- That RNA came before DNA. That rests on one comparison with assumed enthalpies and half-lives (C5 in NOTE-419, weak).
- Where a living replicator ends. “Self-replication is something that happens relative to an observer.” The bound works for any coarse-graining with a reliable population model. It supplies no physical mark of a living replicator, and that framing is an assertion, not a result (C6).
- A cost per operation. Like the other second-law bounds in the record, it is a bound on averages.
Connections
- Landauer’s scope (THEORY-030). The paper calls its bound “closely related to the well-known Landauer bound”. On THEORY-030’s reading, copying onto a blank register has no minimum cost. England’s bound is consistent with that, because what it prices vanishes in the slow limit. Reading England as evidence for a per-bit cost of replication is the mistake THEORY-030’s “What this does not say” warns against.
- THEORY-026. No direct bearing: England’s system is undriven and has no memory variable.
- The fluctuation theorems (THEORY-431) are the family this bound belongs to; its driven case leans on Crooks (LIT-522).
- Dissipative adaptation (THEORY-412), the sequel, LIT-518, takes this macrostate relation, adds a drive and drops replication.
- Metabolism first (THEORY-492). Morowitz and Smith put selection on metabolism before replication. England’s comparison concerns which replicator is thermodynamically possible, not whether replication came first, so the two do not meet head on.
- The demarcation question. Nahas and Sachs’s dispute (LIT-192, NOTE-094) is over whether living systems can be told apart from dissipative structures. This paper takes the deflationary side by locating the “self” in the observer’s coarse-graining. THEORY-072 is the record’s account of where the line falls.