Source
Perunov, Marsland & England (Phys. Rev. X 6:021036, 2016), LIT-518, read in full in NOTE-416 from arXiv v1 (2014). The PRX text was not seen and may differ. The macrostate relation it starts from is England (2013), LIT-515, read in NOTE-419.
NOTE-416 concludes that “no new THEORY is indicated” because the central claim (its C2) “is too weakly supported to source one”. This document files the claim as an account the record does not yet believe, with its weakness stated. It does not dispute that judgement.
Note (2026-10-21): exact part split off
The identity this account starts from, Eq. (8), and the conditional in its title (with the entropy and reversal terms equal, the likelier outcome releases more heat with less variation) are re-derived and filed as THEORY-831, Active. What remains here is the thesis that real driven many-body systems have outcomes for which those terms can be held equal, so that the dissipation term decides and the favoured states look adapted to the drive.
What was actually shown
The exact part (Eq. 8). Prepare a system in macrostate I and drive it with a field λ(t) fixed in advance, with no feedback. For two outcomes II and III, ln[π^fwd(I→II)/π^fwd(I→III)] = Δln Ω_{II,III} + ln[ratio of reversal probabilities] + ΔΨ − ΔΦ. Here Ψ = β⟨ΔQ⟩ is the mean heat released on the way, and Φ = ln⟨e^{−βΔQ}⟩ + Ψ ≥ 0 grows with how much that heat varies between paths, so Ψ − Φ = −ln⟨e^{−βΔQ}⟩. The step to Eq. (5) adds no-correlation and near-uniformity assumptions. Eq. (6) restricts to likely paths, which is stated, not derived, and the cumulant reading (Eq. 7) is called “heuristic”.
The interpretive claim. Hold the entropy and reversal terms equal, and the likelier outcome is the one reached by paths that absorbed more work and shed it more reliably. Such states have “dynamical response properties” tuned to the drive, as a resonant oscillator is tuned to its forcing, and the paper calls them “better adapted”. “The adaptation we predict is expected to take place independent of whether or not there is anything in the system that can copy itself.”
The toy models, checked in NOTE-416.
- Three states (Fig. 3). The barrier is set by hand so return rates are equal. Hops to x₁ release βΔE/2 and hops to x₃ release nothing, and r_max(2→1)/r(2→3) = e^{ΔΨ} (Eq. 12). The cycle-averaged ratio is I₀(βΔE/2), so the identity holds to leading order in βΔE.
- Two paths (Fig. 4). Two barriers driven out of phase give Ψ = 0 and Φ ≃ βΔE/2, and the forward rate falls as ln r ≃ −Φ (Eq. 13). Fluctuating dissipation slows the transition.
- Futile cycles. For paths that return to their start, Ψ = Φ, so heat made in cycles drives no drift.
What is argued, not shown. Some of the ∼10²⁵ directions of a many-body phase space “should” behave the same way (C3, weak). A driven chemical mixture showing “adaptive” resonance is summarised from a study “published in a separate article, rather than shown here” (C6).
What could have come out otherwise. In the toy models the dissipating transition could have failed to dominate. It did, but the comparison was built so that it would.
What this does not say
- That the drifted states are better at anything but absorbing the drive’s work. “Adaptation” is defined by the paper’s own quantity, so calling the states “better adapted” is partly a choice of word (NOTE-416).
- That “all else equal” can be had in a real system. The paper admits it is “not immediately intuitive” what equal reversal probability means in general, and does not show a parcelling of many-body phase space into equal, equally accessible macrostates exists.
- That driven systems maximize dissipation. Maximizing Ψ alone “is not predicted to correspond to a likely outcome”: the fluctuation term and the other two terms count too. This is the paper’s contrast with maximum-entropy-production principles (C7, informal; THEORY-671).
- That an adapted system dissipates more in every sense. In THEORY-026 a system matched to its drive keeps a predictive memory and wastes less work. Here a system matched to its drive absorbs and dissipates more. The two answer different questions and do not contradict each other, but a claim that “adapted systems dissipate more” or “less” has to say which sense it means (NOTE-416).
- That Darwinian and thermodynamic adaptation are one. The paper says they “become one and the same” for self-replicators, supported only by England’s replication bound (C5, assertion).
- Where a living thing ends. The paper calls it “somewhat arbitrary how we have decided to draw a line around a whale”. It offers no criterion that would separate a cell from a hurricane except degree of reliable dissipation.
- Anything with feedback. The drive is fixed in advance.
Connections
- Premises. The Crooks relation (THEORY-431) and England’s macrostate form of it (THEORY-679). If either fails, Eq. (8) loses its derivation.
- Gradient dissipation (THEORY-532). Schneider and Kay make a related claim, that pushed systems develop structures that degrade the applied gradient, as a law of development rather than as statistical mechanics.
- The demarcation question. THEORY-072 holds that organisms differ from dissipative structures by organisational differentiation. This account denies there is a physical line to draw and replaces it with a graded property of histories. The two are not declared rivals: this account’s claim is the tendency, and its denial of a line is a framing, not a result.
- A test the record could run. The primitive metabolic cycles of LIT-036 are self-organized states of a chemically driven system. This account predicts they are the states that absorb and dissipate the drive reliably. LIT-036 does not compute dissipation.