Decoherence and the Emergence of the Classical
退相干与经典之涌现

Part I　The Ontological Identity of Decoherence

— Ledger-merger and phase delocalization; settling as an orthogonal extraction; slot-crossing and settling, two boundaries on one and the same sequence

1. Point of departure

The paper on the Born rule [1] gave the weights; the immediately preceding paper, on measurement [2], gave the single content that gets read out — yet both close their account on a single event of closure. Where the menu before closure comes from, and how the classical world that disperses toward the macroscopic after closure settles into place, neither paper treats; both leave it standing. The prior paper already places decoherence clearly: it belongs to the energy route, and what it hands over is a classical menu of weighted candidates without picking out any one of them. (On where decoherence stands relative to the causal slot, the prior paper's wording admits of refinement: the present paper places the menu that decoherence delivers above the slot, having already crossed that categorial threshold; this is a refinement of the prior paper's slot-crossing wording, ontologically consistent with it — see §4.) As for how the quantum becomes one with its environment, how that menu forms, and how it stabilizes into something many can read, the prior paper expressly leaves to what follows. It is precisely this stretch that the present paper takes up — the third of quantum mechanics' three ontological movements, the turn from the microscopic toward the macroscopic.

It must be said first what this paper does not do, and what it says over and above standard decoherence. The settling that picks out a single content is the prior paper's jurisdiction; this paper does not redo it, and adds to settling no mechanism and no trigger. Strip away the terminology, and the decoherence physics this paper rests on is just the standard system–environment entanglement and reduced density matrix [15][16]; this paper changes not one of its equations. What it adds lies entirely in three ontological readings: coherence is not destroyed but delocalized; that classical menu is not a mixture born of an observer's ignorance but an above-slot, dephased yet still many-valued ρ-OR; its objectivity is not some single outcome but the redundant readability of a pointer index. The floor of the whole is the purely a-posteriori physics that measurement yields information while the quantum (or its record) still persists; this paper changes none of that, only reads it in a different ontological key.

1. Four distinctions

To say clearly what decoherence is, one must first prise apart four things long run together: coherence, decoherence, recoherence, and settling.

Set down first a distinction that runs through the whole paper, lest it trip us repeatedly later:

　　coherence ＝ many-valuedness ＋ local phase.

Many-valuedness — several candidates coexisting — is only the necessary half; coherence also requires that interferable relative phase locked within the home ledger (the off-diagonal entries of the density matrix not vanishing). The e^{iθ} of the paper on complex amplitude [3], and the dual-orientation pairing ψ♯ψ of the paper on the Born rule [1], both speak of this latter half. It is just because the phase is local that the several ρ-OR orientations can pair up and interfere. To split these two halves apart is the key to all that follows: many-valuedness can lose its local phase and remain many-valuedness; coherence cannot.

Decoherence, then, is the merging of one ledger 𝔏_S into a larger, non-factorizable joint ledger 𝔏_SE that includes the environment. This merger actually happens: the quantum truly becomes part of a larger informational whole. What moves in it is exactly that phase-half — phase is delocalized out of the home ledger 𝔏_S and disperses into the correlations of 𝔏_SE (on local read-out, the off-diagonal entries tend to zero). Phase is not destroyed; its ownership is transferred. So what decoherence leaves on the local read-out layer is something that has lost its local phase yet is still many-valued: a classical menu. It has already crossed that readable, categorial threshold — in that sense it has become information; yet it is still many-valued and has generated no single content. This paper henceforth calls it menu-information, giving this cell an explicit name lest it be confused with the coherent ρ-OR that preceded the quantum's merger.

Here the paper joins onto the paper on information [4], and marks that it inherits that paper rather than refining one of its conflations. The paper on information already drew this distinction: its paper on the causal slot states plainly, in §6, that decoherence is the physical process by which the substrate crosses the causal slot and is categorially lifted from the substrate into the 4DD information category — a categorical lift — and says explicitly that it is "not the selection of a pointer state"; while a single definite value, by that paper's §1.3, hangs on measurement, not on decoherence. Decoherence (becoming a readable menu) and measurement (picking out a single definite value) were thus already separated there. This paper starts no new fire; it merely inherits that existing distinction, names the categorized-yet-still-many-valued cell menu-information, and locates the single content at settling (inheriting that paper's clause "measurement yields the definite value"). Let no reader take the paper on information to have said that decoherence yields a single value — it did not; the present paper is continuous with it here.

Recoherence, then, is nothing but decoherence run backward, and it works only at controllable scales: apply a U† — spin echo and photon echo are its instances [17] — to gather the phase that had delocalized into 𝔏_SE back into 𝔏_S, restoring the quantum's local coherence. It is no new category, only the inverse of the same phase delocalization; it need undo no settling, because decoherence performed no settling at all. At the macroscopic scale this inverse no longer works; why it fails is left to §4.

Settling, finally, lies off the road of the three above. It is not the quantum changing its state but an orthogonal extraction: out of that dephased, many-valued menu, a single content is generated on the spot and drawn out, written into a ρ-AND record. Orthogonal to what, and why it is another boundary on the sequence, is the proper business of §4.

This section alone already answers two common misreadings of decoherence. First, decoherence is not the destruction of coherence but the delocalization of phase — it is transferred out of 𝔏_S into the correlations of 𝔏_SE, never vanishing; this is just where the series' remainder-conservation is made good here. Second, that classical menu is not a mixture born of an observer's ignorance but, on the local read-out layer, a readable grammar that has lost any interferable phase yet still has undetermined content. What decoherence delivers is which kind of question may be asked, not what this answer is.

1. Two faces of one merger

This merger has two faces — two views of one and the same thing, not two things, still less two successive steps.

Seen from the joint-ledger side, it is Many-to-One: 𝔏_S and 𝔏_E have merged into one 𝔏_SE. The truth of this joint ledger lies entirely in its non-factorizability — by the criterion the paper on entanglement [5] sets down, 𝔏_SE ≠ 𝔏_S ⊕ 𝔏_E. It is just this that separates "become one" from "merely gained a few more factors": if after merging it were still factorizable, then "become one" would be empty words in new dress; only in being truly non-factorizable is the merger a fact. One thing more must be said — this is also the linchpin by which the whole paper distinguishes the framework from many-worlds: what has become one is the ledger, not the outcome — 𝔏_SE is still a many-valued ledger.

Seen from the local read-out side, the same merger is One-to-Many: the phase in the quantum's single ledger is redundantly dispersed across the environment's manifold fragments. Three things here must not be conflated under the one word "environmental record": first, the trace — that some environmental fragment is correlated with some basis-vector of the quantum; second, the menu — the stable, redundant set of candidates that many observers can read, i.e. the menu-information above; third, the content — that single record written down after settling. Decoherence handles only the first two; the third is settling's affair.

It is for just this reason that einselection [16] selects the basis of the menu — that set of pointer modes, that pointer subalgebra, which survives robustly under the environment's monitoring — and not some one eigenvalue. It fixes which kind of question may be asked, not this answer. The layer of objectivity must also be set straight here, lest selecting-a-basis and selecting-an-outcome again be conflated: what the environment redundantly, readably-by-many encodes is that set of pointer basis, that grammar of the menu — "which question has objectively gone classical," which is exactly the redundancy quantum Darwinism [18] speaks of. It lies on the capacity layer; it is objectivity about "which question," not the realization of "which answer." That many observers read one and the same single answer, their redundant copies agreeing with one another, is a matter of after settling. Two further points must be kept: what the environment redundantly copies is the pointer-correlated branch index, a classical correlation — not the unknown quantum state itself — so no-cloning is not violated; and the correlation it disperses is, by the paper on entanglement, a capacity-layer correlation, capacity and not content. As for how this menu stabilizes into place across the macroscopic through layer upon layer of merger — its cascade carrying a renewal-encapsulated structural form — that is the mechanism Thesis B treats; this paper does not enter the detail of its timescales.

1. Two boundaries on one and the same sequence, and two irreversibilities

The two preceding sections laid out the four distinctions; it must now be made plain: they do not fall on two mutually independent axes but are different observations, different boundaries, of one and the same sequence. This sequence is what one obtains by applying that four-fold methodology once more, to the single rung L₄d — a natural application, at this finer grain, of a recursive skeleton the SAE framework has already established; it is neither newly fabricated by this paper for the sake of decoherence, nor a claim that there must ontologically be just these four. The prior section already divided the interior of L₄ into four — the concept of time, the arrow of time, the direction of time, and information-and-the-causal-law (L₄a–d); the present section subdivides its last rung, L₄d (closure giving information and remainder), one grain further, to see just where slot-crossing occurs. Inheriting the single home-side direction L₄c has already locked, one question remains: what can this direction determine? The answer unfolds along four steps —

　　First, marking out an addressable slot that can be closed (L₄da): the single direction gives a formal locus of "there is a place that can be determined" — it delimits an eigenspectrum / a Hilbert slot that can be closed (closure-addressability: fixing "where it can be determined," while not yet fixing "what is determined") — marking without constructing, with no specific structure yet. 　　Second, the addressabilization of relative phase (L₄db): once the single time-direction is locked, the unitary evolution e^{-iHt/ℏ} acquires a single-valued ruler for its unfolding, along which phase rotates and accumulates, and only then can phase-differences be spoken of as arranged into a comparable, addressable structure — this is exactly the physical link by which a single direction "grows" a determinate phase. What is to be oriented is the relative phase between branches (the recombinability of the local off-diagonal entries ρ_ij), not the unobservable global phase; the superposition of phase vectors in the complex plane is just this layer's "addition gives direction." 　　Third, the phase-possibilities take form as a menu of legitimate candidates — this is menu-level slot-crossing (L₄dc): the tensor product of system and environment composes the phase-possibilities and lays out a menu; slot-crossing occurs here — a set of legitimate candidates has taken objective form, while there is yet no act of "picking one." 　　Fourth, closure extracts and generates a single content — this is settling (L₄dd): the closure-query and the record-register arrive, mesh with the menu, and let one of its contents settle on the spot, written into a ρ-AND record; the structure is information, the remainder is how this content is stored.

〔On the framework's logical skeleton these four steps correspond, in order, to the law of identity, the law of non-contradiction, the interval law, and the synthesis law (the synthesis law showing, at the L₄ layer, as the causal law); the main text of the methodology overview gave these in the 4DD / causal-law context and did not spell out the names of the latter two laws. This law-name mapping — especially "addition gives direction ↔ the law of non-contradiction" — the present paper treats as a set of prospective correspondences awaiting verification, bearing no weight here: the arguments below (closing off pre-inscription, the halting of the recursion) all run only on the already-established capacity/content, L₄dc/L₄dd, ρ-OR/ρ-AND, and do not depend on these law-names. This paper concentrates on the specific domain of decoherence, invokes the four-step skeleton only here, and adds nothing to the overview; spelling out the law-names is left to later methodological work.〕 This subdivision stops here: the product of the closure step is ρ-AND (a single content), already synthesized into one, no longer containing a multiplicity that could have "another layer of candidates laid out" — this is the ontological floor (no many-valued fuel remains to be subdivided, so the recursion halts); and functionally the same: the menu/content distinction is by now closed, and anything further down (the internal storage, read-write retention, and macroscopic stabilization of the content record) belongs to another problem-domain and to later work, no longer within "the distinction of decoherence from settling." So "stopping here" is not a goal-presupposing post-hoc stop but a necessary halt where the ontological floor and the functional stopping-point coincide.

Thus slot-crossing (L₄dc) and settling (L₄dd) are two successive boundaries on this one and the same sequence — not two orthogonal axes. The "two axes" used elsewhere in this paper (the slot-crossing one, the settling one) are precisely the aliases of the two boundaries this single sequence projects onto a coarser grain; as a rough shorthand it is still useful, but ontologically there is only one sequence and two boundaries, not two mutually independent dimensions. The passage from many-valued to single is not a displacement along another transverse axis but the one step of crossing from the boundary L₄dc to the boundary L₄dd. Where the text below occasionally calls settling an "orthogonal" extraction, a line must be drawn here: by "orthogonal" is meant only the heterogeneity of mechanism-type — slot-crossing is supported by unitary system–environment spreading and thermodynamic irreversibility, settling by non-unitary content-extraction and causal irreversibility; it does not mean the two belong to two mutually independent ontological dimensions — on the L₄d subdivision they remain the two successive boundaries of one and the same sequence.

This sequence divides decoherence's before and after into three stages, by a single criterion: whether the terms satisfy the interval law (the readable structure of the Born rule). Before decoherence, below-slot coherence — the terms interfere with one another, phase not yet delocalized, not yet satisfying the interval law, not yet legitimate candidates readable independently; the interference fringes are only the shadow of a "menu about to form," and at this point not even an "answerability" is yet present. After slot-crossing, the above-slot menu — the terms are now a set of legitimate candidates, mutually orthogonal, each bearing its Born weight, fit to be the object of "take one from among"; this is answerability before the query: the menu has taken objective form, while the closure-query has not yet meshed. It is not the "answer" (in the content sense); it is the condition of the answer's possibility — capacity (ρ-OR), not content. After settling, a single content — out of this set of legitimate candidates, one is generated and settled on the spot, written into a ρ-AND record.

Between these three stages there is one clean criterion that divides capacity from content: reversibility. Before settling, L₄dc is reversible — one can recohere, gather the delocalized phase back, and retract the menu into a coherent superposition (in Stern–Gerlach [19], recombining the two paths brings the fringes back; spin echo and photon echo likewise). Whereas a content once generated (ρ-AND) is, once standing, irreversible. Whatever can be undone back to a coherent state of "no answer yet possible" can hold no generated content; hence L₄dc is capacity, not yet crossed into content — reversibility is precisely the criterion that it is still on the capacity layer. This criterion inherits the prior paper: whatever approaches settling without settling is reversible, "what never came to be can of course be taken back"; this paper only locates that reversibility precisely at the boundary L₄dc.

This also closes off, logically, the perplexity of "pre-inscription," and accords with the prior paper's "generation, not selection." What L₄dc (the tensor-product laying-out of the menu) issues is a menu — many-termed, legitimate candidates, capacity; what L₄dd (closure) issues is content — single. Capacity is not actuality, so the two must be two successive boundaries, and the step of "laying out legitimate candidates" cannot, logically, inscribe any definite content; which content gets settled is generated on the spot at the moment of L₄dd, neither pre-stored nor selected from a row of ready-made answers. 〔Put in the language of the laws: the interval law only establishes that the terms are separable and readable, generating no content; only the synthesis law generates content; the distinction of the two laws is just the distinction of the two boundaries.〕 The framework is therefore contextual by nature, and accords with positing no pre-stored values; it is not a hidden-variable theory.

A concrete instance must be given for "above-slot many-valuedness," lest it sound like an empty cell. It is exactly that improper mixture after decoherence and before settling — the objective menu of legitimate candidates. An atom in a Stern–Gerlach apparatus whose which-path information has, as far as local control is concerned, irrecoverably leaked into the environment, yet for which no detector has yet fired, is at this moment just that: phase already delocalized (if one cannot address and invert the relevant environmental degrees of freedom, interference no longer returns), yet no path fixed — dephased, objective, and up-or-down undetermined. A qubit that has irreversibly decohered yet not yet been read out is, at every moment, just like this. It must be cut clean from "single but unknown," and this is exactly the crux of its being many-valued and not single: an improper mixture is globally still entangled, with no single content ontologically realized, all its terms still capacity; whereas a covered coin is a proper mixture — its single content is already realized (already settled), you simply do not know it. The menu is the former, not the latter. An honest addition: an improper mixture and a proper mixture are locally indistinguishable (the local statistics are identical), and this is precisely the core of the measurement problem; so "above-slot many-valuedness" is an ontological category — that improper mixture after decoherence and before settling — not an operational quantity one could single out by a local measurement.

Slot-crossing (L₄dc) is decided by the categorial threshold of the interval law, not by scale — yet category and scale are not unrelated. This threshold falls on the causal spectrum of the paper on information [4] — causality being a continuous gradient of effective causal strength corresponding to the magnitude of substrate aggregation. Substrate aggregation (rising toward the upper end of the causal spectrum) is the dominant natural route to this threshold: the higher the aggregation, the more objective and more redundant the menu (i.e. the readiness for many observers to read it). But scale is not necessary — there is also measurement: environmental coupling can, without scale, force a system across the threshold (the Stern–Gerlach atom is always microscopic, never grows to μm; its slot-crossing is accomplished entirely in flight by the environment's gradual leakage, and only the strike on the screen is that closure-query). Nor is it sufficient — effective causal strength depends on aggregation and coupling, so a large but still coherent object, low in effective causality on account of its coherence, remains below the slot. To ask "how large counts as slot-crossing" is a category mistake: the threshold is categorial, and scale is only one (dominant, not sole) route to it. One must distinguish "route" from "that boundary": aggregation is the route to L₄dc, whereas settling (L₄dd) is orthogonal to this aggregation spectrum — it is not some point on the spectrum but the next boundary, where, after the objective menu already stands, the closure-query meshes and generates content.

From this one can fix the word "macroscopic," and it differs from the "macroscopic" standard quantum mechanics sometimes uses loosely. The macroscopic of this paper is categorial: a being is macroscopic in that it can bear causal information, the criterion being information-bearing and not size — just as the paper on information [4] establishes in §1.5. So the causal slot is not a size threshold but a categorial one. A large but still coherent object — a Bose–Einstein condensate, superfluid helium, a superconducting current, a large molecule in matter-wave interference — is macroscopic by size yet, by the framework's category, below the slot: it is a large-scale sub-causal substrate, its phase still local (coherent many-valuedness), bearing no objective information. There is no "coherent large object above the slot"; once objectified, phase delocalizes, i.e. decoheres, i.e. arrives above the slot. The distinction of capacity from content therefore also divides into three cells: below-slot capacity (coherent ρ-OR, phase present), above-slot menu capacity (menu-information, dephased many-valuedness), and above-slot content (ρ-AND, single). Nailed in one sentence: below-slot/above-slot is a categorial boundary, capacity/content is a read-out state; the menu is an above-slot ρ-OR that has not yet been content-ized.

The two successive boundaries on this sequence each pawn a kind of "no going back," and the two are heterogeneous, not to be run into one.

The irreversibility on the boundary of slot-crossing (spreading) is thermodynamic. Decoherence is the unitary evolution of the joint system+environment — the local reduced state looks mixed because the environment has been traced out, while the global state is always pure and evolves by a unitary U; the framework is non-collapse and adds no random term to the Schrödinger equation, so it retains as-is this unitary dynamics. So reversing the spreading uses U† — a coherent control operation; at small, controllable scales it is empirically reversible (spin echo, photon echo, dynamical decoupling), with no need to cancel any outcome, because no outcome was ever selected. At the macroscopic scale it cannot be reversed — and the reason it "cannot" is that phase has delocalized into 10²³ unaddressable degrees of freedom: the formal U† still exists, it just cannot be reached. This is a thermodynamic / FAPP signature, and it is scale-dependent (small can echo, macroscopic cannot afford it). The One-to-Many irreversibility the paper on entanglement speaks of is, at this macroscopic end, exact; but its root is thermodynamic, not causal — to recall a dispersed objective menu one would have to re-address those manifold degrees of freedom, which is unreachable in computational cost, not forbidden by the causal law.

The irreversibility on the boundary of settling is the causal one. Drawing a single content out of the many-valued is non-unitary — one is realized, the rest (which were capacity) are not, and for this step there is no U† at any scale. The settling of one electron's spin and the settling of a macroscopic pointer are, in "one is realized," equally not reversible, not varying with scale. It does not turn on the difficulty of addressing; there simply is no unitary operator that returns one to "no content realized." This is what one-directionalization is, what causal irreversibility is — to reverse it one would have to reverse that one established fact, i.e. reverse causality.

So the two irreversibilities issue under the one genus of "generating 4DD information," yet are two species: slot-crossing generates a dephased objective menu (unitary, thermodynamic, gated by the slot and by scale, echo-able below the slot); settling generates a single content (non-unitary, causal, scale-independent, no U†). The hat of "causal irreversibility" fits settling alone; the irreversibility of macroscopic many-valuedness is purely thermodynamic. These two ends are each held down by one of the framework's commitments: to say slot-crossing too is non-unitary would be to add a random term to decoherence and sink into an objective-collapse theory, violating non-collapse; to say settling too has a U† and can in principle be reversed would be to have all branches real and slide back into Everett, violating its non-compatibility. What the causal slot gates is at which scale slot-crossing occurs; it cannot move this species-line of unitary/non-unitary. That these two irreversibilities belong to the two successive boundaries of one and the same sequence, with utterly different things pawned (spreading pawns the FAPP entropic barrier of vast addressing, settling pawns the one-directional causal generation), is exactly what the "two axes" spoken of earlier denote on a coarser grain.

The perplexity of macroscopic superposition thereby dissolves of itself. A system after macroscopic decoherence is a dephased objective classical menu, capacity — redundant, objective — and not a superposition already realized, nor a collapse into some single state. That cat, decohering in reality almost instantaneously, is an above-slot improper mixture: both the dead term and the live term are capacity (possibility), not two realized worlds. Settling extracts one true content from it; the rest remain capacity. Many-worlds has every branch realized; the framework realizes only the one extracted, the rest left as possibility. The distinction of possibility from actuality — the distinction of capacity from content — this modal adjudication is the true divide between the framework and many-worlds, not a change of vocabulary. And that a single outcome is true, admitting no unitary many-worlds, is not a separately posited premise but the corollary of settling's non-unitary realization: settling realizes one true content, so there is no layer on which all branches are real. As for just where many-worlds goes wrong, and where its "branches" are to be classified, that is left to Thesis C.

1. A posteriori physics and compatibility

Since this paper only re-reads and issues no prediction of its own, it must show, to an auditable degree, how it meshes seamlessly with standard quantum mechanics.

First align one point with the a-posteriori physics, lest "after measurement the quantum is still a classical ρ-OR" be misread. It does not say that the original many-valuedness sits untouched — were it so, a repeated measurement would give mutually inconsistent results, which is empirically false. The exact reading is: that classical ρ-OR after settling has been conditioned on the record — it is classically correlated with the extracted ρ-AND record — so that reading again yields the same result. Unconditionally, capacity (ρ-OR) still persists, an ontological surplus; conditioned on that extracted record, the prediction agrees with the standard post-projection state, with standard collapse, and is repeatable. The conditioning here is a classical correlation, not quantum coherence, so the two layers do not clash. Decoherence yields the capacity menu, extraction writes down a record, and the record conditions subsequent predictions — together the three reproduce exactly standard quantum mechanics, with no deviation whatever.

The floor of this a-posteriori layer this paper neither adds to nor subtracts from, only restates: measurement yields information while the quantum or its record still persists; a single trial obtains one content, not the complete state. Two cases must be distinguished within this. After a quantum-non-destructive measurement the quantum persists as a classical ρ-OR, still a quantum object; whereas many real measurements are destructive — a photon absorbed by a photomultiplier, a particle annihilated in a calorimeter — and then the original quantum is consumed, and what persists is the extracted record, not the quantum. The framework holds both: in the one, capacity remains; in the other, content has been extracted. Again, measuring one quantum once obtains an eigenvalue, one content, and cannot read out the complete state ψ — a single shot gives only one sample, an ensemble gives the distribution or expectation, neither of which is ψ; this is where the paper on tunneling [6]'s "what is read out is the expectation value, not the state itself" converges with the distinction of capacity from content here.

This paper issues no SAE-proprietary prediction of a decoherence rate — coincide with the standard, and it is not proprietary; diverge from the standard, and it has no a-posteriori support. But issuing no proprietary prediction is not the same as having no failure conditions. This paper's ontological reading is a set of compatibility obligations laid over the standard open-system formalism, each falsifiable: first, the ledger-merger must be rewritable as the standard reduction, ρ_S = Tr_E ρ_SE, or the ontological mapping of merger fails; second, phase delocalization must correspond to the standard off-diagonal suppression, ρ_ij ↦ ρ_ij⟨E_j|E_i⟩, or "delocalization" is mere metaphor; third, that pointer basis must be derivable from the system–environment interaction Hamiltonian, from a robust pointer subalgebra, or the classical menu has no physical anchor; fourth, the One-to-Many redundancy must correspond to quantum Darwinism's fragment redundancy and mutual-information plateau, or the mechanism of the "objective menu" comes to nothing; fifth, and the crux: if this paper's account must introduce some single i₀ as the actual outcome in order to explain the macroscopic record, then it has smuggled in settling, and the ontological boundary fails. These five are not new empirical predictions but structural failure conditions — and daring to lay the crux out in the open is the strongest defense.

1. The ontological place of the arrow of time

Where does all this place the arrow of time? Direction and irreversibility must be kept apart, each set on the framework's own structure.

Direction first. The home-side 4DD time-direction is fixed — this is the source of the arrow's direction, and it is internal to the framework, not an externally imposed past hypothesis. The paper on cosmology [7], in §3.1, already establishes: a 3DD symmetry gives rise to a pair of 4DDs with opposite time-arrows, the home-side parameter order being (T₁, T₂) and the opposite side (T₂, T₁), our Big Bang being the opposite side's Big Crunch; the home side therefore has a fixed 4DD time-orientation. The measure of this clause must be nailed: this pair of 4DD orientations is the framework's cosmological boundary commitment (from the paper on cosmology), not an internal derivation of this paper; it gives the direction of time and does not thereby give "why the starting point on our side is low-entropy" — that question is anchored to the orientation structure, not necessarily thereby dissolved. What is given is only the rooting of the arrow's direction in the framework's own orientation, with no recourse to a thermodynamic low-entropy past hypothesis.

From this one can fix the relation of below/above the causal slot to this orientation. Below the slot, although a being also evolves along the home-side 4DD direction, it is at the same time under the influence of that opposite pair of arrows, hence has no single direction, no causality, only probability (the paper on the Born rule); slot-crossing and settling are the gradual gathering of this bidirectionality onto the home-side single direction. And by §4, this "gathering" divides into two successive boundaries on the sequence: slot-crossing (dephasing, objectification) is the unitary, thermodynamic step, and settling (drawing out a single content) is the non-unitary, truly one-directionalizing step. So the framework's fundamental, causal arrow lies on the side of settling and the 4DD orientation (also the side the paper on entanglement [5] and the series' first paper [8] establish); the One-to-Many spreading is the thermodynamic amplification of this fundamental directionality, through redundant records of astronomical magnitude, into an objective entropy-arrow the whole cosmos cannot disavow. It is amplification, not origin — the directionality of spreading itself still presupposes that fixed 4DD orientation, and to say the arrow originates in spreading would be to turn the direction one set out to explain into one's conclusion. It does not derive the second law; it only gives, at the boundary of quantum and classical, a ledger-mechanism version of the second law, which must be consistent with that fundamental 4DD causal arrow.

Last, the disposition of the remainder, with the measure marked to the end. This passage speaks of settling, not decoherence. Decoherence is unitary; it discards no branch — the delocalized phase turns wholly into the environment's classical correlations, which is the thermodynamic remainder, kept in that still-unitary global ledger. What truly one-directionalizes, discarding the side influenced by the opposite arrow, is settling. The side settling discards is not annihilated; the series' remainder-conservation does not permit its annihilation. As the 4DD information that one-directionalization generates, it goes — by the series' two papers on information [9,10] — to the broadcast channel of the 4DD invariant: gravitational waves (gravitational waves being the dynamical manifestation of 4DD broadcast, propagating along the absolute Planck substrate, their energy by the Mass series [11]'s E=Ic³, c being the breakthrough rate of 4DD information through the substrate). The measure of this clause must be nailed to the death: this is a global ontological accounting, an analogy of a broadcast channel, not a local-emission model. This paper in no way claims that some single local settling event must emit a resolvable gravitational-wave packet; the gravitational broadcast at the microscopic scale of settling is negligible, its magnitude far below any detection. So this paper sets up no measurable gravitational-wave signal of its own for settling, and does not inherit the paper on information's measurable prediction on black-hole ringdown — that is the jurisdiction of macroscopic sources, proper to black holes; and this paper does not weld this local remainder onto any cosmological constant. The framework issues predictions where a difference is detectable (black-hole interior imprints, multi-messenger time delays), and only re-reads where it merely reproduces standard physics; one and the same honest discipline, acting by scale and by domain.

So decoherence, this route, the paper can gather into a few sentences: it is the merging of one ledger into a larger, non-factorizable joint ledger that includes the environment; phase is thereby delocalized, and locally there remains only a dephased yet still many-valued classical menu; this menu can be gathered back by an echo at small scales, but at macroscopic scales it truly irreversibly spreads — and this irreversibility is thermodynamic. What picks out a single content among them is not on this route — that is settling, an orthogonal extraction that draws one on-the-spot-generated content out into a ρ-AND record while capacity is not extinguished; its irreversibility is the causal, one-directionalizing one. Two boundaries on one and the same sequence, two irreversibilities: slot-crossing's dephasing is unitary, thermodynamic (the prior boundary); settling's drawing-out of a single content is non-unitary, causal (the latter boundary). Direction is given by the fixed 4DD orientation; the arrow of time is thermodynamically amplified at the macroscopic scale through spreading; the remainder discarded by one-directionalization is conservatively borne by the gravitational-broadcast ontology. As for why that one content is generated, what its timescale is, and how several old paradoxes are reclassified accordingly, those are the business of the two parts that follow.

Part II　The Cascaded Mechanism and Timescales of Decoherence

— Partial ledger-mergers, a renewal skeleton, and a discipline that proposes no independent decoherence rate

1. Point of departure

The prior part fixed decoherence's ontological identity and set it, with settling, as the two successive boundaries on one and the same sequence: slot-crossing (spreading) is the unitary, dephasing step that lifts the substrate into a dephased yet still many-valued objective menu; settling is the non-unitary step that draws a single content out of the menu. The prior part left two questions plainly open, both falling on the boundary of slot-crossing: that menu is not laid out in one step — it stabilizes rung by rung through layer upon layer of merger; what is its mechanism? And what is the timescale of this process, and what is its relation to that rate in standard decoherence theory? It is these two questions the present part takes up. It must be drawn clear at the outset: this part treats only the mechanism and timescales of the boundary of slot-crossing; settling, that orthogonal, non-unitary extraction, the prior part already established as having no mechanized timescale, and this part does not touch it.

The discipline must be set down first, of one piece with the prior part. This part proposes no independent decoherence rate. A system's actual decoherence speed — its Γ_dec, its T₂, its T_φ — is taken from the standard system–environment Hamiltonian and noise spectrum (the Lindblad equation, master equations, the spin–boson model, that whole apparatus [20]), and the framework does not compute it from scratch. What the framework adds here is a structural re-reading of that cascade — why it is a cascade, what its rung-by-rung structural form is, what its ontological commitment is — and not an independent rate. What coincides with the standard is not SAE-proprietary; what diverges from the standard has no a-posteriori support. This two-faced discipline the part keeps throughout.

1. The cascade: layer upon layer of partial ledger-mergers

The prior part's One-to-Many spreading is not a one-step affair. A quantum's ledger does not merge in a single instant into that joint ledger which includes the whole environment; it merges in layer by layer — phase first delocalizes into the nearest few environmental degrees of freedom, then disperses into a larger ring, and at last into that macroscopic layer of redundancy. Each layer is a partial ledger-merger, a partial dephasing: phase delocalizes rung by rung, and the menu's readability rises with it rung by rung — the more observers can read it, the thicker the redundancy, the more stable and objective this menu. So decoherence in the framework is not a single event but a cascade of partial mergers, in which the menu settles into place rung by rung.

One point of wording must be kept, inheriting the prior part's terminological discipline. This part does not call each layer of this cascade a "closure." The prior part reserved "closure" for that content-closure of settling, and expressly established settling as having no mechanized timescale; were each layer of the cascade also called a "closure" and endowed with the rung-by-rung tempo τ_slot below, that would be to fit a clock to something defined as timescale-less — a self-contradiction. So each layer of the cascade this part uniformly calls a partial ledger-merger (or partial dephasing); these have a timescale (τ_slot) and are matters on the boundary of slot-crossing. "Closure" is reserved for settling alone.

This cascade has a self-similar structure, a form the series' paper on thermodynamics [12] has already given, which this part invokes as a structural form: the merger of the lower layer provides the upper layer with its renewal-slot — that is, one partial merger at the bottom generates the one tempo on which the upper layer counts its steps. So the whole cascade has the shape of a renewal process: each "renewal" is one layer's partial merger, and the upper layer advances on the lower layer's tempo. This is renewal-encapsulation — the cascade is not a structureless continuous decay but a nesting of layer within layer, each taking the lower layer's merger as its tempo.

The measure of this invocation must be nailed. That paper on thermodynamics, in giving the renewal-encapsulation form, attaches a no-go: a purely continuous coupling cannot inherit a well-defined, discrete timescale. This part demotes that no-go to a structural analogy, a suggestive parallel, not an independent dynamical proof. What it suggests is why the cascade carries a renewal structure — why there must be a rung-by-rung tempo rather than a layerless continuum — but it does not independently derive what value this cascade's timescale takes. The value of the timescale still returns to the standard physics of the next section.

The thing this section stabilizes must also be recognized: what the cascade settles into place is that menu — the dephased yet still many-valued objective capacity, the "which question has objectively gone classical." It does not pick out a single content. The settling that draws a single content out of the menu is another matter, orthogonal to this cascade, not waiting at the cascade's end to round it off. Renewal-encapsulation organizes the menu's stabilization, not the generation of content.

1. Two timescales and one overlap

Three quantities must be kept apart in the cascade, long apt to be confused.

First, τ_slot, the rung-by-rung slot-time — the time one partial merger takes, the single-step tempo of this cascade. Second, τ_cascade, the cascade-time — the time the menu takes to stabilize into place through the whole cascade. Third, ρ_ret, the rung-by-rung retention factor, which this part identifies precisely as the overlap of environment states: the overlap among those environment states correlated with the different candidates in the menu. Overlap small, distinguishability good, phase delocalizes fast; overlap near one, distinguishability poor, delocalization slow.

This overlap must be written in general, not as a single constant. It varies by branch-pair, by step, by environmental fragment; written fragment by fragment it is

　　ρ_ret^{(ij,k)} ＝ |⟨E_i^{(k)} | E_j^{(k)}⟩|,

and in a many-fragment environment the total overlap of branch-pair (i,j) is the product over fragments

Written as a product, the One-to-Many redundant copying truly connects to quantum Darwinism's fragment redundancy, rather than to only one global single overlap.

The structural relation of the three must be set out as only a homogeneous renewal skeleton, not a universal decoherence-timescale formula. In a homogeneous renewal simplification — supposing each partial merger multiplies the coherent overlap by an approximately fixed factor ρ_ret — the number of steps to reach some classical-menu threshold ε is

　　N_ε ＝ max(1, ⌈ ln ε / ln ρ_ret ⌉),

and

　　τ_cascade ～ N_ε · τ_slot.

Written in shorthand as τ_slot / (−ln ρ_ret) (with max(1,·) as floor), this is only a structural skeleton under the triple assumption of fixed threshold, fixed step-size, fixed retention. Read as a universal formula it would be cornered: where is the threshold? But the ρ_ret→0 end is already held by that max(1,·) floor — at least one τ_slot, so τ_cascade does not go to zero as ρ_ret→0. These would be the points caught at the formula level — and the answer is that it is not a formula-level assertion at all but a skeleton-level structural reading. Once ρ_ret varies by branch, by step, by fragment, it falls back to a coarse-grained reading of standard open-system dynamics, not an independent timescale.

One point of naming must be made plain, also a deliberate clarification of this part: this part does not use the notation "τ_dec." The τ_dec of the thermodynamics papers denotes a finite causal-settling time, the decay scale of a non-Boltzmann kernel; it is not a quantum decoherence rate; using one term for both breeds ambiguity. So this part calls the SAE-side structural timescales τ_slot, τ_cascade, and leaves the quantum's actual decoherence rate to standard physics' Γ_dec, T₂, T_φ. And identifying ρ_ret as environment-state overlap is also a deliberate alignment: an earlier paper on thermodynamics in the series [13] once identified this retention as copying fidelity f, and this part reinterprets it as that environment-state overlap of standard decoherence theory, merging the two languages here.

1. Meshing with the standard decoherence rate

With the previous section set out, this part's relation to standard decoherence theory is clear: it is a structural re-reading, not an independent rate, nor some derivation that "maps by construction" onto that rate.

A system's actual decoherence rate is determined by the standard system–environment Hamiltonian and noise spectrum — this the part accepts wholesale. The renewal-encapsulation structure the framework adds is a structural reading of this same cascade: where standard open-system dynamics can be coarse-grained into a rung-by-rung decay of fragment overlaps, τ_slot and ρ_ret give its SAE structural reading. It does not usurp the Lindblad equation's computational right, nor compute a rate of its own.

So after the demotion, what exactly is this part's net increment, which must be said plainly lest the doubt "merely standard decoherence in new dress" return. The standard decoherence rate gives a number; renewal-encapsulation gives the organizing principle of why decoherence presents as layered menu-stabilization — the partial merger of lower environmental fragments provides new renewal-slots for the upper redundant menu; plus a commitment to discrete closure: the cascade is not a layerless continuous decay but a nesting of layer within layer, taking the lower merger as its tempo. This part's contribution stops at this organizing principle and this ontological commitment, not the derivation of a rate. Thin, but not empty, provided it is marked thus.

So the part must restate the discipline once more, joining the prior part's falsifiable layout: this part proposes no independent decoherence rate. Coincide with the standard, and it is not SAE-proprietary; diverge from the standard, and it has no a-posteriori support. Everything touching specific values of rate and timescale returns to standard physics; everything the framework says stops at a structural and ontological re-reading.

1. Two matters out of bounds

Finally, two matters adjacent to timescales yet not within this part's jurisdiction must be plainly marked out of bounds.

First, settling has no framework-mechanized timescale. The prior part established: settling is a bare closure, the framework adds to it no mechanism and no trigger; being unmechanized, the framework also gives it no rate, no timescale. "When settling occurs" is that perplexity of timing in the measurement problem — the framework stays neutral among the interpretations, leaves it as a bare closure, and does not here force a timescale upon it. The two timescales this part gives are both timescales of the cascade on the boundary of slot-crossing, having nothing to do with settling, that orthogonal, non-unitary extraction.

Second, the thermodynamic floor R_min(T) = ℏc/(2πk_B T) — by the series' paper on information [4], the minimal scale of the causal slot, the spatial lower bound at which the cascade can stabilize one bit at the thermodynamic floor — this part does not pronounce it the discreteness scale of the probability cloud. To read it as "the pixel size of the probability cloud" has no a-posteriori support, and is of one discipline with this part's not assigning τ a new value: it is the spatial lower bound of cascade stabilization, not the scale of the probability cloud's discreteness. The notion of a cell-tick this part also takes only as a future direction of the framework, not redeemed here.

As for the proper question of how continuity and discreteness are compatible — substrate aggregation being a continuous gradient while the 4DD topological closure emerges sharply at a critical scale, which, by that section of the series' paper on information [4], is the composition of three layers of sharpness, continuous physics + critical topological closure + error suppression — that is left to the next part, treated together with the reclassification of several old paradoxes.

So the mechanism and timescales of decoherence the part can gather into a few sentences: spreading is a cascade of partial ledger-mergers, in which the menu settles into place rung by rung, the whole cascade carrying a renewal-encapsulated nested structure (the lower merger counting steps for the upper); what it stabilizes is that dephased, many-valued menu, not picking out content (the content-picking settling is orthogonal to this and does not round it off at the end). Its structural timescale τ_cascade ～ N_ε·τ_slot is only a homogeneous renewal skeleton under fixed threshold, step-size, retention, the rung-by-rung retention ρ_ret being environment-state overlap, taken as a product over fragments — the framework here only re-reads structurally, giving the organizing principle of layered menu-stabilization and the commitment to discrete closure, proposing no independent rate; coincide with the standard and it is not proprietary, diverge and it has no a-posteriori support. Settling has no mechanized timescale, the thermodynamic floor is only the spatial lower bound of cascade stabilization and not the discreteness scale of the probability cloud; the proper question of continuity-discreteness compatibility is left to the next part.

Part III　Reclassifying the Cruxes

— The place on the interpretive spectrum, the ontological identity of the density matrix, and three old paradoxes

1. Point of departure

The two preceding parts established two things: decoherence's ontological identity — a merger, phase delocalization, a dephased and many-valued objective menu, with settling as an orthogonal, non-unitary extraction of content, the two lodged at the two boundaries of one and the same sequence; and the mechanism and timescales of that cascade. This part turns the apparatus around to face several famous cruxes in the measurement and interpretation of quantum mechanics. The stance of this part must be stated first: this is a reclassification, not a treating of those cruxes as new physics to be solved. The framework issues no proprietary prediction; what it can do is use the two boundaries, capacity and content, slot-crossing and settling, generation-not-selection — this ontological vocabulary of the two preceding parts — to set these cruxes back, one by one, into the places where they belong: pointing out which thereby dissolve, which are merely a different reading, which remain open. What is sought is to disentangle a long-tangled cluster of concepts, not to issue a new falsifiable assertion.

1. The place on the interpretive spectrum

The two preceding parts said piecemeal what the framework is not; here it is set down together, with a clear account of why it falls just here.

It is not a GRW/CSL-type dynamical-collapse theory. GRW [21], CSL [22], Penrose [23] and their kind add to the Schrödinger equation a random, spontaneous localization term, making the wavefunction physically collapse — a new physics that alters the dynamics and is therefore in principle falsifiable. The framework does not do this: the decoherence step is unitary, and the framework retains it as-is; the settling step, though non-unitary, is a bare closure — no threshold, no trigger, no superadded random dynamics, and the framework writes no equation for it. The matter must be put precisely, lest it be misclassified: the framework does have an objective, single settling, but it proposes no new collapse-dynamical equation — hence it is an objective single-outcome ontology, not a dynamical collapse modification. Such dynamical-collapse theories give a new dynamical term; the framework gives an ontological reading. The two are of different categories: a falsifiable physical modification, versus a pure re-reading.

It is not many-worlds. This divide the prior part already noted; here it is taken to the root: it is that modal adjudication of capacity and content. Many-worlds [24] has every branch in the superposition realized, each a parallel real world. The framework realizes only the one content settling extracts; the rest remain capacity — possibility, not realized worlds. So that sentence is the linchpin: what has become one is the ledger, not the outcome; the menu is many-valued objective capacity, not a set of realized worlds. Many-worlds and the framework often coincide formally (the same unitary evolution, the same decoherence); the divide lies entirely in this modal commitment: all branches real, or one realized and the rest left as possibility. In the language of this paper's sequence, many-worlds is precisely stuck at L₄dc and misreads: the menu (legitimate candidates) has taken objective form, but it takes this reversible, contentless capacity menu for a set of content-worlds that have completed L₄dd closure — taking answerability for already-answered, taking the interval law's having-stood for the synthesis law's having-completed.

Yet what many-worlds rests on is a real question: from menu to settling, one is taken; where did the rest of the "options" go? Many-worlds answers, "each becomes real, branching into worlds." The framework must meet this question head-on, not evade it. The rest of the terms are precisely settling's remainder: settling (L₄dd) is closure giving structure and remainder, the structure being that one taken content, the remainder being precisely the disposition of the terms not taken — they do not vanish, nor branch into worlds, but enter, as remainder, the conservation ledger of 4DD broadcast (their energy by the Mass series [11]'s E=Ic³, their dynamics the 4DD broadcast channel the two papers on information [9,10] establish). It must be sealed by the measure of the remainder stated earlier: this is a global-accounting-layer disposition, not a claim that every settling emits a resolvable gravitational wave. So "where did the rest of the options go" has an answer — they went to the remainder, into the conservation ledger, with no need for all worlds to be real. Many-worlds' ailment is precisely that it lacks this remainder-outlet: not knowing where the untaken terms can go, it has no recourse but to let each become real. The measure must be declared on the spot: this is an ontological-layer adjudication, not an empirical verdict — the framework and many-worlds have no non-trivial difference in observable predictions (as declared at the opening of this part, the framework issues no proprietary prediction); what it rejects is not the branch formalism but the ontologization of branches into parallel real worlds. Capacity is not worlds, possibility is not actuality; that a single outcome is true is not a separately posited premise but the corollary of settling's non-unitary realization.

It is not a hidden-variable theory. The framework does not hold that measurement reveals a pre-stored definite value. The prior part's "generation, not selection" says exactly this: those weights are capacity pre-stored in the ρ-OR, but which content is extracted is generated at the moment of settling, neither pre-stored nor selected from a row of ready-made answers. So the framework is contextual by nature and accords with positing no pre-stored values — the fine grain of which is left to §4 on Kochen–Specker.

From this, the framework's place on the spectrum can be fixed in a sentence: it is concordant with Copenhagen — there is a classical/quantum boundary, and the single outcome is generated at that boundary — but it splits the "boundary" Copenhagen passed over in vagueness, by the two boundaries on the sequence, into two separable things: slot-crossing (dephasing, objectification; unitary, thermodynamic) and settling (drawing out a single content; non-unitary, causal). Copenhagen takes measurement as one undifferentiated affair; the framework splits it into two orthogonal steps and points out that objective-collapse theories alter the dynamics of the settling step (the framework does not), and many-worlds denies the singleness of the settling step (the framework does not). The place is fixed thereby.

1. The ontological identity of the density matrix

The density matrix is the tool most apt, in this discussion, to be assigned excess ontological weight; it must be given a precise place.

The reduced density matrix ρ_S ＝ Tr_E ρ_SE has on its diagonal the weights of the menu's terms (the capacity-layer ψ♯ψ) and off-diagonal the local phase (coherence). Decoherence — phase delocalization, i.e. tracing out the environment, i.e. one partial ledger-merger — wipes the off-diagonal toward zero, leaving a diagonal "classical mixture." This whole object, the density matrix, plainly lives on the capacity layer, on the boundary of slot-crossing: it tracks the menu (capacity) and its phase (its standing relative to the causal slot), and about the settling step, about the content boundary, it says nothing.

The most crucial point is exactly the one it cannot speak. The improper mixture after decoherence (the above-slot many-valued menu, capacity) and the proper mixture after settling (a single content realized, merely unknown) are mathematically identical in their density matrices — the density matrix cannot tell the two apart. This is not its defect but its layer: it is a coarse representation of the capacity layer, and is by nature blind to whether a single content has been realized — a matter on the boundary of settling. This is exactly why decoherence alone (diagonalizing the density matrix alone) cannot solve the measurement problem — diagonalization is a matter of the capacity layer, of the boundary of slot-crossing, while the generation of a single content is orthogonal, a matter of the boundary of settling, which the density matrix cannot see. To mistake the density matrix's diagonalization for the completion of measurement is precisely to mistake the objectification on the boundary of slot-crossing for the content-realization on the boundary of settling. The framework's reading is therefore clean: the density matrix is an accounting tool of the capacity layer; slot-crossing (dephasing) drives its off-diagonal to zero, while settling (taking one content) occurs on the boundary it cannot see; the weight it should bear stops at the capacity layer.

1. Reclassifying three old paradoxes

With these two boundaries and capacity/content, several famous paradoxes can be set back, one by one, into place.

Quantum Zeno [25]. Frequent measurement freezes a system's evolution — a watched pot never boils. The framework's reading: each measurement is a settling, extracting a content and conditioning the subsequent state on that record. Frequent settling is the repeated conditioning of the system back onto the same content, and the evolution that would have carried it away from this content is reset time and again — freezing is just this repeated record-conditioning. Capacity (ρ-OR) persists throughout, each settling conditioned on the latest record. There is no new physical mechanism here, and no need for any collapse dynamics: Zeno is just the result of repeatedly applied record-conditioning, in agreement with the standard frequent-projection read-out. (The anti-Zeno [26], where frequent measurement instead accelerates decay in certain regimes, reads as the joint effect of settling and environmental coupling in that regime, equally requiring no new mechanism.)

Wigner's friend [27]. The friend measures a system in a sealed lab, while Wigner outside describes "lab + system" as a whole superposition until he himself reads it. Whose description is right? The framework's reading: the friend has performed a settling — he extracted a content and was conditioned on the record, so the friend holds a definite content (a matter on the boundary of settling). Wigner, outside, describes that lab as an above-slot many-valued menu, an improper mixture (a matter on the capacity layer, on the boundary of slot-crossing), because from his position the lab + friend + system is an entangled menu whose record he has not yet read. The paradox dissolves through a difference of layer, not a relationist "each true for its own": the friend's settled content is objective, and does not retreat to pre-closure capacity because Wigner has not yet read it; what Wigner legitimately holds, externally, is only an unconditional capacity description of "that record-channel he has not yet conditioned on" — this is neither sending an already-completed measurement back into a coherent pre-closure superposition, nor denying that the friend already has content. The two descriptions fall on different layers (capacity versus content) and do not contradict. When Wigner reads the friend's record (he performs his own settling, conditioned on the lab), record-conditioning guarantees he reads out the same content as the friend; this step is his own conditioning, not the generation of the friend's fact. So there is no contradiction and no need for any action-at-a-distance objective collapse: the friend realized a content, and Wigner's capacity menu contains that content as one of its terms, agreeing with the friend once he conditions on the record. That version sharpened into inconsistency [28] (a group of agents chaining inferences that treat one another's unread settlings as definite outcomes) is inconsistent precisely because, among the agents, capacity description and content realization are run together — and the framework's capacity/content distinction is just what blocks this confusion: you cannot, from your position, take another agent's settling that you have not yet read as a definite content; before you condition on its record, it is for you a capacity menu.

Kochen–Specker [29]. The theorem proves that one cannot consistently assign all observables pre-stored, context-independent definite values — quantum mechanics is contextual: the measured value depends on the measurement context (which observables are co-measured). At this the framework is not embarrassed but rather finds it just what it expects. First, there are no pre-stored definite values — generation, not selection: content is generated at the moment of settling, not read out of a pre-stored assignment table. Second, the content generated depends on context — i.e. on which set of pointer basis, which menu, einselection fixes in that measurement. So the framework is contextual by nature: no pre-stored content (generation, not selection), the menu's basis fixed by the measurement context (einselection). Kochen–Specker is therefore not the framework's paradox but its native color: values are not pre-stored, content is generated with the context. The pre-stored, non-contextual assignment that the hidden-variable route wants to keep, the framework never wanted.

1. Continuity and discreteness

Finally, the loose end each of the two preceding parts left must be taken up — how continuity and discreteness are compatible — and the framework's prior here laid plainly on the table.

The causal slot being a categorial threshold, how can crossing it be sharp while the substrate's aggregation under coarse-graining looks continuous? On the effective, coarse-grained layer this is no difficulty at all: sharp structure routinely emerges from seemingly continuous microscopic evolution, a road already trodden by a family of mature physics with rigorous mathematics, with experiment, even with Nobel Prizes; slot-crossing is of the very same make. First, topological phase transitions — in the Berezinskii–Kosterlitz–Thouless transition [30][31], the parameter of temperature varies continuously, yet the binding and unbinding of vortices changes sides sharply at a critical point; the BKT transition is the paradigm case of a topological phase transition, and Kosterlitz and Thouless received the 2016 Nobel Prize in Physics for work on such transitions in two-dimensional systems. Second, the fault-tolerance threshold theorem — the physical error rate is continuous, yet the correctability of the logical state flips knife-edge at a threshold, and this threshold is itself mathematically a phase transition (mapping to the deconfinement transition of the random-bond Ising model); several platforms have lately demonstrated the error correction across this threshold experimentally. Third, the emergence of objectivity — quantum Darwinism and spectrum broadcast structure [32] give a rigorous characterization of how a classical objective state emerges from the quantum world: the system's pointer state is redundantly broadcast among environmental fragments, whereby many observers independently read one and the same objective information. This strand is especially close: it is of one stock with the broadcast ontology the series' two papers on information [9,10] establish, and is exactly the physical floor of that objective menu, that quantum-Darwinist redundancy obligation, of this paper's Part III. So slot-crossing — a sharp topological closure on the continuous substrate spectrum (a topological closure at the critical point, plus the operational sharpness error suppression gives) — is not a lone assertion of the framework but stands upon this same family of phenomena in mature physics; continuity and discreteness do not clash on this effective layer: continuous is the aggregation under coarse-graining, sharp is the closure at the critical scale.

But a deeper layer must be kept clear of this effective one, and the framework's prior here erected firmly. That "continuous gradient" above is an effective-layer description — substrate aggregation under coarse-graining looks continuous, just as a continuous field overlies a discrete lattice, hydrodynamics overlies molecules. On the fundamental layer, the framework's prior is a definite assertion: any physical quantity, at a sufficiently small scale, is discrete (inheriting the series' foundation paper [14]). This is no conjecture awaiting proof, still less an open blank: all is generated by discrete DD, and a continuum is not something the discrete can generate, so "continuous" can, in the framework, only be an effective-layer emergence and can in no way be a fundamental-layer original. The probability cloud likewise — that it is discrete at a sufficiently small scale is a special case of this prior, not an extra assertion awaiting separate support.

Two matters must be cut clean, lest prior and empirical be run together. Discreteness — any physical quantity discrete at the fundamental layer — is that hard prior above. The specific scale of that discreteness (its magnitude, whether there is a minimal tick, what its value is) is to be measured empirically: the framework forces no value, and does not pronounce any existing scale (such as the thermodynamic floor R_min(T), as already stated in Thesis B) the discreteness scale of the probability cloud — that is the spatial lower bound of the causal slot and of cascade stabilization, and to borrow it as the probability cloud's pixel has no ground. So what is open is the value of the scale, not discreteness itself; the former is to be measured, the latter is prior.

And the framework lays this prior so firmly on the table precisely to make it falsifiable. A prior written soft, shiftable at will, nothing can knock over and it can never correct itself — to write a prior soft looks modest but is in fact an evasion of falsification; only by standing as a definite assertion does it give the world a target it can strike. So the framework states plainly here: if this world is continuous at the fundamental layer — if it should one day be shown that some physical quantity, at arbitrarily small scales, is still a true continuum and not an emergence of the discrete — then this discreteness prior of the framework is overturned, and the framework will revise its foundation accordingly. This paper welcomes, and expects, such a falsification. Hard is not for arrogance, but for handing oneself over to be checked; welcoming falsification is exactly what a hard prior properly entails. The falsification-mode of this prior must be stated precisely, lest it be taken for an ordinary experimental proposition: it is not an operable new prediction this paper redeems (this paper points to no experiment that could directly adjudicate it) but a framework-level risk commitment; its observable signatures, and what result would suffice to exclude some class of SAE discrete implementation, are left to a later dedicated paper. Its being overturned is in the sense that "the fundamental layer in principle requires a non-discretizable continuum, and all discrete DD implementations have failed" — not by a single experiment that "measures some physical quantity still continuous at arbitrarily small scales," the latter being operationally unreachable.

1. Conclusion

The three parts together, this movement of quantum mechanics — the turn from the microscopic toward the macroscopic — can now be drawn to a close.

Thesis A gave the ontology: decoherence is a ledger-merger, phase delocalization, leaving locally a dephased yet still many-valued objective menu; settling is an orthogonal, non-unitary extraction that draws a single content out of the menu; the two lodge at the two boundaries of one and the same sequence — slot-crossing's dephasing is the unitary, thermodynamic prior boundary, settling's drawing-out of content the non-unitary, causal latter boundary; the line of reversibility is drawn above/below the causal slot, not at size, the macroscopic being categorial and not size-based; the arrow of time's direction is rooted in the 4DD orientation, its irreversibility dividing into the thermodynamic (the amplification of spreading) and the causal (settling's one-directionalization). Thesis B gave the mechanism: spreading is a cascade of partial ledger-mergers, the menu stabilizing rung by rung, carrying a renewal-encapsulated nested structure; its structural timescale is only a homogeneous renewal skeleton, the framework proposing no independent decoherence rate, giving only the organizing principle of layered menu-stabilization and the commitment to discrete closure. Thesis C gave the classification: the framework is not a GRW/CSL-type dynamical-collapse theory, not many-worlds, not hidden variables; concordant with Copenhagen yet splitting that boundary into two orthogonal steps; the density matrix is a coarse representation of the capacity layer, blind to the content boundary; Zeno, Wigner's friend, and Kochen–Specker are each set back into place by this apparatus; and that all things are discrete at a sufficiently small scale — the probability cloud included — is a hard prior the framework lays openly on the table, the falsification of which this paper welcomes and expects, its specific discreteness scale left to be measured empirically.

This movement therefore begins in microscopic coherence, passes through decoherence's merger and settling's extraction, and ends in the objective standing-into-place of the macroscopic classical world. What it joins ahead is this series' later chapters treating the macroscopic and continuum limits — many-body, the thermal limit, the emergence of classical mechanics — all upon that causal slot already laid. The framework issues predictions where a difference is detectable, and only re-reads where it merely reproduces standard quantum mechanics; this movement belongs to the latter — it contests standard quantum mechanics on not one inch of empirical ground, only setting that same body of experience into an ontology that can tell apart capacity and content, slot-crossing and settling, possibility and actuality.

References

Entries are numbered in order of first appearance in the text; DOIs are per the author's Zenodo records.

[1] Qin, H. (2026). The Born Rule (SAE Quantum Mechanics VI). Zenodo. https://doi.org/10.5281/zenodo.20479658 [2] Qin, H. (2026). Measurement (SAE Quantum Mechanics VII). Zenodo. https://doi.org/10.5281/zenodo.20533295 [3] Qin, H. (2026). Complex Amplitude and e^{iθ} (SAE Quantum Mechanics II). Zenodo. https://doi.org/10.5281/zenodo.20277037 [4] Qin, H. (2026). The Causal Slot of Information (SAE Information Theory III). Zenodo. https://doi.org/10.5281/zenodo.19797457 [5] Qin, H. (2026). Quantum Entanglement (SAE Quantum Mechanics V). Zenodo. https://doi.org/10.5281/zenodo.20455398 [6] Qin, H. (2026). Quantum Tunneling (SAE Quantum Mechanics IV). Zenodo. https://doi.org/10.5281/zenodo.20369138 [7] Qin, H. (2026). Dual-4DD Time-Reversed Arrows; Λ₁+Λ₂=0 (SAE Cosmology V). Zenodo. https://doi.org/10.5281/zenodo.19329771 [8] Qin, H. (2026). Pre-Closure Ontology (SAE Quantum Mechanics I). Zenodo. https://doi.org/10.5281/zenodo.20252029 [9] Qin, H. (2026). Broadcasting and Reception (SAE Information Theory V). Zenodo. https://doi.org/10.5281/zenodo.19968504 [10] Qin, H. (2026). Gravitational-Wave Dynamics as 4DD Broadcast (SAE Information Theory VI). Zenodo. https://doi.org/10.5281/zenodo.20066645 [11] Qin, H. (2026). Mass Convergence: E = Ic³ (SAE Mass series). Zenodo. https://doi.org/10.5281/zenodo.19510868 [12] Qin, H. (2026). Renewal Encapsulation (SAE Thermodynamics VII). Zenodo. https://doi.org/10.5281/zenodo.19673079 [13] Qin, H. (2026). Copying Fidelity / Retention (SAE Thermodynamics VIII). Zenodo. https://doi.org/10.5281/zenodo.19688304 [14] Qin, H. (2026). SAE Foundation v2. Zenodo. https://doi.org/10.5281/zenodo.19361950 [15] Joos, E., & Zeh, H. D. (1985). The emergence of classical properties through interaction with the environment. Zeitschrift für Physik B 59, 223. [16] Zurek, W. H. (2003). Decoherence, einselection, and the quantum origins of the classical. Reviews of Modern Physics 75, 715. [17] Hahn, E. L. (1950). Spin echoes. Physical Review 80, 580. [18] Zurek, W. H. (2009). Quantum Darwinism. Nature Physics 5, 181. [19] Gerlach, W., & Stern, O. (1922). Der experimentelle Nachweis der Richtungsquantelung im Magnetfeld. Zeitschrift für Physik 9, 349. [20] Lindblad, G. (1976). On the generators of quantum dynamical semigroups. Communications in Mathematical Physics 48, 119. [21] Ghirardi, G. C., Rimini, A., & Weber, T. (1986). Unified dynamics for microscopic and macroscopic systems. Physical Review D 34, 470. [22] Ghirardi, G. C., Pearle, P., & Rimini, A. (1990). Markov processes in Hilbert space and continuous spontaneous localization of systems of identical particles. Physical Review A 42, 78. [23] Penrose, R. (1996). On gravity's role in quantum state reduction. General Relativity and Gravitation 28, 581. [24] Everett, H. (1957). "Relative state" formulation of quantum mechanics. Reviews of Modern Physics 29, 454. [25] Misra, B., & Sudarshan, E. C. G. (1977). The Zeno's paradox in quantum theory. Journal of Mathematical Physics 18, 756. [26] Kofman, A. G., & Kurizki, G. (2000). Acceleration of quantum decay processes by frequent observations. Nature 405, 546. [27] Wigner, E. P. (1961). Remarks on the mind–body question. In The Scientist Speculates (I. J. Good, ed.). Heinemann. [28] Frauchiger, D., & Renner, R. (2018). Quantum theory cannot consistently describe the use of itself. Nature Communications 9, 3711. [29] Kochen, S., & Specker, E. P. (1967). The problem of hidden variables in quantum mechanics. Journal of Mathematics and Mechanics 17, 59. [30] Berezinskii, V. L. (1971). Destruction of long-range order in one-dimensional and two-dimensional systems possessing a continuous symmetry group. Soviet Physics JETP 32, 493. [31] Kosterlitz, J. M., & Thouless, D. J. (1973). Ordering, metastability and phase transitions in two-dimensional systems. Journal of Physics C: Solid State Physics 6, 1181. [32] Korbicz, J. K., Horodecki, P., & Horodecki, R. (2014). Objectivity in a noisy photonic environment through quantum state information broadcasting. Physical Review Letters 112, 120402.

Acknowledgments

A draft of this paper was stress-tested point by point through several rounds of AI review (ChatGPT, Claude, Gemini, Grok); it benefited at its cruxes — the assignment of spreading and settling to two boundaries on one and the same sequence, the three stages of capacity and content, the disposition of the remainder, the ontological identity of the density matrix, and the clarification of continuity and discreteness. The sustained critical feedback of Zesi Chen (陈则思) runs throughout this series. All claims herein, and any errors, are the author's sole responsibility.

第一部分　退相干的本体身份

——账本并入、相位离域，落定作为一桩正交的提取；越槽与落定，同一序列的两道分界

一、起点

论玻恩规则的一篇[1]给了权重，论测量的前篇[2]给了那一份被读出的单一内容——可两篇都把话收在了一桩闭合事件上。闭合之前的那份菜单从何而来、闭合之后散向宏观的那个经典世界如何立定，两篇都没有处理，只把它留下。前篇已经把退相干的位置摆得很清楚：它属能量这一路，交出的是一份各有权重的经典菜单，却不点出哪一项（前篇就退相干相对因果槽的确切位置，措辞尚有可细化处——本文将退相干所交付的菜单置于槽上、已跨那道范畴门槛；此为对前篇越槽措辞的细化，二者本体一致，详见第四节）；至于量子如何与环境结为一体、那份菜单如何形成、又如何稳定成一件许多人都读得到的东西，前篇明言留与后续。本文接的，正是这一段——量子力学三大本体乐章里，由微观转向宏观的第三章。

须先讲明本文不做什么，也讲明它在标准退相干之上多说了什么。点出单一内容的那一桩落定，是前篇的辖区，本文不重做；本文也不为落定添加任何机制、任何触发器。剥掉术语，本文所依的退相干物理就是标准的系统—环境纠缠与约化密度矩阵[15][16]，本文不改它一个方程。本文添的，全在三条本体判读：相干不是被毁灭，而是离域；那份经典菜单不是观者无知造成的混合，而是槽上化的、失相而仍多容的 ρ-OR；其客观性不是某个单一结果，而是 pointer 指标的冗余可读。整段的底，是测量给出信息而量子（或其记录）依旧留存这一桩纯粹的后验物理；本文不改它分毫，只换一种本体的读法。

二、四个分际

要说清退相干是什么，须先把四件向来混作一谈的事分开：相干、退相干、再相干、落定。

先立一条贯穿全文的区分，免得后面反复绊倒：

　　相干 ＝ 多容 ＋ 本地相位。

多容——即多个候选项并存——只是必要的一半；相干还要那个锁在本账之内、可干涉的相对相位（密度矩阵的非对角项不为零）。论复振幅的一篇[3]里的 e^{iθ}、论玻恩规则的一篇[1]里的对偶取向配对 ψ♯ψ，说的都是这后一半。正因相位在本地，诸 ρ-OR 取向才配得起对、干涉得起来。把这两半分开，是下文一切的钥匙：多容可以失了本地相位而仍是多容，相干却不能。

所谓退相干，是一本账 𝔏_S 并入了一本连环境也算在内的、不可因子化的更大联合账 𝔏_SE。这桩并入是实际发生的：量子真的成了一个更大信息整体的一部分。其间动的，正是相位那一半——相位从 𝔏_S 这一本里离域，散进了 𝔏_SE 的关联之中（局域读出上，非对角项趋于零）。相位并未毁灭，它转移了归属。所以退相干留在局域读出层上的，是一份失了本地相位、却仍多容的东西：一份经典菜单。它已经跨过了那道可读的、范畴的门槛——就这一层意义说，它已成信息；可它仍是多容的，并未生出任何单一内容。本文此后称它为 menu-information（菜单信息），给这一格一个显名，免得与量子并入之前那本相干的 ρ-OR 相混。

须就此与论信息的一篇[4]接上，并标明本文是继承它、而非精化它一处合并。论信息的一篇早已把这条分际划好：其论因果槽的一篇第六节明写，退相干是衬底越过因果槽、由衬底范畴提升进 4DD 信息范畴的物理过程，是一桩 categorical lift，并明言"不是 pointer 态的选取"；而单一确定值，该篇第一节之三系于测量，不系于退相干。可见退相干（成为可读的菜单）与测量（点出单一确定值）在那一篇里本就分开。本文不另起炉灶，只承这一既有区分，给那范畴化而仍多容的一格命名为 menu-information，并把单一内容定位在落定（承该篇"测量给出确定值"那一句）。读者切莫以为论信息的一篇说过退相干给出单一值——它没有；本文与它在此处是连续的。

所谓再相干，不外是退相干倒着走，且只在可控的尺度上行得通：施一道 U†——自旋回波、光子回波即是其实例[17]——把离域进 𝔏_SE 的相位重新收拢回 𝔏_S，恢复量子的局域相干。它不是新的范畴，只是同一桩相位离域的逆；它无须解开任何落定，因为退相干本就没有做过落定。到了宏观，这道逆便行不通了，何以行不通，留在第四节细说。

所谓落定，不在上面三者那条路上。它不是量子换了状态，而是一桩正交的提取：从那份失相而多容的菜单里，当场生成并抽出单一一份内容，写进一条 ρ-AND 记录。正交于谁、为何是序列上的另一道分界，是第四节的正题。

仅凭这一节，已可答两桩对退相干常有的误读。其一，退相干不是相干的毁灭，而是相位的离域——它从 𝔏_S 转移进了 𝔏_SE 的关联，未曾消失，这正是本系列余项守恒一语在此处的落实。其二，那份经典菜单不是观者无知造成的混合，而是局域读出层上一份已无可干涉相位、却仍未定内容的可读语法。退相干交付的，是哪一种问题可以被问，不是这一次答案是什么。

三、一桩并入的两面

这桩并入有两面，是同一件事的两个看法，不是两件事，更不是先后两步。

从联合账这一面看，它是 Many-to-One：𝔏_S 与 𝔏_E 并成了一本 𝔏_SE。这本联合账之为真，全在它的不可因子化——按论纠缠的一篇[5]所立的判据，𝔏_SE ≠ 𝔏_S ⊕ 𝔏_E。正是这一点把"结为一体"与"只是因子多了些"分开：若并入之后仍可因子化，那"结为一体"便只是一句换了词的空话；唯其真正不可因子化，并入才是一桩实事。须再说明一句，它也是全篇分判本框架与多世界的阵眼：结为一体的是账本，不是结果——𝔏_SE 仍是一本多容的账。

从局域读出这一面看，同一桩并入是 One-to-Many：量子那本单一账里的相位，被冗余地散进了环境的浩繁碎片里。这里有三样东西不可混作"环境记录"一词：其一是痕迹，某一环境碎片与量子的某个基向量相关联；其二是菜单，即那份稳定、冗余、多个观者都读得到的候选项之集，亦即上文的 menu-information；其三是内容，即落定之后写下的那一份单一记录。退相干经手的只是前两样；第三样是落定的事。

也正因如此，einselection[16] 选的是菜单的基——环境监控之下稳健存活的那一组 pointer 模式、那一片 pointer 子代数——而不是某一个本征值。它定的是哪一种问题可以被问，不是这一次的答案。还须把客观性这一层摆正，免得在此处又把选基与选结果混作一谈：环境冗余、举多个观者皆可读地编码下来的，是那一组 pointer 基、那一套菜单的语法——"哪个问题客观地经典了"，这正是量子达尔文主义[18]所说的冗余。它落在容量这一层，是关于"哪个问题"的客观，不是关于"哪个答案"的实现；众观者读到同一份单一答案、其冗余拷贝彼此一致，那是落定之后的事。还须两点谨守：环境冗余复制的，是与 pointer 相关的分支索引、是经典关联，不是那个未知的量子态本身，故不可克隆之律未曾被犯；而它所散开的关联，按论纠缠的一篇，是容量层的关联，是容量，不是内容。至于这份菜单在宏观如何经一层层并入而稳定立定——其级联具有更新封装的结构形式——那是论题乙处理的机制，本文不入其时标的细节。

四、同一序列的两道分界，与两种不可逆

前两节把四个分际摆开了，须就此点明：它们并不落在两条彼此独立的轴上，而是同一序列的不同观察、不同分界。这序列，是把方法论那套四分形再施于 L₄d 一层而得——是 SAE 既立的递归骨架在这一更细颗粒度上的一次自然应用，既非本篇为退相干一事新造，亦不主张本体上必然只此四道。前一节已将 L₄ 内部分作时间概念、时间箭头、时间方向、信息与因果律四道（L₄a–d）；本节把其末一道 L₄d（闭合给信息与余项）再细分一层，看越槽究竟在何处发生。承 L₄c 已锁的那条本侧单一方向，余下一问：这条方向能确定什么？答案沿四道展开——

　　其一，标定一个可被闭合的寻址槽位（L₄da）：单一方向给出"有一处可被确定"的形式位置——界定一个可被闭合的本征谱／Hilbert 槽位（closure-addressability：定下"在何处可被确定"，而尚未定"确定什么"），标而不构，尚无具体结构。 　　其二，相对相位之可寻址化（L₄db）：单一时间方向一旦锁定，幺正演化 e^{-iHt/ℏ} 便有了单值展开的标尺，相位沿之转动、积淀，诸相位差才谈得上被排成可比较、可寻址的结构——这正是单一方向何以"长出"可定相位的物理环。要被定向的，是分支间的相对相位（局域非对角项 ρ_ij 的可重组性），不是不可观测的全局相位；相位矢量在复平面上的叠加，即此一层的"加法给方向"。 　　其三，相位诸可能成形为一份合法候选菜单——是为菜单层越槽（L₄dc）：系统与环境的张量积把诸相位可能复合、铺成一份菜单；越槽在此发生——一组合法候选已客观成形，而尚无"点选"那个动作。 　　其四，闭合提取，生成单一内容——是为落定（L₄dd）：闭合查询与记录槽到来，与菜单咬合，使其中一份内容当场落定，写进一条 ρ-AND 记录；构是信息，余项是这内容如何储存。

〔这四道在本框架的逻辑骨架上，依次对应同一律、矛盾律、间隔律与综合律（综合律在 L₄ 这一层显为因果律）；方法论总论的正文当时以 4DD／因果律语境给出，未把后两道之律名显式写出。此一律名映射——尤其"加法给方向 ↔ 矛盾律"一处——本篇视作一组待验的前瞻对应，不在本篇承重：下文关死预写、递归止步诸论证，皆只跑在已立的容量／内容、L₄dc／L₄dd、ρ-OR／ρ-AND 之上，不依赖这组律名。本文专注于退相干这一具体域，只在此调用该四步骨架，不替总论增补；将其律名显式化，留作日后方法论之事。〕这一细分到此为止：闭合那一步的产物是 ρ-AND（单一内容），已综合成一、不再含可被"再铺一层候选"的多重——这是本体的底（无多容燃料可供再分，递归遂止）；功能上亦同：menu 与 content 的区分至此已闭合，再往下（内容记录的内部储存、读写保持、宏观稳定化）属另一问题域、归后续之事，已不在"退相干与落定之分"的题中。故"在此停"不是预设了目标的 post-hoc 之停，是本体底与功能停点叠合的必停。

于是越槽（L₄dc）与落定（L₄dd），是这同一序列上前后相继的两道分界——不是两条正交的轴。本文他处所用的"两条轴"（越槽那条、落定那条），正是这一条序列在粗颗粒度上投影出的两道分界之别名；作粗略的简写仍有用，但本体上只有一条序列、两道分界，不是两个彼此独立的维。从多容到单一，不是沿另一条横轴的位移，而是从 L₄dc 这道分界跨到 L₄dd 那道分界的一步。下文偶称落定为"正交"的提取，须就此立一道界：所谓正交，只指机制类型的异质——越槽由系统—环境的幺正弥散与热力学不可逆支撑，落定由非幺正的内容提取与因果不可逆支撑；它并不意味二者属于两个彼此独立的本体维，在 L₄d 的细分上，二者仍是同一序列的前后两道分界。

这一序列把退相干前后分作三段，判据只是一条：项是否符合间隔律（玻恩规则的可读结构）。退相干之前，槽下相干——诸项彼此干涉、相位未离域，尚不符合间隔律，还不是可独立读出的合法候选；干涉条纹只是"菜单将成"的影子，此时连"答之可能性"都还没有。越槽之后，槽上菜单——诸项已是一组合法候选，各正交、各带玻恩权重、可作"从中取一"的对象；这是问之前的可回答性：菜单已客观成形，而闭合查询尚未咬合。它不是"答"（内容义），是答之可能性的条件——是容量（ρ-OR），不是内容。落定之后，单一内容——从这组合法候选里，当场生成并落定其中一份，写进一条 ρ-AND 记录。

这三段之间，有一条干净的判据分容量与内容：可逆性。L₄dc 在未落定前可逆——可再相干，把已离域的相位收回、菜单撤回成相干叠加（斯特恩—盖拉赫[19]里把两路重新合束、干涉条纹便复现，自旋回波、光子回波亦然）。而一份已生成的内容（ρ-AND）一旦立起便不可逆。能被撤销回"无答之可能"的相干态者，必不持有任何已生成的内容；故 L₄dc 是容量、未跨入内容——可逆，正是它仍在容量层的判据。这一判据承前篇：凡逼近落定而未落定者皆可逆，"没成的事，自然撤得回"；本文只把这可逆性确切地定位在 L₄dc 这道分界。

这也从逻辑上关死了"预写"之惑，与前篇"生成而非选定"相洽。L₄dc（张量积铺菜单）出的是菜单——多项、合法候选、容量；L₄dd（闭合）出的才是内容——单一。容量不是现实，故二者必是前后两道分界，而"铺成合法候选"这一步在逻辑上不可能写出任何确定内容；究竟哪一份内容被落定，是在 L₄dd 那一刻当场生成的，既非预存、亦非从一排现成答案里选中。〔以律名言：间隔律只立诸项之可分可读、不产内容，综合律才生内容；二律之别即两道分界之别。〕本框架因此天然是语境性的，与不预设预存值相洽；它不是隐变量理论。

须给"槽上多容"一个具体的实例，免得它听着像个空格。它就是退相干之后、落定之前的那个非纯混合（improper mixture）——客观的合法候选菜单。一颗在斯特恩—盖拉赫装置里、which-path 信息已对局域控制而言不可回收地泄漏进环境、却还没有探测器响动的原子，此刻便是：相位已离域（若不能寻址并反演相关环境自由度，干涉便不再恢复），却没有哪一路被定下——已失相、客观，而上下未定。一个已不可逆退相干、却还没被读出的量子比特，每时每刻都是这样。须把它与"单一而未知"切清，这恰是它之为多容、不为单一的关键：非纯混合是全局仍纠缠、没有任何单一内容在本体上被实现，诸项都还是容量；而一枚盖住的硬币是纯混合——单一内容已经实现（已落定），只是你不知道。菜单是前者，不是后者。诚实地补一句：非纯混合与纯混合在局域上区分不开（本地统计一模一样），这正是测量问题的核心；所以"槽上多容"是一个本体范畴——退相干后、落定前的那个非纯混合——不是一个能靠本地测量挑出来的操作量。

越槽（L₄dc）由间隔律这道范畴阈值判定，不由尺度判定，但范畴与尺度并非毫无关系。这道阈值落在论信息的一篇[4]那条因果谱上——因果性是衬底聚合量级所对应之有效因果强度的连续梯度。衬底聚合（升向因果谱上端）是抵达这道阈值的主导自然路径：聚合越高，菜单越客观、越冗余（即众观者可读之就绪）。但尺度不是必要条件——还有测量：环境耦合可不靠尺度、强行把系统推过阈值（斯特恩—盖拉赫的原子始终是微观的、从未长到 μm，越槽全在飞行中由环境的逐步泄漏完成，撞屏才是那一记闭合查询）。它也不是充分条件——有效因果强度取决于聚合与耦合，故一个大而仍相干的物体，因相干而有效因果低，仍在槽下。问"多大才算越槽"是范畴错置：阈值是范畴的，尺度只是抵达它的一条（主导而非唯一）路径。须分清"路径"与"那一道分界"：聚合是通往 L₄dc 的路径，而落定（L₄dd）与这条聚合谱正交——它不是聚合谱上的某一点，而是客观菜单已立之后、由闭合查询咬合而生成内容的下一道分界。

由此可定"宏观"一词，且它与标准量子力学时而松用的"宏观"不同。本文的宏观，是范畴义的：一个存在之为宏观，在于它能承载因果信息，判据是载信息、而非尺寸——此即论信息的一篇[4]第一节之五所立。所以因果槽不是尺寸门槛，是范畴门槛。一个大而仍相干的物体——玻色—爱因斯坦凝聚、超流氦、超导电流、物质波干涉里的大分子——按尺寸宏观，按本框架的范畴却在槽下：它是大尺度的亚因果衬底，相位仍在本地（相干多容），不载客观信息。没有"槽上的相干大物体"；一旦客观化，相位即离域、即去相干、即到了槽上。容量与内容之别，也由此分作三格：槽下容量（相干 ρ-OR，相位在）、槽上菜单容量（menu-information，失相多容）、槽上内容（ρ-AND，单一）。一句话钉死：槽下／槽上是范畴边界，容量／内容是读出状态；菜单是槽上、却尚未内容化的 ρ-OR。

这一序列上前后两道分界，各押着一种"回不去"，二者异质，不可混作一根。

越槽（弥散）这道分界上的不可逆，是热力学的。 退相干是 system+environment 联合的幺正演化——局域约化态看着混了，是因为把环境 trace 掉了，全局态始终是纯态、按一个幺正 U 演化；本框架非坍缩、不往薛定谔方程加任何随机项，故原样保留了这套幺正动力学。所以倒回弥散用的是 U†——一个相干控制操作；在小、可控的尺度上它实测可逆（自旋回波、光子回波、动力学解耦），无须取消任何结果，因为没选过任何结果。到了宏观，倒不回来——可"倒不回"的缘由是相位已离域进 10²³ 个寻址不到的自由度：形式上的 U† 仍然存在，只是够不着。这是热力学／FAPP 的签名，且尺度依赖（小可回波、宏观付不起）。论纠缠的一篇所谓 One-to-Many 之不可逆，落在这宏观一端是确切的；但它的根是热力学，不是因果——散开的客观菜单，要收回得重新寻址那浩繁的自由度，那是算力上够不着，不是因果律禁止。

落定这道分界上的不可逆，才是因果的。 从多容收出单一内容，是非幺正的——一份被实现、其余（本是容量）未实现，这一步任何尺度都没有 U†。一颗电子自旋的落定，与一根宏观指针的落定，"一份被实现"同样倒不回，不随尺度变。它不靠寻址难度，是根本没有退回"无内容被实现"的幺正算符。这才是单向化，才是因果不可逆——要它退回，就得反转那一份已立的事实，即反转因果。

所以两种不可逆同出在"生成 4DD 信息"这一属之下，却是两个种：越槽生成失相的客观菜单（幺正、热力学、受槽与尺度 gate、槽下可回波）；落定生成单一内容（非幺正、因果、尺度无关、无 U†）。"因果不可逆"这顶帽子只扣落定；宏观多容的不可逆是纯热力学。 这两端，各由本框架的一条承诺摁住：若说越槽也非幺正，便是给退相干加了随机项、沦为客观坍缩理论，违非坍缩；若说落定也有 U†、可原则倒回，便是诸支皆实、滑回艾弗雷特，违其不相容。因果槽 gate 的是越槽在哪一尺度发生，挪不动这条幺正／非幺正的种线。这两种不可逆分属同一序列前后两道分界、抵押物截然不同（弥散押 FAPP 的海量寻址熵障，落定押单向的因果生成），便是前文所谓"两条轴"在粗颗粒度上的确切所指。

宏观叠加之惑由此自行消解。宏观退相干之后的系统，是一份失相的客观经典菜单，是容量——冗余、客观——而不是一个已经实现了的叠加，也不是坍缩成了某个单态。那只猫，在现实里几乎瞬间退相干，是一份槽上的非纯混合：死、活两项都是容量（可能），不是两个已实现的世界。落定从中提取出一份真的内容；其余诸项仍是容量。多世界让每一条分支皆得实现；本框架只实现被提取的那一份，其余留作可能。可能与现实之别——容量与内容之别——这一桩模态上的分判，才是本框架与多世界的真分野，不是换了一套词。而单一结果之为真、不容一个幺正多世界，并非另立的前提，正是落定那一步非幺正实现的推论：落定实现了真的一份，故无诸支皆实的层面。至于多世界究竟错在何处、其"分支"该往何处归类，留到论题丙细说。

五、后验与相容

本文既只重读、不另出预言，就得把它与标准量子力学如何严丝合缝接上，讲到可审计的地步。

先把一处与后验物理对齐，免得"测量之后量子依然是经典 ρ-OR"一语被读错。它不是说那份原来的多容原封未动——若如此，重复测量便会给出彼此不一的结果，这是经验上为假的。确切的读法是：落定之后那份经典的 ρ-OR，是被记录条件化了的——它与已被提取的那条 ρ-AND 记录经典地相关联——故而再读一次，给出的是同一个结果。无条件地看，容量（ρ-OR）依旧留存，是本体的盈余；条件于那条已提取的记录，预测便与标准的投影后态、与标准坍缩一致，是可重复的。这里的条件化是经典关联，不是量子相干，两层因此不打架。退相干给出容量菜单，提取写下一份记录，记录再条件化此后的预测——三者合起来，复现的正是标准量子力学，没有任何偏离。

这一层后验的底，本文不增不减，只复述一遍：测量给出信息，而量子或其记录依旧留存；单次只得一份内容，得不到完整的态。其间还须分两种情形。量子非破坏的测量之后，量子作为经典的 ρ-OR 留存，仍是一个量子对象；而许多真实测量是破坏性的——光子为光电管所吸收、粒子在量热计里湮灭——这时原量子被消耗掉了，留存的是那条被提取的记录，不是量子。两种情形本框架都接得住：一者容量仍在，一者内容已提。又，单次测一个量子，所得的是一个本征值、一份内容，读不出完整的态 ψ——单发只给一次采样，系综才给出分布或期望，两者都不是 ψ，此即论隧穿的一篇[6]所立"读出读得到的是期望值，不是态本身"与容量／内容之别在此处的合流。

本文不出 SAE 独有的退相干率预言——与标准重合则非独有，与标准相左则无后验支持。但不出独有预言不等于没有失败条件。本文的本体读法，是套在标准开放系统形式之上的一组相容义务，逐条可证伪：其一，账本并入须能重写为标准的约化，即 ρ_S = Tr_E ρ_SE，否则并入的本体映射失败；其二，相位离域须对应标准的非对角抑制，即 ρ_ij ↦ ρ_ij⟨E_j|E_i⟩，否则"离域"只是隐喻；其三，那份 pointer 基须能由系统—环境相互作用哈密顿量、由稳健的 pointer 子代数接出，否则经典菜单没有物理锚；其四，One-to-Many 的冗余须对应量子达尔文主义的碎片冗余、互信息坪，否则"客观菜单"的机制落空；其五，也是命门：若本文的叙述不得不引入某个单一的 i₀ 作为实际结果、才能解释宏观记录，那它就偷渡了落定，本体边界即告失败。这五条不是新的经验预测，而是结构性的失败条件——敢把命门亮出来，正是最强的防守。

六、时间之箭的本体位置

这一切把时间之箭摆在了何处？须把方向与不可逆分开，分别落在本框架自己的结构上。

先说方向。本侧 4DD 的时间方向是确定的——这是箭头方向之源，且本框架内生，不是外加的过去假说。论宇宙学的一篇[7]第三节之一已立：3DD 对称生出一对时间箭头相反的 4DD，本侧的参数序是 (T₁, T₂)、对面是 (T₂, T₁)，我们的大爆炸即对面的大坍缩；本侧因此有一个确定的 4DD 时间取向。须把这一句的分寸标死：这一双 4DD 取向是本框架的宇宙学边界承诺（出自论宇宙学的一篇），不是本文内部的推导；它给的是时间的方向，并不因此就给出"我们这侧的起点为何低熵"——那一问被锚到了取向结构上，未必就此消解。所给的，只是把箭头方向根在本框架自己的取向上，不假外求于一条热力学的低熵过去假说。

由此可定因果槽上下与这取向的关系。在槽下，存在虽也沿本侧 4DD 方向演化，却同时受那一对相反箭头的影响，故无单一方向、无因果，只是概率（论玻恩规则的一篇）；越槽与落定，是逐步把这一双向收束到本侧单一方向上。而依第四节，这"收束"分作序列上前后两道分界：越槽（去相位、客观化）是幺正的、热力学的那一步，落定（收出单一内容）才是非幺正的、真正单向化的那一步。所以本框架那支根本的、因果的箭，在落定与 4DD 取向这一面（亦是论纠缠的一篇[5]与系列首篇[8]所立的那一面）；One-to-Many 的弥散，是把这一支根本方向性，经由天文量级的冗余记录，热力学地放大成一支客观的、举宇宙皆无从抵赖的熵箭。它是放大、不是起源——弥散的方向性本身仍预设着那个确定的 4DD 取向，若说箭头起源于弥散，便是把所要解释的方向倒过来当作了结论。它不推导第二定律，只在量子与经典的交界处给出第二定律的一个账本机制的版本，且须与那支根本的 4DD 因果之箭相一致。

最后说余项的去向，且把分寸标到底。这一段说的是落定，不是退相干。 退相干是幺正的，它不舍弃任何分支——离域掉的相位，悉数化为环境的经典关联，那是热力学的余项，留在那本仍幺正的全局账里。真正发生单向化、舍弃了那受相反箭头影响一侧的，是落定。落定所舍的那一侧，并未消灭，本系列的余项守恒不容它消灭。它作为单向化所生的 4DD 信息，按本系列论信息的两篇[9,10]，去往 4DD 不变量的广播信道——引力波（引力波是 4DD 广播的动态显现，沿绝对的 Planck 衬底传播，其能量经 Mass 系列[11]的 E=Ic³，c 即 4DD 信息穿衬底的突破率）。须把这一句的分寸标到死：这是全局的本体记账、是广播信道的类比，不是局域发射模型。本文绝不主张某一桩局域落定事件必然发出一个可分辨的引力波包；落定这一微观尺度上的引力广播微不足道，量级远在任何探测之下。故本文不就落定自立任何可测的引力波信号，也不继承论信息的一篇在黑洞 ringdown 上那条可测预言——那是宏观源、黑洞专属的辖区，且本文不把这局域余项焊到任何宇宙学常数上。本框架在能探测到分野处出预言（黑洞内部印记、多信使时差），在只复现标准物理处只重读；同一条诚实的纪律，按尺度与域行事。

于是退相干这一路，本文可收成几句话：它是一本账并入一本更大的、连环境也算在内的不可因子化的联合账，相位由此离域，局域上只余一份失相而仍多容的经典菜单；这份菜单在小尺度上可经回波收回，在宏观尺度上则真不可逆地弥散开来——而这一桩不可逆是热力学的。点出其中单一一份内容的，不在这一路上——那是落定，是一桩正交的提取，把一份当场生成的内容抽离进一条 ρ-AND 记录，而容量不灭；它的不可逆才是因果的、单向化的。同一序列两道分界，两种不可逆：越槽去相位是幺正的、热力学的（前一道分界），落定收单一内容是非幺正的、因果的（后一道分界）。方向由确定的 4DD 取向给定，时间之箭于宏观经弥散热力学地放大，单向化所舍的余项由引力—广播本体守恒地承载；至于那一份内容因何被生成、其时标几何、几桩旧佯谬如何随之重新归类，则是此后两部分的事。

第二部分　退相干的级联机制与时标

——部分账本并入、一个更新骨架，与一条不提出独立退相干率的纪律

一、起点

前一部分把退相干的本体身份摆定了，并把它与落定分作同一序列的前后两道分界：越槽（弥散）是幺正的、去相位的那一步，把衬底抬成失相而仍多容的客观菜单；落定是非幺正的那一步，从菜单里收出单一内容。前一部分明白地留下两问，都落在越槽这道分界上：那份菜单不是一步铺成的——它在一层层的并入里逐级稳定，其机制几何？又，这一过程的时标几何，它与标准退相干理论里的那个率是什么关系？本部分接的正是这两问。须一开头就划清：本部分只处理越槽这道分界的机制与时标；落定那一桩正交的、非幺正的提取，前一部分已立其无机制化的时标，本部分不碰。

须先把纪律立在前头，与前一部分同出一辙。本部分不提出独立的退相干率。一个系统实际的退相干快慢——其 Γ_dec、其 T₂、其 T_φ——取自标准的系统—环境哈密顿量与噪声谱（Lindblad 方程、主方程、自旋—玻色模型那一套[20]），本框架不另从头算它。本框架在此添的，是对那一桩级联的一种结构重读——它何以呈级联、其逐级的结构形式为何、其本体承诺是什么——而不是一个独立的率。与标准重合者，非 SAE 独有；与标准相左者，无后验支持。这条两面的纪律，本部分自始至终守。

二、级联：层层的部分账本并入

前一部分的 One-to-Many 弥散，不是一桩一步到位的事。一个量子那本账，并不在一瞬之间并入那本连整个环境都算在内的联合账；它是一层一层并进去的——相位先离域进最近的几个环境自由度，再散进更大的一圈，终至宏观那一层冗余。每一层都是一桩部分的账本并入、一次部分的去相位：相位逐级离域，菜单的可读性随之逐级抬升——读得到的观者越多、冗余越厚，这份菜单便越稳、越客观。所以退相干在本框架里不是单一的事件，而是一串部分并入的级联，菜单在其中逐级立定。

须就用词谨守一处，它接前一部分的术语纪律。本部分不以"闭合"称这级联的每一层。前一部分把"闭合"专留给落定那一桩内容闭合，且明立落定无机制化的时标；若把级联的逐层也唤作"闭合"，又赋之以下文的逐级节拍 τ_slot，便等于给一个被定义为无时标的东西安了时钟，自相矛盾。故级联的逐层，本部分一律称部分账本并入（或部分去相位）；它们有时标（τ_slot），是越槽这道分界上的事。"闭合"只留给落定。

这一串级联有一种自相似的结构，本系列热力学的一篇[12]已就此给出过形式，本部分作为结构形式援引：下一层的并入，为上一层提供其更新的槽——即底层的一桩部分并入，生成上层赖以计步的那一个节拍。于是整条级联具有更新过程的形状：每一次"更新"是一层的部分并入，上一层在下一层的节拍上推进。这便是更新封装——级联不是一条无结构的连续衰减，而是一层套一层、各自以下层并入为节拍的嵌套。

须把援引的分寸标死。热力学那一篇在给出更新封装形式的同时，附有一条 no-go：纯粹连续的耦合无法继承一个良定的、离散的时标。本部分把这条 no-go 降格为一则结构类比、一处提示性的平行，不作独立的动力学证明。它提示的是级联何以呈更新结构——何以须有逐级的节拍，而非一片无层次的连续——但它并不独立地推导出这条级联的时标取何值。时标之值，仍回到下一节所述的标准物理。

还须把这一节稳定下来的东西认准：级联稳定立定的，是那份菜单——失相而仍多容的客观容量，是"哪个问题客观地经典了"。它不点出单一内容。从菜单里收出单一内容的落定，是正交于这条级联的另一桩事，不在级联的末端等着收尾。更新封装组织的是菜单的稳定化，不是内容的生成。

三、两个时标与一个交叠

级联里须分开三个量，向来易混。

其一是 τ_slot，逐级的槽时——一层部分并入所需的时间，是这条级联里单步的节拍。其二是 τ_cascade，级联时——菜单经整条级联稳定立定所需的时间。其三是 ρ_ret，逐级的保留度，本部分把它确切地认作环境态的交叠：即与菜单中不同候选项相关联的那些环境态彼此之间的交叠。交叠小，区分度好，相位离域得快；交叠近一，区分度差，离域得慢。

须把这交叠写到一般，不可只写成一个单一常数。它依分支对、依步、依环境碎片而变，逐碎片地写是

　　ρ_ret^{(ij,k)} ＝ |⟨E_i^{(k)} | E_j^{(k)}⟩|，

多碎片环境下，分支对 (i,j) 的总交叠是各碎片之积

写成连乘，One-to-Many 的冗余复制才真正接到量子达尔文主义的碎片冗余上，而不是只接一个全局的单一交叠。

三者的结构关系，须摆明它只是一个齐次更新骨架，不是一条普适的退相干时标公式。在一个齐次的更新简化情形里——设每一级部分并入把相干交叠乘以一个近似固定的因子 ρ_ret——达到某个经典菜单阈值 ε 所需的步数是

　　N_ε ＝ max(1, ⌈ ln ε / ln ρ_ret ⌉)，

而

　　τ_cascade ～ N_ε · τ_slot。

简写作 τ_slot / (−ln ρ_ret)（以 max(1,·) 为底），只是在固定阈值、固定步长、固定保留度这三重假设下的结构骨架。把它当普适公式读，便会被问倒：阈值在哪里？而 ρ_ret→0 这一端已由那个 max(1,·) 的底兜住——至少一个 τ_slot，τ_cascade 不会随 ρ_ret→0 而趋零。这些本是公式层会被抓的点——而答案是，它本就不是公式层的断言，是骨架层的结构读法。一旦 ρ_ret 依分支、依步、依碎片而变，它就退回为标准开放系统动力学的一种粗粒化读法，而非一个独立的时标。

须就命名讲明一处，这也是本部分一桩刻意的厘清：本部分不用"τ_dec"这个记号。热力学诸篇里的 τ_dec 指的是一个有限的因果落定时、是非玻尔兹曼核的衰减尺度，它不是量子退相干率；兼称即生歧义。故本部分以 τ_slot、τ_cascade 称 SAE 这一侧的结构时标，而把量子实际的退相干率另留给标准物理的 Γ_dec、T₂、T_φ。而 ρ_ret 认作环境态交叠，亦是一桩刻意的对齐：本系列热力学较早的一篇[13]曾把这一保留度认作复制保真 f，本部分将它重释为标准退相干理论里那个环境态交叠，使两套语言在此处合流。

四、与标准退相干率的对接

把上一节摆明，本部分与标准退相干理论的关系便清楚了：是结构的重读，不是独立的率，也不是"按构造映到"那个率上的某种推导。

一个系统实际的退相干率，由标准的系统—环境哈密顿量与噪声谱决定——这一点本部分照单全收。本框架所添的更新封装结构，是对这同一桩级联的一种结构读法：当标准开放系统动力学可被粗粒化为逐级碎片交叠的衰减时，τ_slot 与 ρ_ret 给出它的 SAE 结构读法。它不抢 Lindblad 方程的计算权，也不另算一个率。

那么降格之后，本部分的净增量到底是什么，须明说，免得"只是把标准退相干换了套词"的质疑回来。标准退相干率给的是数值；更新封装给的是退相干为何呈层级菜单稳定化的组织原则——下层环境碎片的部分并入，为上层冗余菜单提供新的更新槽；再加上一条离散闭合的本体承诺：级联不是一片无层次的连续衰减，而是一层套一层、以下层并入为节拍的嵌套。本部分的贡献，止于这一组织原则与这一本体承诺，不是导出一个率。薄，但不空，前提是这么标。

故本部分须把纪律再申一遍，且与前一部分的可证伪布局接上：本部分不提出独立的退相干率。与标准重合，则非 SAE 独有；与标准相左，则无后验支持。凡涉具体的率与时标之值，皆回标准物理；凡本框架所言，止于结构与本体的重读。

五、界外两处

末了须把两处与时标相邻、却不归本部分定夺的事，明白地标在界外。

其一，落定没有本框架机制化的时标。前一部分已立：落定是一桩裸的闭合，本框架不为它添任何机制、任何触发器；既不机制化，本框架便也不给它一个率、一个时标。"落定何时发生"是测量问题里那桩计时之惑——本框架对各家诠释保持中立，把它留作裸闭合，不在此处强为它派一个时标。本部分所给的两个时标，都是越槽这道分界上级联的时标，与落定那一桩正交的、非幺正的提取无涉。

其二，热力学地板 R_min(T) = ℏc/(2πk_B T)——按本系列论信息的一篇[4]，它是因果槽的最小尺度、是级联在热力学地板上得以稳定一个 bit 的空间下限——本部分不把它断作概率云的离散尺度。把它读成"概率云的像素大小"是没有后验支持的，与本部分不给 τ 派新值是同一条纪律：它是级联稳定化的空间下限，不是概率云离散性的标度。cell-tick 这一概念，本部分亦只作框架未来的方向，不在此处兑现。

至于连续与离散如何相容这桩正题本身——衬底聚合是连续的梯度，4DD 拓扑闭合却在临界尺度上 sharp 地涌现，按本系列论信息的一篇[4]那一节，是连续物理 + 临界拓扑闭合 + 误差抑制三层 sharpness 的合成——则留与下一部分，与几桩旧佯谬的重新归类一并处理。

于是退相干的机制与时标，本部分可收成几句话：弥散是一串部分账本并入的级联，菜单在其中逐级立定，整条级联具更新封装的嵌套结构（下层并入为上层计步）；它稳定的是那份失相而多容的菜单，不点出内容（点内容的落定正交于此、不在末端收尾）。其结构时标 τ_cascade ～ N_ε·τ_slot 只是固定阈值、步长、保留度下的齐次更新骨架，逐级保留度 ρ_ret 即环境态交叠、多碎片取积——本框架在此只作结构的重读，给的是层级菜单稳定化的组织原则与离散闭合的本体承诺，不提出独立的率，与标准重合则非独有、相左则无后验。落定无机制化的时标，热力学地板只是级联稳定化的空间下限而非概率云的离散标度；连续与离散相容的正题，留待下一部分。

第三部分　痛点的重新归类

——在诠释谱上的位置、密度矩阵的本体身份，与三桩旧佯谬

一、起点

前两部分立了两样东西：退相干的本体身份——一桩并入、相位离域、失相而多容的客观菜单，落定作为一桩正交的、非幺正的内容提取，分居同一序列的两道分界；以及这级联的机制与时标。本部分把这副装置转过来，对准量子力学测量与诠释里那几桩著名的痛点。须先讲明本部分的姿态：这是重新归类，不是把那些痛点当作新物理去解。本框架不出独有的预言；它能做的，是用前两部分的两道分界、容量与内容、越槽与落定、生成而非选定这副本体语汇，把这些痛点逐一摆回它们该在的位置——指出哪些因此消解、哪些只是换一种读法、哪些仍悬而未决。所求的，是把一团向来纠缠的概念厘清，不是给出一个可证伪的新断言。

二、在诠释谱上的位置

前两部分零散地说过本框架不是什么；这里一并摆定，并说清它何以落在这一处。

它不是 GRW/CSL 型动力学坍缩理论。 GRW[21]、CSL[22]、彭罗斯[23]那一类，往薛定谔方程里添一个随机的、自发的定域化项，让波函数在物理上真地坍缩——这是一个改动了动力学、因而原则上可证伪的新物理。本框架不做这件事：退相干那一步是幺正的，本框架原样保留；落定那一步虽非幺正，却是一桩裸的闭合——无阈值、无触发器、无外加的随机动力学，本框架不为它写任何方程。须把话说准，免被误归：本框架确有一桩客观的、单一的落定，但它不提出任何新的坍缩动力学方程——故它是一种 objective single-outcome ontology（客观单一结果的本体读法），不是一种 dynamical collapse modification（动力学坍缩修正）。这一类动力学坍缩论给的是一个新的动力学项；本框架给的是一种本体的读法。两者范畴不同：可证伪的物理修正，对纯重读。

它不是多世界。 这一条的分野，前一部分已点出，这里讲到根上：是容量与内容那一桩模态的分判。多世界[24]让叠加里每一条分支皆得实现，成为一个个并行的真实世界。本框架只实现被落定提取的那一份内容，其余诸项仍是容量——是可能，不是已实现的世界。所以那一句是阵眼：结为一体的是账本，不是结果；菜单是多容的客观容量，不是诸个已实现的世界。多世界与本框架在形式上常常一致（同一套幺正演化、同一份退相干），分野全在这一桩模态承诺：诸支皆实，还是一份被实现、其余留作可能。以本文那条序列言，多世界恰是卡在 L₄dc 而误读：菜单（合法候选）已客观成形，可它把这一份可逆的、尚无内容的容量菜单当成了已完成 L₄dd 闭合的诸个内容世界——把可回答性错当作了已回答，把间隔律已立错当作了综合律已成。

可多世界赖以立足的，是一个真问题：菜单到落定，被取的是一份，其余诸"选项"去哪了？多世界答"各自成真、分叉成诸世界"。本框架须正面接此问，而非回避。其余诸项即落定之余项：落定（L₄dd）是闭合给构与余项，构是被取那一份内容，余项正是未取诸项之去向——它们不消失、也不分叉成世界，而是作为余项进入 4DD 广播的守恒账（其能量经 Mass 系列[11]的 E=Ic³，其动力学即论信息的两篇[9,10]所立之 4DD 广播渠道）。须按前文余项之纪律收口：这是全局记账层的去向，不主张每一桩落定都发出一份可分辨的引力波。于是"其余选项去哪了"有了答案——去了余项、入守恒账，无须诸世界皆实。多世界之病，正在它缺这一条余项出口：不知未取诸项能往何处去，只好令其各自成真。须就地申明分寸：这是一桩本体层的判定，不是经验上的裁断——本框架与多世界在可观测预言上并无非平凡的分别（一如本部分开篇所申，本框架不出独有的预言），所拒者非分支这套形式（branch formalism），而是把分支本体化为并行真实世界。容量不是世界，可能不是现实；单一结果之为真，不是另立的前提，是落定那一步非幺正实现的推论。

它不是隐变量理论。 本框架不认为测量是在揭示一个预存的确定值。前一部分立的"生成而非选定"说的正是这个：那些权重是预存于 ρ-OR 的容量，可究竟哪一份内容被提取，是在落定那一刻生成的，既非预存、亦非从一排现成答案里选中。所以本框架天然是语境性的，与不预设预存值这一立场相洽——这一点的细处，留到第四节论 Kochen–Specker 时再说。

由此，本框架在谱上的位置可一句话定：它与哥本哈根相协——经典/量子有一道边界，单一结果在那道边界处生成——但把哥本哈根含混带过的那道"边界"，用序列上的两道分界拆成了两件分得开的事：越槽（去相位、客观化，幺正、热力学）与落定（收出单一内容，非幺正、因果）。哥本哈根把测量当一桩浑然的事；本框架把它分成正交的两步，并指出客观坍缩论改的是其中落定那一步的动力学（本框架不改），多世界否的是其中落定那一步的单一性（本框架不否）。位置由此而定。

三、密度矩阵的本体身份

密度矩阵是这场讨论里最易被赋予过多本体分量的工具，须给它一个准确的位置。

约化密度矩阵 ρ_S ＝ Tr_E ρ_SE，其对角是菜单各项的权重（容量层的 ψ♯ψ），其非对角是本地相位（相干）。退相干——相位离域、即对环境取迹、即一桩部分账本并入——把非对角抹向零，留下一个对角的"经典混合"。可见密度矩阵这整个对象，活在容量这一层、越槽这道分界上：它追踪的是菜单（容量）及其相位（相对于因果槽的归属），对落定那一步、对内容那道分界，它什么也没说。

最要紧的一处，正是它说不出的那一处。退相干之后那个非纯混合（improper mixture，槽上的多容菜单、容量），与落定之后那个纯混合（proper mixture，单一内容已实现、只是未知），其密度矩阵数学上全同——密度矩阵分不开这两者。这不是它的缺陷，是它的层级：它是容量层的粗表示，对"是否已有单一内容被实现"这一桩落定那道分界上的事天生是盲的。这恰恰说明了，为什么单凭退相干（单凭把密度矩阵对角化）解不了测量问题——对角化是容量层、越槽这道分界上的事，而单一内容的生成是正交的、落定那道分界上的事，密度矩阵看不见它。把密度矩阵的对角化误当成测量之完成，正是把越槽那道分界上的客观化，错认作了落定那道分界上的内容实现。本框架的读法因此干净：密度矩阵是容量层的记账工具，越槽（去相位）使其非对角趋零、落定（取一份内容）则在它看不见的那道分界上发生；它该担的分量，到容量层为止。

四、三桩旧佯谬的重新归类

有了这两道分界与容量／内容，几桩著名的佯谬可以逐一摆回位置。

量子芝诺[25]。 频繁测量冻住一个系统的演化——看着的水壶不开。本框架的读法：每一次测量是一桩落定，提取一份内容、并把此后的态条件化在那条记录上。频繁落定，便是把系统反复条件化回同一份内容，那本会把它带离这份内容的演化，于是被一次次重置——冻结，就是这反复的记录条件化。容量（ρ-OR）始终留存，每一次落定条件于最新的记录。这里没有新的物理机制，也无须任何坍缩动力学：芝诺只是记录条件化反复施加的结果，与标准的频繁投影读出一致。（频繁测量在某些区段反而加速衰减的反芝诺[26]，则读作落定与环境耦合在那一区段的合力，同样不需新机制。）

维格纳的朋友[27]。 朋友在封闭实验室里测了一个系统，室外的维格纳却把"实验室＋系统"整体描述为一个叠加，直到他自己去读。谁的描述对？本框架的读法：朋友做了一桩落定——他提取了一份内容、被记录条件化，故朋友手里是一份确定的内容（落定那道分界上的事）。维格纳在室外，把那间实验室描述为一份槽上的多容菜单、一个非纯混合（容量层、越槽那道分界上的事），因为从他的位置看，实验室＋朋友＋系统是一本他尚未读取其记录的纠缠菜单。佯谬之所以化解，在一个层级差，而非一种关系主义的"各自为真"：朋友处的落定内容是客观的，不因维格纳尚未读取而退回闭合前的容量；维格纳在外部合法拥有的，只是对"他尚未条件化的那条记录通道"的无条件容量描述——这既不是把一桩已完成的测量重新送回相干的前闭合叠加，更不是否认朋友处已有内容。两个描述落在不同的层（容量对内容），并不矛盾。待维格纳去读朋友的记录（他做自己的、条件于实验室的那一桩落定），记录条件化保证他读出与朋友同一份内容；这一步是他自己的条件化，而非朋友那桩事实的生成。所以无矛盾、无须任何超距的客观坍缩：朋友实现了一份内容，维格纳的容量菜单把那份内容含作其中一项，待他条件于记录便与朋友一致。那桩被加强成不自洽的版本[28]（一群行动者把彼此尚未读取的落定当作确定结果来链接推理），其不自洽恰恰源于在行动者之间把容量描述与内容实现混作一谈——本框架的容量／内容之分，正是堵住这一混淆：你不能把另一行动者尚未被你读取的那一份落定，从你的位置当作一份确定的内容；在你条件于其记录之前，它对你是一份容量菜单。

Kochen–Specker[29]。 该定理证明，无法给所有可观测量一致地指派预存的、与语境无关的确定值——量子力学是语境性的：测得的值依赖于测量的语境（与之同测的是哪些可观测量）。本框架对此非但不窘，反而正是它所预期的。其一，没有预存的确定值——生成而非选定：内容是在落定那一刻被生成的，不是从一份预存的赋值表里读出的。其二，被生成的那份内容依赖于语境——即依赖于 einselection 在那一桩测量里所定的是哪一组 pointer 基、哪一份菜单。所以本框架天然语境性：无预存内容（生成非选定），菜单的基由测量语境所定（einselection）。Kochen–Specker 因此不是本框架的佯谬，是它的本色：值不预存、内容随语境生成。隐变量那一路想保住的预存非语境赋值，本框架本就不要。

五、连续与离散

末了，须接前两部分各自留下的那个尾——连续与离散如何相容，并把本框架在此的先验明白地摆上台面。

因果槽既是一道范畴门槛，越它何以能是 sharp 的，而粗粒化下衬底的聚合看着却是连续的？在有效的、粗粒化的那一层上，这本不是难题：sharp 的结构常规地从看似连续的微观演化里涌现，这是一族有严格数学、有实验、乃至有诺奖的成熟物理早已趟过的路，越槽这一步正与它们同一机杼。其一，拓扑相变——Berezinskii–Kosterlitz–Thouless 转变[30][31]里，温度这个参数连续地变，涡旋的束缚与解束缚却在一个临界点 sharp 地易帜；BKT 转变是拓扑相变的典型范例，Kosterlitz 与 Thouless 因二维系统中这类相变的工作获 2016 年诺贝尔物理学奖。其二，容错阈值定理 (fault-tolerance threshold theorem)——物理错误率连续，逻辑的可纠错性却在一个阈值上 knife-edge 地翻转，而这道阈值本身在数学上即是一桩相变（映到随机键 Ising 模型的退禁闭转变）；近年多个平台已实验演示了越过这道阈值的纠错。其三，客观性的涌现——量子达尔文主义与谱广播结构 (quantum Darwinism / spectrum broadcast structure)[32] 给出了经典客观态如何从量子世界涌现的严格刻画：系统的 pointer 态在环境碎片间冗余地广播，多个观者由此独立地读到同一份客观信息。这一支尤其切近：它与本系列论信息的两篇[9,10]所立的广播本体同出一辙，也正是本文第三节那份客观菜单、那条量子达尔文主义冗余义务的物理底。所以越槽——连续衬底谱上一桩 sharp 的拓扑闭合（临界处拓扑地闭合，再加误差抑制给出的操作 sharpness）——不是本框架孤悬的一句声明，而是站在这片成熟物理的同一族现象之上；连续与离散在这一有效层上并不打架：连续的是粗粒化下的聚合，sharp 的是临界尺度上的闭合。

可须把更深的一层与这有效层分清，并把本框架在此的先验硬硬地立出来。上面那个"连续的梯度"，是有效层的描述——粗粒化之下衬底的聚合看着连续，正如连续的场叠在离散的格点之上、流体力学叠在分子之上。在基本层上，本框架的先验是一句明确的断言：任何物理量，在足够小的尺度上，都是离散的（承本系列地基一篇[14]）。这不是一桩待证的猜想、更不是一处悬而未决的留白：一切由离散的 DD 生成，而连续统不是离散所能生出之物，故"连续"在本框架里只能是有效层的浮现，断不能是基本层的本原。概率云亦然——它在足够小的尺度上离散，是这条先验的一个特例，不是一个另待支持的额外断言。

须把两件事切清，免得把先验与经验混作一谈。离散性——任何物理量在基本层离散——是上述那条硬先验。那个离散的具体尺度（其量级、是否有一个最小节拍、其值几何）则是经验待测的：本框架不强指一个值，也不把任何已有的尺度（如热力学地板 R_min(T)，此点论题乙已申）认作概率云的离散标度——那是因果槽与级联稳定化的空间下限，借作概率云的像素是没有依据的。所以开放的是尺度之值，不是离散性本身；前者待测，后者是先验。

而本框架把这条先验如此硬地摆上台面，恰恰是为了让它可被证伪。一条写软了的、随时可挪的先验，谁也撞不倒它，它也永远修正不了自己——把先验写软看似谦虚，实则是逃避证伪；唯有立成一条明确的断言，才给了世界一个能撞上的靶。所以本框架在此明言：若这世界在基本层是连续的——若有朝一日表明，某个物理量在任意小的尺度上仍是真正的连续统，而非离散之浮现——那么本框架这条离散的先验即被推翻，本框架将据此修正其地基。本文欢迎、并期待这样的证伪。硬，不是为傲慢，是把自己交出去受检；欢迎证伪，正是一条硬先验的题中应有之义。须就此把这条先验的证伪形态说准，免被当成一个寻常的实验命题：它不是本篇兑现的一个可操作新预言（本篇不指认任何可直接判决它的实验），而是一条框架级的风险承诺；其可观测签名、以及何种结果足以排除某类 SAE 离散实现，留待后续专文。它之被推翻，是在"基本层在原则上要求一个不可离散化的连续统、且一切离散 DD 实现皆告失败"这个意义上——而非靠一次"测得某物理量在任意小尺度上仍连续"的实验，后者在操作上本不可达。

六、收束

三个部分合起来，量子力学这一由微观转向宏观的乐章可以收束了。

论题甲给了本体：退相干是一桩账本并入、相位离域，局域上留一份失相而仍多容的客观菜单；落定是一桩正交的、非幺正的提取，从菜单里收出单一内容；两者分居同一序列的两道分界——越槽去相位是幺正的、热力学的前一道，落定收内容是非幺正的、因果的后一道；可逆界画在因果槽上下、不画在大小，宏观是范畴义的而非尺寸义的；时间之箭的方向根在 4DD 取向，其不可逆分作热力学（弥散的放大）与因果（落定的单向化）两种。论题乙给了机制：弥散是一串部分账本并入的级联，菜单逐级稳定，具更新封装的嵌套结构；其结构时标只是一个齐次更新骨架，本框架不提出独立的退相干率，只给层级菜单稳定化的组织原则与离散闭合的本体承诺。论题丙给了归类：本框架不是 GRW/CSL 型动力学坍缩论、不是多世界、不是隐变量，与哥本哈根相协而把那道边界拆成正交两步；密度矩阵是容量层的粗表示、对内容那道分界是盲的；芝诺、维格纳的朋友、Kochen–Specker 各经这副装置摆回了位置；而万物在足够小尺度皆离散——含概率云——是本框架明确摆上台面的一条硬先验，本文欢迎并期待对它的证伪，其具体的离散尺度则经验待测。

这一乐章因此始于微观的相干、经由退相干那一桩并入与落定那一桩提取、终于宏观经典世界之客观立定。它向前接的，是这一系列此后处理宏观与连续极限的篇章——多体、热极限、经典力学之浮现，皆在那条已经铺好的因果槽之上。本框架在能探测到分野处出预言，在只复现标准量子力学处只重读；这一乐章属后者，它不与标准量子力学争一寸经验，只把那同一套经验，摆进一个分得清容量与内容、越槽与落定、可能与现实的本体里。

引用文献

本表以数字顺序制、按正文出现先后编号。方括号内为暂拟之英文描述题名，于英文母语重写时一并转为正式题名；DOI 概以作者 Zenodo 记录为准。出版年以各篇 Zenodo 记录为准（早期诸篇或为 2025 年）。

[1] Qin, H. (2026). [SAE QM VI — The Born Rule｜玻恩规则]. Zenodo. https://doi.org/10.5281/zenodo.20479658 [2] Qin, H. (2026). [SAE QM VII — Measurement｜测量]. Zenodo. https://doi.org/10.5281/zenodo.20533295 [3] Qin, H. (2026). [SAE QM II — Complex Amplitude / e^{iθ}｜复振幅]. Zenodo. https://doi.org/10.5281/zenodo.20277037 [4] Qin, H. (2026). The Causal Slot of Information (SAE Information Theory III｜信息的因果槽). Zenodo. https://doi.org/10.5281/zenodo.19797457 [5] Qin, H. (2026). [SAE QM V — Quantum Entanglement｜量子纠缠]. Zenodo. https://doi.org/10.5281/zenodo.20455398 [6] Qin, H. (2026). [SAE QM IV — Quantum Tunneling｜量子隧穿]. Zenodo. https://doi.org/10.5281/zenodo.20369138 [7] Qin, H. (2026). [SAE Cosmology V — Dual-4DD Time-Reversed Arrows; Λ₁+Λ₂=0｜双 4DD]. Zenodo. https://doi.org/10.5281/zenodo.19329771 [8] Qin, H. (2026). [SAE QM I — Pre-Closure Ontology｜前闭合本体]. Zenodo. https://doi.org/10.5281/zenodo.20252029 [9] Qin, H. (2026). [SAE Information Theory V — Broadcasting and Reception｜广播与接收本体]. Zenodo. https://doi.org/10.5281/zenodo.19968504 [10] Qin, H. (2026). [SAE Information Theory VI — Gravitational-Wave Dynamics as 4DD Broadcast｜引力波动力学＝4DD 广播]. Zenodo. https://doi.org/10.5281/zenodo.20066645 [11] Qin, H. (2026). [SAE Mass Convergence — E = Ic³｜质量收敛]. Zenodo. https://doi.org/10.5281/zenodo.19510868 [12] Qin, H. (2026). [SAE Thermodynamics VII — Renewal Encapsulation｜更新封装]. Zenodo. https://doi.org/10.5281/zenodo.19673079 [13] Qin, H. (2026). [SAE Thermodynamics VIII — Copying Fidelity / Retention｜复制保真]. Zenodo. https://doi.org/10.5281/zenodo.19688304 [14] Qin, H. (2026). [SAE Foundation v2｜基础]. Zenodo. https://doi.org/10.5281/zenodo.19361950 [15] Joos, E., & Zeh, H. D. (1985). The emergence of classical properties through interaction with the environment. Zeitschrift für Physik B 59, 223. [16] Zurek, W. H. (2003). Decoherence, einselection, and the quantum origins of the classical. Reviews of Modern Physics 75, 715. [17] Hahn, E. L. (1950). Spin echoes. Physical Review 80, 580. [18] Zurek, W. H. (2009). Quantum Darwinism. Nature Physics 5, 181. [19] Gerlach, W., & Stern, O. (1922). Der experimentelle Nachweis der Richtungsquantelung im Magnetfeld. Zeitschrift für Physik 9, 349. [20] Lindblad, G. (1976). On the generators of quantum dynamical semigroups. Communications in Mathematical Physics 48, 119. [21] Ghirardi, G. C., Rimini, A., & Weber, T. (1986). Unified dynamics for microscopic and macroscopic systems. Physical Review D 34, 470. [22] Ghirardi, G. C., Pearle, P., & Rimini, A. (1990). Markov processes in Hilbert space and continuous spontaneous localization of systems of identical particles. Physical Review A 42, 78. [23] Penrose, R. (1996). On gravity's role in quantum state reduction. General Relativity and Gravitation 28, 581. [24] Everett, H. (1957). "Relative state" formulation of quantum mechanics. Reviews of Modern Physics 29, 454. [25] Misra, B., & Sudarshan, E. C. G. (1977). The Zeno's paradox in quantum theory. Journal of Mathematical Physics 18, 756. [26] Kofman, A. G., & Kurizki, G. (2000). Acceleration of quantum decay processes by frequent observations. Nature 405, 546. [27] Wigner, E. P. (1961). Remarks on the mind–body question. In The Scientist Speculates (I. J. Good, ed.). Heinemann. [28] Frauchiger, D., & Renner, R. (2018). Quantum theory cannot consistently describe the use of itself. Nature Communications 9, 3711. [29] Kochen, S., & Specker, E. P. (1967). The problem of hidden variables in quantum mechanics. Journal of Mathematics and Mechanics 17, 59. [30] Berezinskii, V. L. (1971). Destruction of long-range order in one-dimensional and two-dimensional systems possessing a continuous symmetry group. Soviet Physics JETP 32, 493. [31] Kosterlitz, J. M., & Thouless, D. J. (1973). Ordering, metastability and phase transitions in two-dimensional systems. Journal of Physics C: Solid State Physics 6, 1181. [32] Korbicz, J. K., Horodecki, P., & Horodecki, R. (2014). Objectivity in a noisy photonic environment through quantum state information broadcasting. Physical Review Letters 112, 120402.

致谢

本文初稿经多轮 AI 评审（ChatGPT、Claude、Gemini、Grok）逐条压测；于命门（弥散与落定之分属同一序列的两道分界）、容量与内容之三段、余项之归属、密度矩阵之本体身份、连续与离散诸处的厘清，皆得其益。陈则思（Zesi Chen）持续的批评性反馈贯穿本系列。文中一切立论与可能之误，皆由作者独负其责。