Complexification as an Articulation of L₂: Covering Structure, the Regular Period Core, and the Irregular Boundary
L₂ 的复化阐述——覆叠结构、正则周期核心与不规则边界

1. Introduction and Positioning

1.1 Where this paper stands

Paper 3 gave a complete four-Main-Step articulation of the quantitative dimension $L_1$, and stated in its § 6.4 that how the various post-$L_1$ articulation candidates are to be placed is not presupposed there but left to subsequent layer papers. The present paper is one of those subsequent papers — the one that treats the complexification main trajectory.

What this paper does for $L_2$ is what Paper 3 did for $L_1$: distinguish substrate from articulation mode, articulate the mode's unfolding through four Main Steps and their inner sub-articulation, give the capture range, and register the remainder ledger.

One difference in format must be declared at the outset: the registration class of Main Step 4. By the three-way classification established in Paper 3 § 1.3 and § 6.1, Main Step 4 is a meta-level closure whose sub-articulation is a remainder ledger structure and does not follow the four-step fractal pattern. This paper follows that registration, and carries it in the layout itself: Main Step 4 spans § 6, § 7 and § 8, while Main Steps 1 through 3 occupy one section each (§ 2.1).

This paper is not a restatement of Paper 1 § 5. Paper 1 articulated the four steps of $L_2$ and the six components of $\omega_\rho^{(2)}$ at the architectural level; the present paper develops the specialist treatment, and applies four upstream treatments to Paper 1 (§ 11.1), one of which is explicitly labelled a charitable reconstruction rather than a recovery of the original intent.

1.2 The four inherited debts

First, the composition law of Paper 1 § 11.1. That paper deferred the composition law for the six components of $\omega_\rho^{(2)}$ to a subsequent $L_2$ specialist paper, observing that without it the six-tuple is only a list. This paper makes a partial payment — a minimal typed composition skeleton (§ 9) — and registers the unified categorical encapsulation as an outstanding debt (§ 13.1).

Second, the Cayley–Dickson question of Paper 1 § 11.8. That paper asked whether the alternative closures provide another realization of $L_2$, and how the sufficiency criterion applies to them. This paper answers (§ 10), and in doing so revises the classification in Paper 1 § 5.1 under which the members of that sequence were listed as $L_2$ articulation candidates.

Third, the remainder ledger of Paper 3 § 6. That paper established three classes of remainder and the criterion distinguishing cross-level forcing from lateral-branch forcing. This paper rearranges the ledger at $L_2$ (§ 8.2), and articulates why the first class stands differently in the two layers: $i$ was forced out by an interrogation internal to $L_1$, whereas $\mathbb{H}$ has no corresponding source.

Fourth, the monolithic $\infty$ of the ZFCρ line. The question raised there is that the monolithic $\infty$ of $L_1$ cannot resolve fine ratios. This paper settles the order half (§ 6.2); the interface for the coefficient half is located at the second stratum of the Newton polygon but is not connected (§ 7.4). The debt is re-registered from "no interface" to "interface located but not connected", and may not be cited as settled.

A further open question, transferred from Paper 3 § 6.4 — the placement of the effective-Cauchy articulation mode — is decided here as a side-branch parallel to the complexification trajectory (§ 11.3).

1.3 What this paper does not claim

Seven items are explicitly disclaimed. Several are expanded in the reader's misreading table (§ 2.6) and registered in the three-tier table (§ 12.1) as not to be drawn as a conclusion of this paper.

First, this paper does not derive complex analysis. The stratification into divisors, residues, periods and monodromy, and Stokes data with asymptotic growth is an existing result of classical complex analysis. What this paper claims is the position of that stratification within a cross-layer comparison, not the stratification itself.

Second, this paper does not claim that $\mathcal{P}$ exhausts the $L_2$ substrate. $\mathcal{P}$ is countable while the substrate is not; and $\mathcal{P}$ does not even exhaust the computable reals. Further, the analytic–covering spine on curves along which this paper works is not claimed to generate all of $\mathcal{P}$ (§ 2.3, the two-spine declaration).

Third, this paper neither proves nor refutes $e \notin \mathcal{P}$. The proposition is an open conjecture. The paper's load does not rest on it (§ 7.1); the list of sentences requiring change should it be refuted is given in § 12.3.

Fourth, this paper does not deliver the full composition law of $\omega_\rho^{(2)}$. See the first debt above.

Fifth, this paper does not retract Paper 1's architectural articulation of $L_2$. All four upstream treatments are registered as refinement, charitable reconstruction, or revision of a candidate classification, not as retraction (§ 11.1).

Sixth, this paper does not expel the quaternions from $L_2$. § 10 concludes only that $\mathbb{H}$ does not share the signature and defect system of this trajectory; on its position within the full space of $L_2$ trajectories the paper makes no judgment.

Seventh, this paper does not presuppose the trajectory of $L_3$. § 13.2 offers a T3 programmatic prediction concerning the type of pressure release, together with an operationalized falsification condition and a registered counter-candidate; the prediction is not about the content of $L_3$.

1.4 Organization

§ 2 establishes four firewalls: substrate versus articulation mode (including the registration declaration for Main Step 4); numerical objects versus functions versus operators; expression syntax versus evaluation image versus expression equivalence; and numerical propositions versus proof-theoretic status versus algorithmic decidability. A terminology table and a reader's misreading table follow. Every substantive claim in the remainder of the paper traces back to one of these four.

§ 3 through § 5 articulate Main Steps 1 through 3. Of these, § 5 (the multiplicative path) is the mathematical spine: it gives the asymmetry of reverse interrogation (§ 5.4) and the double identity of $2\pi i$ (§ 5.5), and the entire technical content of § 6 through § 8 has its reason for being articulated there.

§ 6 through § 8 are the three sections of Main Step 4: the regular period core, the irregular boundary, and closure with the remainder ledger. Their relation is core and enrichment, not mutually exclusive binning — irregular connections still have divisors, residues and monodromy; what § 7 adds are exponential factors, directions, and Stokes data.

§ 9 delivers the minimal typed composition of the operational defect algebra. § 10 treats Cayley–Dickson. § 11 registers the relations to Papers 1, 2, 3, to the ZFCρ line, and to SAE Physics Foundation v2. § 12 gives the three-tier claim table, the jurisdiction statement, a two-way falsifiability registration, the anti-immunization declaration, and the prediction discipline. § 13 lists open problems and the subsequent trajectory.

---

2. Four Firewalls

This section is the paper's general firewall. Paper 3 § 2 established the distinction between substrate and articulation mode as that paper's general firewall; the present paper lifts that distinction to $L_2$ and adds three more. The four are not rhetorical disclaimers: every substantive claim in § 3 through § 9 traces back to one of them, and if any one of them is read through, the corresponding claims fail with it.

The four distinguish, respectively: substrate from articulation mode (§ 2.1); numerical objects from functions from operators (§ 2.2); expression syntax from evaluation image from expression equivalence (§ 2.3); and numerical propositions from proof-theoretic status from algorithmic decidability (§ 2.4). § 2.5 gives the terminology table and § 2.6 the reader's misreading table.

2.1 Substrate and articulation mode

Paper 3 § 2 articulates: the quantitative-dimension substrate and the $L_1$ articulation mode are two distinct objects and may not be conflated. The substrate is the "totality of quantities" articulated along the quantitative dimension, realized at the level of classical mathematical construction, via ZFC together with Dedekind cuts, as the real line $\mathbb{R}$ taken as a completed object. The $L_1$ articulation mode is the set of finite algebraic tools under the SAE operational-articulation view, capturing the real algebraic numbers $\mathbb{Q}^{\rm alg}_{\mathbb{R}}$. That paper's firewall statement is: the classical construction is not retracted; what is articulated is that, from the SAE operational-articulation view, no single articulation mode exhausts the real-line substrate.

That statement is universal, not enumerative. It is earned in Paper 3 § 6.3.2 by a cardinality argument: for any countable formal language, the totality of formulas in that language is at most $\aleph_0$, while the real line has cardinality $2^{\aleph_0}$, so an undefinable remainder exists for every such mode. What matters is that the argument does not depend on the choice of language — change the language, extend the language, and so long as it remains countable the conclusion stands. This paper inherits the universal form and does not re-prove it; § 8.2 exhibits that form once more at $L_2$.

This paper lifts the same distinction to $L_2$.

The $L_2$ substrate is the complex plane and the Riemann sphere. At the level of classical construction, $\mathbb{C}$ is realized as a completed object via $\mathbb{R}^2$ with a multiplication, or via $\mathbb{R}[x]/(x^2+1)$; $\hat{\mathbb{C}}$ via one-point compactification. These constructions are not retracted here.

The $L_2$ articulation mode is algebraic de Rham–Betti integration together with analytic continuation. The core of its capture range is the ring of ordinary periods $\mathcal{P}$ (defined in § 2.3, articulated substantively in § 6). The paper does not claim that $\mathcal{P}$ exhausts the $L_2$ substrate (§ 2.6, second item).

That the capture criterion differs from layer to layer is itself one of this paper's identifications. Paper 3's $L_1$ mode uses finite algebraic constraint as its capture criterion: a real number is captured just in case it satisfies a finite polynomial constraint over $\mathbb{Q}$. This paper's $L_2$ mode uses algebraic integral representation: a complex number is captured just in case it can be written as the value of an absolutely convergent integral of algebraic data over a $\mathbb{Q}$-semi-algebraic region. These are not two instances of one kind of criterion. The first adjudicates a relation between a number and a family of equations; the second, between a number and a family of geometric integral expressions. § 6.1 articulates this type shift substantively under the name carrier migration; here it is only registered.

Registration declaration for Main Step 4. Paper 3 § 1.3 and § 6.1 divide within-layer sub-articulation into three classes and state explicitly: cross-level threshold sub-articulation (Main Step 1); within-layer fractal sub-articulation (Main Steps 2 and 3); and Main Step 4, of which that paper writes that it is a meta-level closure whose sub-articulation is a remainder ledger structure and does not follow the four-step fractal pattern. Each of the three classes carries its own registration standard.

This paper follows the same registration. Main Step 1 is registered as cross-level threshold (§ 3); Main Steps 2 and 3 as within-layer fractal (§ 4, § 5); Main Step 4 as remainder ledger structure (§ 6, § 7, § 8), organized as capture core → unaptured ledger → trigger condition for the next layer, not as a re-enactment of the four phases.

The four stages onto which the six components of $\omega_\rho^{(2)}$ map — pointwise, local-monetization, global-cycle, boundary-sector — form a scale progression within a type stratification, not a re-enactment of the four phases inside Main Step 4. This paper carries that registration through the asymmetry of its layout: Main Step 4 spans three sections while Main Steps 1 to 3 occupy one each. Form is the declaration.

An inherited open question. Paper 3 § 6.4 registers the effective-Cauchy articulation mode as a post-$L_1$ analytic-completion branch, and explicitly does not presuppose whether it functions as a sub-articulation within the $L_2$ main trajectory, as an $L_{1\to 2}$ bridge, or as a side-branch parallel to the complexification trajectory; that adjudication is left to subsequent layer papers. This paper is such a paper.

The substantive material it can supply is a comparison of capture ranges. The effective-Cauchy mode uses algorithmic approximability as its criterion and captures the computable reals; the $L_2$ mode uses algebraic integral representation, and its capture core on the real axis is $\mathcal{P} \cap \mathbb{R}$. The relation is that $\mathcal{P} \cap \mathbb{R}$ is strictly contained in the computable reals — every period can be approximated to arbitrary precision, while Yoshinaga has exhibited a computable real that is not a period (§ 8.2). The effective-Cauchy mode therefore captures strictly more on the real axis than the $L_2$ period core, and the three capture criteria (finite algebraic constraint, algebraic integral representation, algorithmic approximability) are none of them a special case of another.

§ 11.3 will register it as a side-branch parallel to this paper's complexification trajectory; the present section only announces the comparative material and does not open the argument early.

2.2 Numerical objects, functions, operators

The articulated content of $L_2$ contains three kinds of object at once, and in ordinary mathematical usage the three often share a name. This firewall separates them.

Numerical objects: $\pi$, $e$, $2\pi i$, a given period, the value of a residue. These are outputs of evaluation maps.

Functions: $\exp$, $\log$, a given meromorphic function, a branch of an analytic continuation. These are what the $L_2$ articulation mode acts upon.

Operators: $\mathrm{div}(\cdot)$, $\mathrm{Res}_p(\cdot)$, analytic continuation along a path, $\oint_\gamma(\cdot)$. These send functions to numerical objects, or to other functions.

This is not a pedantic distinction. One of the paper's load-bearing decisions rests directly on it: the boundary remainder of the $L_2$ regular period core is the exponential map $\exp$, a function, and not the constant $e$, a numerical object (§ 7.5). The number $e = \exp(1)$ is merely the value of $\exp$ at 1. Placing the load on the numerical object would make the paper's structure depend on an open conjecture about that object ($e \notin \mathcal{P}$); moving it from the constant onto maps and expression classes means the structure depends instead on a definitional syntactic extension and a theorem-level classification of connections, and not on any open conjecture about the number $e$. The two placements differ in epistemic tier, and the difference becomes visible only once numerical objects and functions have been separated.

The same distinction dissolves a legacy confusion in reading. A residue is not a primitive object of the numerical category: $\mathrm{Res}_p$ is an operator acting on the local data of a meromorphic function at $p$ and returning a number. When § 9 articulates the internal operations of each component, the type of each component (group, bilinear pairing, representation, groupoid) is specified precisely along this distinction.

2.3 Expression syntax, evaluation image, expression equivalence

This firewall separates three objects and carries the whole of the paper's definitional commitment concerning periods.

Expression syntax. Write $\mathsf{Int}_{\rm alg}$ for the class of ordinary period expressions: integrands are algebraic differential forms, domains of integration are $\mathbb{Q}$-semi-algebraic regions. Write $\mathsf{Int}_{\exp}$ for the class of exponential integral expressions, which further admit factors of the form $e^{-f}$ with $f$ algebraic. Both are syntactic objects — sets of integral expressions one may write down, independent of what those expressions evaluate to.

Evaluation image. The evaluation map $\mathrm{ev}: \mathsf{Int}_{\rm alg} \to \mathbb{C}$ sends an expression to its value. Its image

$$\mathcal{P} = \mathrm{im}(\mathrm{ev})$$

is the Kontsevich–Zagier ring of ordinary periods. Precisely: a complex number is a period if its real and imaginary parts are both values of absolutely convergent integrals of $\mathbb{Q}$-rational functions over $\mathbb{Q}$-semi-algebraic regions. Correspondingly $\mathcal{P}^e = \mathrm{im}(\mathrm{ev}_{\exp})$ is the set of exponential periods.

The four objects form a commuting square, presented here as a table.

Table 1 · Expression classes and evaluation images

Syntactic side (expression class)	Evaluation map	Value side (numerical set)
$\mathsf{Int}_{\rm alg}$ algebraic integrand data, $\mathbb{Q}$-semi-algebraic domain	ev ⟶	$\mathcal{P}$ ordinary period ring
↓ syntactic extension: admit $e^{-f}$ factors		↓ inclusion
$\mathsf{Int}_{\exp}$ additionally admits factors $e^{-f}$, $f$ algebraic	ev$_{\exp}$ ⟶	$\mathcal{P}^e$ exponential periods

The commutativity of the square says: an ordinary period expression, evaluated and then included among the exponential periods, agrees with the same expression regarded as an exponential integral expression with $f = 0$ and then evaluated. In short, the syntactic extension is compatible with evaluation. The statement looks trivial, but it is precisely what allows the paper to place its load on the left column (the syntactic extension) rather than the right (a strict inclusion between numerical sets): the left extension is constructive and decidable, whereas whether $\mathcal{P} \subsetneq \mathcal{P}^e$ strictly is an open problem (§ 2.6, third item).

Every claim in the paper about capture range can be located at one corner or one arrow of this table. For ease of reference: upper left and lower left are syntactic objects, upper right and lower right numerical; the two horizontal arrows are evaluation maps; the left vertical arrow is the syntactic extension (where the load rests, § 7.1); the right vertical arrow is a numerical inclusion whose strictness is undecided (§ 8.2).

Expression equivalence. That two expressions take the same value is a relation lying in $\ker(\mathrm{ev})$. The three Kontsevich–Zagier rules — additivity, change of variables, and Stokes' formula — supply a family of relations that certainly lie in $\ker(\mathrm{ev})$; and whether those relations generate all expression equivalence is precisely the content of the Kontsevich–Zagier conjecture, which is unproved.

This distinction must be held strictly. The three rules are not three forward numerical operations generating all periods from nothing, and they are not the definition of $\mathcal{P}$. That definition is as given above, resting on the algebraicity of the integrand data and the domain, and containing no inverse-function operator. Wherever this paper states a definitional boundary, it states it on that definition, and therefore does not depend on the Kontsevich–Zagier conjecture at all. To attach a definitional boundary to the three rules would be to attach a definitional statement to an open conjecture; the paper does not do this. (The corresponding false-friend registration appears in § 2.6.)

Two-spine declaration. This paper has two intersecting but non-identical spines.

The first is the analytic–covering spine on curves: $\mathbb{C}$, $\mathbb{P}^1$, $\mathbb{C}^\times$, universal coverings, monodromy, Stokes data. It is one-complex-variable, and is the actual working ground of § 5 and § 7.

The second is the de Rham–Betti evaluation spine in arbitrary-dimensional algebraic geometry: the definition of $\mathcal{P}$ admits algebraic varieties of any finite dimension and semi-algebraic domains of any dimension.

This paper does not claim that the first spine generates all of $\mathcal{P}$. Here $\mathcal{P}$ serves as a general evaluation core, not as a totality of numbers derived from the Riemann sphere alone. The two intersect in the periods actually used here (small-loop periods, elliptic periods, and so on), but intersection is not identity. Without this declaration, the capture-range statements of § 6 would read as claiming that one-variable contour integration exhausts the periods — a claim this paper neither makes nor could make.

2.4 Numerical propositions, proof-theoretic status, algorithmic decidability

This firewall governs the question: at which level is a given open problem open?

The remainder ledger of $L_2$ contains several undecided propositions, of which the most conspicuous is whether $e$ belongs to $\mathcal{P}$. Without distinguishing the level at which such a proposition is open, it is easily misread as evidence for or against a layer boundary. This paper registers each item along three orthogonal coordinates.

Coordinate	Possible states
Numerical membership status	proved to belong / proved not to belong / unknown
Proof-theoretic status	proved / conjectural / relatively independent / unknown
Algorithmic status	decidable / semi-decidable / decidable on a restricted class / open in general

The three are orthogonal: the proof-theoretic status of a proposition does not determine its numerical membership status, and numerical membership status does not determine algorithmic status.

A further column registers the role a given item plays in this paper.

SAE role	Meaning
Load-bearing	independent of every open conjecture; the paper's structural claims rest on it
Witness	consistent with the framework, but not counted as support for it; see the anti-immunization declaration, § 12.4
Interface	lifts a question at the level of values into one at the level of expressions, rules, or algorithms

Two different boundaries must be kept apart.

The regular/irregular divide is a boundary of the $L_2$ capture range, lying within the layer. It is supported by formal classification theorems (§ 7.1) and is the paper's theorem-level load.

The reframing from the level of values to the level of rules is the $L_2$/$L_3$ layer boundary. Paper 1 § 5.5 already established the trigger criterion: what triggers $L_3$ is not the difficulty of $L_2$ itself, but the reframing of the problem into one about rules and decidability. § 8.4 gives the substantive articulation.

Conflating these two produces errors in both directions. One is to read the Stokes phenomenon as the $L_2/L_3$ boundary — the Stokes phenomenon is a structural fact internal to the $L_2$ mode, not a layer boundary. The other is to read the proof-theoretic status of some numerical proposition as determining the layer to which an object belongs: independence is a property of propositions, not of objects. Even were "$e \notin \mathcal{P}$" proved relatively independent of some formal system, what enters the rule layer is the provability of that proposition in that system; the number $e$ itself does not thereby change layer (§ 12.3, second counterfactual).

This firewall is placed in § 2 rather than in the closure sections because it is the common premise of every capture-range statement in § 6 through § 8, and not merely a closing device for § 8.

2.5 Terminology

The following distinctions are enforced throughout. The first four are commonly conflated under the single word "period", and one of the paper's load-bearing statements falls exactly on their difference (§ 7.5).

Object	Term used here	Example
A period in the sense of $\mathcal{P}$ (a number)	period	$\pi$, $\zeta(3)$, $\log 2$
A function invariant under some translation	periodic function	$\exp$ (period $2\pi i$)
A function taking period values at algebraic points	period-valued function	$\log$
An element of $\mathcal{P}^e$	exponential period	$e$, $\gamma$

The unqualified phrase "periodic function" is not used anywhere in this paper. Note in particular that $\log$ is not a periodic function while $\exp$ is; and that $\log$ takes period values at algebraic points. Each role appears once, in § 5.5 and § 7.5 respectively, and confusing them renders both arguments unreadable.

As to notation, the following three are three distinct objects and do not share a symbol.

Object	Notation	Source
Growth at infinity (order and type of an entire function, with Phragmén–Lindelöf)	$G_\infty$	Paper 1 Appendix B.5; symbol retained unchanged
The slope filtration	$\mathrm{Slope}_\infty$	This paper, § 7.2
The local differential Galois group	$G_{\mathrm{diff,loc}}$	This paper, § 7.3, § 9.2

What Paper 1's $G_\infty$ carries is a scalar asymptotic invariant, not a group object; the $G$ in the notation should not be read as denoting a group. This paper neither changes the symbol nor replaces the object, but registers in § 7.3 the candidate comparison interface that Newton–Turrittin data provide for it (making explicit that identity has not been proved).

Six further symbols are kept apart to avoid changing sense within a single subsection:

Use	Notation
Slope of a Newton polygon	$\lambda$ (§ 7.2)
Horizontal length of a Newton edge	$m$ (§ 7.2)
Ramification index	$\nu$ (§ 7.1, § 7.2)
Order of an entire function	$\rho$ (Paper 1 Appendix B.5; homographic with the subscript in the series symbol $\omega_\rho^{(2)}$ but distinct in sense — the latter is an established series notation and is not changed)
Modulus	$r$ (§ 5)
A point of the base space	$p$ (§ 6, § 9)

2.6 Reader's misreading table

The paper's claims come in five kinds of differing strength. Each row of the claim table (§ 12.1) notes which kind it belongs to.

Kind	Meaning
Standard theorem	an existing theorem of mathematics; cited, not proved
Standard definition	an existing definition; adopted, not altered
SAE registration	an identificatory adjudication made within the SAE framework; not a mathematical theorem
Structural analogy	a structural correspondence; identity not claimed
Open conjecture	undecided in mathematics; the paper places no load on it

Six misreadings are anticipated and addressed in turn.

First, this paper does not derive complex analysis. The four stages onto which the six components of $\omega_\rho^{(2)}$ map — pointwise, local-monetization, global-cycle, boundary-sector — constitute an existing stratification in classical complex analysis: divisors, residues, periods and monodromy, Stokes data and asymptotic growth, each with technical content established by mathematical work of the nineteenth and twentieth centuries. What this paper claims is the position of that stratification within a cross-layer comparison (namely the asymmetry of reverse interrogation articulated in § 5.4), not the stratification itself. Tabular presentation tends to suggest that the four stages generate the stratification; that suggestion is explicitly denied here. 〔Kind: attribution declaration〕

Second, this paper does not claim that $\mathcal{P}$ exhausts the $L_2$ substrate. $\mathcal{P}$ is countable, the substrate is not; and $\mathcal{P}$ does not exhaust even the computable reals (§ 8.2). That $\mathcal{P}$ is the evaluation core of the regular algebraic-integral articulation of $L_2$, and that it may be taken as the principal capture range of $L_2$, is an SAE registration. 〔Kind: SAE registration〕

Third, this paper does not prove $e \notin \mathcal{P}$, nor does it place load on it. The proposition is an open conjecture. The paper's load rests on the two extensions of § 7.1, not on it. Should it be refuted, the list of sentences requiring change is given in § 12.3. 〔Kind: open conjecture〕

Fourth, this paper does not retract Paper 1's architectural articulation of $L_2$. Four upstream treatments are applied (the modality of boundary closure, the Cayley–Dickson candidate classification, the placement of Euler's formula, and the comparison interface for $G_\infty$), all registered in § 11.1 as refinement, charitable reconstruction, or revision of a candidate classification, and none as retraction. Of these, the $G_\infty$ item is explicitly registered as a charitable reconstruction and not Paper 1's original intent. 〔Kind: SAE registration〕

Fifth, this paper does not expel the quaternions from $L_2$. § 10 adjudicates that $\mathbb{H}$ is not an equivalent parallel realization of this paper's commutative complex-period trajectory; the paper makes no judgment on this basis about its ultimate position within the full space of $L_2$ trajectories. 〔Kind: SAE registration〕

Sixth, this paper does not claim that the ZFCρ monolithic-$\infty$ debt is settled. § 6.2 articulates how valuation equips that $\infty$ with an integer scale, settling the order half; the coefficient half — the fine ratio — remains outstanding, its interface located at the Newton polygon in § 7.4, and an interface located is not an interface connected. 〔Kind: structural analogy plus outstanding-debt registration〕

False-friend registrations. The following two pairs share a name and nothing else, and no bridge may be built across either of them anywhere in this paper.

First, the Stokes' formula appearing as the third Kontsevich–Zagier rule is Stokes' theorem (the relation between the integral of a differential form and that of its exterior derivative); the Stokes data of the fourth stage of $\omega_\rho^{(2)}$ is the Stokes phenomenon for divergent series at sector boundaries. The two carry the same person's name and are otherwise unrelated.

Second, the three Kontsevich–Zagier rules concern relations among period values; the composition law sought in Paper 1 § 11.1 concerns relations among the six components of $\omega_\rho^{(2)}$. The two are related through the Riemann–Hilbert correspondence but are not the same thing (§ 9.3, § 13.1).

---

3. Main Step 1: Mark-Without-Construct of $i$

3.1 Main Step 1 and Inevitability 1

Paper 3 § 1.2 establishes that Inevitability 1 corresponds to Main Step 1, and that its form is mark-without-construct: a mark is fixed as a public token and its operational standing is settled, while its construction as an object is not given. At $L_1$, what is marked without being constructed is the unit "1".

At $L_2$, what is marked without being constructed is $i$.

Registration class: cross-level threshold (§ 2.1). What this Main Step articulates is not an unfolding internal to $L_2$ but the $L_1 \to L_2$ transition itself. Its inner step 1 takes its input from the layer below, not from anywhere inside this Main Step — and that is precisely the registration difference between the cross-level threshold class and the within-layer fractal class.

3.2 Inner steps

Inner step 1: $L_1$'s closure remainder resists one-dimensional inclusion

Paper 3 articulates the second-level reverse interrogation of $L_1$ Main Step 3 inner step 4 as closing on the real closed field $\mathbb{Q}^{\rm alg}_{\mathbb{R}}$. But

$$x^2 + 1 = 0$$

has no solution in a real closed field.

The structure here deserves attention. The equation is posed by $L_1$'s own interrogation — finding roots of polynomials is exactly the content of $L_1$ Main Step 3's reverse interrogation — while $L_1$'s substrate, the real line, does not contain the answer. Hence $i$ is an substrate-external remainder in the sense of Paper 3 § 6: not "the $L_1$ mode failed to capture something lying inside the substrate", but "what is being asked for is not in the substrate at all".

This distinction determines the type of forcing. An substrate-external remainder is pressed out by the layer's own internal interrogation, and it therefore forces a crossing of levels rather than forcing a lateral branch. On this basis Paper 3 registers $i$ as cross-level forcing.

The same distinction has a second consequence in the present paper. The substrate-external remainder of $L_2$ — the $\mathbb{H}$ direction — has no corresponding internal interrogation pressing it out: $\mathbb{C}$ is algebraically closed, and root-finding no longer presses. The first class therefore stands differently in the two layers (§ 8.2), and it is on this basis that § 10 adjudicates Cayley–Dickson a cost branch.

Inner step 2: cross-subject propagation

In solving the cubic, Cardano met square roots of negative numbers, called them sophistic quantities, and refused to grant them reality — while continuing to use them in computation. Bombelli then supplied rules of operation for these quantities, so that they could be carried consistently through a computation.

What matters is that propagation precedes acceptance. Before the token was taken to be an object, it was already circulating between subjects as a sign transmissible within computation. An intermediate step containing a square root of a negative number, written down by one user, could be picked up by another, carried forward, and made to yield the correct real solution.

This is of the same type as the cross-subject propagation of Paper 3 Main Step 1 inner step 2: the publicity of a mark is secured not by its ontological standing but by its transmissibility in computation. A sign becomes public because it passes between subjects without failing, not because the subjects have reached agreement about what it is.

Inner step 3: public binding of the symbol

Euler introduces the symbol $i$.

The effect of the binding is that the quantity ceases to be a descriptive phrase — "a square root of some negative number" — and becomes a single public token. A descriptive phrase must have its reference unfolded afresh at every use; a single token need not.

The two faces of the binding (following the discipline of Paper 3 § 4.2.3, no value judgment is made):

• one face is the processing capacity articulated: $i$ can be manipulated in a formula as a single symbol, standing alongside real numbers within one and the same expression;
• the other face is the remainder articulated: the symbol conceals what it is. "$\sqrt{-1}$" at least still carries a description, unsatisfiable though that description is; "$i$" carries not even a description.

Inner step 4: construct and remainder

Construct: the public token $i$, together with the relation $i^2 = -1$.

This construct is a mark-without-construct: $i$ is fixed as a public token and its rules of operation are settled, while the space of habitation in which it exists as an object has not been constructed. At the close of this Main Step, $i$ has complete algebraic properties and no geometric position whatsoever.

Remainder: $i$ has no space of habitation.

This remainder is the input to Main Step 2 (§ 4.2).

3.3 Alignment with upstream

Paper 1 § 5.1 articulates $L_2$ Main Step 1 as the marking of $i$, and lists the terms of the Cayley–Dickson sequence as candidates for the articulation of $L_2$. § 10 of the present paper revises that candidate classification (§ 11.1, second upstream registration).

Paper 3 § 6 supplies the registration of substrate-external remainders and the criterion for cross-level forcing; inner step 1 above is the receiving end of that criterion on the $L_2$ side.

The cross-level ledger table of SAE Foundation v1 records $i$ as remainder one of $L_1 \to L_2$ without giving a reason. § 8.3 of the present paper supplies it: $i$ is the input to the transition — it is not produced by $L_2$; it is taken in by this Main Step from the layer below.

---

4. Main Step 2: The Additive Path — Two-Dimensionalization

4.1 Main Step 2 and Inevitability 2

Paper 3 § 1.2 establishes that Inevitability 2 corresponds to Main Step 2, the additive path, and that its form is the iteration of a mark together with the reverse interrogation of that iteration. At $L_1$, iteration of the unit mark gives $\mathbb{N}$, and the reverse interrogation gives $\mathbb{Z}$.

At $L_2$, the remainder left by Main Step 1 is: $i$ has no space of habitation. This Main Step constructs that space and closes it along the additive path.

Registration class: within-layer fractal (§ 2.1).

4.2 Inner steps

Inner step 1: a sub-mark for the orthogonal direction

Through Via Negativa, the subject refuses to let articulation halt at the state of a purely algebraic sign, and articulates a direction orthogonal to the real axis as a sub-mark.

It should be noted that this sub-mark is not derived from within the real line. Nothing in the structure of the real line singles out an orthogonal direction; the direction is posited in order to house $i$. This differs from the situation at $L_1$ Main Step 2 inner step 1: there, the direction of iteration is given by the unit mark itself and requires no separate positing.

Historically this step was carried out through the geometric representations of Wessel, Argand, and Gauss, some two centuries after Euler's binding of the symbol. The interval itself shows that the sub-mark is not an immediate corollary of the symbol.

Inner step 2: iteration of componentwise addition

$$(a+bi) + (c+di) = (a+c) + (b+d)i.$$

Addition proceeds componentwise, each component carrying over the additive path already articulated at $L_1$. Iteration yields the whole plane — the Argand plane.

What this iteration articulates is translation, not rotation. This is the source of the remainder of this Main Step (inner step 4).

Inner step 3: binding the ordered pair into a single token

Cost: if every complex number is written as the ordered pair "real part $a$, imaginary part $b$", then every operation must be carried out in pairs, articulation length doubles, and each operation must specify which component it acts on.

Binding: $z = a + bi$ as a single token. Under this token an operation can be applied to $z$ as a whole, without decomposition into components at every step.

The two faces:

• processing capacity: the complex number becomes a single object that can be operated on as a whole, so that the multiplicative path which follows (§ 5) can be articulated at the level of $z$ rather than remaining permanently at the level of components;
• remainder: the token conceals the fact that it is an ordered pair. When components are needed the token must be decomposed again, and decomposition is not an operation the token carries with it.

This occupies the same cell as the $N$-ary positional binding of $L_1$ Main Step 2 inner step 3: $N$-ary positional notation binds repeated unit marks into a compact sign; here an ordered pair is bound into a single sign. Both are cross-iteration bindings that manage cost.

Inner step 4: construct and remainder

Reverse interrogation: reversing addition gives subtraction; and $\mathbb{C}$ is closed under subtraction — componentwise it inherits the closure already achieved by $\mathbb{Z}$ at $L_1$, and no new demand for expansion arises.

The reverse interrogation of this Main Step is therefore single, not a nested cascade. This is one end of the sub-fractal depth asymmetry articulated in § 5.6: the additive path is shallow, the multiplicative path deep, and this holds at both layers.

Construct: $\mathbb{C}$ as a two-dimensional additive closure; the boundary mark $\hat\infty$ (see § 4.3).

Remainder: the additive path does not articulate rotation.

This remainder is the input to Main Step 3 (§ 5.2.1).

4.3 $\hat\infty$ is a chosen canonical closure

Paper 1 § 5.2 articulates the identification, in the passage to $L_2$, of $L_1$'s $+\infty$ and $-\infty$ into the single $\hat\infty$, and says that this identification is forced by topology.

The present paper refines this.

Topology does not uniquely force one-point compactification. A locally compact Hausdorff space admits several compactifications — the one-point, the Stone–Čech, the end compactification, and others — which are mutually non-homeomorphic, and topology by itself does not choose among them.

The one-point compactification is the canonical closure chosen within the meromorphic and conformal context. The reasons for choosing it are specific: it gives meromorphic functions a well-defined behaviour at $\infty$ (the $\mathrm{ord}_\infty$ of § 6.2 depends on this); it makes $z \mapsto 1/z$ an automorphism of the sphere (the boundary pair $0 \leftrightarrow \hat\infty$ of § 5.3 depends on this); and it makes $\mathbb{P}^1(\mathbb{C})$ a compact Riemann surface (the global residue theorem of § 6.3 depends on this).

The choice has a cost: every direction of approach to infinity is collapsed into a single point, and directional resolution is lost.

Within the present paper this cost has a definite date at which it falls due. Directional resolution is required at an irregular singularity (§ 7.4), and one-point compactification has already collapsed it; a real oriented blow-up must therefore be performed at that point, re-expanding what was collapsed to a point into a circle of directions. § 8.1 articulates the two-level closure structure that results.

The refinement the present paper makes to Paper 1 § 5.2 thus has two parts: it restates "forced by topology" as "canonical choice within a context", and it tracks the cost of that choice to the place where the cost falls due. This is refinement and not retraction — the identification Paper 1 articulates is itself unchanged; what changes is the modality of that identification (forced versus chosen) and the visibility of its cost.

〔Kind: upstream refinement. § 11.1, first item〕

---

5. Main Step 3: The Multiplicative Path — Covering and Rotational Memory

5.1 Main Step 3 and Inevitability 3

Paper 3 § 1.2 establishes that Inevitability 3 — comparison cannot remain isolated, it propagates; cannot but develop further — corresponds to Main Step 3, the multiplicative path. At $L_1$ the substantive content of that correspondence is this: once the additive path has been articulated, development turns to a higher-order articulation of repeated addition, articulated as the multiplicative path by way of the (count, unit) binding.

At $L_2$, Main Step 2 closes on $\mathbb{C}$, and its remainder is: the additive path does not articulate rotation.

An attribution declaration. The "rotational memory" of this Main Step's title belongs to the covering, not to multiplication. Multiplication in $\mathbb{C}^\times$ composes rotations, but it does not record winding number — winding number is a quantity available only in the covering space (§ 5.3). Inner steps 2 and 3 deliver the composition and the decomposition of rotation; only inner step 4 delivers memory. Componentwise addition $(a+bi)+(c+di) = (a+c)+(b+d)i$ iterates component by component; it articulates translation in the plane, not turning in the plane. Main Step 3 is the articulated response to that remainder.

Registration class: within-layer fractal (§ 2.1). As with $L_1$ Main Step 3 in Paper 3, the inner step 4 of this Main Step contains a nested cascade, so its sub-fractal is one degree deeper than that of Main Step 2; § 5.6 articulates the asymmetry.

5.2 Inner steps 1 through 3

5.2.1 Inner step 1: the rotation remainder forces a multiplicative sub-mark

The remainder left by Main Step 2 inner step 4 is: the additive path does not articulate rotation. The subject, as a finite existence, through Via Negativa, refuses to let articulation halt at a state possessing only translation, and articulates "$\times$" as a sub-mark.

It must be noted that multiplication on $\mathbb{C}$ is not a direct continuation of $L_1$ multiplication. The multiplicative path of $L_1$ acts on a one-dimensional line and permits dilation only; multiplication on $\mathbb{C}$ permits both dilation and turning. The articulation content borne by the new sub-mark is therefore strictly greater than that of its namesake at $L_1$.

5.2.2 Inner step 2: multiplicativity of the modulus, composition of rotations

Resource discipline: this inner step invokes purely algebraic means only. The real trigonometric functions, the exponential function, and any additive angular parameter are not invoked here — the former are objects $L_1$ has not captured (§ 8.2, classes two and three; § 11.3), and the latter is a product of inner step 4 (§ 5.3). Both of the deliverables of this inner step follow directly from the algebraic structure on $\mathbb{C}$.

First, multiplicativity of the modulus. From $|z|^2 = z\bar{z}$ and the multiplicativity of conjugation one obtains directly

$$|zw| = |z|\,|w|.$$

The modulus $|z|$ carries over the multiplicative path of $L_1$ Main Step 3 (magnitudes multiply) and brings nothing new.

Second, composition of rotations. Write $S^1 = \{w : |w| = 1\}$. By the identity above, $S^1$ is closed under multiplication and contains inverses, hence is a multiplicative group. And for $|w| = 1$ the map $z \mapsto wz$ is an $\mathbb{R}$-linear map preserving the modulus, whose determinant is $|w|^2 = 1$; it therefore lies in $SO(2)$.

Rotation is obtained from linear algebra and requires no trigonometry. This is where the substantively new content of $L_2$ lies: multiplication at $L_1$ permits dilation only, whereas multiplication on $\mathbb{C}$ splits dilation and composition of rotations in $SO(2)$ into two independent accounts.

Under this articulation, the constant $i$ is best understood not as "an imaginary number unreal in some fashion" but as an element of order four in $S^1$: $|i| = 1$, so $i \in S^1$, and multiplication by $i$ sends $1 \mapsto i \mapsto -1 \mapsto -i \mapsto 1$; the relation $i^2 = -1$ is the algebraic statement of that order-four property. The language of "a quarter turn" is not needed here — order four is a group-theoretic fact, whereas angle is a parameter available only at inner step 4. The unreality of $i$ on the real line is a downstream consequence of its function as a generator of rotation: $L_1$ cannot articulate rotation, so a generator of rotation cannot lie within the articulation scope of $L_1$.

This inner step leaves a cost remainder. If every multiplication is expanded in component form,

$$(a+bi)(c+di) = (ac-bd) + (ad+bc)i,$$

then a single multiplication requires four real multiplications and two real additions; the articulation length of $k$ iterations grows linearly in $k$, and at every step the two accounts, modulus and rotation, must be separated afresh. This is of the same type as the articulation cost of repeated-multiplication expansion articulated in Paper 3 § 5.2.2: it is not a logical remainder — component form is logically closed within $\mathbb{C}$ — but a remainder of articulation-cost character.

5.2.3 Inner step 3: the polar-decomposition binding

So that iterated multiplication may be articulated compactly, the subject, through Via Negativa, refuses this articulation cost and articulates the polar decomposition

$$z = r\,u, \qquad r = |z| \in \mathbb{R}_{>0},\quad u = z/|z| \in S^1$$

as a binding. The two accounts already separated — modulus and rotation — are composed by this form into a single token; under it, multiplication is $(r_1 r_2,\ u_1 u_2)$: two independent multiplications, with no expansion into components and no angular parameter required.

This is a cross-iteration binding, occupying the same cell as the power notation of $L_1$ Main Step 3 inner step 3 (cross-multiplications binding). Paper 3 § 5.2.3 articulates the power-notation binding as the concrete instantiation of Via Rho at the site of binding, the binding operation producing two inseparable faces; the polar binding does the same:

• one face is the processing capacity the binding articulates: powers are given in one stroke, $z^n = r^n u^n$; the accounting of modulus and the accounting of rotation are separated completely, each proceeding within its own group;
• the other face is the remainder the binding articulates: the rotation group $S^1$ has not yet been additively parametrized. $S^1$ is a multiplicative group — its elements can be multiplied with one another, but there is no additive measure of by how much they differ. The iteration $u^n$ can be performed, and yet the questions "how many times around?" and "how far apart are $u_1$ and $u_2$?" cannot be answered.

Both faces are substantive content produced by the operation of the binding; the present paper makes no value judgment between them (following the discipline of Paper 3 § 4.2.3).

This remainder is handled at inner step 4, and it is handled at one stroke. The second-level reverse interrogation of inner step 4 produces the covering map $\exp$, and $\exp$ delivers three things simultaneously: the additive parametrization of $S^1$ (the angle $\theta$ acquires its definition only here), the $2\pi$ indeterminacy of that parametrization, and the deck transformation group $2\pi i\mathbb{Z}$ (§ 5.3).

This junction is the substantive link of the sub-fractal internal to this Main Step: the remainder left by the binding (the missing additive parameter) and the structure produced by the inversion (the covering) are the two ends of one and the same thing. Additive angle has exactly one source in this paper.

An ordering registration. The binding at this inner step takes the polar decomposition $z = ru$, and not the trigonometric form $r(\cos\theta + i\sin\theta)$, nor the exponential form $re^{i\theta}$. This is not notational preference but resource discipline (§ 5.2.2):

• the complex exponential is a product of the second-level reverse interrogation at inner step 4 (§ 5.3), and is not yet available at this inner step;
• the real trigonometric functions are likewise unavailable. They are not products of the finite algebraic mode of $L_1$ (§ 2.1); and $\pi$, on the ledger of Paper 3 § 6, is an substrate-internal transcendental remainder that $L_1$ can name but cannot capture (§ 8.3), while the effective-Cauchy mode that responds to the completeness remainder is adjudicated a side-branch parallel to the complexification trajectory, and does not lie on this one (§ 11.3). To import onto the mathematical spine an object that the paper elsewhere adjudicates uncaptured would deprive the core statement of § 5.4 of its independence.

This discipline has one direct consequence: $2\pi$ has exactly one source in this paper, namely the kernel of the $\exp$ of inner step 4 (§ 5.3), and is not imported from the periodicity of $\cos$ and $\sin$. Both the double identity of $2\pi i$ in § 5.5 and the articulation of $\pi$ as a deck transformation group in § 8.3 depend on that single source.

5.3 Inner step 4: a nested cascade of two-level reverse interrogation

Main Step 3 reaches closure through Inevitability 4 (cannot but be interrogated). In the context of the multiplicative path, interrogation naturally takes the form of reverse interrogation; and as with $L_1$ Main Step 3 in Paper 3, the reverse interrogation here is not single but a nested cascade: the first level of reverse interrogation produces a new articulation space, that space still contains an unclosed remainder, and further reverse interrogation produces a second layer of structure.

First level: can multiplication be reversed?

It can: division is articulated. Division is not closed at $0$, so the articulation space is

$$\mathbb{C}^\times = \mathbb{C}\setminus\{0\}.$$

A phenomenon with no counterpart at $L_1$ appears here in the boundary structure: under the transformation $z \mapsto 1/z$, $0$ and $\hat\infty$ are exchanged with one another, forming a two-periodic boundary pair for multiplicative inversion on the Riemann sphere (Paper 1 § 5.2). The boundary mark of the multiplicative path of $L_1$ is $0$, and the boundary marks $\pm\infty$ of its additive path stand in no duality with it; the two-dimensionalization at $L_2$ pairs these two boundary marks.

The first level closes on $\mathbb{C}^\times$.

Second level: can powers be reversed?

They can, but the inversion splits into two branches whose remainders are of different types. This is the substantive content of the present section.

First, the finite algebraic inversion: inverting $z \mapsto z^n$.

The $n$-th root. Every nonzero complex number has $n$ $n$-th roots, so the inversion is $n$-valued. The branch points are $0$ and $\hat\infty$; going once around a branch point permutes the roots, and the monodromy group is finite (the cyclic group $\mathbb{Z}/n$).

The decisive point is that all roots remain within $\mathbb{C}$. By the fundamental theorem of algebra, every non-constant polynomial over $\mathbb{C}$ has a root. Hence the reverse interrogation of polynomial root-finding no longer forces an enlargement of the scalar substrate at $L_2$ — in contrast with the corresponding step at $L_1$ (§ 5.4).

But the inversion still requires structure. To single out one among the $n$ roots, and to keep that selection consistent under continuation, a covering or branch structure must be introduced. The inversion does not expand the substrate, and yet it demands a layer of structure over the substrate — this is the first sign that the pressure is changing direction.

Second, the transcendental inversion: inverting $\exp$.

The logarithm. The map $\exp: \mathbb{C} \to \mathbb{C}^\times$ is surjective but not injective:

$$\exp(z) = \exp(z + 2\pi i k), \quad k \in \mathbb{Z}.$$

Hence $\log z$ is infinitely many-valued: $\log z + 2\pi i k$. This is exactly how the remainder left by inner step 3 — the want of an additive parametrization of the rotation group — is handled: $\exp$ supplies the additive parameter of $S^1$, and the cost of that parameter is precisely its $2\pi$ indeterminacy. The additive angle and its indeterminacy are born at the same moment; it is not that the angle exists first and its multivaluedness is discovered afterwards.

The complete form of this structure is the exact sequence of groups

$$0 \longrightarrow 2\pi i\mathbb{Z} \longrightarrow (\mathbb{C}, +) \stackrel{\exp}{\longrightarrow} (\mathbb{C}^\times, \cdot) \longrightarrow 1.$$

This exact sequence articulates five things at once: the source of $2\pi i$ (it is the kernel); the multivaluedness of the logarithm ($\exp$ is not injective); the monodromy group as $\mathbb{Z}$ (the kernel is isomorphic to $\mathbb{Z}$); the $2\pi i k$ memory produced by circulation (the action of the kernel on a fibre); and covering in place of substrate enlargement ($\mathbb{C}$ is simply connected, so $\exp$ is the universal cover of $\mathbb{C}^\times$).

Construct and remainder

Construct $=$ the universal cover of $\mathbb{C}^\times$, with deck transformation group $2\pi i\mathbb{Z}$.

Remainder $=$ path dependence. The result of analytic continuation depends on the homotopy class of the path. This remainder cannot be dissolved further within this Main Step — reverse interrogation has been exhausted within $\mathbb{C}$. It is the object that the periods and monodromy of Main Step 4 are to articulate (§ 6.4).

5.4 Core structural statement: algebraic closure removes the scalar gap, not the representation gap

This section compares § 5.3 with Paper 3 § 5.2.4 at the same cell. Both are the second-level reverse interrogation of Main Step 3 inner step 4, but the types of their products differ.

$L_1$: the inversion of powers — extraction of roots and polynomial root-finding — is not closed within $\mathbb{Q}$. The pressure is released by enlarging the scalar substrate: $\mathbb{Q} \to \mathbb{Q}^{\rm alg}_{\mathbb{R}}$. Main Step 3 closes on the real closed field. The substrate has grown larger.

$L_2$: the inversion of powers, so far as polynomial root-finding is concerned, is already closed within $\mathbb{C}$ — the fundamental theorem of algebra. The pressure has nowhere outward to be released, because the scalar substrate is already full in the algebraic sense. The mode of release turns upward: a layer of covering is added.

Hence the structural asymmetry the present paper identifies:

> Algebraic closure removes the scalar gap but not the gaps of representation, path, and uniformization.

Two points require comment.

First, the asymmetry is not a quantitative difference in depth but a difference in the type of product. The reverse interrogation of $L_1$ produces a larger number field — a scalar object. The reverse interrogation of $L_2$ produces a covering space and a group — a topological object and a representation-theoretic object. To ask only which layer's reverse interrogation is the more complicated is to read the asymmetry as a matter of degree; the actual difference is that the pressure is released onto objects of a different kind.

Second, this asymmetry is the structural condition under which the operational defect algebra $\omega_\rho^{(2)}$ becomes necessary work. Had the pressure again been released by enlarging the substrate, Main Step 4 would need only to register a new number field, and no full operational defect algebra would be required. What is stated here is a necessary condition, not a causal explanation: the existence of the asymmetry makes the defect algebra a meaningful work content; it is not that "because there is an asymmetry, there must be a defect algebra". The defect algebra is the core work content of $L_2$ (Paper 1 § 5.4) precisely because the pressure has been lifted onto that family of objects — coverings, monodromy groups, periods, and Stokes data. In other words, the entire technical content of § 6 through § 8 has its reason for being articulated in the present section.

The articulation chain:

$$i \ \longrightarrow\ \mathbb{C} \ \longrightarrow\ \mathbb{C}^\times \ \longrightarrow\ \widetilde{\mathbb{C}^\times} \ \longrightarrow\ 2\pi i\mathbb{Z} \ \longrightarrow\ \mathrm{Mon}$$

from the complexifying seed of Main Step 1, through two-dimensionalization and the multiplicative path, to the universal cover, the deck transformation group, and then the monodromy of Main Step 4. The chain does not step outside $\mathbb{C}$ at any point.

Registration consequence. $\exp$ is here articulated as the covering map produced by Main Step 3 inner step 4, that is, as an internal product of $L_2$. Accordingly, within the present trajectory the paper adopts the internal-product reading: the exponential direction as a closure remainder of $L_2$ (§ 7.5), rather than as a remainder of the $L_1 \to L_2$ transition map. The transition reading is no longer the primary registration of this paper; the paper does not thereby judge the standing of the transition reading within other trajectories (following the anti-essentialization discipline of Paper 2).

5.5 The double identity of $2\pi i$

$2\pi i$ occurs at two places in this paper. The two are not an analogy; they are two presentations of one and the same statement.

Main Step 3: $2\pi i\mathbb{Z}$ is the kernel of $\exp: \mathbb{C} \to \mathbb{C}^\times$, that is, the deck transformation group of the universal cover.

Main Step 4, inner step 2: $2\pi i$ is the factor in the normalization of residues, $\oint_{|z|=1} dz/z = 2\pi i$ (§ 6.3).

Their identity: going once counterclockwise around the unit circle, $\log z$ increases by $2\pi i$. Read from the left, this is the action of a generator of the deck transformation group on a fibre; read from the right, it is the integral of $dz/z$ along that loop. They are two notations for one and the same thing.

This identity has two uses.

First, it locks § 5 and § 6 together by an identity rather than by mere adjacency of sections. Paper 1 § 4.5 identifies $2\pi i$ as the articulative unit of $L_2$'s exchange law — which the present paper calls the within-layer articulative unit of the exchange law, so as to keep it distinct from the transition signature slot (see § 11.5.3); the present section articulates that this unit is not merely a product of residue calculus but a product of the covering structure as a whole. The structural ground on which commutativity holds at $L_2$ lies in Main Step 3.

Second, it strengthens the criterion applied to Cayley–Dickson in § 10. If this within-layer unit were no more than a factor occurring in the residue theorem, then the claim that paying away commutativity is paying away the unit would be an observation about one formula; but since the unit is at the same time a generator of the deck transformation group, what the criterion bears on is the covering structure of the entire layer, and not a single integral formula.

The position of Euler's formula. Euler's formula

$$\cos\theta + i\sin\theta = e^{i\theta}$$

has its two sides falling on the two sides of substrate and mode, produced along routes of different origin.

The left-hand side is a substrate-level classical construction: the real trigonometric functions are pre-existing objects of analysis on $\mathbb{R}$, and are not the product of any cell on the trajectory of this paper (the resource discipline of § 5.2.3 explicitly declines to invoke them on the spine).

The right-hand side is the restriction to the imaginary axis of the covering map $\exp$ articulated by the second-level reverse interrogation at inner step 4: that is, the universal covering parametrization $\mathbb{R} \to S^1$, with deck transformation group $2\pi\mathbb{Z}$.

Euler's formula states that these two objects are equal. And $e^{i\pi} + 1 = 0$ is the value of that equality at $\theta = \pi$, together with the remainder $\pi$ of $L_1$ Main Step 4 and the two boundary tokens of $L_1$.

This is the structural reason why Euler's formula is a superposition placement and not a single-point placement: it does not reside at any cell of this trajectory; it is an identification between a substrate-level classical construction and a mode-level product. What that identification crosses is the general firewall of this paper between substrate and mode (§ 2.1). The present paper therefore registers it as a substrate–mode superposition placement, and not as a cross-cell identification within the layer. An inventory of constituents alone — that it contains $e, \pi, i, 1, 0$ all at once — yields no more than a description of position; what the present section supplies is a structural reason.

Placement and closure equation are two different predicates, and they do not compete. Placement answers "where does it reside"; closure equation answers "what does it close". Euler's formula is both at once: as regards placement it is a substrate–mode superposition placement (present section); as regards closure it is the closure equation of the $L_1 \to L_2$ transition, closing that transition's two remainders $i$ and $\pi$ (§ 8.3).

Note that "closure" here is not the closure of a layer. The closure of the layer $L_2$ is coarse closure together with directional refinement (§ 8.1), which is a different object from transition closure. Paper 1 § 5.3 articulates Euler's formula as best understood as a closure instantiation and not the full closure of $L_2$; its precise content is just this — it is an instance of transition closure, not of layer closure. The full articulation of placement is given in § 8.3.

Note also that it is not a section of the covering. Were there a global section $s: \mathbb{C}^\times \to \mathbb{C}$ with $\exp \circ s = \mathrm{id}$, that would be a globally single-valued logarithm, and precisely such a thing does not exist — which is the substantive content of the second-level reverse interrogation of § 5.3. Euler's formula is a canonical parametrization witness for the covering structure, not a section. Its complete articulation as a substrate–mode superposition placement is given in § 8.3.

5.6 Sub-fractal depth asymmetry

Paper 3 § 5.3 articulates a sub-fractal depth asymmetry at $L_1$: the inner step 4 of Main Step 3 contains a nested cascade (division $\to \mathbb{Q}$; extraction of roots and polynomial root-finding $\to \mathbb{Q}^{\rm alg}_{\mathbb{R}}$), whereas the inner step 4 of Main Step 2 has only a single reverse interrogation (subtraction $\to \mathbb{Z}$). The sub-fractal of the multiplicative path is therefore one degree deeper than that of the additive path. That paper articulates this depth asymmetry as no artefact of articulation but a substantive fact: within the complete precisification path, the greatest substantive content of algebraic articulation occurs at the multiplicative path.

$L_2$ reproduces the same asymmetry, and at the same cell. Main Step 2 inner step 4 closes in a single move on $\mathbb{C}$; Main Step 3 inner step 4 is a nested cascade (division $\to \mathbb{C}^\times$; extraction of roots and the logarithm $\to$ covering).

But what lies at the depth differs between the two layers. $L_1$ is deep in enlarging the scalar substrate; $L_2$ is deep in lifting into a covering (§ 5.4). The form of the depth asymmetry therefore recurs at both layers, while the substantive content that the asymmetry carries varies with the layer. This agrees with Paper 3's reading of the asymmetry: it is not formal self-similarity but substantive content at its thickest at that cell. Form here is an indicator of substantive content, not its source.

This is in pattern consistent with the $r \approx 5$ topological asymmetry articulated in Methodology 6: Phase-Transition Windows: the four steps are not even; the middle steps (Main Steps 2 and 3) open up the articulation space, while the final step (Main Step 4) closes it. The multiplicative path opens up further at the middle-step position, commensurate with the substantive content it carries.

---

6. Main Step 4(a): The Regular Period Core

6.0 The organization of Main Step 4

Main Step 4 is the closure of $L_2$ together with the registration of its remainder. By the registration declaration of § 2.1 its class is remainder ledger structure, not within-layer fractal; internally it is organized as capture core → uncaptured ledger → trigger condition for the next layer, and not as a re-enactment of the four phases.

Main Step 4 spans three sections: the present one (§ 6) articulates the regular period core, § 7 the irregular boundary, § 8 closure and the remainder ledger. That these three sections carry one Main Step while Main Steps 1 through 3 carry one section each is itself the registration just stated (§ 2.1).

The relation between § 6 and § 7 is core and enrichment, not mutually exclusive binning. This must be said first, or the present section will be read as claiming that divisors, residues, periods, and monodromy exist only in the regular case. They do not: an irregular connection still has poles and divisors, still has formal residues, still has a monodromy group, still has global period data. What § 7 adds is exponential factors, directions, and Stokes data. The relation is

$$\{\text{regular core}\} \hookrightarrow \{\text{irregular enrichment}\},$$

not two boxes standing side by side. This section articulates the core; § 7 articulates what the enrichment adds.

6.1 Carrier migration

Before the technical content of the four stages, one change of type must be registered. It is not a technical premise of this section; it is one of the paper's identifications.

Paper 3 articulates the capture range of the $L_1$ mode as the real closed field $\mathbb{Q}^{\rm alg}_{\mathbb{R}}$: a scalar field closed under algebraic operations. The present paper articulates the capture core of the $L_2$ mode as the ring of ordinary periods $\mathcal{P}$: the image of a class of integral expressions under an evaluation map (§ 2.3). These are not objects of the same kind.

> Carrier migration: between $L_1$ and $L_2$, the carrier of the capture range migrates from the algebraic closure of a scalar field to the evaluation image of a geometric integral language.

The tier of this identification must be stated at once. It is a contrastive reframing of two existing mathematical definitions — the real closed field, and Kontsevich–Zagier periods — and it is therefore an SAE registration, not a mathematical theorem; nor does it claim that any ontological leap has been discovered. The paper nevertheless lists it as a core identification because its three consequences are in force at the technical level throughout, not because it constitutes new mathematical content.

The migration has three consequences, on which several later sections depend.

First, the word "saturation" no longer applies. A scalar field can be said to saturate: it is closed under a family of operations and is the smallest such field containing a given starting point. An evaluation image has no corresponding closure property. What $\mathcal{P}$ is, is settled on the syntactic side by $\mathsf{Int}_{\rm alg}$, not by any operational closure of $\mathcal{P}$ itself. The paper accordingly says evaluation core throughout and never says saturation point (§ 2.6, second item).

Second, the type of the capture criterion has changed. The $L_1$ criterion asks what equations a number satisfies; the $L_2$ criterion asks what representations it admits (§ 2.1). The first adjudicates a relation between a number and a family of equations, the second between a number and a family of geometric integral expressions. The difference resurfaces in § 8.2 when both are set beside the algorithmic-approximability criterion of the effective-Cauchy mode: none of the three criteria is a special case of another.

Third, the difference between a ring and a field is not an oversight. $\mathbb{Q}^{\rm alg}_{\mathbb{R}}$ is a field; $\mathcal{P}$ is a ring, and whether it is closed under inversion is undecided — whether $1/\pi \in \mathcal{P}$ is not known, which is why the literature introduces the extension $\hat{\mathcal{P}} = \mathcal{P}[1/\pi]$. A reader will naturally ask: Main Step 3 has already articulated division and closed on $\mathbb{C}^\times$, so why does the evaluation core stop at a ring?

The answer lies in the distinction of § 2.2. The closure under division of Main Step 3 is closure at the level of functions and paths; whether $\mathcal{P}$ is closed under inversion is a question at the level of numerical sets. The two belong to different levels, and the first does not entail the second. The paper does not claim that closure under division at the level of functions forces a set of numbers to be closed under inversion.

A reminder about the two spines. What this section articulates — divisor, residue, period, monodromy — is the analytic–covering spine on curves, whereas the definition of $\mathcal{P}$ admits algebraic varieties of any finite dimension and semi-algebraic domains of any dimension. By the two-spine declaration of § 2.3, the paper does not claim that the former generates all of $\mathcal{P}$. What this section articulates is the position of the four stages within a cross-layer comparison, not the manner in which $\mathcal{P}$ is generated.

6.2 Pointwise: divisors and valuation

The action of the first stage is to register the defect left by multivaluedness and by singularities as an integer sitting at a discrete point.

Valuation. For a meromorphic function $f$ and a point $p$, $\mathrm{ord}_p(f)$ is the order of the zero (positive) or of the pole (negative) of $f$ at $p$. It is a $\mathbb{Z}$-valued local quantity.

Divisor. $\mathrm{div}(f) = \sum_p \mathrm{ord}_p(f)\cdot p$ packages the valuations at the several points into an element of the free abelian group $\mathrm{Div}(X)$. The principal divisors form a subgroup, and the quotient $\mathrm{Cl}(X)$ is the divisor class group (Paper 1 Appendix B.1).

Note that valuation and divisor are objects at different levels: a valuation is an integer attached to a point, a divisor is a formal sum of them. This distinction is used in the next passage and again in the type table of § 9.1.

The first half of the ZFCρ debt.

The $\infty$ of $L_1$ is monolithic: the $L_1$ articulation mode has no tool for telling different divergences apart. Main Step 2 identified $L_1$'s $+\infty$ and $-\infty$ into the single $\hat\infty$ (§ 4.3), and that step did not reduce the monolith but reinforced it — two boundary marks were merged into one.

Valuation now equips $\hat\infty$ with a scale. Since $\mathrm{ord}_\infty$ is $\mathbb{Z}$-valued, $\hat\infty$ is no longer a point without internal structure but carries an integer grading. The order of the ratio of two meromorphic functions at $\infty$ is $\mathrm{ord}_\infty(f) - \mathrm{ord}_\infty(g)$: the $\infty/\infty$ that collapsed at $L_1$ has become a subtraction of integers.

But this settles only the order half. What ZFCρ is owed is the fine ratio, and a fine ratio is a quantity at the level of coefficients; valuation resolves order and does not resolve coefficients. Two functions of the same order at $\infty$ have a ratio of asymptotic coefficients that $\mathrm{ord}_\infty$ does not resolve.

The interface for the coefficient half is located at the Newton polygon in § 7.4. But by § 2.6, sixth item, an interface located is not an interface connected; the paper does not claim the debt settled, and only re-registers it from "no interface" to "interface located but not connected".

A registration in passing: what interfaces with ZFCρ is the valuation, not the divisor — the divisor merely packages the valuations at the several points. To say that the divisor equips $\infty$ with a scale would be to mistake an act of packaging for an act of grading.

6.3 Local: residues and $2\pi i$

The action of the second stage is to convert the pointwise integer scale into a numerical quantity that can be integrated.

Residue. $\mathrm{Res}_p(\omega)$ is a linear functional acting on the local data of a meromorphic differential form at the point $p$ and returning a number (§ 2.2, the distinction between operators and numerical objects). On a compact Riemann surface the global residue theorem imposes the constraint $\sum_p \mathrm{Res}_p(\omega) = 0$.

The interface with the first stage. Logarithmic differentiation gives

$$\mathrm{Res}_p\!\left(\frac{df}{f}\right) = \mathrm{ord}_p(f),$$

whence

$$\deg\,\mathrm{div}(f) = \sum_p \mathrm{ord}_p(f) = \sum_p \mathrm{Res}_p\!\left(\frac{df}{f}\right) = 0.$$

That is: the vanishing of the degree of a principal divisor is a specialization of the global residue theorem to the logarithmic differential $d\log$. Note that this interface holds only through that one map; it may not be said in general that all divisor theorems and all residue theorems are two notations for one conservation law. § 9.2 registers $d\log$ as one of the cross-component interfaces.

$2\pi i$ as the within-layer articulative unit of the exchange law. The rate at which residues are converted is

$$\oint_{|z|=1} \frac{dz}{z} = 2\pi i.$$

§ 5.5 has already articulated that this and the deck transformation group $2\pi i\mathbb{Z}$ of $\exp$ are two presentations of one statement — going once counterclockwise around the unit circle, $\log z$ increases by $2\pi i$. The rate of conversion at this stage is therefore not a constant internal to residue calculus but the presentation, in the language of integration, of the covering structure produced by Main Step 3.

Paper 1 § 4.5 establishes $2\pi i$ as the articulative unit of $L_2$'s exchange law. At this stage that commutation takes the concrete form of the residue theorem and the argument principle. From

$$\frac{1}{2\pi i}\oint_\gamma \frac{f'}{f}\,dz = N - P,$$

where $N$ and $P$ count, with multiplicity, the zeros and poles of $f$ inside $\gamma$, two different articulation routes — integrating along a closed path, and counting zeros and poles — can be exchanged in order, and the cost of the exchange is exactly the unit $2\pi i$. This is precisely the form of exchange law Paper 1 § 4.5 articulates: not that two arbitrary operations commute, but that two articulation routes are exchangeable under some unit.

The word "local-monetization" is not a metaphor here. The first stage produces integers, namely orders; this stage converts them into integral quantities at the rate $2\pi i$; and that rate is at the same time the within-layer articulative unit of the exchange law of this paper's trajectory. An exchange, a rate of exchange, and a rate that is itself the identity of the layer — all three are present.

(Notational discipline: this paper has two signature slots, the transition signature and the within-layer articulative unit of the exchange law; see § 11.5.3. Nowhere does the paper speak simply of "the signature of $L_2$".)

6.4 Global: periods and monodromy

The action of the third stage is the passage from local data to global data.

Monodromy. $\mathrm{Mon}$ is a representation of the fundamental group $\pi_1$. The remainder left by Main Step 3 inner step 4 — path dependence (§ 5.3) — is here articulated as an object: analytic continuation along paths in different homotopy classes yields results whose difference is characterized by that representation. For $\log$ the representation is $\pi_1(\mathbb{C}^\times) = \mathbb{Z}$ acting on the fibre by translation by $2\pi i\mathbb{Z}$; for the $n$-th root it is $\mathbb{Z}$ acting through $\mathbb{Z}/n$. The two levels of reverse interrogation of § 5.3 each receive their representation here.

Period. $\mathrm{Per}$ is the pairing between de Rham cohomology and Betti homology, $\int_\gamma \omega$ with $\gamma \in H_1$ and $\omega \in H^1_{\rm dR}$. This is the geometric source of the evaluation core $\mathcal{P}$.

The interface with the second stage, and its direction. The pairing on a small loop gives

$$\oint_{\gamma_p}\omega = 2\pi i\,\mathrm{Res}_p(\omega).$$

That is: a small-loop period is $2\pi i$ times a residue. The direction must be held. One may not divide this identity by $2\pi i$ and infer that residues lie in $\mathcal{P}$, because whether $1/\pi \in \mathcal{P}$ is undecided (§ 6.1, third consequence). If $\omega$ is an algebraic meromorphic differential form defined over $\overline{\mathbb{Q}}$ and $p$ is an algebraic point, then $\mathrm{Res}_p(\omega)$ is an algebraic number and therefore lies in $\mathcal{P}$ — but that is a separate reason, running through $\overline{\mathbb{Q}} \subset \mathcal{P}$, and is not obtained by dividing the identity above.

This discipline about direction is the first place where the third consequence of § 6.1 — that $\mathcal{P}$ is a ring and not a field — is actually in force at the technical level.

$\mathcal{P}$ as evaluation core. Every algebraic number lies in $\mathcal{P}$, so under inclusion $\mathcal{P}$ sits above $\mathbb{Q}^{\rm alg}_{\mathbb{R}}$. But by § 6.1 this sitting-above is set inclusion only, not the continuation of an object of the same type: the first is an algebraically closed scalar field, the second an evaluation image. The carrier migration takes place inside this very inclusion, and the inclusion itself conceals it.

Three boundaries must be registered together:

• $\mathcal{P}$ is countable while the $L_2$ substrate is not, so $\mathcal{P}$ does not exhaust the substrate;
• $\mathcal{P}$ does not exhaust even the computable reals (§ 8.2);
• the curve spine of this section does not generate all of $\mathcal{P}$ (§ 2.3, the two-spine declaration).

The remainder of this stage. In the regular case, $\mathrm{Per}$ and $\mathrm{Mon}$ jointly articulate path dependence: the monodromy characterizes the effect of going around, the period gives the value of the integral. So far as regular singularities are concerned, the remainder left by Main Step 3 has here been articulated as an object.

But this covers regular singularities only. At an irregular singularity the formal solutions carry exponential factors, and the monodromy group does not suffice to characterize their behaviour — one and the same formal solution, as the singularity is encircled, has its asymptotic validity fail to match across different sectors. This is the entrance to § 7, and it is what the fourth stage adds.

---

7. Main Step 4(b): The Irregular Boundary

7.0 What this section adds

The close of § 6 articulated that, in the regular case, periods and monodromy jointly articulate the path-dependence remainder left by Main Step 3. This section treats what that articulation does not cover.

By the relation registered in § 6.0, this section is not a second box standing beside § 6. An irregular connection still has poles and divisors, still has formal residues, still has a monodromy group, still has global period data. What this section adds is three things: exponential factors, directions, and Stokes data.

$$\{\text{regular core}\} \hookrightarrow \{\text{irregular enrichment}\}$$

This section is at the same time where the paper's load rests. § 2.6, third item, has already declared that no load is placed on $e \notin \mathcal{P}$; the load is carried jointly by the two extensions of § 7.1 — the syntactic side is definitional, the function side is theorem-level — while the registration that reads the two sides as one and the same boundary is listed separately as a T2-conditional item.

7.1 The two extensions, of expression classes and of connections

The paper's T2 load consists of the following two extensions.

Syntactic side:

$$\mathsf{Int}_{\rm alg} \hookrightarrow \mathsf{Int}_{\exp}$$

that is, the integrand is further allowed to carry factors of the form $e^{-f}$ with $f$ algebraic. This is the left vertical arrow of Table 1 (§ 2.3).

Function side:

$$\{\text{connections with regular singularities}\} \hookrightarrow \{\text{connections admitting irregular singularities}\}$$

The classification of the latter is given by the Levelt–Turrittin theorem: any connection on the formal punctured disc decomposes as a direct sum of elementary connections each carrying a single exponential factor — comprising a ramified cover of ramification index $\nu$, an exponential part of the form $E^\varphi = (\mathbb{C}((t)),\, d + d\varphi)$, and a regular connection $R$.

The two extensions differ in epistemic tier and must be registered in three tiers.

Tier	Content	Character
Definitional load	ordinary integral syntax extends to exponential integral syntax ($\mathsf{Int}_{\rm alg} \hookrightarrow \mathsf{Int}_{\exp}$)	definition / syntax; given directly by the definitions of $\mathcal{P}$ and $\mathcal{P}^e$
Theorem-level load	regular connections extend to connections admitting irregular singularities, the latter requiring exponential factors, directions, and Stokes data	theorem (Levelt–Turrittin formal classification, and related results)
Conditional bridging	the two extensions above are the numerical-side and function-side presentations of one and the same SAE boundary	T2-conditional; a research-level correspondence belonging to the exponential-motives line

The three tiers may not be written together. In particular: what Levelt–Turrittin supports is the second tier, not the first, and still less the third. The first tier is given by definition and needs no theorem; the third is an alignment registered by the paper within the SAE framework, and is not a mathematical conclusion derived from the second.

The paper's T2 load is carried jointly by the first and second tiers; the third is listed separately as a conditional item, and should that alignment be refuted in future, what must change is the paper's registration reading the two sides as one boundary, and not the two extensions themselves.

This distinction is the crux of the paper's load-bearing structure. That $e \notin \mathcal{P}$ is an open conjecture; were the load placed on it, a refutation would bring the structure down. With the load on the extensions, any outcome for that conjecture leaves this section standing (§ 12.3).

7.2 Newton–Turrittin data as the formal organizer of the irregular boundary

The formal data at an irregular singularity can be organized by the Newton polygon. But the scope of that organization must be drawn precisely: the Newton polygon organizes formal classification; it does not settle analytic classification. This section articulates three strata, and the three are three distinct levels of determination, not a causal chain.

First stratum · the coarse formal stratum: slopes

The slopes of a connection are given by the slopes of the boundary of the Newton polygon of its differential operator, and are independent of the choice of operator. This yields the paper's most basic criterion:

> All slopes zero is a regular singularity; a positive slope present is an irregular singularity.

This is the criterial form of the function-side extension of § 7.1.

The slopes of the several edges are exactly the degrees of the determining polynomials (including the trivial polynomial $q \equiv 0$, of degree zero); if the edge of slope $\lambda$ has horizontal length $m$, there are $m$ determining polynomials of degree $\lambda$.

This stratum yields: the divide between regular and irregular, the coarse-grained degree of irregularity, the ramification order, the Gevrey level, and the order of growth of solutions at the singularity.

Second stratum · the fine formal stratum: full exponential factors

The full form of an exponential factor is a Puiseux polynomial

$$q = \sum_{k=1}^{s} a_k\, z^{-k/\nu}, \qquad a_s \neq 0.$$

Slopes govern only the exponents $k/\nu$; the coefficients $a_k$ lie outside the slopes. This is the section's central distinction.

Two things are settled by this stratum and not by the first:

• The type of growth. The first stratum gives the order, determined by degree; the type depends on the leading coefficient $|a_s|$, and slopes do not settle it.
• The candidate Stokes directions. Directions are settled by the difference $q_\alpha - q_\beta$ of exponential factors, and depend in particular on the leading coefficient of that difference and on its argument — that is, on where $\mathrm{Re}(q_\alpha - q_\beta)$ changes sign. Correspondingly the level datum is defined as the set of nonzero slopes of the differences $q_0 - q_j$ of the Galois conjugates; these are the slopes of differences, not the slopes of $q$ itself.

The identification the paper earns from this split: type of growth and candidate Stokes directions both require the coefficient data of the second stratum, and neither is settled by slopes alone. Under a two-stratum account that runs the Newton polygon straight into Stokes data, this shared feature is invisible.

(The word "cognate" is not used here. Stokes directions are settled by the difference $q_\alpha - q_\beta$ of exponential factors, so individual leading terms may cancel, and the two need not be settled by the same leading coefficient. What the paper earns is that they belong to the same second stratum, both lying beyond the coarse stratum of slopes.)

Third stratum · the analytic stratum: Stokes multipliers

The isomorphism of formal meromorphic connections guaranteed by the Levelt–Turrittin theorem cannot in general be lifted to a global meromorphic isomorphism; it lifts only locally, on arc neighbourhoods of the singularity (Hukuhara–Turrittin). The formal data describe the asymptotic behaviour as the singularity is approached, and the precise description of that behaviour changes over the course of going once around the singularity.

Hence: with the formal normal form fixed, the Stokes matrices still supply further complete analytic invariants. Stokes multipliers are analytic moduli lying beyond formal classification.

The three strata do not compress into a chain

> Slopes do not settle Stokes multipliers.

To write "slopes → exponential factors → Stokes directions and Stokes multipliers" as a single causal string is to merge formal classification with analytic classification in one step. The first two strata belong to formal classification and the third does not; and between the first and the second there is no relation of determination either, but a relation of coarse to fine — slopes settle degrees, not coefficients.

The paper therefore registers Newton–Turrittin data as the formal organizer of the irregular boundary, and not as the common generator of stage 4.

7.3 The candidate comparison interface for $G_\infty$

Paper 1 Appendix B.5 articulates $G_\infty$ as growth at infinity: the order $\rho$ and type $\sigma$ of an entire function, together with the Phragmén–Lindelöf principle. That object is a scalar asymptotic invariant; the $G$ in the notation should not be read as denoting a group (§ 2.5).

The first stratum of § 7.2 gives the order of growth of solutions at a singularity, and the second gives the type. This agrees in form with the content of order and type, and suggests that a comparison interface between the two may exist.

But the paper registers a candidate interface only, and claims no identity. The reason: Paper 1's $G_\infty$ is an asymptotic invariant of entire functions in general and is not specific to solutions of differential equations, whereas $\mathrm{Slope}_\infty$ is a formal invariant of meromorphic connections. Whether the two are the same invariant is not proved here, and no statement of that identity has been found in the literature.

The registration is therefore:

> Newton–Turrittin data supply a candidate comparison interface for Paper 1's $G_\infty$; identity is not proved. Slopes give a candidate correspondent for the order of growth; type depends on the leading coefficient of the exponential factor and is not something slopes can settle.

§ 11.1 lists this registration as a charitable reconstruction and not Paper 1's original intent — the content Paper 1 attached to $G_\infty$ when it wrote the symbol was scalar, and to say that the notation anticipated a group which the present paper completes is this paper's reconstruction, not a recovery.

Notational discipline (§ 2.5): $G_\infty$ is kept as Paper 1's object; the slope filtration is written $\mathrm{Slope}_\infty$; the local differential Galois group is written $G_{\mathrm{diff,loc}}$ (§ 9.2). The three do not share a symbol.

7.4 Sectors, real oriented blow-up, and the second half of the ZFCρ debt

The circle of directions

The directional structure at an irregular singularity cannot be carried by a point. The standard treatment is to perform a real oriented blow-up at the singularity: the preimage is a circle, points of the circle correspond to directions around the singularity, and open sets of the circle correspond to germs of open sectors there. Exponential factors are then regarded as sections of a local system on that circle of directions (the Deligne–Malgrange formulation).

This is the technical content of the coarse closure plus directional refinement of § 8.1; the present section supplies the object of that refinement, and § 8.1 articulates what it means for the closure structure of $L_2$.

Sector mismatch

One and the same formal solution has different asymptotic validity on different sectors. As one travels around the circle of directions, the correspondence between actual solutions and formal solutions changes — this is the Stokes phenomenon. The monodromy group does not suffice to characterize it: monodromy records the overall effect of one circuit, whereas what is needed here is sector transition data ordered by direction.

This locates precisely the remainder of § 6.4: periods and monodromy articulated path dependence in the regular case, whereas path dependence in the irregular case contains directional resolution, which exceeds what monodromy can record.

The second half of the ZFCρ debt

§ 6.2 articulated that valuation equips the monolithic $\infty$ with an integer scale, settling the order half. What ZFCρ is owed is the fine ratio, and a fine ratio is a quantity at the level of coefficients.

By the three strata of § 7.2, quantities at the level of coefficients fall in the second stratum: the Newton polygon gives a stratification by slope, each slope carries a further characteristic equation, and a fine ratio is a quantity at the level of the coefficients of that characteristic equation. The location of the interface for this debt can therefore be given precisely as the second stratum, and not the first.

But an interface located is not an interface connected. The paper gives no explicit map, conserved quantity, or functor between ZFCρ's fine ratio and any coefficient of any characteristic equation. The status of the debt is accordingly re-registered from "no interface" to "interface located but not connected" (§ 2.6, sixth item), and may not be written up as settled.

One internal consistency is worth registering: the half that is not connected falls precisely on the stratum this layer has not captured — the second and third strata are exactly what § 7.5 adjudicates as marked out but not entered. The unpaid half of the debt and the unreached part of the capture are two faces of one boundary.

7.5 $\mathcal{P}^e$ is marked out and not entered

Everything articulated in this section so far — slopes, exponential factors, the circle of directions, Stokes data — lies on the function side. The corresponding object on the numerical side is the exponential periods $\mathcal{P}^e$ (Table 1, § 2.3, lower right).

Their relation is the conclusion of this section:

> The fourth stage marks out the exponential direction by means of Stokes data, and does not enter $\mathcal{P}^e$.

"Does not enter" requires an operational definition: it means that $\mathcal{P}^e$ is not raised to the evaluation core of this paper's current trajectory — the evaluation core here is $\mathcal{P}$ (§ 6.4), and the function-side work of stage 4 does not alter that selection. It does not mean that the paper cannot define, discuss, or display exponential periods: the present section defines $\mathcal{P}^e$ and states that $e \in \mathcal{P}^e$ is proved. There is no contradiction.

Further, the tier of this statement is the third tier, conditional bridging, per the table of § 7.1 — it is the paper's registration of an alignment between function side and numerical side, and not a mathematical conclusion derived from Levelt–Turrittin alone.

This is the recurrence, at the close of the layer, of the mark-without-construct of Main Step 1. Main Step 1 marked out $i$ without constructing its space of habitation (§ 3); Main Step 4 marks out the exponential direction without constructing its values. The same act occurs once at the cross-level threshold and once at the closing of the layer.

The closure remainder of $L_2$ is therefore: the exponential map and the direction of exponential integration.

The standing of $e$

$e$ is the canonical numerical witness for that direction; it is not a load-bearing item. Three facts must be stated separately.

First, $e = \exp(1)$. And $\exp$ is the covering map produced by Main Step 3 inner step 4 (§ 5.3), that is, an internal product of $L_2$. So $e$ is the value of an internal product at one point.

Second, $e$ certainly lies in $\mathcal{P}^e$. Taking $f(x) = -x$, $\omega = dx$, and the semi-algebraic domain $[0,1]$,

$$e = 1 + \int_0^1 e^x\,dx.$$

This is a fact at the level of definition and computation, not a conjecture.

Third, whether $e$ lies in $\mathcal{P}$ is not known. This is an open conjecture (§ 2.6, third item).

The precise form of the remainder ledger is therefore:

Question	Status
$e \in \mathcal{P}^e$	proved
$e \in \mathcal{P}$	unknown
Is the exponential expression class strictly larger than the ordinary period expression class?	strictly larger syntactically (§ 7.1, definitional tier)
Is the corresponding inclusion of evaluation images witnessed as strict by $e$?	unknown

It may not be written that $e$ is a remainder within the substrate that goes uncaptured — that would run the proved row together with the unknown row. The precise formulation is: membership status in the ordinary period core is unknown; capture within the exponential period extension is established; $e$ serves only as a candidate strictness witness.

Inverse-solution specification, and its limit

$e$ admits one further representation bearing on the paper's load:

$$\int_1^e \frac{dx}{x} = 1.$$

That is, $e$ is specified implicitly by an equation in a period-valued function. This mode of specification does not belong to ordinary period expression syntax — it is by inverse solution, and the definition of $\mathsf{Int}_{\rm alg}$ contains no inverse-function operator (§ 2.3). And this integral is $\log$, which is exactly the articulative unit Paper 1 § 4.5 establishes for $L_1$'s exchange law. So it may be stated:

> The articulative unit of $L_1$'s exchange law is realized at $L_2$ as a period-valued function; $e$ is its inverse solution; and specification by inverse solution falls outside ordinary period expression syntax.

But the discipline of § 2.3 must be held here. The identity above says that this expression, the one we have written down, specifies $e$ by inverse solution; it cannot rule out that $e$ has some other, ordinary period representation — and whether it does is exactly what is undecided in $e \in \mathcal{P}$. Every statement of this subsection is therefore at the level of expressions, and may not be read as an assertion about numerical membership status. The distinction between forward capture and inverse-solution specification is a distinction at the level of expression formation, not a theorem-level substitute for $e \notin \mathcal{P}$. The paper places no load on it, and uses it only to articulate where $e$ sits within the articulation structure of this layer.

Locating the load

In summary, the paper's load is distributed as follows:

• Definitional load: $\mathsf{Int}_{\rm alg} \hookrightarrow \mathsf{Int}_{\exp}$, given directly by the definitions of $\mathcal{P}$ and $\mathcal{P}^e$;
• Theorem-level load: regular connections extend to connections admitting irregular singularities, supported by Levelt–Turrittin and related formal classification;
• T2-conditional item: the registration of the two sides as the numerical-side and function-side presentations of one boundary;
• T3 witness: $e \notin \mathcal{P}$.

$e$ bears no load-bearing role. Specifically: it does not bear the load of the regular/irregular boundary, which is borne by the two extensions of § 7.1; nor does it bear the load of strictness for the evaluation core, which is an undecided T3. It retains two non-load-bearing registrations: the transition signature inherited from Foundation v2 (§ 11.5.3), and this paper's candidate numerical witness.

Should $e$ be proved to be a period, the only things requiring change are the witness row of this section and one line of the ledger in § 8.2; the two extensions of § 7.1, the three strata of § 7.2, and the core structural statement of § 5.4 all stand (§ 12.3).

---

8. Main Step 4(c): Closure and the Remainder Ledger

8.1 Coarse closure and directional refinement

Main Step 2 has already given the one-point compactification of $\mathbb{C}$, merging the two boundary marks of $L_1$ into the single $\hat\infty$; § 4.3 declared at the same time that this merging is a chosen canonical closure and not something topology forces. The present section articulates where the cost of that choice falls due.

The circle of directions of § 7.4 conflicts directly with that one-point compactification: the former requires that the direction of approach to the singularity be preserved, the latter collapses all directions to a point. The two cannot hold at the same resolution.

The resolution is not to choose between them but to work at two levels.

Coarse closure: the Riemann sphere. This is the canonical closure in the meromorphic and conformal context. It makes the two-periodic boundary pair $0 \leftrightarrow \hat\infty$ hold (§ 5.3), makes divisors defined at $\infty$ (§ 6.2), and gives the global residue theorem a compact base space (§ 6.3). This section does not retract it.

Directional refinement: real oriented blow-up. The irregular remainder forces the directions that were collapsed away to reopen into a circle of directions (§ 7.4).

From this the paper obtains an identification about the closure structure of $L_2$:

> The closure of $L_2$ is not a single compactification. The coarse closure collapses all directions to a point; the irregular remainder then forces that point to reopen into a circle of directions.

The shape is worth pointing out on its own: closure → remainder → the reopening of the closure. Closure is not a terminus; the closure itself produces the remainder that demands it be opened again. And the reopening does not cancel the closure — it is a second level of resolution of that same closure.

A discipline: the reopening does not retract the coarse closure. The two are not competing closures but two levels of resolution of one closure. Divisors, residues, and periods are defined on the coarse closure and are not invalidated by directional refinement; the refinement takes effect only where directional resolution is required, that is, at irregular singularities.

The relation to upstream: Paper 1 § 5.2 articulates the identification of $\pm\infty$ as forced by topology. The present paper refines this to: topology does not uniquely force one-point compactification, and one-point compactification is the canonical closure chosen within the meromorphic and conformal context; the cost of that choice is directional resolution, and the cost falls due at the irregular boundary. § 11.1 lists this as the first upstream refinement.

8.2 The remainder ledger

Following the three-way division of Paper 3 § 6, rearranged at $L_2$.

First class: substrate-external remainder

At $L_1$ the instance of this class is $i$, which $L_2$ Main Step 1 takes in (§ 3). At $L_2$ the class points in the direction outside $\mathbb{C}$, namely the Cayley–Dickson sequence $\mathbb{H}$, $\mathbb{O}$.

But there is one crucial difference from the situation at $L_1$. $i$ was pressed out by an interrogation internal to $L_1$: $x^2+1=0$ has no root in $\mathbb{R}$, and root-finding is $L_1$ Main Step 3's own reverse interrogation. Paper 3 registers it on that basis as cross-level forcing and not as a lateral branch.

$\mathbb{H}$ has no such source. $\mathbb{C}$ is algebraically closed and root-finding no longer presses (§ 5.4); no interrogation internal to $L_2$ presses $\mathbb{H}$ out. Cayley–Dickson is a construction that may be applied, not a product that must be accepted.

The standing of the first class at $L_2$ therefore differs from its standing at $L_1$: it is not a forced remainder. § 10 adjudicates it on this basis as a cost branch — obtained by paying away commutativity, rather than by treating a remainder of this layer.

This difference is a downstream consequence of the asymmetry of § 5.4. In removing the scalar gap, algebraic closure also removed the mechanism that produces forced substrate-external remainders.

Second class: structures over the substrate, not captured by the mode

The member of this class is the exponential direction, a structural object, and not the number $e$. The "not captured" in the name of the class is a syntactic boundary: ordinary period expression syntax $\mathsf{Int}_{\rm alg}$ does not admit exponential factors, so the ordinary-period articulation mode this paper has selected does not construct the exponential direction (§ 7.1, definitional tier). Whether the evaluation images accordingly form a strict extension, that is

$$\mathcal{P} \subsetneq \mathcal{P}^e,$$

remains an undecided question on the numerical side. The two belong respectively to the syntactic side and the evaluation side of § 2.3, and may not be put in one sentence.

One generalized definition is also required: "over the substrate" here generalizes to "defined over that substrate and taking values still in that substrate", and does not assert that the object belongs to $\mathbb{C}$ as an element — the exponential direction is a structural object, not a number. The name of the class follows the wording of the three-way classification of Paper 3 § 6, and the generalization is made only to accommodate structural objects.

The exponential direction (§ 7.5). Its numerical witness is $e$.

Question	Status
$e \in \mathcal{P}^e$	proved
$e \in \mathcal{P}$	unknown
Is the exponential expression class strictly larger than the ordinary period expression class?	strictly larger syntactically (§ 7.1, definitional tier)
Is the corresponding inclusion of evaluation images witnessed as strict by $e$?	unknown

The precise formulation is: membership status in the ordinary period core is unknown; capture within the exponential period extension is established; $e$ serves only as a candidate strictness witness. It may not be written that $e$ is a remainder within the substrate that goes uncaptured — that would run the proved row together with the unknown row.

Third class: uncomputable, undefinable

Inherited from $L_1$. Paper 3 § 6.3.2 earns it by a cardinality argument: for any countable formal language the definable reals are at most countable, while the real line has cardinality $2^{\aleph_0}$ (§ 2.1). The class remains uncaptured at $L_2$.

At $L_2$ there is a further and stronger fact: $\mathcal{P}$ does not exhaust even the computable reals — Yoshinaga has exhibited a computable real that is not a period. A whole stratum therefore still separates the evaluation core of $L_2$ from the third class.

The capture chain and the remainder left by several modes together

Setting the ranges earned by the three criteria in one line:

$$\mathbb{Q}^{\rm alg}_{\mathbb{R}} \subsetneq \mathcal{P}\cap\mathbb{R} \subsetneq \text{computable reals} \subsetneq \text{definable reals} \subsetneq \mathbb{R}$$

The three criteria are finite algebraic constraint (the $L_1$ mode), algebraic integral representation (the $L_2$ mode), and algorithmic approximability (the effective-Cauchy articulation mode). By § 2.1, none of the three is a special case of another — they are not three instances of one kind of criterion but three different types of standard of adjudication.

From this a reading follows:

> Each of the three criteria earns one stratum, and their union still stops at the definable reals. This is the strongest form the paper can display: not the remainder of a single mode, but the remainder left after several modes with mutually distinct criteria have acted together.

Jurisdiction statement. The paper does not claim on this basis that the union of all mathematical tools fails to exhaust the substrate. Three is not all; and such a claim would first require drawing the boundary of what counts as a "tool" — tools may carry parameters and may form uncountable families, so the cardinality argument of Paper 3 § 6.3.2 cannot be transplanted directly — and drawing that boundary lies outside this paper's jurisdiction.

The universally quantified form, that no framework exhausts, is a meta-proposition already laid down in Paper 1's abstract (the meta-level half of the twofold inexhaustibility), and earning it substantively is work for a framework-level paper. This paper displays and points forward; it does not endorse.

〔Kind: structural display plus jurisdiction statement〕

In passing: the position of the effective-Cauchy articulation mode in the chain above is the material for the open question left by Paper 3 § 6.4 (§ 2.1). That mode captures strictly more on the real axis than the period core of $L_2$, while its criterion is of a different kind from those of both $L_1$ and $L_2$. Its placement is adjudicated in § 11.3.

8.3 Euler's formula: superposition placement and transition closure equation

Euler's formula carries two different predicates in this paper. They do not compete, and they answer different questions.

As placement

§ 5.5 has given the structural reason:

$$\cos\theta + i\sin\theta = e^{i\theta}$$

The two sides fall on the two sides of substrate and mode — the left is a substrate-level classical construction of analysis (the real trigonometric functions; by the resource discipline of § 5.2.3 the paper does not invoke them on the spine), the right a mode-level product of Main Step 3 inner step 4 ($\exp$ restricted to the imaginary axis). The formula states that the two are equal.

It therefore resides at no cell of this trajectory: it is an identification between a substrate-level construction and a mode-level product. The paper accordingly registers it as a substrate–mode superposition placement — what it crosses is the general firewall between substrate and mode (§ 2.1), not two cells within the layer.

An inventory of constituents may serve as corroboration but is not the reason: in $e^{i\pi}+1=0$, $\pi$ is the remainder of $L_1$ Main Step 4, $1$ and $0$ are the two boundary tokens of $L_1$, $i$ is the product of $L_2$ Main Step 1, and $e$ is the numerical witness of the closure remainder of $L_2$. Five items across two layers and four positions. But the inventory yields only a description of position; the structural reason lies in the identification of two articulation routes.

As closure equation

The cross-level ledger table of SAE Foundation v1 records the closure equation of $L_1 \to L_2$ as $e^{i\pi}+1=0$ and its two remainders as $i$ and $\pi$. That table gives no reason why these two in particular. The present paper gives one.

$i$ is the input to the transition. It is the substrate-external remainder of $L_1$, taken into $L_2$ by the mark-without-construct of Main Step 1 (§ 3). It is not produced by $L_2$; the transition brings it in.

$\pi$ is the first structural product of this layer. It is the substrate-internal transcendental remainder of $L_1$ (the ledger of Paper 3 § 6), which $L_1$ can name but cannot capture. $L_2$ first articulates it at Main Step 3 inner step 4 as the deck transformation group $2\pi i\mathbb{Z}$ (§ 5.3), and then captures it as a period at Main Step 4 inner step 3 (§ 6.4).

What the closure equation binds is therefore: the input to the transition, and the first structural product of this layer.

And the item by which they are bound — $e$ — is captured within the exponential period extension, while its membership status in the ordinary period core remains undetermined (§ 7.5). The equation may accordingly be read as: two remainders already captured in the ordinary period core, passing through the one whose membership status is undetermined, return to $L_1$'s own two boundary tokens.

The relation between the two predicates

Placement answers "where does it reside"; closure equation answers "what does it close". One object can be both, and the two do not compete.

Note further that it is the closure equation of the transition, not of the layer. The closure of the layer $L_2$ is coarse closure together with directional refinement (§ 8.1), a different object from transition closure.

Paper 1 § 5.3 articulates Euler's formula as best understood as a closure instantiation rather than the full closure of $L_2$. Its precise content is just this: it is an instance of transition closure, not of layer closure. § 11.1 lists this as the third upstream refinement.

For the full reception and rearrangement of that row of Foundation v1 — including the change of its "behaviour" column from the exponential map to complexification — see § 11.5.4.

8.4 The $L_2 \to L_3$ trigger: the level of values and the level of rules

Paper 1 § 5.5 establishes the trigger criterion: what triggers $L_3$ is not any difficulty of $L_2$ itself, but the reframing of the problem into one about rules and decidability. This section registers two questions along the three coordinates of § 2.4, so as to show that they belong to different levels.

Question	Numerical membership status	Proof-theoretic status	Algorithmic status	Layer attribution
$e \in \mathcal{P}$?	unknown	conjectural	—	undecided within $L_2$
Is the general period-expression identity problem decidable?	—	—	open in general	an $L_3$ phenomenon

The first row is at the level of values: the form of the question is a membership relation between one specific number and one specific set. The question stays in $L_2$.

(The paper does not adjudicate the type of obstruction for this question. At the outline stage an inference was drawn from the proved function-field analogue to the conclusion that the obstruction is arithmetic difficulty rather than logical independence; that inference is invalid and has been withdrawn, and no wording here brings it back.)

The second row is at the level of rules, and the level of rules itself contains two distinct slots, which may not be conflated.

Slot	Question	Coordinate
Derivational completeness	Do the three Kontsevich–Zagier transformations completely generate all ordinary period expression equivalence?	rules / proof system
Decision procedure	Is there a totally terminating procedure deciding general ordinary period expression identity?	algorithmic decidability

Completeness alone does not entail decidability. That a family of rules completely characterizes an equivalence relation does not by itself supply a procedure that always terminates; a termination guarantee, or a finite certificate for inequality, is needed in addition.

For periods that additional element happens to be available: periods are computable, so for any two expressions one may run two branches in parallel — numerical evaluation to increasing precision (if the two are unequal they eventually separate, giving a certificate of inequality), and enumeration of chains of Kontsevich–Zagier transformations (if the two are equal and the conjecture holds, a chain is eventually found). One of the two branches must halt. Hence

$$\text{KZ conjecture} + \text{computability of periods} \ \Longrightarrow\ \text{general period-expression identity is decidable}.$$

The second premise must be written out explicitly: to infer decidability from completeness alone is a leap, and the paper does not make it.

The $L_2 \to L_3$ trigger is therefore not in two stages but in three:

$$\text{value question} \longrightarrow \text{complete characterization of the equivalence relation} \longrightarrow \text{decision procedure}$$

The last two both belong to the level of rules, but to different slots within it. The rule level is accordingly not a single "decidability" slot.

A precisification: one must write "the general period-expression identity problem" and not loosely "equality of periods". Restricted classes already admit algorithmic results — transcendence and the computation of linear relations for 1-periods; what is undecided is the general problem.

Two prohibitions against layer-mixing

First, independence is a property of propositions, not of objects. Even were "$e \notin \mathcal{P}$" proved independent of some formal system, what enters the rule level is the provability of that proposition in that system; the number $e$, the period expressions, and the membership question itself do not thereby change layer. Violating this destroys on the spot the separation of columns this section has established (§ 12.3, second counterfactual).

Second, the divide between regular and irregular is not a layer boundary. It is a boundary of the capture range of $L_2$, lying within the layer (§ 2.4). The Stokes phenomenon is a structural fact internal to the $L_2$ mode, not a divide between $L_2$ and $L_3$.

The form of the trigger

When the question turns from "what is this value, does this number belong to that set" into "can this class of equalities be decided by rules", articulation has left the object level of $L_2$ and entered the level of rules. That reframing is not progress within $L_2$; it is a transfer of layer.

The full formalization of the composition law for $\omega_\rho^{(2)}$ falls here as well: a unified encapsulation of the six components is a meta-level statement about how this layer's own defect apparatus composes, its full formalization requires category theory, and Paper 1 § 11.4 has already established that adequate engagement with categorical structuralism must be handled at both $L_2$ and $L_3$. § 9 delivers a minimal composition skeleton, and registers the full encapsulation as an outstanding debt (§ 13.1).

The paper does not presuppose the trajectory of $L_3$ (§ 1.3). The interface with the $L_3$ paper is § 13.2.

---

9. The Operational Defect Algebra: Minimal Typed Composition

9.0 What this section delivers and what it does not

Paper 1 § 11.1 deferred the composition law for the six components of $\omega_\rho^{(2)}$ to a subsequent $L_2$ specialist paper, and articulated that without a composition law the six-tuple is only a list, not yet enough to be called a complete operational defect algebra.

This paper does not pay the debt in full. But neither does it diagnose the failure to pay as a finding — that would violate the anti-immunization declaration of § 12.4: a debt may not be converted into an identification. The treatment here is partial payment plus explicit registration of the debt.

Delivered: a minimal typed composition skeleton, namely two tables and four chains — the first table gives the type and internal operations of each component (§ 9.1), the second gives the cross-component interfaces (§ 9.2).

Not delivered: a unified categorical encapsulation of the six kinds of data (§ 9.3).

The boundary of "partial payment" must be nailed down: what is delivered is skeleton-level — it makes the six-tuple no longer merely a list, but it does not make it a complete algebra. The word "partial" is therefore not droppable: the debt still stands, and only what is owed has narrowed, from the whole composition law to the unified categorical encapsulation (§ 13.1). A skeleton delivered and a debt still standing are not in contradiction.

The name: the name "operational defect algebra" is retained. Paper 1's abstract already designates $\omega_\rho^{(2)}$ by that name, and to demote the layer's core object in the $L_2$ specialist paper would be self-dismantling rather than refinement. But this section must state plainly: what is delivered is a minimal composition skeleton, not a complete algebra — retaining the name is not the same as the content being sufficient.

Attribution: everything in the two tables and the four chains is a classical result. What this paper does is organize (§ 9.4).

9.1 The first table: component types and internal operations

Following the definitions of the six components in Paper 1 Appendix B, each is registered with its type — the distinction of § 2.2 among numerical objects, functions, and operators is in force here — and its internal operations.

Component	Type	Internal operations
$\mathrm{Div}$	free abelian group $\mathrm{Div}(X)$	addition of divisors; principal divisors form a subgroup, with quotient the divisor class group $\mathrm{Cl}(X)$
$\mathrm{Res}$	local linear functional	addition of residue vectors; global constraint $\sum_p \mathrm{Res}_p(\omega) = 0$
$\mathrm{Per}$	bilinear pairing $H_1 \times H^1_{\rm dR} \to \mathbb{C}$	bilinearity of the pairing; values land in $\mathcal{P}$
$\mathrm{Mon}$	representation of $\pi_1$	concatenation of paths corresponds to composition of matrices
$\mathrm{Stokes}$	family of operators ordered by direction	composition of sector transitions along the circle of directions (the Stokes groupoid)
$G_\infty$	scalar asymptotic invariant (order $\rho$ and type $\sigma$)	no group operation; a partial order on order and type

On the last row. The type of $G_\infty$ must be written out, because the $G$ in the notation invites the association with a group. What Paper 1 Appendix B.5 attaches to it is the order and type of an entire function together with the Phragmén–Lindelöf principle, that is, a scalar asymptotic invariant and not a group object (§ 2.5, § 7.3). The table registers it faithfully as a scalar invariant.

This produces one unevenness of type: of the two components at stage 4, $\mathrm{Stokes}$ is a family of matrices while $G_\infty$ is a scalar, and the two do not compose in any group-theoretic sense. The three-strata analysis of § 7.2 supplies the treatment — re-register the objects of that stage by stratum:

Object	Type	Position
$\mathrm{Slope}_\infty$	family of rational numbers (the slope filtration)	§ 7.2, first stratum
exponential torus $\mathcal{T}$	algebraic torus	§ 7.2, second stratum
$\mathrm{Stokes}$	family of operators ordered by direction	§ 7.2, third stratum
$G_{\mathrm{diff,loc}}$	algebraic group	Zariski closure of the three strata (§ 9.2, chain 4)

One precise characterization of the type of the exponential torus deserves separate registration: it is the smallest subtorus whose Lie algebra contains all the coefficients of the exponential factors $Q$. It is therefore settled by the second stratum (coefficients) and not by the first (slopes). This corroborates the stratification of § 7.2 independently, from the side of differential Galois theory — the stratification is not something the paper imposed for convenience of organization; here another theory restates it in its own language.

9.2 The second table: cross-component interfaces

Interface	Map	Statement
1↔2	$d\log$	$\mathrm{Res}_p(df/f) = \mathrm{ord}_p(f)$
2↔3	restriction to small loops	$\oint_{\gamma_p}\omega = 2\pi i\,\mathrm{Res}_p(\omega)$
3↔4	(wild) Riemann–Hilbert	connections with regular singularities $\leftrightarrow$ local systems and monodromy data; irregular connections $\leftrightarrow$ monodromy together with Stokes enrichment data
4 internal	Ramis density	the subgroup generated by the formal monodromy, the exponential torus, and the Stokes operators is Zariski dense in $G_{\mathrm{diff,loc}}$

Each of the four chains has one point of precision to hold.

Chain 1 · $d\log$. The residue of a logarithmic differential equals the order given by valuation, whence

$$\deg\,\mathrm{div}(f) = \sum_p \mathrm{ord}_p(f) = \sum_p \mathrm{Res}_p\!\left(\frac{df}{f}\right) = 0.$$

That is, the vanishing of the degree of a principal divisor is a specialization of the global residue theorem to $df/f$. This interface holds only through that one map; it may not be said in general that all divisor theorems and all residue theorems are two notations for one conservation law (§ 6.3).

Chain 2 · restriction to small loops. A small-loop period is $2\pi i$ times a residue. The direction is not reversible: one may not divide by $2\pi i$ and infer that residues lie in $\mathcal{P}$, because whether $1/\pi \in \mathcal{P}$ is undecided (§ 6.1, § 6.4). Under algebraic data a residue lies in $\mathcal{P}$ for the separate reason of being algebraic, which runs through $\overline{\mathbb{Q}} \subset \mathcal{P}$.

Chain 3 · Riemann–Hilbert. The irregular side still retains monodromy; Stokes is an addition, not a replacement. The difference between classical RH and wild RH lies exactly in the Betti data being enriched by Stokes information.

And the density theorems themselves exhibit this structure: Schlesinger's theorem, in the regular-singular case, uses monodromy alone (monodromy is dense in the differential Galois group); Ramis's theorem, in the irregular case, adds the exponential torus and the Stokes operators. The core $\hookrightarrow$ enrichment relation established in § 6.0 is therefore not a reading imposed by this paper but the relation between this pair of theorems themselves.

Chain 4 · Ramis density, and its two levels. One must write Zariski dense and not "topologically generates" — the latter belongs to the parametrized version (the Kolchin topology), whereas the non-parametrized Ramis theorem gives Zariski density.

This chain in fact has two levels, and the first is a genuine composition law rather than a mere density statement:

• Formal level: the local irregular Riemann–Hilbert correspondence gives an equivalence of categories at the formal level, and the corresponding group is the semidirect product

$$G^{\rm formal} = \mathcal{T} \rtimes \overline{\langle \hat\gamma \rangle}^{\rm Zar},$$

the semidirect product of the exponential torus with the Zariski closure of the group generated by the formal monodromy $\hat\gamma$. This is a determinate composition relation between two objects internal to stage 4.

Note that the closure may not be omitted: $\mathcal{T} \rtimes \mathbb{Z}$ is the abstract generated group and $\mathbb{Z}$ is not an algebraic group; $G^{\rm formal}$ is its Zariski closure. The discipline of writing Zariski density was just established in the first lines of this chain, and the same discipline is followed here.

• Non-formal level: on passing to convergent Laurent series, the universal differential Galois group acquires additional generators, which depend on the resummation of divergent series and are tied to the Stokes phenomenon. Taking the Zariski closure gives $G_{\mathrm{diff,loc}}$.

Composition internal to stage 4 is therefore at two levels: the semidirect product at the formal level, plus the Stokes generators at the non-formal level. This two-level structure and the formal/analytic divide of § 7.2 are one and the same divide, presented in the language of groups.

9.3 What remains unpaid

What the paper does not deliver is registered explicitly.

Not delivered: a unified categorical encapsulation of the six kinds of data — that is, organizing $\mathrm{Div}$, $\mathrm{Res}$, $\mathrm{Per}$, $\mathrm{Mon}$, $\mathrm{Stokes}$, and $G_\infty$ into a single categorical object in which every interface of § 9.2 becomes a morphism.

The obstruction is not a lack of categorical tools. Full formalizations of the Riemann–Hilbert correspondence and of its irregular (wild) extension already exist. The obstruction is that the six components have not yet been mapped onto Riemann–Hilbert data; the four chains of § 9.2 are local fragments of that map, not the whole picture. In particular, chains 3 and 4 have already entered the language of Riemann–Hilbert, while chains 1 and 2 remain in the language of classical complex analysis, and this paper gives no translation between the two.

The paper does not diagnose the failure to deliver as a finding. § 8.4 has already articulated that such an encapsulation is meta-level in character — it is a statement about how this layer's defect apparatus composes, and by the trigger criterion of Paper 1 § 5.5 statements of that kind belong to the level of rules. But two things must be kept apart: "the encapsulation is meta-level in character" is a description of that object, and "this paper cannot supply it" is another matter; the former is no excuse for the latter. By § 12.4, a debt is registered as a debt.

Registration: outstanding debt, pushed forward to the $L_2 \to L_3$ interface paper (§ 13.1).

9.4 Attribution declaration

Everything in the two tables and the four chains is a classical result: the divisor group and the divisor class group, the residue theorem, the de Rham–Betti pairing, monodromy representations, the sector composition of Stokes data, the residue of a logarithmic differential, the small-loop integral formula, the Riemann–Hilbert correspondence and its wild extension, and the Ramis density theorem. This paper cites these results and does not prove them.

What this paper does is organize: it places the six components at determinate positions, following the three strata of § 7.2 and the core–enrichment relation of § 6.0, gives the type of each component, and marks which interfaces exist and which do not.

The paper does not claim to have discovered these interfaces, nor that this manner of organization is the only one. The organization itself is an SAE registration, and its value lies in making the unpaid part visible — in the six-tuple-as-list form of Paper 1, what is missing cannot be seen.

〔Kind: standard theorem (the content of the tables and chains) plus SAE registration (the manner of organization)〕

---

10. Side Branch and Alternative Closure: Cayley–Dickson

10.1 The adjudication

Paper 1 § 11.8 left an open question: whether the Cayley–Dickson sequence supplies an alternative closure at $L_2$, and how the sufficiency criterion applies to its members. This section answers it.

Adjudication: $\mathbb{H}$ and $\mathbb{O}$ are not equivalent parallel realizations of this paper's commutative complex-period trajectory.

The criterion is the within-layer articulative unit of the exchange law. The within-layer articulative unit of the exchange law of this paper's complexification trajectory is $2\pi i$, and the concrete form of that commutation is the residue theorem and the argument principle (§ 6.3); both require commutativity. What $\mathbb{H}$ pays away is precisely commutativity.

(This criterion involves the within-layer slot only, not the transition signature slot. It is therefore unaffected by Foundation v2's assignment of $e$ as the transition signature of the mathematical $L_1 \to L_2$; see § 11.5.3.)

By one of the second-level nesting rules established in Physics Paper 1 — the second level issues no new signature, sharing the first level's — whatever pays that unit away cannot count as a parallel realization of the same layer.

The force of this criterion comes from § 5.5. If $2\pi i$ were merely a factor in the residue theorem, then "paying away commutativity is paying it away" would be an observation about one formula. But § 5.5 has articulated that $2\pi i$ is at the same time a generator of the deck transformation group of $\exp$, and that this and the normalization of residues are two presentations of one statement. What paying away commutativity pays away is therefore the presentation, in the language of integration, of the covering structure of the whole layer, and not one integral formula.

10.2 The conclusion must be narrowed

The adjudication of this section cannot be pushed far. Two points must be made explicit.

First, the vertical-iteration argument is withdrawn. One natural argument runs: Cayley–Dickson is a vertical iteration, $\mathbb{H} = \mathrm{CD}(\mathbb{C})$ takes $\mathbb{C}$ as input, so $\mathbb{H}$ was never a sibling standing at the $L_1 \to L_2$ threshold. That argument does not hold — because $\mathbb{C} = \mathrm{CD}(\mathbb{R})$ holds equally. If vertical iteration meant "the next level up", the argument would yield "$\mathbb{H}$ is $L_3$" rather than "$\mathbb{H}$ is a side branch". It proves too much. The paper does not use it.

Second, quaternionic analysis does not lose all Cauchy-type structure. Slice-regular theory recovers structure of complex-analytic type on each complex slice

$$\mathbb{C}_J = \{x + Jy : x, y \in \mathbb{R}\}, \qquad J \in \mathbb{S}^2,$$

with a corresponding Cauchy kernel and integral formula. The change is therefore not that the signature vanishes altogether, but that a single global $i$ is replaced by a family of imaginary units $J \in \mathbb{S}^2$.

What this section can conclude is therefore only:

> $\mathbb{H}$ does not share this paper's single $2\pi i$ signature, and does not share the same six-component defect system.

It may not be concluded from this that $\mathbb{H}$ is expelled from $L_2$ as a whole. That would require a judgment about the entire space of $L_2$ trajectories, and this paper has studied one of them. By the anti-essentialization discipline of Paper 2, using the signature of the current trajectory to exclude other trajectories conflicts directly with the position that complexification is only one articulation of $L_2$.

〔Kind: not to be drawn as a conclusion of this paper — see § 12.1〕

10.3 Cost branch and remainder branch

This section establishes a further classificatory criterion, to deal with a problem that would otherwise arise.

If $\mathbb{H}$ is something "after $L_2$", and the level of rules (§ 8.4) is also something "after $L_2$", then the paper has two competing successor candidates while the prediction of § 13.2 speaks of only one.

Criterion: distinguish branches by their source.

• Cost branch: obtained by paying away a signature. $\mathbb{H}$ and $\mathbb{O}$ belong here. They are not a treatment of any remainder of this layer; they are what is bought by relinquishing one of this layer's structural constraints.
• Remainder branch: obtained by treating a remainder of this layer. The exponential direction (§ 7.5) and the level of rules (§ 8.4) belong here.

The two have different sources and therefore do not compete for the same position. $L_3$ arrives along the remainder branch; Cayley–Dickson departs along the cost branch. They do not lie on one lineage, and the problem of "two successor candidates" does not arise.

This criterion and the articulation of the first class of remainder in § 8.2 support each other. § 8.2 articulates that $i$ is a substrate-external remainder pressed out by an interrogation internal to $L_1$ and therefore forces a crossing of levels, whereas $\mathbb{C}$ is algebraically closed and no interrogation internal to $L_2$ presses $\mathbb{H}$ out. In the language of this section: $L_2$ has no forced cost branch, only a cost branch that may be applied.

〔Kind: an SAE T3 classificatory proposal, not a classical mathematical conclusion〕

10.4 Revision of an upstream candidate classification

Paper 1 § 5.1 lists the members of the Cayley–Dickson sequence as candidates for the articulation of $L_2$. This paper reclassifies them as candidates for a lateral trajectory.

It should be said that this is not merely the payment of the open question of Paper 1 § 11.8 but the revision of an upstream candidate classification. The first is answering a question already posed; the second is changing a classification already made. The paper registers it as the second (§ 11.1, second upstream registration).

Direct response to Paper 1 § 11.8: the alternative closures do not support the same $\omega_\rho^{(2)}$ — the signature differs, and the types of the six components differ with it. But "does not support the same one" is not "does not support any defect system": slice-regular theory has its own structure, and what its defect apparatus is has not been studied here and is not prejudged.

---

11. Cross-Paper and Cross-Series Relations

11.1 Four upstream treatments applied to Paper 1

This paper applies four upstream treatments to Paper 1: three refinements or charitable reconstructions, and one revision of a candidate classification. By the discipline of this series, a refinement is written out and never made silently; and each must state what changes and what does not.

#	Paper 1 as articulated	Treatment here	Location	Kind
1	§ 5.2: the identification of $\pm\infty$ is forced by topology	one-point compactification is the canonical closure chosen in the meromorphic and conformal context; the cost of that choice is directional resolution, falling due at the irregular boundary	§ 4.3, § 8.1	refinement
2	§ 5.1, § 11.8: the members of the Cayley–Dickson sequence are listed as candidates for the articulation of $L_2$	reclassified as candidates for a lateral trajectory; and § 11.8 is answered: the alternative closures do not support the same $\omega_\rho^{(2)}$	§ 10.4	revision of a candidate classification
3	§ 5.3: Euler's formula is best understood as a closure instantiation and not the full closure of $L_2$	precise content: it is the closure equation of the transition, not of the layer; and separately a substrate–mode superposition placement	§ 8.3	refinement
4	Appendix B.5: $G_\infty$ as the order and type of an entire function with Phragmén–Lindelöf	Newton–Turrittin data supply a candidate comparison interface for it; identity not proved	§ 7.3	charitable reconstruction

First: the identification itself is unchanged — $\pm\infty$ are indeed merged into $\hat\infty$; what changes is the modality of that identification (forced → chosen within a context) and the visibility of its cost.

Second: this is not merely the payment of a question already posed but the changing of a classification already made. It is registered as the latter.

Third: Paper 1's judgment in that sentence is unchanged — Euler's formula is indeed not the complete closure of $L_2$; what changes is that an account is now given of what it is.

The fourth must be marked specially as a charitable reconstruction. What Paper 1 attached to $G_\infty$ when it wrote the symbol was a scalar asymptotic invariant; to say that the notation anticipated a group and that this paper completes it is this paper's reconstruction and not Paper 1's original intent. The paper retains the symbol $G_\infty$ and its original object, and introduces $\mathrm{Slope}_\infty$ and $G_{\mathrm{diff,loc}}$ separately (§ 2.5), rather than putting a new object under an old symbol.

Beyond these, the paper discharges part of the composition-law debt left by Paper 1 § 11.1 (§ 9), registering the unpaid part explicitly as an outstanding debt (§ 13.1); and § 6.3 supplies the concrete form of the $L_2$ exchange law established in Paper 1 § 4.5, namely the residue theorem and the argument principle.

11.2 Relation to Paper 2

Paper 2 establishes that $L_1$ is not a single foundational level but a product of dimension and path, and establishes the anti-essentialization discipline.

The $L_2$ of this paper is likewise one trajectory, not the only $L_2$. That discipline is executed concretely at three places here:

• the title does not say "the complexification layer $L_2$" — complexification is one articulation of $L_2$, not the definition of $L_2$;
• the registration consequence of § 5.4 is not written as a structural exclusion — within the current trajectory the paper adopts the internal-product reading and the transition reading is no longer its primary registration, but the paper passes no judgment on the standing of the transition reading in other trajectories;
• the conclusion of § 10.2 is narrowed — only that $\mathbb{H}$ does not share the signature and defect system of this paper's trajectory, and not that it is expelled from $L_2$ as a whole.

On Paper 2's three closure paths: this paper follows the complete precisification path (Paper 2's English renders this complete exactification). The effective-Cauchy articulation mode is a mode on another path; its placement is given in § 11.3.

11.3 Relation to Paper 3

Format inherited: four Main Steps each containing four inner steps, following Paper 3's articulation format.

But the registration class of Main Step 4 differs — by the three-way classification established in Paper 3 § 1.3 and § 6.1, Main Step 4 is a meta-level closure whose sub-articulation is a remainder ledger structure and does not follow the four-step fractal pattern. This paper follows that registration and carries it in the layout asymmetry of Main Step 4 spanning § 6, § 7, and § 8 (§ 2.1).

The distinction between substrate and articulation mode: inherited from Paper 3 § 2 and lifted to $L_2$ (§ 2.1). The paper further articulates that this distinction undergoes a change of type between the two layers — carrier migration (§ 6.1).

The three classes of the remainder ledger: inherited from Paper 3 § 6 and rearranged at $L_2$ (§ 8.2). Of these the first class, substrate-external, stands differently in the two layers: $i$ was pressed out by an interrogation internal to $L_1$, while $\mathbb{H}$ has no corresponding source.

Sub-fractal depth asymmetry: Paper 3 § 5.3 articulates that the multiplicative path of $L_1$ is one degree deeper than the additive path; § 5.6 here articulates that $L_2$ reproduces the same asymmetry at the same cell, while what lies at the depth differs between the layers.

The third nesting rule under stress test: Physics Paper 1 establishes that the sub-step's d product is the parent layer's remainder. The $L_2$ of this paper satisfies it — the exponential direction produced by the final stage of Main Step 4 (boundary-sector, § 7) is exactly the closure remainder of the whole layer (§ 7.5). The satisfaction is automatic; nothing was arranged for it.

Use of the first nesting rule: the second level issues no new signature, sharing the first level's. The paper uses it at § 10.1.

The placement of the effective-Cauchy articulation mode

Paper 3 § 6.4 registers the effective-Cauchy articulation mode as a post-$L_1$ analytic-completion branch, and explicitly does not presuppose whether it functions as a sub-articulation within the $L_2$ main trajectory, as an $L_{1\to 2}$ bridge, or as a side-branch parallel to the complexification trajectory; that adjudication was left to subsequent layer papers. This paper is such a paper.

The material assembled here (§ 8.2):

First, a comparison of capture ranges. The effective-Cauchy mode uses algorithmic approximability as its criterion and captures the computable reals, while $\mathcal{P} \cap \mathbb{R}$ is strictly contained in the computable reals. It therefore captures strictly more on the real axis than the period core of $L_2$.

Second, the type of the criteria. The three criteria — finite algebraic constraint, algebraic integral representation, algorithmic approximability — are none of them a special case of another.

Third, the remainders they respond to differ. Complexification responds to the substrate-external remainder $i$ of $L_1$ (§ 3.2); the effective-Cauchy mode responds to the completeness remainder of $L_1$ — the finite algebraic mode does not articulate the convergence of Cauchy sequences.

The adjudication: relative to this paper's complexification trajectory, the effective-Cauchy articulation mode is registered as a parallel side-branch responding to a different remainder of $L_1$. The reasons follow.

• It is not a sub-articulation internal to the ordinary-period capture mode studied here: were it so, its capture range would be contained in $\mathcal{P} \cap \mathbb{R}$, whereas it captures strictly more on the real axis. Note that this excludes only its falling within that capture mode, and does not exclude its falling within some other $L_2$ sub-articulation — the paper states repeatedly that $\mathcal{P}$ is an evaluation core and does not exhaust $L_2$ (§ 2.6, second item), so one may not infer from "more than the core" to "not within any $L_2$".
• It is not a necessary bridge for the complexification route constructed here: the $L_1 \to L_2$ route articulated in § 3 through § 5 — mark-without-construct of $i$ → two-dimensionalization → covering — does not pass through it. This excludes only its being a necessary bridge for this route, and does not exclude its being a bridge for some other $L_1 \to L_2$ trajectory.
• It is a parallel side-branch: by the criterion of § 10.3, complexification and the effective-Cauchy mode are both remainder branches — both obtained by treating some remainder of $L_1$ — but the remainders they treat differ. They are two responses at the same layer to different remainders, and are therefore parallel and not competing.

On this basis the open question left by Paper 3 § 6.4 is closed here.

The scope of the adjudication must be limited. What the paper adjudicates is the relation between the effective-Cauchy mode and this paper's complexification trajectory, not its position within the whole space of $L_2$ trajectories — the same limit observed for Cayley–Dickson at § 10.2. And the adjudication is an SAE registration, not a mathematical conclusion: the inclusion relations among the three capture ranges are standard results, while adjudicating a branching relation among articulation modes from those inclusions is an identification made by this paper within the SAE framework.

〔Kind: SAE registration, resting on T2 facts (the inclusion relations among capture ranges) plus the classificatory criterion of § 10.3〕

11.4 Relation to the ZFCρ line

The question the ZFCρ line raises is that the monolithic $\infty$ of $L_1$ cannot resolve fine ratios.

This paper's treatment falls into two halves with different statuses:

Half	Location	Status
Order	§ 6.2	settled by the valuation $\mathrm{ord}_p$. $\hat\infty$ acquires a $\mathbb{Z}$-valued grading; the order of $\infty/\infty$ becomes a subtraction of integers
Coefficient (fine ratio)	§ 7.4	not settled. The interface is located at the second stratum of the Newton polygon (the leading coefficients of the exponential factors), but is not connected

Two points to hold:

First, what interfaces with ZFCρ is the valuation, not the divisor — the divisor merely packages the valuations at the several points (§ 6.2). To say that the divisor equips $\infty$ with a scale is to mistake an act of packaging for an act of grading.

Second, an interface located is not an interface connected. The paper gives no explicit map, conserved quantity, or functor between the fine ratio and any coefficient of any characteristic equation. The status of the debt is re-registered from "no interface" to "interface located but not connected", and may not be cited as settled (§ 2.6, sixth item; § 12.1).

Papers on the ZFCρ line citing the present paper should cite the status column of this table, and should not cite it as the debt having been resolved.

11.5 Relation to SAE Physics Foundation v2

11.5.1 What v2 already states

Foundation v2 is a systematic restatement on the physics side, but it contains several statements about the mathematical ladder that bear directly on this paper:

• § 2.1, the parallel-ladder comparison table: the mathematical $L_1 \to L_2$ is dualization closure, with closure equation $e^{i\pi}+1=0$, remainders $i$ and $\pi$, and signature $e$;
• § 7.3: mathematical $e$ is the scalar-level closure signature, mathematical $\pi$ the scalar-level closure remainder, and physical $\hbar$ the operator-level closure signature;
• § 7.2: what mathematical $L_2$ closes is a number field, with a pure-number substrate; what physical $L_2$ closes is conjugate readout, carrying one further level of state dependence;
• § 1.2: the two ladders run parallel, isomorphic in structure and independent in content, assisting each other through cross-ladder ontological correspondence — for example, $i$ arises natively on each ladder, and physics does not borrow the symbol from mathematics.

11.5.2 Three points of agreement

The remainders $i$ and $\pi$: § 8.3 here agrees with v2's assignment. This paper additionally supplies why these two in particular — $i$ is the input to the transition (§ 3), $\pi$ is the first structural product of this layer (the deck cell of § 5.3, the period capture of § 6.4). v2 gives no such reason.

$i$ is native on each ladder: § 3.2 articulates that $i$ is pressed out by $L_1$'s own reverse interrogation, polynomial root-finding, and is the substrate-external remainder of $L_1$. This agrees with v2's position that it is native on each ladder, and supplies a concrete articulation of that nativity from the mathematical side.

The substrate of mathematical $L_2$ is the pure-number level: in agreement with § 2.1 here. But substrate and articulation mode must be kept apart — what v2 speaks of is the substrate ($\mathbb{C}$ closing a number field), whereas the carrier migration of § 6.1 speaks of the capture core of the mode ($\mathcal{P}$ is an evaluation image, not a scalar field). The two do not conflict; a reader taking the two papers together without drawing this distinction would suppose that they had adjudicated the same object in different terms.

11.5.3 Two signature slots

The signature v2 assigns to the mathematical $L_1 \to L_2$ is $e$; the articulative unit Paper 1 § 4.5 assigns to $L_2$'s exchange law is $2\pi i$. This paper establishes: these are two different slots, not two candidates for one slot.

Slot	Content	What it signs	Location
Transition signature	$e$	the crossing itself: binding the two remainders in $e^{i\pi}+1=0$	§ 8.3
Within-layer articulative unit of the exchange law	$2\pi i$	the commutability of two articulation routes within this layer: the residue theorem and the argument principle	§ 6.3

One structural consequence deserves registration: $e$ is both the transition signature and the numerical witness of this layer's closure remainder (§ 7.5). That is — the item that signs this crossing lies determinately within the exponential period extension, while its membership status in this layer's ordinary period core remains undetermined.

(The wording "the item this layer cannot capture" is deliberately avoided: at the numerical level that would be equivalent to asserting $e \notin \mathcal{P}$, which is undecided. That it lies determinately within the extension is a theorem; that it lies outside the core is a conjecture; the structural tension is carried by the former.)

This double identity is of the same form as the double identity of $2\pi i$ articulated in § 5.5, but runs in a different direction: the double identity of $2\pi i$ is one object at two places within the layer (the deck cell and the normalization of residues); the double identity of $e$ is one object at the transition and within the layer.

Effect on § 10: the Cayley–Dickson criterion of § 10 uses the within-layer articulative unit $2\pi i$, not the transition signature $e$. The criterion is therefore unaffected by v2's assignment of the signature.

11.5.4 This paper's rearrangement of v2's mathematics row

In articulating symplectic structuring, v2 § 4 uses "the mathematical $L_1 \to L_2$ exponential map" as an act type. § 5.3 and § 5.4 here articulate $\exp$ as an internal product of $L_2$ — the covering map produced by Main Step 3 inner step 4 — and therefore restate the cross-level act as complexification (mark-without-construct of $i$, § 3).

Registration: the principal work of Foundation v2 lies on the physics side, and the rows of the mathematical ladder appear there as a parallel comparison rather than as specialist articulation. The present paper, as the $L_2$ specialist paper of the mathematics series, rearranges that row within the jurisdiction of the mathematics series. It is recommended that the rearrangement be absorbed into Foundation v3.

The paper takes the modality of explicit revision, not that of a compatibility reading. That is: Paper 4 revises the placement of the act type in the mathematics row of Foundation v2, reclassifying $\exp$ from a cross-level act type to an internal product of $L_2$. The revision affects no argument on the physics side of Foundation v2, and it is recommended that Foundation v3 absorb it.

The reading not taken — that v2 speaks of a closure relation at coarse resolution while this paper speaks of the address of generation, so that the two are compatible — would indeed let both texts stand, but it is a reinterpretation made by this paper on v2's behalf, and v2's own text draws no such distinction. By the discipline of this series, reinterpreting upstream on its behalf is less transparent than revising explicitly.

11.5.5 The vocabularies are not unified, and this is deliberate

v2 § 6.3 assigns the closure of mathematical $L_2$ a pair of terms: the bare-readout form ($e^{i\pi}+1=0$) and the invariant form ($e^{ix} = \cos x + i\sin x$, along with the Cauchy–Riemann equations, the spectral theorem, and so on).

This paper does not adopt that pair, and uses the mathematics series' own vocabulary instead: placement, transition closure equation, superposition placement.

The reason comes from v2 itself: v2 § 6.4 explicitly confines the universality of the double-articulation pattern to the articulation scope of the physical-quantity ladder, and does not extend it in advance into a universal theorem over all possible ladders; and § 1.2 establishes that the two ladders are independent in content. Articulating the mathematical ladder by the mathematics series' own organization is therefore following v2's limitation rather than departing from it.

But a comparison must be given, or a reader taking the two together will suppose them to be two unrelated descriptions:

v2 (physical-quantity ladder vocabulary)	This paper (mathematics series vocabulary)	Same object?
bare-readout form $e^{i\pi}+1=0$	transition closure equation (§ 8.3)	yes
invariant form $e^{ix} = \cos x + i\sin x$	the identification of a substrate-level classical construction with a mode-level product (§ 5.5)	yes
—	superposition placement (§ 8.3)	a third predicate added by this paper

One convergence deserves registration: § 5.5 here derives the structural standing of $\cos\theta + i\sin\theta = e^{i\theta}$ from the equality of a substrate-level classical construction with a mode-level product (the covering map of inner step 4 restricted to the imaginary axis); v2 § 6.3 independently assigns the same object to the invariant form. Two routes point to one place. The convergence is not repetition but mutual corroboration — this paper arrives from the internal structure of the articulation process, v2 from a cross-ladder formal comparison.

11.5.6 What is not welded

$\hbar = h/2\pi$ and $[\hat x, \hat p] = i\hbar$ both contain $2\pi$ and $i$, and are formally close to the $2\pi i$ of mathematical $L_2$. This is a tempting weld.

This paper does not weld.

v2 § 7.3 has already given the reason why $\hbar$ is not a copy of mathematical $e$ or $\pi$: mathematics is scalar-level closure, physics is operator-level closure, and the latter carries one further level of state dependence than the former. This paper accepts that distinction and welds no further.

The reason: welding would require an argument of cross-ladder ontological correspondence. v2 § 1.2 supplies such an argument for $i$ alone — $i$ as a universal feature of closure structure arising natively on each ladder — and supplies none for $2\pi$. This paper does not supply one on its behalf.

〔Registered as a T3 interface, § 12.1〕

11.5.7 Mathematical $L_2 \to L_3$

v2 § 2.1 assigns the mathematical $L_2 \to L_3$ to diagonalization, with signature $\Omega_U$. This paper does not treat that row (§ 12.2, jurisdiction).

One item is registered as pending: whether the migration prediction of § 13.2 is compatible with that assignment is work for Paper 5 ($L_3$). This paper claims neither compatibility nor incompatibility.

---

12. Claim Tiers, Jurisdiction, Falsifiability, and Prediction Discipline

12.1 The three-tier table

By the three tiers established in Physics Paper 1 — T2 lockable, T3 open, and not to be drawn as a conclusion of this paper — the tiers of the paper's principal statements are as follows.

Statement	Tier	Ground / location
$\mathsf{Int}_{\rm alg} \hookrightarrow \mathsf{Int}_{\exp}$	definitional T2	the definitions of the ordinary and exponential period expression classes; § 7.1
regular connections extend to connections admitting irregular singularities	theorem-level T2	Levelt–Turrittin and related formal classification; § 7.1
all slopes zero $\Leftrightarrow$ regular singularity	T2	standard slope theory; § 7.2, first stratum
slopes are the degrees of the determining polynomials	T2	as above
type of growth depends on $\lvert a_s\rvert$, not on slopes	T2	§ 7.2, second stratum
candidate Stokes directions depend on the leading coefficient and argument of the difference $q_\alpha - q_\beta$, not on slopes	T2	§ 7.2, second stratum
Stokes multipliers are analytic moduli beyond formal classification	T2	Hukuhara–Turrittin; § 7.2, third stratum
the four chains of § 9.2	T2	each is a classical result; § 9.4, attribution declaration
$\overline{\mathbb{Q}} \subset \mathcal{P}$; $\mathcal{P}$ is countable; a computable non-period real exists	T2	standard results and Yoshinaga; § 8.2
$e \in \mathcal{P}^e$	T2	definition and computation; § 7.5
the alignment "regular $\leftrightarrow \mathcal{P}$, irregular $\leftrightarrow \mathcal{P}^e$"	T2-conditional	a research-level correspondence on the exponential-motives line; § 7.1
$e \in \mathcal{P}$	T3	open conjecture; serves as candidate strictness witness
decidability of the general period-expression identity problem	T3	open; § 8.4
the comparison interface between $G_\infty$ and $\mathrm{Slope}_\infty$	T3	candidate interface, identity not proved; § 7.3
carrier migration	SAE registration	§ 6.1
$\mathcal{P}$ as the principal capture range of $L_2$	SAE registration	§ 6.4; exhaustion not claimed
the asymmetry of Main Step 3's reverse interrogation	SAE registration (resting on T2 facts)	§ 5.4
the cost-branch / remainder-branch criterion	an SAE T3 classificatory proposal	§ 10.3
the migration prediction for $L_3$	T3 programmatic prediction	§ 13.2
the ZFCρ monolithic-$\infty$ debt is settled	not to be drawn as a conclusion of this paper	only the order half is settled; § 6.2, § 7.4
$\mathbb{H}$ is expelled from $L_2$ as a whole	not to be drawn as a conclusion of this paper	§ 10.2
this paper derives complex analysis	not to be drawn as a conclusion of this paper	§ 2.6, first item
this paper proves or refutes $e \notin \mathcal{P}$	not to be drawn as a conclusion of this paper	§ 2.6, third item
the union of all mathematical tools fails to exhaust the substrate	not to be drawn as a conclusion of this paper	§ 8.2, jurisdiction statement

12.2 Jurisdiction statement

Within jurisdiction: the capture range and remainder ledger of the $L_2$ articulation mode; the sub-articulation of the four Main Steps and their registration classes; the types of the six components of $\omega_\rho^{(2)}$ and the interfaces that exist among them; and the relations of the foregoing to Paper 1, Paper 2, Paper 3, and Foundation v1.

Outside jurisdiction: attribution and priority for specific results in the arithmetic of periods; attribution for the technical theorems of complex analysis; the full composition law for $\omega_\rho^{(2)}$ (§ 9.3, § 13.1); the trajectory of $L_3$ (§ 1.3); drawing the boundary of what counts as a "mathematical tool" (§ 8.2); a mathematical placement coordinate system (§ 13.3); the defect structure of quaternionic analysis itself (§ 10.4).

12.3 Falsifiability: a two-way registration

Counterfactual one: should $e$ be proved to be a period ($e \in \mathcal{P}$), the only items requiring structural rewriting are the witness role of $e$ in § 7.5 and its classification in the ledger of § 8.2; the statements of fact throughout the paper recording that proposition as unknown or open (the abstract, § 1.3, § 2.6 third item, the three-tier table of § 12.1, and so on) require mechanical synchronization. The load-bearing structure does not move: the two extensions of § 7.1, the three strata of § 7.2, the core structural statement of § 5.4, and the two tables and four chains of § 9.

This item holds on the premise that the paper's load has been moved off the constant $e$ and onto $\exp$ and the extension of expression classes (§ 7.1). Were the load still pressing on $e \notin \mathcal{P}$, the counterfactual would not hold — what then required change would include the whole articulation of remainder.

Counterfactual two: should "$e \notin \mathcal{P}$" be proved independent of some formal system $F$, what requires updating is the proof-theoretic status of that proposition in $F$ (the second of the three coordinates of § 2.4). That update does not automatically settle the algorithmic status of the general period-expression identity problem, and does not change the object type of the number $e$ — independence is a property of propositions, not of objects (§ 8.4, first prohibition against layer-mixing). The $L_2$/$L_3$ separation of columns is therefore unaffected.

Why the second must be registered alongside the first. In a paper that makes an anti-immunization declaration, a falsifiability section registering only counterfactuals harmless to itself turns that declaration into decoration. The drafting history of this item is recorded in Appendix A, "Version and Review Provenance Record".

12.4 Anti-immunization declaration

Following Paper 1 Appendix A.8.5: one may not routinely reclassify every counter-example as merely an update opportunity — that would convert the schema into an immunization clause and destroy its falsifiability.

The paper has two places where this discipline must be executed.

First, the long-standing failure to refute $e \notin \mathcal{P}$ is recorded as consistency only, not as support. That the proposition has not been refuted since Kontsevich and Zagier posed it is a fact consistent with the framework here; but it does not constitute evidence in support of the framework. The argument that "had it been a period, a representation would have been found by now" is a plausibility argument, belongs in the ontological-boundary column (§ 7.5), and does not enter the load.

Second, the debt on the composition law is registered as a debt and not converted into a finding. That the full composition law has not been given is not by itself any support for the framework — it is a debt, registered as an outstanding debt per § 9.3.

In particular, that debt may not be restated as an identification on the ground that the encapsulation is meta-level in character (§ 8.4): the properties of an object and whether this paper supplies it are two different matters.

The paper has two records of executing this discipline during drafting, one of them a self-correction; see Appendix A, "Version and Review Provenance Record". The credibility of an anti-immunization discipline depends on its being executed inside the paper as well, and not only invoked when judging others; the record is therefore not omitted, only placed elsewhere than in this section.

12.5 Prediction discipline

This series issues predictions. Predictions may be wrong. When falsified, they are conceded.

This stance and the anti-immunization declaration of § 12.4 are two sides of one thing: the cost of not being immune is that one may be wrong; and refusing to issue predictions is itself a form of immunity — a framework that never makes a testable statement is equally beyond refutation.

Paper 1 set the precedent: it writes out that the exchange-law pattern is a non-trivial cross-layer prediction of the schema, and lists the outcome of testing it layer by layer.

> 〔Revised in v2〕 Version 1 continued: "The prediction of § 13.2 here is registered in the same way: the prediction is stated, an operationalized falsification condition is given, and the known counter-candidates are identified." The premise of that sentence has lapsed — the falsification condition of § 13.2 is withdrawn (its adjudicative reading is empty), and the first row of the counter-candidate table is withdrawn with it. § 13.2 is now registered as a T3 explanatory proposal without an operationalized falsification condition. > > Also: Paper 1's own exchange-law prediction has been reduced in its v2 — after the two-ladder split, only three items remain on the mathematical side, the third undetermined, and the three are the construction set rather than a test set. The "precedent" invoked here must therefore be read together with that reduction.

Chronological credential. This paper is published before the $L_3$ paper. The version-level DOI of this signed-off version, together with its archive page and public record, constitutes a third-party verifiable credential of public priority: it establishes that § 13.2 entered the public record before the $L_3$ paper was published. It does not establish the order of drafting in the author's private workspace, and it does not constitute an independent test of the prediction.

Citation and version discipline: the version-level DOI of this signed-off version should be cited, and not the concept DOI, whose target moves with each new version. Should this paper later be modified on account of the $L_3$ paper, the modification should be published as a new version and should not overwrite the present one; the present version carries the public priority of § 13.2.

An author-supplied date is deliberately not made load-bearing here: a self-supplied date is a self-declaration and can establish nothing to an outside reader; were the load placed on it, the prediction discipline would rest on self-report, forming a double standard with the load discipline this paper observes for $e \notin \mathcal{P}$ (§ 7.1). What bears the load is the publication record itself, not any statement of this paper.

But two limits must be registered at the same time, and neither has anything to do with chronology, so neither is lifted by publishing first.

First, the subsequent $L_3$ paper does not constitute an independent test of § 13.2. A sequel written by the same author within the same framework is not a test of the prediction, whatever it turns out to say. An independent test must come from mathematics outside. This limit is unaffected by the downgrade.

> 〔Revised in v2〕 Version 1 continued: "Under the operationalized condition given in § 13.2, what can actually trigger falsification is an extension of objects that is conservative over consequences in the old language, and the instance now registered is the conservative class-theoretic extension; large cardinal extensions and forcing extensions remain legitimate objects of stress testing, but under the present falsification condition they do not constitute a path to falsification." > > That passage is withdrawn in full: (i) a conservative object extension cannot adjudicate any sentence undecidable in the old language (§ 13.2 (a)); (ii) once the closure target is made precise, the condition is empty for every effectively axiomatized extension, conservative or not (§ 13.2 (b)). The two-column division into "can trigger" and "cannot trigger" therefore does not stand. Large cardinal and forcing extensions remain legitimate objects of stress testing — that clause is retained, but no longer set against a "path to falsification".

Second, the paper may therefore claim that the prediction is prior, and may not claim that the prediction has been tested. The difference is not small: the first is warranted by the publication record, the second requires a tester independent of this framework. Should a subsequent $L_3$ paper agree with the prediction, this paper may not on that basis claim the prediction confirmed.

> 〔Revised in v2〕 Version 1 closed: "The testability of § 13.2 is accordingly carried mainly by its operationalized falsification condition; the publication record only guarantees that it is a prediction and not a backfill, and does not guarantee that it has been tested." > > The bearer named in the first half has lapsed: that condition is withdrawn, § 13.2 now has no operationalized falsification condition, and its testability accordingly has no bearer at present. The second half stands: the publication record only guarantees that it is a proposal and not a backfill.

One case not previously provided for must therefore be added 〔addition〕.

Version 1 writes: this series issues predictions; predictions may be wrong; when falsified, they are conceded. But what has occurred here is not falsification — the prediction was not overturned by any counterexample; its falsification condition was found to be empty. That case is unprovided for. Proposed addition:

> If the falsification condition of a prediction is subsequently shown to be structurally unsatisfiable, that condition lapses, and the prediction must be downgraded to an explanatory proposal or rebuilt; it may not be retained on the ground that it has not yet been falsified.

The reason: a discipline that provides only for "own up when wrong" is silent in the face of "the condition could never have been triggered" — and the latter is the more concealed form of immunisation. The addition therefore strengthens this section rather than weakening it.

The word "structurally" has a theorem behind it: by § 13.2 (b), once the closure target is made precise as $C_T^\vdash=C_{\mathbb{N}}^\vDash$, the condition is empty for every effectively axiomatized extension. This is not a combinatorial observation but a direct consequence of the obstruction-to-full-closure theorem.

The state of execution after revision: the anti-immunization discipline of § 12.4 is now carried in this section by three things — (i) the prediction is downgraded rather than quietly retained; (ii) the grounds and course of the downgrade are registered item by item (the v2 revision of § 13.2); (iii) the addition above brings the case of "an empty condition" within the discipline. It is no longer carried by "§ 13.2 has an operationalized falsification condition".

---

13. Open Problems and the Subsequent Trajectory

13.1 The full composition law for $\omega_\rho^{(2)}$

Status: outstanding debt.

Against the six-component composition law left open by Paper 1 § 11.1, this paper delivers a minimal typed composition skeleton (§ 9.1, § 9.2) and does not deliver a unified categorical encapsulation of the six kinds of data. The obstruction is not a lack of categorical tools but that the six components have not been fully mapped onto Riemann–Hilbert data — chains 3 and 4 of § 9.2 have entered that language while chains 1 and 2 remain in the language of classical complex analysis, and the translation between them is not given here (§ 9.3).

Pushed forward to: the $L_2 \to L_3$ interface paper. The encapsulation is meta-level in character (§ 8.4), and by Paper 1 § 11.4 adequate engagement with categorical structuralism must be handled at both $L_2$ and $L_3$.

13.2 The migration prediction for $L_3$

Prediction: within the complete precisification trajectory selected here, once the enlargement of the object domain is exhausted, the pressure of reverse interrogation passes into the space of representations and rules.

Ground: the asymmetry of § 5.4 ($L_1$ enlarges the scalar substrate; $L_2$, being algebraically closed, turns instead to covering) and the form of the trigger in § 8.4 (the level of values turning into the level of rules). Both display the same direction: when the pressure is blocked on the object side, the mode of release shifts toward representations and rules.

Falsification condition 〔withdrawn in v2; see the end of this section〕:

> If the principal closing act of $L_3$ can be accomplished solely by adding new kinds of object to the language of the theory of objects, without any change to the rules of equivalence, proof, or decidability, then the asymmetric migration from object domain to rule domain proposed here is falsified.

Version 1 held that formulating it this way gives both "enlargement of the object domain" and "release of the pressure" an operable standard of adjudication. That judgement does not stand — the condition never defines what "the principal closing act of $L_3$" refers to, and so describes the extension mechanism in detail while supplying no closure target. See the v2 revision at the end of this section.

Version 1 registered counter-candidates in two columns according to whether they can trigger the falsification condition. The first row of that table is now withdrawn:

Candidate	v1 verdict	v2 verdict
conservative class-theoretic extension (e.g. NBG)	can trigger	cannot trigger — see (a) below
non-conservative class-theoretic extension (e.g. MK)	no	no (but the reason must change, see (b))
large cardinal extensions	no	no (but the reason must change, see (b))
forcing axioms (MA, PFA, and so on)	no	no (as above)

From that table version 1 concluded that "there is therefore only one path to falsification, confined to the conservative branch", and that since conservativity is a mathematical property with a standard of adjudication, the falsifiability of the prediction is a concrete and checkable question. Both conclusions are withdrawn.

Paper 1's stress-test list already registers forcing and large cardinal extensions as open candidates; both remain legitimate objects of stress testing, though they are no longer set against a "path to falsification" (see (b)).

---

The v2 revision of § 13.2

This revision originates in SAE Mathematics Paper 5 (10.5281/zenodo.21538494) § 11.2, in four items.

(a) Withdrawal of the registration that NBG can trigger falsification 〔correction of error〕

Let $S$ conservatively extend $T$ over the old language $L$, and let $G$ be an $L$-sentence with $T\nvdash G$ and $T\nvdash\neg G$. If $S\vdash G$ then $T\vdash G$ by conservativity — a contradiction; similarly for $\neg G$. Hence

> A conservative object extension cannot adjudicate any sentence undecidable in the old language.

The error in the first row of the v1 table lies here: it verified that $S/T$ is conservative and adds a language of classes, but not that $S$ accomplishes the principal closure of $L_3$. These are not the same thing.

(b) Withdrawal of "only one path to falsification", and the emptiness extends further 〔correction of error〕

By (a), the conservative branch is empty. And by Corollary 7.2 of Paper 5 § 7.1, the emptiness extends to all effective extensions, conservative or not:

Type of extension	Why it does not trigger
conservative (NBG-type)	adjudicates no old-language independent sentence, by (a)
non-conservative but still effective (large cardinals, forcing axioms)	adjudicates some, but never attains full closure (Corollary 7.2)

Note that version 1's reasons for the last two rows are also inaccurate: it says large cardinals "necessarily change provability and are therefore incapable of meeting 'without any change to the rules of proof'", which runs together changing provable consequences with changing rules of proof — these belong respectively to axes R3 and R0 of the five listed in (d) below.

But one still may not write "the original condition is empty under every reading": what has been shown is emptiness under the adjudicative reading; the expressive reading is a separate matter.

(c) The standing of the original falsification condition 〔substantive revision〕

The original condition does not define the closure target. It describes the extension mechanism in detail (what may be added, what may not be touched) while never saying what "the principal closing act of $L_3$" refers to.

Paper 5 makes the closure target precise as the equality of the syntactic consequence closure with the semantic consequence closure over the standard model, $C_T^\vdash=C_{\mathbb{N}}^\vDash$. What results is not a new falsification condition but two theorems:

• Obstruction to full closure: there is no effectively axiomatized theory $T$ with $C_T^\vdash=C_{\mathbb{N}}^\vDash$ (soundness is not a hypothesis);
• Threshold for partial closure: if an effective, $\Gamma$-sound and $\Gamma$-complete theory exists then $\mathrm{True}_\Gamma$ is c.e.; the threshold breaks between $\Sigma_1$ and $\Sigma_2$ along $\Sigma$ and is already in force at $\Pi_1$ along $\Pi$, with no middle ground.

These theorems do not establish the migration prediction of this section. They cut along the axis effective / non-effective, whereas the migration prediction cuts along object / rule — two different axes. Indeed $\mathrm{Th}(\mathbb{N})$ attains full closure while being no effective rule system, and it need not be described as "migration into rules" either: it simply places complete semantic truth directly into the axiom set.

(d) The standing of the migration statement 〔substantive revision〕

The tier is unchanged and remains T3 — version 1 already registers it so. What changes is the type of the claim:

$$\text{T3 programmatic prediction (with an operationalized falsification condition)}\ \longrightarrow\ \text{T3 explanatory proposal (without one)}.$$

What happened must be stated accurately: one may not write "the prediction was subsequently proved to be a theorem" — that would conceal a change of object. The accurate registration is a conjunction of three:

$$\text{withdraw the operationalisation of the old prediction}\ +\ \text{add two theorems}\ +\ \text{downgrade the migration reading to a proposal}.$$

One further term must be unpacked 〔clarification〕: "rule" in the original condition covers at least five axes — R0 the logical calculus · R1 language/signature · R2 non-logical axioms · R3 consequences in the old language · R4 algorithmic structure. The original wording "rules of equivalence, proof, or decidability" reads like R0, whereas the two-column table in fact applies R3 (taking conservativity as the criterion). Expressive power of the metalanguage belongs to R1, not R2. Any successor condition must be written as a condition vector and may not use the bare word "rule" again.

What remains of this section after revision: the migration observation itself stands — the asymmetry of § 5.4 and the form of the trigger in § 8.4 are both earned textual facts. What changes is that this section no longer claims the observation to be equipped with an operable falsification condition.

Tier: T3 explanatory proposal (version 1: T3 programmatic prediction; the tier is unchanged, the claim type changes — see (d)). It is not a prediction of any mathematical theorem — the paper predicts no specific theorem to hold or fail, and offers only a typological reading of the mode of release of the pressure.

The paper does not presuppose the trajectory of $L_3$ (§ 1.3). The proposal above concerns the type of the mode of release, not the content of $L_3$; for the latter see Paper 5.

On chronology: this paper is published before the $L_3$ paper, and the version-level DOI and archival record of this signed-off version are the external credential of the public priority of this section — not priority in writing, since a publication record cannot establish the order of drafting in a private workspace. But the paper does not claim the prediction tested on the ground that a subsequent paper agrees with it: a sequel by the same author within the same framework does not constitute an independent test. See § 12.5.

Added in v2: after the downgrade, what is prior must be stated in two parts — the observation has priority (that pressure appears to pass into rule space was indeed published before the $L_3$ paper), while the predictive claim is withdrawn (its operationalisation does not stand). Failing to separate them yields "retain the priority and withdraw the content".

13.3 A mathematical placement coordinate system

Physics Paper 1 establishes the residence criterion: an object resides at the highest root layer among the resources required to articulate it. The mathematical counterpart of that criterion — a mathematical placement coordinate system — is not undertaken here.

The paper uses the vocabulary of placement locally, where needed (the superposition placement of § 8.3), and establishes no systematic coordinate system. A comprehensive placement — where each constant, each theorem, each construction resides — is left to be decided when the series is assembled.

13.4 Other interfaces

The placement of the effective-Cauchy articulation mode: the open question left by Paper 3 § 6.4; material at § 8.2, adjudication at § 11.3.

The logic specialist paper, the aesthetic mathematics specialist paper, and the relational mathematics specialist paper: interfaces unchanged. This paper does not touch those three lines and does not alter the registrations Paper 2 made for them.

The ZFCρ line: the status of the two halves of the debt is given at § 6.2 and § 7.4. The unsettled half, the fine ratio at the level of coefficients, has its interface located at the second stratum of the Newton polygon and not yet connected.

The Foundation line: this paper receives and rearranges the $L_1 \to L_2$ row of the cross-level ledger table of Foundation v1; see § 8.3 and § 11.5.4.

---

Appendix A · Version and Review Provenance Record

This paper passed through internal review by four seats (Claude, ChatGPT, Gemini, Grok, one seat each), the outline running v1 → v2 → v2.1 redline over three rounds, with the ChatGPT seat returning Major Revision on v1.

What is called review here is a model-assisted internal structural review organized by the author, and is not equivalent to journal peer review.

Two corrections bearing directly on the anti-immunization declaration of § 12.4 are registered here.

First. Outline v1 diagnosed the failure to give a full composition law as an instance of the $L_2 \to L_3$ trigger — that is, it converted a debt into an identification. That treatment forms a double standard with the same discipline this paper executes for $e \notin \mathcal{P}$. It was corrected after review identified it; the debt is now registered as an outstanding debt per § 9.3, with a minimal composition skeleton delivered per § 9.1 and § 9.2 as partial payment.

Second. The falsifiability registration of § 12.3 contained, in its first draft, only the benign counterfactual (should $e$ be proved a period). In a paper that makes an anti-immunization declaration, a falsifiability section registering only counterfactuals harmless to itself turns that declaration into decoration. Further, the second counterfactual first drafted read "the remainder itself becomes an $L_3$ object" — a formulation that would have destroyed the value/rule separation of columns established in § 8.4, because independence is a property of propositions, not of objects. § 12.3 now carries a two-way registration, with the second counterfactual in its corrected form.

Three further load-bearing decisions were downgraded in tier or rearranged during review, each registered at the corresponding place in the body: the capture range was downgraded from "saturates in $\mathcal{P}$" to "the evaluation core of the regular period articulation" (§ 6.1); the bearer of the remainder load was changed from the constant $e$ to the exponential map $\exp$ together with the extension of expression classes (§ 7.1, § 7.5); and the Newton polygon was changed from "the common generator of stage 4" to "the formal organizer", with its relation to the $G_\infty$ of Paper 1 Appendix B.5 downgraded from "placed" to "a candidate comparison interface, identity not proved" (§ 7.2, § 7.3).

What this appendix carries is traceability of provenance, not a warrant that the paper is therefore credible. The internal evidence for the anti-immunization declaration of § 12.4 is the body itself: $e \notin \mathcal{P}$ bears no load (§ 7.1), the composition-law debt is not converted into a finding (§ 9.3), and § 12.3 registers a counterfactual unfavourable to the paper. This appendix only supplements the traceability of the version chain.

---

References

Periods and transcendence

1. M. Kontsevich, D. Zagier, "Periods", in B. Engquist, W. Schmid (eds.), Mathematics Unlimited — 2001 and Beyond, Springer, Berlin, 2001, pp. 771–808. — Definition of the ordinary period ring $\mathcal{P}$; the three rules of additivity, change of variables, and Stokes' formula, together with the period conjecture; exponential periods; the expectation that $e$ and $\gamma$ are not ordinary periods.
2. M. Yoshinaga, "Periods and elementary real numbers", arXiv:0805.0349 [math.AG], 2008. — Periods are effectively approximable by elementary rational Cauchy sequences; construction of a computable real that is not a period.
3. J. Ayoub, "Une version relative de la conjecture des périodes de Kontsevich–Zagier", Annals of Mathematics 181 (2015), 915–992. — The relative, function-field version of the Kontsevich–Zagier conjecture.
4. J. Fresán, "Une introduction aux périodes", arXiv:2210.03407 [math.AG], 2022. — Survey article arising from talks at the Journées X-UPS Périodes et transcendance, École polytechnique. Period theory; the de Rham–Betti pairing.
5. J. Ax, "On Schanuel's conjectures", Annals of Mathematics 93 (1971), no. 2, 252–268. — The Ax–Schanuel theorem, the differential and function-field analogue of Schanuel's conjecture.
6. Y. André, Une introduction aux motifs (motifs purs, motifs mixtes, périodes), Panoramas et Synthèses 17, Société Mathématique de France, 2004. — Periods and motivic Galois theory; the Grothendieck period conjecture.
7. J. Fresán, P. Jossen, Exponential motives, manuscript, available from the first author's homepage. — The motivic framework for exponential periods; rapid decay cohomology and the oriented real blow-up. Not yet published in book form as of this writing.

Irregular singularities, formal classification, and the Stokes phenomenon

1. H. L. Turrittin, "Convergent solutions of ordinary linear homogeneous differential equations in the neighborhood of an irregular singular point", Acta Mathematica 93 (1955), 27–66.
2. A. H. M. Levelt, "Jordan decomposition for a class of singular differential operators", Arkiv för Matematik 13 (1975), no. 1–2, 1–27. DOI: 10.1007/BF02386195.

> Items 8 and 9, together with the earlier work of M. Hukuhara, are known jointly as the Hukuhara–Levelt–Turrittin theorem: a connection on the formal punctured disc, after a suitable finite ramification, decomposes as a direct sum of elementary connections each carrying a single exponential factor. The theorem-level load of § 7.1 and § 7.2 comes from this. Hukuhara's original papers (Mémoires of the Faculty of Science, Kyushu University, early 1940s) have not been consulted directly; for the statement of the theorem and for the attribution, see item 14 below.

1. W. Wasow, Asymptotic Expansions for Ordinary Differential Equations, Interscience, New York, 1965. — Asymptotic expansions at irregular singularities; order and type of growth; Stokes sectors.
2. M. Loday-Richaud, "Stokes phenomenon, multisummability and differential Galois groups", Annales de l'Institut Fourier 44 (1994), no. 3, 849–906. — The relation among the Stokes phenomenon, multisummability, and differential Galois groups.
3. C. Sabbah, Introduction to Stokes Structures, Lecture Notes in Mathematics 2060, Springer, 2013. — Stokes structures; real oriented blow-up and local systems on the circle of directions.
4. K. S. Kedlaya, "Good formal structures for flat meromorphic connections, I–III", 2010–2015. — Higher-dimensional generalization of the Hukuhara–Levelt–Turrittin decomposition; irregularity and turning loci.

Riemann–Hilbert and differential Galois theory

1. M. van der Put, M. F. Singer, Galois Theory of Linear Differential Equations, Grundlehren der mathematischen Wissenschaften 328, Springer, Berlin, 2003. — Standard reference for differential Galois theory; Schlesinger density in the regular case; the Ramis density theorem in the irregular case (the subgroup generated by the formal monodromy, the exponential torus, and the Stokes operators is Zariski dense in the local differential Galois group); the formal-level structure $\mathcal{T} \rtimes \overline{\langle\hat\gamma\rangle}^{\rm Zar}$. Chain 4 of § 9.2 comes from this.
2. P. P. Boalch, "Geometry and braiding of Stokes data; Fission and wild character varieties", Annals of Mathematics 179 (2014), 301–365. arXiv:1111.6228 [math.AG]. — Wild character varieties; the irregular Riemann–Hilbert correspondence.
3. T. Mochizuki, Wild Harmonic Bundles and Wild Pure Twistor D-modules, Astérisque 340, Société Mathématique de France, 2011. arXiv:0803.1344 [math.DG]. — Higher-dimensional theory of irregular connections.
4. N. M. Katz, "Nilpotent connections and the monodromy theorem: applications of a result of Turrittin", Publications Mathématiques de l'IHÉS 39 (1970), 175–232.

Quaternionic analysis

1. G. Gentili, D. C. Struppa, "A new theory of regular functions of a quaternionic variable", Advances in Mathematics 216 (2007), no. 1, 279–301. DOI: 10.1016/j.aim.2007.05.010. — The founding paper of slice regular function theory: following an idea of Cullen, the restriction of a regular function to each complex slice $\mathbb{C}_J$ is holomorphic. The first half of § 10.2 comes from this.

1. G. Gentili, C. Stoppato, D. C. Struppa, Regular functions of a quaternionic variable, Springer Monographs in Mathematics, Springer, Berlin–Heidelberg, 2013; 2nd ed., 2022. — Systematic development of the theory, including the Cauchy kernel and integral formula on the slices. The Cauchy-type structure invoked in § 10.2 is cited from here, not from item 18.

Classical complex analysis

Divisors and the divisor class group, the residue theorem, the argument principle, the de Rham–Betti pairing, monodromy representations, the fundamental theorem of algebra, one-point compactification and the Riemann sphere. The technical content of these results was established by mathematical work of the nineteenth and twentieth centuries and is found in any standard textbook. By § 2.6, first item, this paper cites them and does not claim to derive them.

SAE series

1. Han Qin, The Architecture of SAE Mathematics: A Layer Articulation Schema and the Twofold Inexhaustibility of Mathematics and Its Frameworks (Paper 1). DOI: 10.5281/zenodo.20153791
2. Han Qin, SAE Mathematics Paper 2: $L_0$ Mathematicality — Inevitability and Multi-Path Realisation. DOI: 10.5281/zenodo.20199082
3. Han Qin, Finite Algebraic Articulation of the Quantitative Dimension $L_1$ (Paper 3). DOI: 10.5281/zenodo.20266351
4. Han Qin, SAE Foundation v1. DOI: 10.5281/zenodo.19361951 — The cross-level ledger table; this paper receives and rearranges its $L_1 \to L_2$ row (§ 8.3, § 11.5.4).
5. Han Qin, SAE Physics Foundation v2. DOI: 10.5281/zenodo.20322460 — The parallel-ladder comparison table; the distinction between operator-level and scalar-level closure (§ 11.5).
6. Han Qin, Four-fold Nesting of the Physical-Quantity Ladder and a Placement Coordinate for Particles and Concepts (SAE Physics Paper 1). DOI: 10.5281/zenodo.21227329 — The second-level nesting rules; single-point and superposition placement; ontological boundary and definitional boundary; the three claim tiers.
7. Han Qin, Methodology 6: Phase-Transition Windows. DOI: 10.5281/zenodo.19464507
8. Han Qin, Methodology 00: Via Rho. DOI: 10.5281/zenodo.19657440

---

1. 引言与定位

1.1 本文的位置

Paper 3 对 $L_1$ 的量维度作了完整的四主步骤阐述, 并在该文 § 6.4 明确: 后 $L_1$ 的诸阐述候选如何安放, 该文不预设, 留后续层级论文. 本文是该后续层级论文中处理复化主轨的一篇.

本文对 $L_2$ 所做的, 与 Paper 3 对 $L_1$ 所做的是同一件事: 区分基底与阐述模式, 按四主步骤及其内层 sub-articulation 阐述该模式的展开, 给出捕获范围, 登记余项账本.

但有一处格式差别须先声明: 主步骤 4 的登记类别. 按 Paper 3 § 1.3 与 § 6.1 所立的三分类, 主步骤 4 是元层级闭合, 其 sub-articulation 是余项账本结构, 不遵循四步分形模式. 本文遵循该登记, 并以主步骤 4 横跨 § 6、§ 7、§ 8 三节的版面不对称承载之 (§ 2.1).

本文不是 Paper 1 § 5 的重述. Paper 1 在建筑学层面阐述了 $L_2$ 的四步与 $\omega_\rho^{(2)}$ 六分量; 本文作专题展开, 并对 Paper 1 有四项上游处理 (§ 11.1), 其中一处标注为善意重构而非原意还原.

1.2 承接的四笔上游债务

本文承接四笔明确的上游债务, 各有其偿付状态.

其一, Paper 1 § 11.1 的合成律. 该文把 $\omega_\rho^{(2)}$ 六分量的合成律挂账给后续的 $L_2$ 专题论文, 并阐述: 若无合成律, 六元组只是清单. 本文作部分偿付——交付最小 typed composition 骨架 (§ 9), 未交付统一范畴封装, 后者登记为未还挂账 (§ 13.1).

其二, Paper 1 § 11.8 的 Cayley–Dickson. 该文问替代闭合是否提供 $L_2$ 的另一种实现, 以及充分性标准如何适用. 本文作答 (§ 10), 并修订 Paper 1 § 5.1 把该序列各项列为 $L_2$ 阐述候选的分类.

其三, Paper 3 § 6 的余项账本. 该文立三类余项与跨层 forcing 的判据. 本文在 $L_2$ 重排该账本 (§ 8.2), 并阐述第一类 (基底外) 在两层的地位不同——$i$ 由 $L_1$ 内部追问逼出, $\mathbb{H}$ 无对应来源.

其四, ZFCρ 线的单块 $\infty$. 该线所提的问题是 $L_1$ 的单块 $\infty$ 不能分辨 fine ratio. 本文清其阶的一半 (§ 6.2), 系数一半的接口定位于 Newton 多边形第二层但未接通 (§ 7.4); 债务状态从"未接口"改登记为"有接口而未接通", 不得引作清账.

另有一笔由 Paper 3 § 6.4 转来的开放问题——有效 Cauchy 阐述模式的落点——本文判定为与复化主轨并行的 side-branch (§ 11.3).

1.3 本文不主张

以下七条, 本文显式不主张. 其中若干在 § 2.6 的读者防误解表中另有展开, 在 § 12.1 的三档表中登记为"不得作本篇结论".

其一, 本文不推导复分析. 除子、留数、周期与单值群、Stokes 数据与渐近增长这一分层, 是经典复分析中既有的结果. 本文主张的是该分层在跨层比较中的位置, 不是该分层本身.

其二, 本文不主张 $\mathcal{P}$ 穷尽 $L_2$ 基底. $\mathcal{P}$ 可数而基底不可数; 且 $\mathcal{P}$ 连可计算实数都不穷尽. 更进一步, 本文所工作的曲线上的解析-覆叠脊柱, 亦不主张生成全部 $\mathcal{P}$ (§ 2.3 二轨声明).

其三, 本文不证明也不反驳 $e \notin \mathcal{P}$. 该命题是开放猜想. 本文承重不落于其上 (§ 7.1); 若它被证否, 须改动的句子清单见 § 12.3.

其四, 本文不给出 $\omega_\rho^{(2)}$ 的完整合成律. 见上文其一.

其五, 本文不撤回 Paper 1 对 $L_2$ 的建筑学阐述. 四项均登记为精化、善意重构或候选分类修订, 不作撤回 (§ 11.1).

其六, 本文不把四元数逐出 $L_2$. § 10 只得出 $\mathbb{H}$ 不共享本文这条 trajectory 的签名与缺陷系统; 关于它在全部 $L_2$ trajectory 空间中的位置, 本文不判断.

其七, 本文不预设 $L_3$ 的 trajectory. § 13.2 给出一条关于压力释放方式类型的 T3 纲领性预测, 并附操作化的证伪条件与已注册的反例候选; 该预测不是关于 $L_3$ 内容的.

1.4 章节安排

§ 2 立四条防火墙: 基底与阐述模式 (含主步骤 4 的登记声明)、数值与函数与算子、表达式语法与取值像与表达式等价、数值命题与证明论状态与算法可判定性; 另附术语表与读者防误解表. 本文其后的每一项实质主张均可回溯到这四条之一.

§ 3 至 § 5 阐述主步骤 1 至 3. 其中 § 5 (乘法路径) 是全文的数学脊柱: 它给出反向追问的不对称 (§ 5.4)、$2\pi i$ 的双重身份 (§ 5.5), 而 § 6 至 § 8 的全部技术内容, 其存在理由在该节被阐述.

§ 6 至 § 8 是主步骤 4 的三节: 正则周期核心、不规则边界、闭合与余项账本. 三者的关系是核心与增强, 不是互斥分箱——不规则联络同样有除子、留数、单值群, § 7 所增加的是指数因子、方向与 Stokes 数据.

§ 9 交付操作缺陷代数的最小 typed composition. § 10 处理 Cayley–Dickson. § 11 登记与 Paper 1、Paper 2、Paper 3、ZFCρ 线及 SAE Physics Foundation v2 的关系. § 12 给出 claim 三档表、辖区声明、双向可证伪性登记、反免疫声明与预测纪律. § 13 列开放问题与后续 trajectory.

---

2. 四条防火墙

本节是全文的总防火墙. Paper 3 § 2 建立了基底与阐述模式的区分, 作为该文的总防火墙; 本文把该区分抬到 $L_2$, 并另立三条. 四条防火墙不是修辞性的免责声明, 而是本文承重结构所依赖的区分: § 3 至 § 9 的每一项实质主张, 都可回溯到这四条之一. 若四条中任一条被读穿, 相应的主张随之失效.

四条防火墙分别区分: 基底与阐述模式 (§ 2.1)、数值与函数与算子 (§ 2.2)、表达式语法与取值像与表达式等价 (§ 2.3)、数值命题与证明论状态与算法可判定性 (§ 2.4). § 2.5 给出术语表, § 2.6 给出读者防误解表.

2.1 基底与阐述模式

Paper 3 § 2 阐述: 量维度基底与 $L_1$ 阐述模式是两个不同对象, 不得混同. 基底是沿量维度阐述的"诸量整体", 在经典数学构造层面经 ZFC 加 Dedekind 分割实现为实直线 $\mathbb{R}$ 作为完成对象; $L_1$ 阐述模式是 SAE 操作阐述视角下的有限代数工具集, 捕获实代数数 $\mathbb{Q}^{\rm alg}_{\mathbb{R}}$. 该文的防火墙陈述是: 本文不撤回经典构造; 本文阐述的是, 从 SAE 操作阐述视角看, 没有任何单一阐述模式穷尽实直线基底.

该陈述是全称的而非枚举的. 它在 Paper 3 § 6.3.2 由基数论证挣得: 对任何可数形式语言, 该语言中的公式总数至多 $\aleph_0$, 而实直线基数为 $2^{\aleph_0}$, 故不可定义余项对一切此类模式皆存在. 关键在于该论证不依赖于具体语言的选择——换一种语言、扩张语言, 只要仍可数, 结论不变. 本文继承该全称形式, 不重证; § 8.2 另给出该形式在 $L_2$ 处的一项展示.

本文把同一区分抬到 $L_2$.

$L_2$ 基底是复平面与 Riemann 球面. 在经典数学构造层面, $\mathbb{C}$ 经 $\mathbb{R}^2$ 加乘法结构、或经 $\mathbb{R}[x]/(x^2+1)$ 实现为完成对象, $\hat{\mathbb{C}}$ 经一点紧化实现. 本文不撤回这些构造.

$L_2$ 阐述模式是代数 de Rham–Betti 积分与解析延拓. 其捕获范围的核心是普通周期环 $\mathcal{P}$ (§ 2.3 给出定义, § 6 给出实质阐述). 本文不主张 $\mathcal{P}$ 穷尽 $L_2$ 基底 (见 § 2.6 第二条).

捕获判据逐层不同, 这一点本身是本文的一项识别. Paper 3 的 $L_1$ 模式以有限代数约束为捕获判据: 一个实数被捕获, 当且仅当它满足 $\mathbb{Q}$ 上的有限多项式约束. 本文的 $L_2$ 模式以代数积分表示为捕获判据: 一个复数被捕获, 当且仅当它可表为代数数据在 $\mathbb{Q}$-半代数区域上的绝对收敛积分之值. 两条判据不是同一种判据的两个实例: 前者判定的是一个数与一族方程的关系, 后者判定的是一个数与一族几何积分表达式的关系. 这一类型跃迁在 § 6.1 以承载体迁移之名作实质阐述, 此处只作登记.

主步骤 4 的登记类别声明. Paper 3 § 1.3 与 § 6.1 把层内 sub-articulation 分为三类, 并明确: 跨层阈口 sub-articulation (主步骤 1)、层内分形 sub-articulation (主步骤 2 与 3)、以及主步骤 4——该文原文阐述"主步骤 4 是元层级闭合, 其 sub-articulation 是余项账本结构, 不遵循四步分形模式". 三类各有自己的登记标准.

本文遵循同一登记. 主步骤 1 登记为跨层阈口类 (§ 3); 主步骤 2 与 3 登记为层内分形类 (§ 4、§ 5); 主步骤 4 登记为余项账本结构 (§ 6、§ 7、§ 8), 按"捕获核心 → 未捕获账本 → 下一层触发条件"组织, 不按四相重演组织.

$\omega_\rho^{(2)}$ 六分量映射到的四阶段 (点态、局部-货币化、整体-圈、边界-扇区) 是类型分层上的尺度递进, 不是四相在主步骤 4 内部的再现. 本文以主步骤 4 横跨三节而主步骤 1 至 3 各占一节的版面不对称承载此登记: 形式即声明.

一笔继承的开放问题. Paper 3 § 6.4 把有效 Cauchy 阐述模式登记为 post-$L_1$ 的解析完备化分支, 并明确不预设它究竟作为 $L_2$ 主轨内的 sub-articulation、作为 $L_{1\to 2}$ 桥、还是作为与复化主轨并行的 side-branch, 该判定留给后续层级论文. 本文是该后续层级论文.

本文所能给出的实质材料是捕获范围的比较. 有效 Cauchy 阐述模式以算法可逼近性为捕获判据, 捕获可计算实数; $L_2$ 模式以代数积分表示为捕获判据, 其实轴上的捕获核心是 $\mathcal{P} \cap \mathbb{R}$. 二者的关系是: $\mathcal{P} \cap \mathbb{R}$ 严格含于可计算实数之中 (周期可任意精度逼近; 而 Yoshinaga 已给出可计算但非周期的实数, 见 § 8.2). 因此有效 Cauchy 在实轴上捕获得严格多于 $L_2$ 的周期核心, 而三条捕获判据 (有限代数约束、代数积分表示、算法可逼近性) 互不为对方的特例.

本文将在 § 11.3 把它登记为相对于本篇复化 trajectory 的并行 side-branch; 本节只预告比较材料, 不提前展开论证.

2.2 数值、函数与算子

$L_2$ 的阐述内容同时含三类对象, 而三类对象在通行数学语言中常共用同一个名字. 本节把它们分开.

数值: $\pi$、$e$、$2\pi i$、某个周期、某个留数值. 数值是取值映射的输出.

函数: $\exp$、$\log$、某个亚纯函数、某个全纯延拓的分支. 函数是 $L_2$ 阐述模式的作用对象.

算子: $\mathrm{div}(\cdot)$、$\mathrm{Res}_p(\cdot)$、沿路径的解析延拓、$\oint_\gamma(\cdot)$. 算子把函数送到数值或送到别的函数.

这条区分不是学究式的. 本文的一项承重决定直接落在它上面: $L_2$ 正则周期核心的边界余项是指数映射 $\exp$ 这一函数, 而不是常数 $e$ 这一数值 (§ 7.5). $e = \exp(1)$ 只是 $\exp$ 在 1 处的规范值. 把承重放在数值上, 会使全文结构依赖于一条关于该数值的开放猜想 ($e \notin \mathcal{P}$); 把承重从常数移到映射与表达式类上, 结构依赖的是定义级的语法扩张与定理级的联络分类, 而不依赖关于数值 $e$ 的开放猜想. 两种放法的认识论档次不同, 而它们的区别只有在数值与函数被分开之后才看得见.

同一条区分也用于消解一处历史遗留的读法混淆. 留数不是一个数值范畴的原始对象: $\mathrm{Res}_p$ 是算子, 它作用于亚纯函数在点 $p$ 处的局部数据, 输出一个数值. § 9 阐述各分量的内部运算时, 分量的类型 (群、双线性配对、表示、群胚) 正是在这条区分上被指定的.

2.3 表达式语法、取值像与表达式等价

本条防火墙区分三个对象, 并给出本文关于周期的全部定义性承诺.

表达式语法. 记 $\mathsf{Int}_{\rm alg}$ 为普通周期表达式类: 被积数据为代数微分形式, 积分域为 $\mathbb{Q}$-半代数区域. 记 $\mathsf{Int}_{\exp}$ 为指数积分表达式类: 在被积函数中另允许 $e^{-f}$ 型因子, $f$ 为代数函数. 二者是语法层面的对象——它们是可以写下来的积分表达式的集合, 与这些表达式取什么值无关.

取值像. 取值映射 $\mathrm{ev}: \mathsf{Int}_{\rm alg} \to \mathbb{C}$ 把表达式送到它的值. 其像

$$\mathcal{P} = \mathrm{im}(\mathrm{ev})$$

即 Kontsevich–Zagier 普通周期环. 精确地说: 一个复数是周期, 当其实部与虚部均可表示为 $\mathbb{Q}$-有理函数在 $\mathbb{Q}$-半代数区域上的绝对收敛积分之值. 相应地 $\mathcal{P}^e = \mathrm{im}(\mathrm{ev}_{\exp})$ 为指数周期.

四者构成一个交换方图, 以表格呈现. 表 1 呈现该方图: 左列是语法侧, 右列是取值侧, 上行是普通情形, 下行是指数扩张情形.

表 1 · 表达式类与取值像的交换方图

语法侧 (表达式类)	取值映射	取值侧 (数值集)
$\mathsf{Int}_{\rm alg}$ 被积数据代数, 积分域 $\mathbb{Q}$-半代数	ev ⟶	$\mathcal{P}$ 普通周期环
↓ 语法扩张: 另允许 $e^{-f}$ 因子		↓ 含入
$\mathsf{Int}_{\exp}$ 另允许 $e^{-f}$ 型因子, $f$ 代数	ev$_{\exp}$ ⟶	$\mathcal{P}^e$ 指数周期

方图的交换性内容是: 一个普通周期表达式, 先取值再含入指数周期, 与先视作 $f = 0$ 的指数积分表达式再取值, 结果相同. 换言之, 语法扩张与取值映射相容. 这一相容性看似平凡, 但它正是本文得以把承重放在左列 (语法扩张) 而非右列 (数值集之间的严格包含关系) 的前提: 左列的扩张是构造性的、可判定的; 右列的 $\mathcal{P} \subsetneq \mathcal{P}^e$ 是否严格, 则是开放问题 (§ 2.6 其三).

本文全部关于捕获范围的陈述均可定位到此表的某一角或某一箭头. 便于检索起见, 四角与三箭头的登记如下: 左上角与左下角是语法对象, 右上角与右下角是数值对象; 两条横箭头是取值映射, 左竖箭头是语法扩张 (T2 承重所在, § 7.1), 右竖箭头是数值含入 (其严格性未决, § 8.2).

表达式等价. 两个表达式取同一值, 是 $\ker(\mathrm{ev})$ 中的一条关系. Kontsevich–Zagier 的加性、变量替换与 Stokes 公式三条规则给出一族确定属于 $\ker(\mathrm{ev})$ 的标准关系; 而这些关系是否生成全部表达式等价, 正是 Kontsevich–Zagier 猜想的内容, 尚未证明.

这一区分必须严格守住. 三条规则不是从零生成所有周期数的三个正向数值运算, 也不是 $\mathcal{P}$ 的定义. $\mathcal{P}$ 的定义如上一段所给, 落在被积数据与积分域的代数性上, 定义中不含反函数算子. 本文凡涉及定义边界之处, 均改述在该定义上, 因而完全不依赖 Kontsevich–Zagier 猜想. 把定义边界接到三条规则上, 等于把一条定义性陈述接到一条开放猜想上; 本文不这样做. (相应的假朋友登记见 § 2.6.)

二轨声明. 本文有两条相交但不等同的脊柱.

第一条是曲线上的解析-覆叠脊柱: $\mathbb{C}$、$\mathbb{P}^1$、$\mathbb{C}^\times$、万有覆叠、单值群、Stokes 数据. 它是一复变量的, 是 § 5 与 § 7 的实际工作场地.

第二条是任意维代数几何中的 de Rham–Betti 取值脊柱: $\mathcal{P}$ 允许任意有限维代数簇与任意维半代数积分域.

本文不主张第一条脊柱生成全部 $\mathcal{P}$. $\mathcal{P}$ 在本文中充当一般取值核心, 而不是由 Riemann 球面单独推出的数值全集. 二者相交于本文实际使用的那些周期 (小环路周期、椭圆周期等), 但相交不是等同. 若无此声明, § 6 的捕获范围陈述会被读成"一复变量围道积分穷尽了周期", 那是一条本文不作、也无法作的主张.

2.4 数值命题、证明论状态与算法可判定性

本条防火墙管的是: 一个开放问题开放在哪一个层面上.

$L_2$ 的余项账本中含若干未决命题, 其中最显眼的是 $e$ 是否属于 $\mathcal{P}$. 若不区分开放的层面, 这类命题极易被误读为层边界的证据或反证. 本文按三条正交坐标登记每一项.

坐标	可能状态
数值成员状态	已证属于 / 已证不属于 / 未知
证明论状态	已证 / 猜想 / 相对独立 / 未知
算法状态	可判定 / 半可判定 / 受限类可判定 / 一般问题未知

三条坐标正交: 一项命题的证明论状态不决定其数值成员状态, 数值成员状态也不决定算法状态.

另设一栏登记该项在本文中承担的角色.

SAE 角色	含义
承重	不依赖任何开放猜想; 本文的结构主张压在其上
见证	与框架一致, 但不作为对框架的支持; 见 § 12.4 反免疫声明
接口	把值层面的问题提升为表达式、规则或算法层面的问题

两条不同的边界必须分开.

正则与不规则的分界, 是 $L_2$ 捕获范围的边界, 位于层内. 它由形式分类定理支撑 (§ 7.1), 是本文的 T2 承重.

值层面向规则层面的重框, 才是 $L_2$ 与 $L_3$ 的层边界. Paper 1 § 5.5 已阐述该触发判词: 触发 $L_3$ 的不是 $L_2$ 的难题本身, 而是问题被重框为规则与可判定性. 本文 § 8.4 作实质阐述.

混同这两条边界会产生两个方向的错误. 一是把 Stokes 现象读成 $L_2/L_3$ 的层边界——Stokes 现象是 $L_2$ 模式内部的结构性事实, 不是层边界. 二是把某个数值命题的证明论状态读成对象的层归属——独立性是命题的属性, 不是对象的属性. 即便"$e \notin \mathcal{P}$"被证明相对某形式系统独立, 进入规则层的是该命题在该系统中的可证性; 数值 $e$ 本身不因此改变层归属 (§ 12.3 第二条反事实).

本条防火墙置于 § 2 而非置于闭合部分, 是因为它是 § 6 至 § 8 全部捕获范围陈述的共同前提, 而非仅仅是 § 8 的收尾装置.

2.5 术语表

以下用词在本文中强制区分. 其中前四项在通行中文数学写作中常共用"周期"一词, 而本文的一项承重陈述恰落在它们的区别上 (§ 7.5).

对象	本文用词	例
$\mathcal{P}$ 意义上的周期 (数值)	周期	$\pi$、$\zeta(3)$、$\log 2$
有平移不变性的函数	周期函数	$\exp$ (周期 $2\pi i$)
在代数点取周期值的函数	取周期值的函数	$\log$
$\mathcal{P}^e$ 的元素	指数周期	$e$、$\gamma$

全文禁用未限定的"周期函数"一词. 特别注意: $\log$ 不是周期函数, $\exp$ 才是; 而 $\log$ 在代数点取周期值. 二者的角色在 § 5.5 与 § 7.5 各出现一次, 若混用则该两处论证均不可读.

记号方面, 以下三者是三个不同对象, 不得共用记号.

对象	记号	出处
无穷处增长 (整函数的阶与型, 加 Phragmén–Lindelöf)	$G_\infty$	Paper 1 附录 B.5, 记号保留不改
斜率过滤	$\mathrm{Slope}_\infty$	本文 § 7.2
局部微分 Galois 群	$G_{\mathrm{diff,loc}}$	本文 § 7.3、§ 9.2

Paper 1 的 $G_\infty$ 所配的内容是标量渐近不变量, 不是群对象; 记号中的 $G$ 不应被读为群. 本文不改该记号, 亦不替换该对象, 而在 § 7.3 为它登记 Newton–Turrittin 数据所提供的候选比较接口 (并明确该接口尚未证明同一性).

另有六组记号须分家, 以免在同一小节内换义:

用途	本文记号
Newton 多边形的斜率	$\lambda$ (§ 7.2)
Newton 边的水平长度	$m$ (§ 7.2)
分歧指标	$\nu$ (§ 7.1、§ 7.2)
整函数的阶	$\rho$ (Paper 1 附录 B.5; 与框架记号 $\omega_\rho^{(2)}$ 的下标同形而不同义, 后者是系列既定记号, 不改)
模长	$r$ (§ 5)
底空间上的点	$p$ (§ 6、§ 9)

2.6 读者防误解表

本文的主张分五种, 强度各不相同. 表中每一行注明该项属于哪一种.

类型	含义
标准定理	数学界既有定理, 本文引用不证明
标准定义	数学界既有定义, 本文采用不改动
SAE 登记	本文在 SAE 框架内作出的识别性判定, 非数学定理
结构类比	结构上的对位, 未主张同一性
开放猜想	数学界未决, 本文不承重

以下六条是本文预判最可能出现的误读, 逐条澄清.

其一, 本文不推导复分析. $\omega_\rho^{(2)}$ 六分量映射到的四阶段 (点态、局部-货币化、整体-圈、边界-扇区), 是经典复分析中既有的分层——除子、留数、周期与单值群、Stokes 数据与渐近增长, 各自的技术内容均由 19 世纪至 20 世纪的数学工作确立. 本文主张的是该分层在跨层比较中的位置 (即 § 5.4 所阐述的反向追问不对称), 不是该分层本身. 表格化的呈现方式容易暗示四阶段生成了该分层; 此处显式否认该暗示. 〔类型: 归属声明〕

其二, 本文不主张 $\mathcal{P}$ 穷尽 $L_2$ 基底. $\mathcal{P}$ 可数, 而 $L_2$ 基底不可数; 且 $\mathcal{P}$ 连可计算实数都不穷尽 (§ 8.2). $\mathcal{P}$ 是 $L_2$ 正则代数积分阐述的取值核心, 把它认作 $L_2$ 的主要捕获范围是 SAE 登记. 〔类型: SAE 登记〕

其三, 本文不证明 $e \notin \mathcal{P}$, 也不以之承重. 该命题是开放猜想. 本文的 T2 承重落在表达式类与联络的扩张上 (§ 7.1), 不落在该命题上. 若该命题被证否, 本文须改动的句子清单见 § 12.3. 〔类型: 开放猜想〕

其四, 本文不撤回 Paper 1 对 $L_2$ 的建筑学阐述. 本文对 Paper 1 有四项上游处理 (边界闭合的强制性、Cayley–Dickson 候选分类、欧拉公式的落点、$G_\infty$ 的比较接口), 均在 § 11.1 如实登记为精化、善意重构或候选分类修订, 不作撤回. 其中 $G_\infty$ 一处, 本文明确登记为善意重构而非 Paper 1 原意. 〔类型: SAE 登记〕

其五, 本文不把四元数逐出 $L_2$. § 10 判定 $\mathbb{H}$ 不是本文这条可交换复周期 trajectory 的等价并列实现; 本文不据此判断它在全部 $L_2$ trajectory 空间中的最终位置. 〔类型: SAE 登记〕

其六, 本文不宣称 ZFCρ 的单块 $\infty$ 债务已清. § 6.2 阐述赋值给该 $\infty$ 装上整数刻度, 清的是阶那一半; 系数层面的 fine ratio 那一半仍欠, 其接口在 § 7.4 定位到 Newton 多边形, 但有接口不等于已接通. 〔类型: 结构类比 + 未清债务登记〕

假朋友登记. 以下两组同名不同物, 本文全程不得据以搭桥.

其一, Kontsevich–Zagier 猜想第三条规则中的 Stokes 公式是 Stokes 定理 (微分形式的积分与其外微分的关系); $\omega_\rho^{(2)}$ 第四阶段的 Stokes 数据是发散级数在扇区边界上的 Stokes 现象. 二者以同一人命名, 内容无关.

其二, Kontsevich–Zagier 的三条规则是周期数值之间的关系; Paper 1 § 11.1 所求的合成律是 $\omega_\rho^{(2)}$ 六分量之间的关系. 二者经 Riemann–Hilbert 对应相关, 但不同一 (§ 9.3、§ 13.1).

---

3. 主步骤 1: 标而不构 $i$

3.1 主步骤 1 与 inevitability 1

Paper 3 § 1.2 立: inevitability 1 对应主步骤 1, 其形式是标而不构——标记被固定为公众 token, 其运算地位被确定, 而它作为对象的构造并未给出. 在 $L_1$, 被标而不构者是单位 "1".

在 $L_2$, 被标而不构者是 $i$.

登记类别: 跨层阈口类 (§ 2.1). 本主步骤阐述的不是 $L_2$ 内部的展开, 而是 $L_1 \to L_2$ 过渡本身. 其内层 1 的输入来自上一层, 而非来自本主步骤内部——这是跨层阈口类与层内分形类的登记差别.

3.2 内层步骤

内层 1: $L_1$ 的闭合余项抵抗一维纳入

Paper 3 阐述 $L_1$ 主步骤 3 内层 4 的第二级反向追问闭合于实闭域 $\mathbb{Q}^{\rm alg}_{\mathbb{R}}$. 但

$$x^2 + 1 = 0$$

在实闭域中无解.

须注意此处的结构: 提出该方程的是 $L_1$ 自己的追问——多项式求根正是 $L_1$ 主步骤 3 的反向追问内容; 而 $L_1$ 的基底 (实直线) 不含答案. 故 $i$ 是 Paper 3 § 6 意义上的基底外余项: 不是"$L_1$ 模式未能捕获基底内的某物", 而是"所求之物不在基底之内".

这条区分决定了 forcing 的类型. 基底外余项由本层内部的追问逼出, 故它迫使跨层, 而非迫使一条 lateral branch. Paper 3 据此把 $i$ 登记为跨层 forcing.

该区分在本文另有一处后果: $L_2$ 的基底外余项 ($\mathbb{H}$ 方向) 没有对应的内部追问逼出它——$\mathbb{C}$ 代数闭, 求根不再逼人. 故 $L_2$ 的第一类余项与 $L_1$ 的第一类余项地位不同 (§ 8.2), § 10 据此判 Cayley–Dickson 为代价支.

内层 2: 跨主体传播

Cardano 在三次方程的求解中遇到负数的平方根, 称之为"诡辩量"而拒绝承认其实在性——但仍在计算中使用它们. Bombelli 随后给出这些量的运算规则, 使它们可以在计算中一致地传递.

关键在于传播先于承认. 该 token 在被认作对象之前, 已经作为运算中可传递的记号在主体之间流通. 一个使用者写下的含负数平方根的中间步骤, 另一个使用者可以接着算下去并得到正确的实数解.

这与 Paper 3 主步骤 1 内层 2 的跨主体传播同型: 标记的公众性不由其本体地位保证, 而由其在运算中的可传递性保证. 一个记号成为公众的, 是因为它在主体之间传递而不失效, 不是因为主体们就它是什么达成了一致.

内层 3: 符号的公众绑定

Euler 引入记号 $i$.

绑定的效果是: 该量从"某个负数的平方根"这一描述性短语, 变为单一的公众 token. 描述性短语在每次使用时都要重新展开其所指; 单一 token 不必.

绑定的两面 (承 Paper 3 § 4.2.3 的纪律, 不作价值判断):

• 一面是所阐述的处理能力: $i$ 可以在公式中作为单一符号被操作, 与实数并列出现在同一表达式中;
• 另一面是所阐述的余项: 记号掩盖了它究竟是什么. "$\sqrt{-1}$" 至少还携带一个 (虽然不可满足的) 描述; "$i$" 连描述也不携带.

内层 4: 构与余项

构: 公众 token $i$, 连同关系 $i^2 = -1$.

该构是标而不构的: $i$ 被固定为公众 token, 其运算规则被确定, 而它作为对象的栖居空间没有被构造. 在本主步骤结束时, $i$ 有完整的代数性质, 没有任何几何位置.

余项: $i$ 无栖居空间.

该余项是主步骤 2 的输入 (§ 4.2).

3.3 与上游的对位

Paper 1 § 5.1 阐述 $L_2$ 主步骤 1 为 $i$ 的标记, 并把 Cayley–Dickson 序列的各项列为 $L_2$ 阐述的候选. 本文 § 10 修订该候选分类 (§ 11.1 第二项上游登记).

Paper 3 § 6 给出基底外余项的登记与跨层 forcing 的判据; 本节的内层 1 是该判据在 $L_2$ 侧的接收端.

SAE Foundation v1 的跨层账本表把 $i$ 记为 $L_1 \to L_2$ 的余项一, 未给出理由. 本文 § 8.3 给出: $i$ 是跃迁的输入——它不是 $L_2$ 产生的, 是本主步骤从上一层接进来的.

---

4. 主步骤 2: 加法路径——二维化

4.1 主步骤 2 与 inevitability 2

Paper 3 § 1.2 立: inevitability 2 对应主步骤 2 (加法路径), 其形式是标记的迭代与该迭代的反向追问. 在 $L_1$, 单位标记的迭代给出 $\mathbb{N}$, 反向追问给出 $\mathbb{Z}$.

在 $L_2$, 主步骤 1 留下的余项是: $i$ 无栖居空间. 本主步骤为它构造栖居空间, 并沿加法路径闭合之.

登记类别: 层内 sub-fractal (§ 2.1).

4.2 内层步骤

内层 1: 正交方向的 sub-mark

主体经 Via Negativa 拒绝阐述停在纯代数记号的状态, 阐述一个与实轴正交的方向作为 sub-mark.

须注意该 sub-mark 不是从实直线内部导出的. 实直线上没有任何结构指定一个正交方向; 该方向是为安置 $i$ 而设的. 这与 $L_1$ 主步骤 2 内层 1 的情形不同——在 $L_1$, 迭代方向由单位标记自身给出, 不需另设.

历史上这一步以 Wessel、Argand、Gauss 的几何表示实现, 距 Euler 的记号绑定约两个世纪. 时间差本身说明该 sub-mark 不是记号的直接推论.

内层 2: 分量加法的迭代

$$(a+bi) + (c+di) = (a+c) + (b+d)i.$$

加法逐分量进行, 每一分量沿用 $L_1$ 已阐述的加法路径. 迭代给出整个平面——Argand 平面.

该迭代阐述的是平移, 不是转动. 这一点是本主步骤余项的来源 (内层 4).

内层 3: 有序对绑成单一 token

代价: 若每个复数都写作"实部 $a$, 虚部 $b$"这一有序对, 则所有运算都须成对进行, 阐述长度加倍, 且每一次运算都要指明它作用于哪一分量.

绑定: $z = a + bi$ 作为单一 token. 在该 token 下, 运算可以整体施加于 $z$, 不必每次分解为分量.

两面:

• 处理能力: 复数成为可以整体操作的单一对象; 后续的乘法路径 (§ 5) 得以在 $z$ 的层面上阐述, 而不必始终停留在分量层面;
• 余项: token 掩盖了它是有序对这一事实. 需要分量时须重新分解, 而分解不是 token 自带的操作.

这与 $L_1$ 主步骤 2 内层 3 的进制绑定同位: 进制把重复的单位标记绑成紧凑记号, 此处把有序对绑成单一记号. 两者都是跨迭代的代价管理绑定.

内层 4: 构与余项

反向追问: 加法反过来是减法; 而 $\mathbb{C}$ 对减法闭合——逐分量继承 $L_1$ 中 $\mathbb{Z}$ 已经闭合的结果, 不产生新的扩张要求.

故本主步骤的反向追问是单次的, 不是嵌套级联. 这是 § 5.6 所阐述的 sub-fractal 深度不对称的另一端: 加法路径浅, 乘法路径深, 两层皆然.

构: $\mathbb{C}$ 作为二维加法闭合; 边界标记 $\hat\infty$ (见 § 4.3).

余项: 加法路径不阐述旋转.

该余项是主步骤 3 的输入 (§ 5.2.1).

4.3 $\hat\infty$ 是被选择的规范闭合

Paper 1 § 5.2 阐述: $L_1$ 的 $+\infty$ 与 $-\infty$ 在过渡到 $L_2$ 时被等同为单一的 $\hat\infty$, 并说该等同"由拓扑学强制".

本文精化这一点.

拓扑并不唯一强制一点紧化. 一个局部紧 Hausdorff 空间可以有多种紧化——一点紧化、Stone–Čech 紧化、端点紧化等——它们彼此不同构, 而拓扑学本身不在其中作出选择.

一点紧化是在亚纯与共形语境中被选择的规范闭合. 选择它的理由是具体的: 它使亚纯函数在 $\infty$ 处有良定义的行为 (§ 6.2 的 $\mathrm{ord}_\infty$ 依赖于此), 使 $z \mapsto 1/z$ 成为球面的自同构 (§ 5.3 的 $0 \leftrightarrow \hat\infty$ 边界对依赖于此), 使 $\mathbb{P}^1(\mathbb{C})$ 成为紧 Riemann 面 (§ 6.3 的全局留数定理依赖于此).

该选择有代价: 所有趋向无穷的方向被压成一点, 方向分辨率丢失.

该代价在本文有确定的到期时间. 不规则奇点处需要方向分辨率 (§ 7.4), 而一点紧化已经把它压掉了; 故须在该点作实定向吹起, 把被压成一点者重新展开为方向圆. § 8.1 阐述由此得到的两级闭合结构.

故本文对 Paper 1 § 5.2 的精化包含两部分: 把"拓扑强制"改述为"语境中的规范选择", 并把该选择的代价追踪到它到期的地方. 这是精化而非撤回——Paper 1 所阐述的等同本身不变, 变的是该等同的模态 (强制 vs 选择) 与其代价的可见性.

〔类型: 上游精化. § 11.1 第一项.〕

---

5. 主步骤 3: 乘法路径——覆叠与旋转记忆

5.1 主步骤 3 与 inevitability 3

Paper 3 § 1.2 立: inevitability 3 (比较不能孤立, 它传播; 不能不进一步发展) 对应主步骤 3 (乘法路径). 在 $L_1$, 该对应的实质内容是: 加法路径阐述完毕之后, 发展转向重复加法的更高阶阐述, 经 (计数, 单位) 绑定阐述为乘法路径.

在 $L_2$, 主步骤 2 闭合于 $\mathbb{C}$, 其余项是: 加法路径不阐述旋转.

一条归属声明: 本主步骤标题中的"旋转记忆"归覆叠, 不归乘法. $\mathbb{C}^\times$ 的乘法合成旋转, 但不记录绕数——绕数是覆叠空间才有的量 (§ 5.3). 内层 2 与内层 3 交付的是旋转的合成与分解, 内层 4 交付的才是记忆. 分量加法 $(a+bi)+(c+di) = (a+c)+(b+d)i$ 逐分量迭代; 它阐述平面上的平移, 不阐述平面上的转动. 主步骤 3 是对该余项的阐述回应.

登记类别: 层内 sub-fractal (§ 2.1). 与 Paper 3 的 $L_1$ 主步骤 3 一样, 本主步骤的内层 4 含嵌套级联, 故其 sub-fractal 深度较主步骤 2 深一档; § 5.6 阐述该不对称.

5.2 内层步骤 1 至 3

5.2.1 内层 1: 旋转余项迫使乘法 sub-mark

主步骤 2 内层 4 留下的余项是: 加法路径不阐述旋转. 主体作为有限存在, 经 Via Negativa 拒绝阐述停在只有平移的状态, 阐述 "$\times$" 作为 sub-mark.

此处须注意 $\mathbb{C}$ 上的乘法不是 $L_1$ 乘法的直接延续. $L_1$ 的乘法路径作用于一维直线, 只允许伸缩; $\mathbb{C}$ 的乘法允许伸缩与转动两者. 新的 sub-mark 承载的阐述内容因此严格多于它在 $L_1$ 的同名者.

5.2.2 内层 2: 模长相乘、旋转合成

资源纪律: 本内层只调用纯代数手段. 实三角函数、指数函数与任何加性角度参数, 皆不在此处调用——前者是 $L_1$ 未捕获的对象 (§ 8.2 第二、三类; § 11.3), 后者是内层 4 的产物 (§ 5.3). 本内层所交付的两项, 都由 $\mathbb{C}$ 上的代数结构直接得出.

其一, 模长相乘. 由 $|z|^2 = z\bar{z}$ 与共轭的乘法性, 直接得

$$|zw| = |z|\,|w|.$$

模长 $|z|$ 延续 $L_1$ 主步骤 3 的乘法路径 (幅度相乘), 无新内容.

其二, 旋转合成. 记 $S^1 = \{w : |w| = 1\}$. 由上式, $S^1$ 在乘法下闭合且含逆, 故是乘法群. 而对 $|w| = 1$, 映射 $z \mapsto wz$ 是保模的 $\mathbb{R}$-线性映射, 其行列式为 $|w|^2 = 1$, 故属 $SO(2)$.

旋转由线性代数得出, 不需三角. 这是 $L_2$ 处真正新颖的内容: $L_1$ 的乘法只允许伸缩, 而 $\mathbb{C}$ 的乘法把伸缩与 $SO(2)$ 中的旋转合成分作两笔独立的账.

常数 $i$ 在此阐述下, 最好不被理解为"以某种方式不实在的虚数", 而是 $S^1$ 中的一个四阶元: $|i| = 1$ 故 $i \in S^1$, 且乘以 $i$ 把 $1 \mapsto i \mapsto -1 \mapsto -i \mapsto 1$; $i^2 = -1$ 是该四阶性的代数陈述. 此处不需要"四分之一转"这一角度语言——四阶性是群论事实, 角度是内层 4 才有的参数. $i$ 在实直线上的非实在性, 是其作为旋转生成元这一功能的下游后果——$L_1$ 不能阐述旋转, 故旋转生成元不能位于 $L_1$ 的阐述范围之内.

本内层留下代价余项. 若每次乘法都在分量形式下展开,

$$(a+bi)(c+di) = (ac-bd) + (ad+bc)i,$$

则单次乘法需四次实乘法与两次实加法; 迭代 $k$ 次的阐述长度随 $k$ 线性增长, 且每一步都须重新分离模长与旋转两笔账. 这与 Paper 3 § 5.2.2 阐述的重复乘法展开代价同型: 不是逻辑余项 (分量形式在 $\mathbb{C}$ 内逻辑闭合), 而是阐述代价性质的余项.

5.2.3 内层 3: 极分解的绑定

为使乘法迭代可被紧凑阐述, 主体经 Via Negativa 拒绝该阐述代价, 阐述极分解

$$z = r\,u, \qquad r = |z| \in \mathbb{R}_{>0},\quad u = z/|z| \in S^1$$

作为绑定. 两笔已分离的账 (模长与旋转) 经该形式合成单一 token; 在该形式下乘法即 $(r_1 r_2,\ u_1 u_2)$——两个独立的乘法, 不再展开分量, 亦不需要任何角度参数.

这是跨迭代绑定, 与 $L_1$ 主步骤 3 内层 3 的幂记号 (跨重复乘法绑定) 同位. Paper 3 § 5.2.3 把幂记号绑定阐述为 Via Rho 在绑定处的实例化, 绑定运作产生不可分的两面; 极坐标绑定同样:

• 一面是绑定所阐述的处理能力: 幂在该形式下一步给出, $z^n = r^n u^n$; 模长的会计与旋转的会计彻底分离, 各自在自己的群中进行;
• 另一面是绑定所阐述的余项: 旋转群 $S^1$ 尚未被加性参数化. $S^1$ 是乘法群——它的元素之间可以相乘, 但没有"相差多少"这一加性度量. 迭代 $u^n$ 可以做, 却无法回答"绕了几圈"或"$u_1$ 与 $u_2$ 相隔多远".

两面都是绑定运作所产生的实质内容; 本文不作价值判断 (承 Paper 3 § 4.2.3 纪律).

该余项在内层 4 被处理, 且是被一次处理的. 内层 4 的第二级反向追问产出覆叠映射 $\exp$, 而 $\exp$ 一次同时给出三样: $S^1$ 的加性参数化 (角度 $\theta$ 由此才有定义)、该参数化的 $2\pi$ 不确定性、以及覆叠变换群 $2\pi i\mathbb{Z}$ (§ 5.3).

这条接续是本主步骤内部 sub-fractal 的实质连接: 绑定所留的余项 (缺加性参数) 与反解所产出的结构 (覆叠) 是同一件事的两端. 加性角度在本文中只有这一个来源.

一条顺序登记. 本内层的绑定取极分解 $z = ru$ 而非三角形式 $r(\cos\theta + i\sin\theta)$, 亦非指数形式 $re^{i\theta}$. 这不是记法偏好, 而是资源纪律 (§ 5.2.2):

• 复指数是内层 4 第二级反向追问的产物 (§ 5.3), 在本内层尚不可用;
• 实三角函数同样不可用. 它们不是 $L_1$ 有限代数模式的产物 (§ 2.1), 而 $\pi$ 按 Paper 3 § 6 的账本是 $L_1$ 能命名不能捕获的基底内超越余项 (§ 8.3); 响应完备性余项的有效 Cauchy 模式则被判为并行 side-branch, 不在本条 trajectory 上 (§ 11.3). 在脊柱上进口本文别处判为未捕获的对象, 会使 § 5.4 的核心陈述失去独立性.

该纪律有一项直接后果: $2\pi$ 在本文只有一个来源, 即内层 4 的 $\exp$ 之核 (§ 5.3), 而非从 $\cos$、$\sin$ 的周期性进口. § 5.5 的 $2\pi i$ 双重身份与 § 8.3 把 $\pi$ 阐述为覆叠变换群, 皆依赖此唯一来源.

5.3 内层 4: 嵌套级联的两级反向追问

主步骤 3 经 inevitability 4 (不能不被追问) 达成闭合. 在乘法路径的语境中, 追问的具体形式自然取反向追问的形式; 而与 Paper 3 的 $L_1$ 主步骤 3 一样, 此处的反向追问不是单次而是嵌套级联: 第一级反向追问产出一个新的阐述空间, 而该空间内仍有未闭合的余项, 进一步的反向追问产出第二层结构.

第一级反向追问: 乘法能否反过来

可以, 阐述除法. 除法在 $0$ 处不闭合, 故阐述空间为

$$\mathbb{C}^\times = \mathbb{C}\setminus\{0\}.$$

边界结构在此出现一项 $L_1$ 无对应者的现象: 在变换 $z \mapsto 1/z$ 下, $0$ 与 $\hat\infty$ 被彼此交换, 形成 Riemann 球面上乘法求逆的二周期边界对 (Paper 1 § 5.2). $L_1$ 的乘法路径边界标记是 $0$, 而其加法路径边界标记 $\pm\infty$ 与之无对偶关系; $L_2$ 的二维化把这两个边界标记配成一对.

第一级闭合于 $\mathbb{C}^\times$.

第二级反向追问: 幂能否反过来

可以, 但反解分两支, 二者的余项类型不同. 这是本节的实质内容.

其一, 有限代数反向: $z \mapsto z^n$ 的反解.

$n$ 次根. 每个非零复数有 $n$ 个 $n$ 次根, 故反解是 $n$ 值的. 分支点在 $0$ 与 $\hat\infty$; 绕分支点一周, 诸根之间发生置换, 单值群是有限群 (循环群 $\mathbb{Z}/n$).

关键之处在于: 所有根仍在 $\mathbb{C}$ 内. 由代数基本定理, $\mathbb{C}$ 上任何非常数多项式有根. 故多项式求根这一反向追问, 在 $L_2$ 不再迫使标量基底扩张——这与 $L_1$ 的同位步骤形成对照 (§ 5.4).

但反解仍需结构. 要在 $n$ 个根中选定一个并使选择在延拓下保持一致, 必须引入覆叠或分支结构. 反解不扩张基底, 却要求在基底之上加一层结构——这是压力改变方向的第一处征兆.

其二, 超越反向: $\exp$ 的反解.

对数. 映射 $\exp: \mathbb{C} \to \mathbb{C}^\times$ 是满射但非单射:

$$\exp(z) = \exp(z + 2\pi i k), \quad k \in \mathbb{Z}.$$

故 $\log z$ 是无限多值的: $\log z + 2\pi i k$. 这正是内层 3 所留余项 (旋转群缺加性参数化) 被处理的方式: $\exp$ 给出 $S^1$ 的加性参数, 而该参数的代价就是它的 $2\pi$ 不确定性. 加性角度与其不确定性同时诞生, 不是先有角度再发现它多值.

该结构的完整形式是群的正合列:

$$0 \longrightarrow 2\pi i\mathbb{Z} \longrightarrow (\mathbb{C}, +) \stackrel{\exp}{\longrightarrow} (\mathbb{C}^\times, \cdot) \longrightarrow 1.$$

这条正合列一次阐述五件事: $2\pi i$ 的来源 (核); 对数的多值性 ($\exp$ 非单); 单值群为 $\mathbb{Z}$ (核同构于 $\mathbb{Z}$); 绕行产生 $2\pi i k$ 记忆 (核在纤维上的作用); 以及覆叠取代基底扩张 ($\mathbb{C}$ 单连通, 故 $\exp$ 是 $\mathbb{C}^\times$ 的万有覆叠).

构与余项

构 $=$ $\mathbb{C}^\times$ 的万有覆叠, 覆叠变换群 $2\pi i\mathbb{Z}$.

余项 $=$ 路径依赖. 解析延拓的结果依赖于路径的同伦类. 该余项不是本主步骤内部可再化解的——反向追问在 $\mathbb{C}$ 内已经用尽. 它是主步骤 4 的周期与单值群所要阐述的对象 (§ 6.4).

5.4 核心结构陈述: 代数闭性消除标量缺口, 不消除表示缺口

本节把 § 5.3 与 Paper 3 § 5.2.4 放在同一格位上比较. 两处都是主步骤 3 内层 4 的第二级反向追问, 但产物类型不同.

$L_1$: 幂的反解 (开方与多项式求根) 在 $\mathbb{Q}$ 内不闭合. 压力的释放方式是扩张标量基底: $\mathbb{Q} \to \mathbb{Q}^{\rm alg}_{\mathbb{R}}$. 主步骤 3 闭合于实闭域. 基底变大了.

$L_2$: 幂的反解, 就多项式求根而言, 在 $\mathbb{C}$ 内已经闭合——代数基本定理. 压力无处向外释放, 因为标量基底在代数意义上已满. 压力的释放方式转为向上: 加一层覆叠.

由此得到本文识别的结构不对称:

> 代数闭性消除了标量缺口, 但没有消除表示、路径与单值化缺口.

两点须说明.

其一, 不对称不是深度上的量差, 而是产物类型上的差别. $L_1$ 的反向追问产出一个更大的数域——标量对象. $L_2$ 的反向追问产出一个覆叠空间与一个群——拓扑对象与表示对象. 若只看"哪一层的反向追问更复杂", 会把这条不对称读成程度问题; 实际的差别在于压力被释放到了不同种类的对象上.

其二, 这条不对称是 $\omega_\rho^{(2)}$ 成为必要工作内容的结构条件. 若 $L_2$ 的反向追问也以扩张基底告终, 则主步骤 4 只需登记一个新的数域, 无须一整套操作缺陷代数. 此处是必要条件而非因果解释: 不对称的存在使缺陷代数成为有意义的工作内容, 而非"因为有不对称, 所以必须有缺陷代数". 缺陷代数之所以是 $L_2$ 的核心工作内容 (Paper 1 § 5.4), 正因为压力被抬成了覆叠、单值群、周期与 Stokes 数据这一族对象. 换言之, § 6 至 § 8 的全部技术内容, 其存在理由在本节被阐述.

阐述链条:

$$i \ \longrightarrow\ \mathbb{C} \ \longrightarrow\ \mathbb{C}^\times \ \longrightarrow\ \widetilde{\mathbb{C}^\times} \ \longrightarrow\ 2\pi i\mathbb{Z} \ \longrightarrow\ \mathrm{Mon}$$

从主步骤 1 的复化种子, 经二维化与乘法路径, 到万有覆叠、覆叠变换群, 再到主步骤 4 的单值群. 该链条不跨出 $\mathbb{C}$ 一步.

登记后果. $\exp$ 在此被阐述为主步骤 3 内层 4 所产出的覆叠映射, 即 $L_2$ 的内部产物. 据此, 本文在当前 trajectory 内采用内部产物读: 指数方向作为 $L_2$ 的闭合余项 (§ 7.5), 而非作为 $L_1 \to L_2$ 跃迁映射的余项. 跃迁读不再作为本篇的主登记; 本文不据此判断跃迁读在其他 trajectory 中的地位 (承 Paper 2 反本质化纪律).

5.5 $2\pi i$ 的双重身份

$2\pi i$ 在本文中出现于两处. 二者不是类比, 是同一条陈述的两种呈现.

主步骤 3: $2\pi i\mathbb{Z}$ 是 $\exp: \mathbb{C} \to \mathbb{C}^\times$ 的核, 即万有覆叠的覆叠变换群.

主步骤 4 内层 2: $2\pi i$ 是留数归一中的因子, $\oint_{|z|=1} dz/z = 2\pi i$ (§ 6.3).

二者的同一性: 沿逆时针单位圆绕行一周, $\log z$ 增加 $2\pi i$. 左边读作覆叠变换群的一个生成元在纤维上的作用, 右边读作 $dz/z$ 沿该圈的积分. 是同一件事的两种记法.

这条同一性有两处用处.

其一, 它把 § 5 与 § 6 用一条恒等式锁住, 而不是靠章节顺序并置. Paper 1 § 4.5 把 $2\pi i$ 识别为 $L_2$ 的阐述交换单位 (本文称层内交换律阐述单位, 以与跃迁签名槽相区别, 见 § 11.5.3); 本节阐述该单位不只是留数演算的产物, 而是整个覆叠结构的产物. 交换律之所以能在 $L_2$ 成立, 其结构根据在主步骤 3.

其二, 它加强 § 10 对 Cayley–Dickson 的判据. 若签名只是留数定理中的一个因子, 则"付掉可交换性即付掉签名"是一条关于某个公式的观察; 而既然签名同时是覆叠变换群的生成元, 该判据所涉及的就是整层的覆叠结构, 而不只是某一条积分公式.

欧拉公式的位置. 欧拉公式

$$\cos\theta + i\sin\theta = e^{i\theta}$$

两边分属基底与模式两侧, 经两条来源不同的路线产出.

左边是基底级的经典构造: 实三角函数是 $\mathbb{R}$ 上分析的既有对象, 不是本文这条 trajectory 上任何一格的产物 (§ 5.2.3 资源纪律明确不在脊柱上调用它们).

右边是内层 4 经第二级反向追问阐述出的覆叠映射 $\exp$ 在虚轴上的限制: 即 $\mathbb{R} \to S^1$ 的万有覆叠参数化, 覆叠变换群 $2\pi\mathbb{Z}$.

欧拉公式陈述这两个对象相等. 而 $e^{i\pi} + 1 = 0$ 是该等式在 $\theta = \pi$ 处的取值, 另加 $L_1$ 主步骤 4 的余项 $\pi$ 与 $L_1$ 的两个边界 token.

这是欧拉公式为叠加落点而非单点落点的结构理由: 它不住本 trajectory 的任何一格, 它是基底级经典构造与模式级产物之间的一次认同. 该认同所跨的是基底与模式这条本文的总防火墙 (§ 2.1). 本文因此把它登记为基底—模式叠加落点, 而非层内的跨格位认同. 仅以成分清点为据 (它同时含 $e, \pi, i, 1, 0$) 只能得到位置描述; 本节给出的是结构理由.

落点与闭合方程是两个不同的谓词, 不相竞争. 落点回答"它住哪"; 闭合方程回答"它闭合什么". 欧拉公式同时是两者: 就落点而言, 它是基底—模式叠加落点 (本节); 就闭合而言, 它是 $L_1 \to L_2$ 跃迁的闭合方程, 闭合该跃迁的两项余项 $i$ 与 $\pi$ (§ 8.3).

须注意这里的"闭合"不是层的闭合. $L_2$ 这一层的闭合是粗闭合加方向细化 (§ 8.1), 与跃迁闭合是不同对象. Paper 1 § 5.3 阐述欧拉公式"是闭合的实例, 而非 $L_2$ 的完整闭合", 其准确内容即在此: 它是跃迁闭合的实例, 不是层闭合. 完整落点阐述在 § 8.3.

须注意它不是覆叠的截面. 若存在全局截面 $s: \mathbb{C}^\times \to \mathbb{C}$ 使 $\exp \circ s = \mathrm{id}$, 那就是一个全局单值的对数, 而它恰恰不存在——这正是 § 5.3 第二级反向追问的实质内容. 欧拉公式是覆叠结构的规范参数化见证, 不是截面. 其基底—模式叠加落点的完整阐述见 § 8.3.

5.6 sub-fractal 深度不对称

Paper 3 § 5.3 阐述 $L_1$ 的一项 sub-fractal 深度不对称: 主步骤 3 内层 4 含嵌套级联 (除法 → $\mathbb{Q}$; 开方与多项式求根 → $\mathbb{Q}^{\rm alg}_{\mathbb{R}}$), 而主步骤 2 内层 4 只有单次反向追问 (减法 → $\mathbb{Z}$). 故乘法路径的 sub-fractal 较加法路径深一档. 该文阐述此深度不对称不是阐述的人为产物, 而是实质事实: 代数阐述在完整精确化路径中的最大实质内容出现在乘法路径.

$L_2$ 重现同一不对称, 且格位相同. 主步骤 2 内层 4 单次闭合于 $\mathbb{C}$; 主步骤 3 内层 4 是嵌套级联 (除法 → $\mathbb{C}^\times$; 开方与对数 → 覆叠).

但两层深处的内容不同. $L_1$ 深在扩张标量基底; $L_2$ 深在抬成覆叠 (§ 5.4). 故深度不对称这一形式在两层重现, 而不对称所承载的实质内容随层变化. 这与 Paper 3 对该不对称的读法一致: 它不是形式自相似, 而是实质内容在该格位上最厚. 形式在此是实质内容的指示器, 不是实质内容的来源.

此与 Methodology 6: 相变窗口 阐述的 $r \approx 5$ 拓扑不对称在模式上一致: 四步不均匀, 中间步骤 (主步骤 2 与 3) 打开阐述空间, 末步 (主步骤 4) 闭合之. 乘法路径在中间步骤的位置进一步打开, 与它所承载的实质内容相称.

---

6. 主步骤 4(a): 正则周期核心

6.0 主步骤 4 的组织

主步骤 4 是 $L_2$ 的闭合与余项登记. 按 § 2.1 的登记声明, 它的类别是余项账本结构, 不是层内 sub-fractal; 其内部按"捕获核心 → 未捕获账本 → 下一层触发条件"组织, 不按四相重演组织.

主步骤 4 横跨三节: 本节 (§ 6) 阐述正则周期核心, § 7 阐述不规则边界, § 8 阐述闭合与余项账本. 三节各占一节而主步骤 1 至 3 各占一节, 这一版面不对称本身即上述登记 (§ 2.1).

§ 6 与 § 7 的关系是核心与增强, 不是互斥分箱. 这一点须先说明, 否则本节的阐述会被读成"除子、留数、周期、单值群只在正则情形存在". 实际上不规则联络同样有极点与除子、有形式留数、有单值群、有整体周期数据; § 7 所增加的是指数因子、方向与 Stokes 数据. 关系是

$$\{\text{正则核心}\} \hookrightarrow \{\text{不规则增强}\},$$

而不是两个并列的箱子. 本节阐述核心, § 7 阐述增强所加者.

6.1 承载体迁移

在进入四阶段的技术内容之前, 须先登记一项类型上的变化. 它不是本节的技术前提, 而是本文的一项识别.

Paper 3 阐述 $L_1$ 模式的捕获范围是实闭域 $\mathbb{Q}^{\rm alg}_{\mathbb{R}}$: 一个在代数运算下闭合的标量域. 本文阐述 $L_2$ 模式的捕获核心是普通周期环 $\mathcal{P}$: 积分表达式经取值映射所得的像 (§ 2.3). 二者不是同一种对象.

> 承载体迁移: $L_1 \to L_2$ 之间, 捕获范围的承载体从标量域的代数闭合, 迁移到几何积分语言的取值像.

该识别的档次须先标明. 它是对两个既有数学定义 (实闭域; Kontsevich–Zagier 周期) 的对照性重新框定, 属 SAE 登记, 不是数学定理, 亦不主张发现了任何本体论跃迁. 本文之所以仍把它列为核心识别, 是因为它的三项后果在全文技术层面实际生效 (见下), 而非因为它本身构成新的数学内容.

这一迁移有三项后果, 全文多处依赖之.

其一, "饱和"一词不再适用. 标量域可以说饱和——它对某族运算封闭, 且是含给定起点的最小者. 取值像没有对应的封闭性: $\mathcal{P}$ 是什么, 由语法侧 $\mathsf{Int}_{\rm alg}$ 决定, 不由 $\mathcal{P}$ 自身的运算封闭性决定. 故本文一律说取值核心, 不说饱和点 (§ 2.6 其二).

其二, 捕获判据的类型变了. $L_1$ 的判据是"满足什么方程", $L_2$ 的判据是"有什么表示" (§ 2.1). 前者是一个数与一族方程的关系, 后者是一个数与一族几何积分表达式的关系. 这条差别在 § 8.2 与有效 Cauchy 的算法可逼近性判据并列时再次出现: 三条判据互不为对方的特例.

其三, 环与域的差别不是疏忽. $\mathbb{Q}^{\rm alg}_{\mathbb{R}}$ 是域; $\mathcal{P}$ 是环, 而它是否对取逆封闭未决——$1/\pi \in \mathcal{P}$ 至今未知, 故文献另设扩张 $\hat{\mathcal{P}} = \mathcal{P}[1/\pi]$. 读者会自然追问: 主步骤 3 已经阐述了除法并闭合于 $\mathbb{C}^\times$, 为何取值核心却止于一个环?

答案在 § 2.2 的区分: 主步骤 3 的除法闭合是函数与路径层面的闭合; $\mathcal{P}$ 对取逆是否封闭是数值集合层面的问题. 二者分属不同层面, 前者不推出后者. 本文不主张函数层面的除法闭合迫使数值集合对取逆封闭.

二轨提醒. 本节所阐述的除子—留数—周期—单值群是曲线上的解析-覆叠脊柱; 而 $\mathcal{P}$ 的定义允许任意有限维代数簇与任意维半代数区域. 按 § 2.3 的二轨声明, 本文不主张前者生成全部 $\mathcal{P}$. 本节阐述的是四阶段在跨层比较中的位置, 不是 $\mathcal{P}$ 的生成方式.

6.2 点态: 除子与赋值

第一阶段的动作是: 把多值与奇点所留下的缺陷, 登记为离散点上的整数.

赋值. 对亚纯函数 $f$ 与点 $p$, $\mathrm{ord}_p(f)$ 是 $f$ 在 $p$ 处的零点阶 (正) 或极点阶 (负). 这是一个 $\mathbb{Z}$ 值的局部量.

除子. $\mathrm{div}(f) = \sum_p \mathrm{ord}_p(f)\cdot p$ 把各点的赋值打包为自由阿贝尔群 $\mathrm{Div}(X)$ 的一个元素. 主除子构成子群, 商群 $\mathrm{Cl}(X)$ 是除子类群 (Paper 1 附录 B.1).

须注意赋值与除子是不同层级的对象: 赋值是点态的整数, 除子是它们的形式和. 这条区分在下一段与 § 9.1 的类型表中都要用到.

ZFCρ 债务的第一半.

$L_1$ 的 $\infty$ 是单块的: $L_1$ 阐述模式没有工具分辨不同的发散. 主步骤 2 把 $L_1$ 的 $+\infty$ 与 $-\infty$ 等同为单一的 $\hat\infty$ (§ 4.3), 这一步不但没有减少单块性, 反而加强了它——两个边界标记被合并为一个.

赋值在此给 $\hat\infty$ 装上刻度. $\mathrm{ord}_\infty$ 是 $\mathbb{Z}$ 值的, 故 $\hat\infty$ 不再是无内部结构的一点, 而带有一个整数分级. 两个亚纯函数在 $\infty$ 处之比, 其阶为 $\mathrm{ord}_\infty(f) - \mathrm{ord}_\infty(g)$——曾在 $L_1$ 处导致塌缩的 $\infty/\infty$, 在此成为整数减法.

但这只清了阶那一半. ZFCρ 所欠的是 fine ratio, 而 fine ratio 是系数层面的量; 赋值分辨阶, 不分辨系数. 两个在 $\infty$ 处同阶的函数, 其渐近系数之比不被 $\mathrm{ord}_\infty$ 分辨.

系数那一半的接口在 § 7.4 定位到 Newton 多边形. 但按 § 2.6 其六, 有接口不等于已接通; 本文不宣称该债务已清, 只把它从"未接口"改登记为"有接口而未接通".

附带登记: 与 ZFCρ 真正接口的是赋值, 不是除子——除子只是把各点的赋值打包. 本文若说"除子给 $\infty$ 装刻度", 是把打包动作误认为分级动作.

6.3 局部: 留数与 $2\pi i$

第二阶段的动作是: 把点态的整数刻度兑换为可积分的数值.

留数. $\mathrm{Res}_p(\omega)$ 是作用于亚纯微分形式在点 $p$ 处局部数据的线性泛函, 输出一个数值 (§ 2.2 算子与数值的区分). 紧 Riemann 面上的全局留数定理给出约束 $\sum_p \mathrm{Res}_p(\omega) = 0$.

与第一阶段的接口. 对数微分给出

$$\mathrm{Res}_p\!\left(\frac{df}{f}\right) = \mathrm{ord}_p(f),$$

故

$$\deg\,\mathrm{div}(f) = \sum_p \mathrm{ord}_p(f) = \sum_p \mathrm{Res}_p\!\left(\frac{df}{f}\right) = 0.$$

即: 主除子次数为零, 是全局留数定理对对数微分 $d\log$ 的一个特殊化. 须注意这条接口只经 $d\log$ 这一张映射成立; 不可笼统说所有除子定理与所有留数定理是同一守恒的两种记法. § 9.2 把 $d\log$ 登记为跨分量接口之一.

$2\pi i$ 作为层内交换律阐述单位. 留数的兑换率是

$$\oint_{|z|=1} \frac{dz}{z} = 2\pi i.$$

§ 5.5 已阐述: 这与 $\exp$ 的覆叠变换群 $2\pi i\mathbb{Z}$ 是同一条陈述的两种呈现——沿逆时针单位圆绕行一周, $\log z$ 增加 $2\pi i$. 故本阶段的兑换率不是留数演算内部的常数, 而是主步骤 3 所产出的覆叠结构在积分语言中的呈现.

Paper 1 § 4.5 立 $2\pi i$ 为 $L_2$ 的阐述交换单位. 该交换律在本阶段的具体形式是留数定理与辐角原理: 由

$$\frac{1}{2\pi i}\oint_\gamma \frac{f'}{f}\,dz = N - P,$$

"沿闭路积分"与"清点零点与极点"这两条不同的阐述路线可以交换次序, 而交换的代价恰是 $2\pi i$ 这一单位. 这正是 Paper 1 § 4.5 所阐述的交换律形式: 不是任意两个运算可交换, 而是两条阐述路线在某个单位下可交换.

"局部-货币化"一词在此不是比喻. 第一阶段产出的是整数 (阶); 本阶段把它兑换为积分量, 兑换率为 $2\pi i$; 而该兑换率同时是本文这条 trajectory 的层内交换律阐述单位. 兑换、汇率、以及汇率即本层身份, 三者齐备.

(记号纪律: 本文有两个签名槽——跃迁签名与层内交换律阐述单位, 见 § 11.5.3. 全篇不裸称"$L_2$ 的签名".)

6.4 整体: 周期与单值群

第三阶段的动作是: 从局部数据过渡到整体数据.

单值群. $\mathrm{Mon}$ 是基本群 $\pi_1$ 的表示. 主步骤 3 内层 4 留下的余项——路径依赖 (§ 5.3)——在此被阐述为对象: 沿不同同伦类的路径解析延拓, 结果之差由该表示刻画. 对 $\log$, 该表示是 $\pi_1(\mathbb{C}^\times) = \mathbb{Z}$ 在纤维上以 $2\pi i\mathbb{Z}$ 平移作用; 对 $n$ 次根, 是 $\mathbb{Z}$ 经 $\mathbb{Z}/n$ 作用. § 5.3 的两级反向追问在此各自得到其表示。

周期. $\mathrm{Per}$ 是 de Rham 上同调与 Betti 同调之间的配对: $\int_\gamma \omega$, 其中 $\gamma \in H_1$, $\omega \in H^1_{\rm dR}$. 这是本文取值核心 $\mathcal{P}$ 的几何来源.

与第二阶段的接口, 及其方向. 小环路上的配对给出

$$\oint_{\gamma_p}\omega = 2\pi i\,\mathrm{Res}_p(\omega).$$

即: 小环路周期是 $2\pi i$ 乘以留数. 方向须守住——不可由此式除以 $2\pi i$ 反推留数属于 $\mathcal{P}$, 因为 $1/\pi \in \mathcal{P}$ 未决 (§ 6.1 其三). 若 $\omega$ 是在 $\overline{\mathbb{Q}}$ 上定义的代数亚纯微分形式且 $p$ 是代数点, 则 $\mathrm{Res}_p(\omega)$ 是代数数, 因而属于 $\mathcal{P}$——但那是经 $\overline{\mathbb{Q}} \subset \mathcal{P}$ 的另一条理由, 不是从上式除得的.

这条方向纪律是 § 6.1 第三项后果 (环非域) 在技术层面的第一次实际生效.

$\mathcal{P}$ 作为取值核心. 全部代数数属于 $\mathcal{P}$, 故 $\mathcal{P}$ 在包含关系上接在 $\mathbb{Q}^{\rm alg}_{\mathbb{R}}$ 之上. 但按 § 6.1, 这个"接在之上"只是集合包含, 不是同类型对象的延续: 前者是代数闭合的标量域, 后者是取值像. 承载体迁移正发生在这个包含关系之内, 而包含关系本身会掩盖它.

三条边界须同时登记:

• $\mathcal{P}$ 可数, 而 $L_2$ 基底不可数, 故 $\mathcal{P}$ 不穷尽基底;
• $\mathcal{P}$ 连可计算实数都不穷尽 (§ 8.2);
• 本节的曲线脊柱不生成全部 $\mathcal{P}$ (§ 2.3 二轨声明).

本阶段的余项. 在正则情形下, $\mathrm{Per}$ 与 $\mathrm{Mon}$ 联合阐述了路径依赖: 单值群刻画绕行的效果, 周期给出积分的取值. 就正则奇点而言, 主步骤 3 留下的余项在此已被阐述为对象.

但这只覆盖正则奇点. 在不规则奇点处, 形式解含指数因子, 而单值群不足以刻画其行为——同一个形式解在绕行奇点时, 其渐近有效性会在不同扇区之间失配. 这是 § 7 的入口, 也是第四阶段所增加者.

---

7. 主步骤 4(b): 不规则边界

7.0 本节所增加者

§ 6 末阐述: 在正则情形下, 周期与单值群联合阐述了主步骤 3 留下的路径依赖余项. 本节处理该阐述所不覆盖的部分.

按 § 6.0 的关系登记, 本节不是与 § 6 并列的另一个箱子. 不规则联络同样有极点与除子、有形式留数、有单值群、有整体周期数据; 本节所增加的是三样: 指数因子、方向、以及 Stokes 数据.

$$\{\text{正则核心}\} \hookrightarrow \{\text{不规则增强}\}$$

本节同时是本文承重的所在. § 2.6 其三已声明本文不以 $e \notin \mathcal{P}$ 承重; 承重由 § 7.1 的两条扩张共同承担——语法侧是定义级, 函数侧是定理级; 两侧被读作同一边界这一登记, 另列为 T2-条件项.

7.1 表达式类与联络的两条扩张

本文的 T2 承重是以下两条扩张.

语法侧:

$$\mathsf{Int}_{\rm alg} \hookrightarrow \mathsf{Int}_{\exp}$$

即在被积函数中另允许 $e^{-f}$ 型因子 ($f$ 代数). 这是 § 2.3 表 1 的左竖箭头.

函数侧:

$$\{\text{正则奇点联络}\} \hookrightarrow \{\text{允许不规则奇点的联络}\}$$

后者的分类由 Levelt–Turrittin 定理给出: 形式穿孔圆盘上的任一联络, 可分解为只含单个指数因子的初等联络之直和——含一个分歧指标为 $\nu$ 的分歧覆盖、形如 $E^\varphi = (\mathbb{C}((t)),\, d + d\varphi)$ 的指数部分, 以及一个正则联络 $R$.

两条扩张的认识论档次不同, 须分三档登记.

档	内容	性质
定义级承重	普通积分语法可扩张为指数积分语法 ($\mathsf{Int}_{\rm alg} \hookrightarrow \mathsf{Int}_{\exp}$)	定义／语法; 由 $\mathcal{P}$ 与 $\mathcal{P}^e$ 的定义直接给出
定理级承重	正则联络可扩展到允许不规则奇点的联络; 后者需要指数因子、方向与 Stokes 数据	定理 (Levelt–Turrittin 形式分类等)
条件性桥接	上述两条扩张是同一 SAE 边界的数值侧与函数侧呈现	T2-条件项; 属 exponential motives 一线的研究级对应

三档不可混写. 特别是: Levelt–Turrittin 支撑的是第二档, 不是第一档, 更不是第三档. 第一档由定义给出, 无须定理; 第三档是本文在 SAE 框架内作出的对齐登记, 不是由第二档推出的数学结论.

本文的 T2 承重由第一档与第二档共同承担; 第三档单列为条件项, 若该对齐在未来被证伪, 须改动的是本文把两侧读作同一边界这一登记, 而非两条扩张本身.

这条区分是本文承重结构的关键. $e \notin \mathcal{P}$ 是开放猜想, 若以之承重, 猜想被证否则结构随之倒塌. 承重放在扩张上, 则该猜想的任何结局都不动摇本节 (§ 12.3).

7.2 Newton–Turrittin 数据作为不规则边界的形式组织器

不规则奇点处的形式数据可由 Newton 多边形组织. 但组织的范围必须精确划定: Newton 多边形组织形式分类, 不决定解析分类. 本节按三层阐述, 三层是三个不同的规定层级, 不构成一条因果链.

第一层 · 形式粗层: 斜率

联络的斜率由其微分算子 Newton 多边形边界的斜率给出, 且与算子的选取无关. 由此得到本文最基本的一条判据:

> 所有斜率为零者即正则奇点; 存在正斜率者即不规则奇点.

这正是 § 7.1 那条函数侧扩张的判据形式.

各边的斜率恰是决定多项式的次数 (含平凡多项式 $q \equiv 0$, 其次数为零); 若斜率为 $\lambda$ 的边水平长度为 $m$, 则有 $m$ 个次数为 $\lambda$ 的决定多项式.

本层给出: 正则与不规则的分界、不规则度的粗粒度、分歧次数、Gevrey 层级, 以及解在奇点处的增长阶.

第二层 · 形式细层: 完整指数因子

指数因子的完整形式是 Puiseux 多项式

$$q = \sum_{k=1}^{s} a_k\, z^{-k/\nu}, \qquad a_s \neq 0.$$

斜率只管指数 $k/\nu$; 系数 $a_k$ 在斜率之外. 这是本节的核心区分.

由本层而非第一层决定的有两样:

• 增长的型. 第一层给出阶 (由次数决定); 型依赖首项系数 $|a_s|$, 斜率不决定它.
• 候选 Stokes 方向. 方向由指数因子之差 $q_\alpha - q_\beta$ 决定, 特别依赖该差的首项系数及其辐角——即 $\mathrm{Re}(q_\alpha - q_\beta)$ 在何处变号. 相应地, level datum 定义为各 Galois 共轭之差 $q_0 - q_j$ 的非零斜率之集; 它是差的斜率, 不是 $q$ 自身的斜率.

本文由此拆分得到的一项识别: 型与候选 Stokes 方向都需要第二层的系数数据, 均不由斜率单独决定. 在把 Newton 多边形直接连到 Stokes 数据的两层说法下, 这条共同性是看不见的.

(此处不写作"同源": Stokes 方向取决于指数因子之差 $q_\alpha - q_\beta$, 个别首项可能发生抵消, 故二者不必由同一个首项系数决定. 本文所挣得的是它们同属第二层数据, 皆超出斜率粗层.)

第三层 · 解析层: Stokes 乘子

Levelt–Turrittin 定理所保证的形式亚纯联络同构, 一般不能提升为整体的亚纯联络同构; 它只能在奇点的弧邻域上局部提升 (Hukuhara–Turrittin). 形式数据描述趋近奇点时的渐近行为, 而该行为的精确描述在绕奇点一周的过程中会改变.

因此: 固定形式正规形之后, Stokes 矩阵仍提供额外的完整解析不变量. Stokes 乘子是形式分类之外的解析模量.

三层不可压成一条链

> 斜率不决定 Stokes 乘子.

把"斜率 → 指数因子 → Stokes 方向与 Stokes 乘子"写成一串因果, 是把形式分类与解析分类并成一步. 前两层同属形式分类, 第三层不属之; 第一层与第二层之间也不是决定关系, 而是粗与细的关系——斜率决定次数, 不决定系数.

本文因此把 Newton–Turrittin 数据登记为不规则边界的形式组织器, 不登记为阶段 4 的共同发生器.

7.3 $G_\infty$ 的候选比较接口

Paper 1 附录 B.5 阐述 $G_\infty$ 为无穷处增长: 整函数的阶 $\rho$ 与型 $\sigma$, 配以 Phragmén–Lindelöf 原理. 该对象是标量渐近不变量; 记号中的 $G$ 不应被读为群 (§ 2.5).

§ 7.2 第一层给出解在奇点处的增长阶, 第二层给出型. 这与阶和型的内容在形式上相合, 提示二者之间可能存在比较接口.

但本文只登记为候选接口, 不主张同一性. 理由: Paper 1 的 $G_\infty$ 是关于整函数一般的渐近不变量, 不专指微分方程解; 而 $\mathrm{Slope}_\infty$ 是亚纯联络的形式不变量. 二者是否为同一个不变量, 本文未证, 亦未在文献中找到该等同的陈述.

故登记为:

> Newton–Turrittin 数据为 Paper 1 的 $G_\infty$ 提供一个候选比较接口, 尚未证明同一性. 斜率给出增长阶的候选对应; 型依赖指数因子的首项系数, 非斜率所能决定.

§ 11.1 把该登记列为对上游的善意重构而非 Paper 1 原意——Paper 1 写下 $G_\infty$ 时所配的是标量内容, 说"记号预示群、本文补齐"是本文的重构, 不是还原.

记号纪律 (§ 2.5): $G_\infty$ 保留为 Paper 1 原对象; 斜率过滤记 $\mathrm{Slope}_\infty$; 局部微分 Galois 群记 $G_{\mathrm{diff,loc}}$ (§ 9.2). 三者不共用记号.

7.4 扇区、实定向吹起与 ZFCρ 债务的第二半

方向圆

不规则奇点处的方向结构不能由一点承载. 标准处理是在奇点处作实定向吹起: 其原像是一个圆, 圆上的点对应奇点周围的方向, 圆上的开集对应该处开扇区的芽. 指数因子于是被视为该方向圆上一个局部系统的截面 (Deligne–Malgrange 式表述).

这是 § 8.1 "粗闭合加方向细化" 的技术内容; 本节给出该细化的对象, § 8.1 阐述它对 $L_2$ 闭合结构的意义.

扇区失配

同一个形式解在不同扇区上的渐近有效性不同. 沿方向圆绕行时, 精确解与形式解之间的对应关系发生改变——这是 Stokes 现象. 单值群不足以刻画它: 单值群记录绕行一周的整体效果, 而此处需要的是按方向排序的扇区过渡数据.

这一点标明了 § 6.4 余项的确切位置: 周期与单值群阐述了正则情形的路径依赖, 而不规则情形的路径依赖含有方向分辨率, 超出单值群的记录能力.

ZFCρ 债务的第二半

§ 6.2 阐述: 赋值给单块 $\infty$ 装上整数刻度, 清的是阶那一半. ZFCρ 所欠的是 fine ratio, 而 fine ratio 是系数层面的量.

按 § 7.2 的三层, 系数层面的量落在第二层: Newton 多边形给出斜率分层, 各斜率上另有特征方程, 而 fine ratio 是特征方程系数层面的量. 故该债务的接口位置可以精确到第二层, 不在第一层.

但有接口不等于已接通. 本文没有给出 ZFCρ 的 fine ratio 与某个特征方程系数之间的显式映射、保持量或函子. 债务状态因此从"未接口"改登记为"有接口而未接通" (§ 2.6 其六), 不得写作清账.

一条内部一致性值得登记: 未接通的那一半, 恰好落在本层未捕获的那一层上——第二层与第三层正是 § 7.5 判为"标出而不进入"的部分. 债务的未清与捕获的未及, 是同一条边界的两面.

7.5 $\mathcal{P}^e$ 被标出而不被进入

本节至此阐述的全部内容——斜率、指数因子、方向圆、Stokes 数据——都在函数侧. 与之对应的数值侧对象是指数周期 $\mathcal{P}^e$ (§ 2.3 表 1 右下角).

二者的关系是本节的结论:

> 第四阶段以 Stokes 数据标出指数方向, 而不进入 $\mathcal{P}^e$.

"不进入"须作操作定义: 它指的是不把 $\mathcal{P}^e$ 提升为本文当前 trajectory 的取值核心——本文的取值核心是 $\mathcal{P}$ (§ 6.4), 而阶段 4 的函数侧工作不改变该选定. 它不指本文不能定义、讨论或展示指数周期: 本节正定义了 $\mathcal{P}^e$, 并陈述 $e \in \mathcal{P}^e$ 已证. 二者不矛盾.

且该陈述的档次是第三档 (条件性桥接), 见 § 7.1 三档表——它是本文对函数侧与数值侧的对齐登记, 不是由 Levelt–Turrittin 单独推出的数学结论.

这是主步骤 1 "标而不构" 在层末的再现. 主步骤 1 标出 $i$ 而不构成它的栖居空间 (§ 3); 主步骤 4 标出指数方向而不构成其取值. 同一动作在跨层阈口处与在层末闭合处各出现一次.

$L_2$ 的闭合余项因此是: 指数映射与指数积分方向.

$e$ 的地位

$e$ 是该方向的规范数值见证, 不是承重项. 三条事实须分开陈述.

其一, $e = \exp(1)$. 而 $\exp$ 是主步骤 3 内层 4 所产出的覆叠映射 (§ 5.3), 即 $L_2$ 的内部产物. 故 $e$ 是一个内部产物在一点的取值.

其二, $e$ 确定属于 $\mathcal{P}^e$. 取 $f(x) = -x$、$\omega = dx$、积分域 $[0,1]$ (半代数), 则

$$e = 1 + \int_0^1 e^x\,dx.$$

这是定义与计算层面的事实, 不是猜想.

其三, $e$ 是否属于 $\mathcal{P}$ 未知. 这是开放猜想 (§ 2.6 其三).

余项账本的精确形式因此是:

问题	状态
$e \in \mathcal{P}^e$	已证
$e \in \mathcal{P}$	未知
指数表达式类是否严格大于普通周期表达式类	语法上为严格扩张 (§ 7.1, 定义级)
其数值像是否由 $e$ 见证为严格包含	未知

不得写作"$e$ 是基底内不捕获的余项"——那会把已证的一栏与未知的一栏混为一谈. 精确的写法是: 在普通周期核心中的成员状态未知; 在指数周期扩张中已捕获; $e$ 只作候选严格性见证.

反解指定, 及其限度

$e$ 另有一条与本文承重相关的表示:

$$\int_1^e \frac{dx}{x} = 1.$$

即 $e$ 由一个取周期值的函数的方程隐式指定. 该指定方式不属于普通周期表达式语法——它是反解式的, 而 $\mathsf{Int}_{\rm alg}$ 的定义中不含反函数算子 (§ 2.3). 而该积分即 $\log$, 正是 Paper 1 § 4.5 所立 $L_1$ 的阐述交换单位. 于是可以陈述:

> $L_1$ 的交换律单位在 $L_2$ 被实现为取周期值的函数; $e$ 是它的反解; 而反解式的指定方式落在普通周期表达式语法之外.

但此处必须守住 § 2.3 的纪律. 上式说的是"我们写下的这一个表达式以反解方式指定 $e$"; 它不能排除 $e$ 另有某个普通周期表示——而后者正是 $e \in \mathcal{P}$ 尚未决的内容. 故本小节的全部陈述都在表达式层面, 不得读作对数值成员状态的断言. 正向与反解的区分是表达式形成层面的区分, 不是 $e \notin \mathcal{P}$ 的定理级替代. 本文不以它承重, 只以它阐述 $e$ 在本层阐述结构中的位置.

承重定位

综上, 本文承重的分布是:

• 定义级承重: $\mathsf{Int}_{\rm alg} \hookrightarrow \mathsf{Int}_{\exp}$——由 $\mathcal{P}$ 与 $\mathcal{P}^e$ 的定义直接给出;
• 定理级承重: 正则联络扩展为允许不规则奇点的联络——由 Levelt–Turrittin 等形式分类支撑;
• T2-条件项: 两侧被登记为同一边界的数值侧与函数侧呈现;
• T3 见证: $e \notin \mathcal{P}$.

$e$ 不承担任何承重角色. 具体地: 它不承担正则／不规则边界的承重 (那由 § 7.1 两条扩张承担), 也不承担取值核心严格性的承重 (那是未决的 T3). 它保留两项非承重登记: 承 Foundation v2 的跃迁签名 (§ 11.5.3), 与本文的候选数值见证.

若 $e$ 被证明是周期, 须改动的只有本节的见证一栏与 § 8.2 的账本一行; § 7.1 的两条扩张、§ 7.2 的三层、以及 § 5.4 的核心结构陈述皆不动 (§ 12.3).

---

8. 主步骤 4(c): 闭合与余项账本

8.1 粗闭合与方向细化

主步骤 2 已给出 $\mathbb{C}$ 的一点紧化, 把 $L_1$ 的两个边界标记合并为单一的 $\hat\infty$; § 4.3 同时声明该合并是被选择的规范闭合, 不是拓扑学强制. 本节阐述该选择的代价在何处到期.

§ 7.4 的方向圆与该一点紧化直接冲突: 前者要求保存趋近奇点时的方向, 后者把所有方向压成一点. 二者不能同时成立于同一分辨率上.

解决方式不是二选一, 而是两级.

粗闭合: Riemann 球面. 这是亚纯与共形语境中的规范闭合. 它使 $0 \leftrightarrow \hat\infty$ 的二周期边界对成立 (§ 5.3), 使除子在 $\infty$ 处有定义 (§ 6.2), 使全局留数定理有紧致的底空间 (§ 6.3). 本节不撤回它.

方向细化: 实定向吹起. 不规则余项迫使被压掉的方向重新展开为方向圆 (§ 7.4).

由此得到本文对 $L_2$ 闭合结构的一项识别:

> $L_2$ 的闭合不是单一 compactification. 粗闭合把方向压成一点; 不规则余项又迫使该点重新展开为方向圆.

这个形状值得单独指出: 闭合 → 余项 → 闭合的再展开. 闭合不是终点; 闭合本身产生了要求把它重新打开的余项. 而再展开并不取消闭合——它是同一闭合的第二级分辨率.

一条纪律: 再展开不撤回粗闭合. 二者不是两个竞争的闭合, 而是同一闭合的两级分辨率. 除子、留数、周期在粗闭合上定义, 不因方向细化而失效; 方向细化只在需要方向分辨率的地方生效, 即不规则奇点处.

与上游的关系: Paper 1 § 5.2 阐述 $\pm\infty$ 的等同"由拓扑学强制". 本文精化为: 拓扑并不唯一强制一点紧化, 一点紧化是在亚纯与共形语境中被选择的规范闭合; 而该选择的代价是方向分辨率, 该代价在不规则边界处到期. § 11.1 把此列为第一项上游精化.

8.2 余项账本

承 Paper 3 § 6 的三类划分, 在 $L_2$ 重排.

第一类: 基底外余项

在 $L_1$ 处, 该类的实例是 $i$——它被 $L_2$ 主步骤 1 接入 (§ 3). 在 $L_2$ 处, 该类指向 $\mathbb{C}$ 之外的方向, 即 Cayley–Dickson 序列 $\mathbb{H}$、$\mathbb{O}$.

但与 $L_1$ 的情形有一项关键差别. $i$ 是被 $L_1$ 内部的追问逼出来的: $x^2+1=0$ 在 $\mathbb{R}$ 无根, 而求根正是 $L_1$ 主步骤 3 自己的反向追问. Paper 3 据此把它登记为跨层 forcing, 不是 lateral branch.

$\mathbb{H}$ 则没有这样的来源. $\mathbb{C}$ 代数闭, 求根不再逼人 (§ 5.4); 没有任何 $L_2$ 内部的追问逼出 $\mathbb{H}$. Cayley–Dickson 是可以施加的构造, 不是被迫接受的产物.

故第一类在 $L_2$ 的地位与在 $L_1$ 不同: 它不是被迫的余项. § 10 据此把它判为代价支——由付掉可交换性而得, 而非由处理本层余项而得.

这条差别是 § 5.4 不对称的下游后果. 代数闭性消除标量缺口的同时, 也消除了生产被迫的基底外余项的那个机制.

第二类: 基底上的结构、模式不捕获

该类的成员是指数方向这一结构对象, 不是数值 $e$. 类名中的"不捕获"是一条语法边界: 普通周期表达式语法 $\mathsf{Int}_{\rm alg}$ 不允许指数因子, 故本文选定的普通周期阐述模式不构造指数方向 (§ 7.1 定义级). 其取值像是否因此形成严格扩张, 即

$$\mathcal{P} \subsetneq \mathcal{P}^e,$$

则仍是数值侧的未决问题. 二者分属 § 2.3 的语法侧与取值侧, 不得并作一句.

另须一条推广定义: 此处"基底内"推广为"定义在该基底之上且取值仍落于该基底", 不表示该对象作为元素属于 $\mathbb{C}$——指数方向是结构对象, 不是数. 类名沿用 Paper 3 § 6 的三分类措辞, 该推广仅为容纳结构对象.

指数方向 (§ 7.5). 数值见证为 $e$.

问题	状态
$e \in \mathcal{P}^e$	已证
$e \in \mathcal{P}$	未知
指数表达式类是否严格大于普通周期表达式类	语法上为严格扩张 (§ 7.1, 定义级)
其数值像是否由 $e$ 见证为严格包含	未知

精确的写法是: 在普通周期核心中的成员状态未知; 在指数周期扩张中已捕获; $e$ 只作候选严格性见证. 不得写作"$e$ 是基底内不捕获的余项"——那会把已证的一栏与未知的一栏混为一谈.

第三类: 不可计算、不可定义

承自 $L_1$. Paper 3 § 6.3.2 由基数论证挣得: 对任何可数形式语言, 可定义实数至多可数, 而实直线基数为 $2^{\aleph_0}$ (§ 2.1). 该类在 $L_2$ 仍不被捕获.

$L_2$ 处另有一项更强的事实: $\mathcal{P}$ 连可计算实数都不穷尽——Yoshinaga 已给出可计算但非周期的实数. 故 $L_2$ 的取值核心与第三类之间还隔着一层.

捕获链与多模式合力余项

把三条判据各自挣得的范围排在一处:

$$\mathbb{Q}^{\rm alg}_{\mathbb{R}} \subsetneq \mathcal{P}\cap\mathbb{R} \subsetneq \text{可计算实数} \subsetneq \text{可定义实数} \subsetneq \mathbb{R}$$

三条判据分别是: 有限代数约束 ($L_1$ 模式)、代数积分表示 ($L_2$ 模式)、算法可逼近性 (有效 Cauchy 阐述模式). 按 § 2.1, 三者互不为对方的特例——它们不是同一种判据的三个实例, 而是三种不同类型的判定标准.

由此得到一项读法:

> 三条判据各自挣得一层, 而其并集仍止于可定义实数. 这是本文所能展示的最强形式: 不是单一模式的余项, 而是判据互异的多个模式合力之后的余项.

辖区声明. 本文不据此主张所有数学工具的并集皆不穷尽基底. 三个不是全部; 该主张须先划定"工具"的边界——工具可以带参数, 可以构成不可数族, 故 Paper 3 § 6.3.2 的基数论证不能直接移植——而划定该边界超出本文辖区.

全称形式的"框架不可穷尽"是 Paper 1 摘要已埋的元命题 (双重不可穷尽性的元层级一半), 其实质挣得属框架层论文的工作. 本文只作展示与前指, 不背书.

〔类型: 结构展示 + 辖区声明〕

附带: 上式中有效 Cauchy 阐述模式的位置, 是 Paper 3 § 6.4 所留开放问题的材料 (§ 2.1). 该模式在实轴上捕获得严格多于 $L_2$ 的周期核心, 而其判据与 $L_1$、$L_2$ 两者皆不同类. 其落点判定在 § 11.3.

8.3 欧拉公式: 叠加落点与跃迁闭合方程

欧拉公式在本文中承担两个不同的谓词. 二者不相竞争, 且回答的是不同的问题.

作为落点

§ 5.5 已给出结构理由:

$$\cos\theta + i\sin\theta = e^{i\theta}$$

两边分属基底与模式两侧——左边是基底级的经典分析构造 (实三角函数; 按 § 5.2.3 的资源纪律, 本文不在脊柱上调用它们), 右边是主步骤 3 内层 4 的模式级产物 ($\exp$ 在虚轴上的限制). 公式陈述二者相等.

故它不住本 trajectory 的任何一格: 它是基底级构造与模式级产物之间的一次认同. 本文因此把它登记为基底—模式叠加落点——所跨的是基底与模式这条总防火墙 (§ 2.1), 不是层内两格.

成分清点可作为佐证, 但不是理由: $e^{i\pi}+1=0$ 中, $\pi$ 是 $L_1$ 主步骤 4 的余项, $1$ 与 $0$ 是 $L_1$ 的两个边界 token, $i$ 是 $L_2$ 主步骤 1 的产物, $e$ 是 $L_2$ 闭合余项的数值见证. 五项分属两层四处. 但清点只给出位置描述; 结构理由在两条阐述路线的认同.

作为闭合方程

SAE Foundation v1 的跨层账本表把 $L_1 \to L_2$ 的闭合方程记为 $e^{i\pi}+1=0$, 两项余项记为 $i$ 与 $\pi$. 该表未给出为何恰是这两项. 本文给出:

$i$ 是跃迁的输入. 它是 $L_1$ 的基底外余项, 经主步骤 1 标而不构接入 $L_2$ (§ 3). 它不是 $L_2$ 产生的, 是跃迁带进来的.

$\pi$ 是本层的第一个结构产物. 它是 $L_1$ 的基底内超越余项 (Paper 3 § 6 账本), $L_1$ 能命名而不能捕获. $L_2$ 先在主步骤 3 内层 4 把它阐述为覆叠变换群 $2\pi i\mathbb{Z}$ (§ 5.3), 再在主步骤 4 内层 3 作为周期捕获它 (§ 6.4).

故闭合方程所绑定的是: 跃迁的输入, 与本层的第一个结构产物.

而绑定所用的那一项——$e$——在指数周期扩张中已被捕获, 而在普通周期核心中的成员状态至今未定 (§ 7.5). 于是该方程可以读作: 两个已在普通周期核心中被捕获的余项, 经那个成员状态未定者, 回到 $L_1$ 自己的两个边界 token.

两个谓词的关系

落点回答"它住哪"; 闭合方程回答"它闭合什么". 一个对象可以同时是两者, 二者不相竞争.

且须注意: 它是跃迁的闭合方程, 不是层的闭合方程. $L_2$ 这一层的闭合是粗闭合加方向细化 (§ 8.1), 与跃迁闭合是不同对象.

Paper 1 § 5.3 阐述欧拉公式"最好被理解为闭合的实例, 而非 $L_2$ 的完整闭合". 其准确内容即在此: 它是跃迁闭合的实例, 不是层闭合. § 11.1 把此列为第三项上游精化.

关于 Foundation v1 该行的完整承接与重组 (含"行为"一列由指数映射改为复化), 见 § 11.5.4.

8.4 $L_2 \to L_3$ 触发: 值的层面与规则的层面

Paper 1 § 5.5 立触发判词: 触发 $L_3$ 的不是 $L_2$ 的难题本身, 而是问题被重框为规则与可判定性. 本节按 § 2.4 的三轴登记两个问题, 以显示它们分属不同层面.

问题	数值成员状态	证明论状态	算法状态	层归属
$e \in \mathcal{P}$?	未知	猜想	—	$L_2$ 内部未决
一般周期表达式等同性问题是否可判定?	—	—	一般问题未知	$L_3$ 现象

第一行是值的层面: 其问题形式是一个具体数值与一个具体集合之间的成员关系. 该问题留在 $L_2$.

(本文不判定该问题的障碍类型. 大纲阶段曾由"函数域类比已证"推"障碍属算术困难而非逻辑独立", 该推理无效已撤回, 此处不以任何措辞将其带回.)

第二行是规则的层面, 而规则层面本身含两个不同的槽, 不可混同.

槽	问题	坐标
推导完备性	KZ 三类变换是否完备生成全部普通周期表达式等价?	规则／证明系统
决策程序	一般普通周期表达式等同性是否存在总终止的判定程序?	算法可判定性

完备性不单独推出可判定性. 一族规则完备刻画某等价关系, 不自动给出一个总会终止的判定程序; 还需另有终止性保证或对"不相等"的有限证书.

就周期而言, 该额外要素恰好可得: 周期是可计算的, 故对任意两个表达式可并行执行两支——数值求值至递增精度 (若二者不等, 终将分离, 给出不等证书), 与 KZ 变换链的枚举 (若二者相等且猜想成立, 终将找到链). 两支必有一支停机. 故

$$\text{KZ 猜想} + \text{周期的可计算性} \ \Longrightarrow\ \text{一般周期表达式等同性可判定}.$$

须显式写出第二个前提: 由完备性单独推可判定性是一步跳跃; 本文不作该跳跃.

由此, $L_2 \to L_3$ 的触发不是两段而是三段:

$$\text{值问题} \longrightarrow \text{等价关系的完备刻画} \longrightarrow \text{判定程序}$$

后两段皆属规则层面, 但分属不同的槽. 规则层因此不是单一的"可判定性"槽.

精确化: 须写"一般周期表达式等同性问题", 不可笼统写"周期相等性". 受限类别已有算法性结果 (1-periods 的超越性与线性关系计算); 未决的是一般问题.

两条混层禁令

其一, 独立性是命题的属性, 不是对象的属性. 即便"$e \notin \mathcal{P}$"被证明相对某形式系统独立, 进入规则层的是该命题在该系统中的可证性; 数值 $e$、周期表达式、以及成员关系问题本身不因此改变层归属. 违反此条会当场破坏本节所建立的分栏 (§ 12.3 第二条反事实).

其二, 正则与不规则的分界不是层边界. 它是 $L_2$ 捕获范围的边界, 位于层内 (§ 2.4). Stokes 现象是 $L_2$ 模式内部的结构性事实, 不是 $L_2$ 与 $L_3$ 之间的分界.

触发的形式

当问题从"这个值是多少 / 这个数属不属于某集合"转为"这一类等式能否由规则判定"时, 阐述已经离开 $L_2$ 的对象层面而进入规则层面. 该重框不是 $L_2$ 内部的进展, 是层的转移.

$\omega_\rho^{(2)}$ 合成律的完整形式化亦落在此处: 六分量之间的统一封装是关于本层自身缺陷装置如何复合的元层级陈述, 其完整形式化要求范畴论, 而 Paper 1 § 11.4 已立范畴论结构主义的充分接洽须在 $L_2$ 与 $L_3$ 两处处理. § 9 交付最小合成骨架, 完整封装登记为未还挂账 (§ 13.1).

本文不预设 $L_3$ 的 trajectory (§ 1.3). $L_3$ 论文的接口在 § 13.2.

---

9. 操作缺陷代数: 最小 typed composition

9.0 本节交付什么、不交付什么

Paper 1 § 11.1 把 $\omega_\rho^{(2)}$ 六分量的合成律挂账给后续的 $L_2$ 专题论文, 并阐述: 若无合成律, 六元组只是清单, 尚不足以称为完整的操作缺陷代数.

本文不还完整账. 但也不把"给不出"诊断为一项发现——那会违反 § 12.4 的反免疫声明: 一笔欠账不得转成一项识别. 本文的处理是部分偿付加明确挂账.

交付: 最小 typed composition 骨架, 即两张表与四条链——表一给出各分量的类型与内部运算 (§ 9.1), 表二给出跨分量接口 (§ 9.2).

不交付: 六类数据的统一范畴封装 (§ 9.3).

"部分偿付"的边界须写死: 本文所交付者是骨架级——它使六元组不再只是清单, 但不使它成为完整代数. 故"部分"二字不可省: 挂账仍然存在, 只是欠的内容由"整条合成律"缩小为"统一范畴封装"(§ 13.1). 交付了骨架而仍挂账, 二者不矛盾.

名称: "操作缺陷代数"之名保留. Paper 1 摘要已以此名指称 $\omega_\rho^{(2)}$; 在 $L_2$ 的专题论文中把该层的核心对象降名, 是自拆而非精化. 但本节须明写: 交付的是最小合成骨架, 不是完整代数——名称保留不等于内容已足.

归属: 两张表与四条链的全部内容均为经典结果. 本文所做的是组织 (§ 9.4).

9.1 表一: 各分量的类型与内部运算

承 Paper 1 附录 B 的六分量定义, 逐项登记其类型 (§ 2.2 数值、函数、算子的区分在此生效) 与内部运算.

分量	类型	内部运算
$\mathrm{Div}$	自由阿贝尔群 $\mathrm{Div}(X)$	除子加法; 主除子构成子群, 商为除子类群 $\mathrm{Cl}(X)$
$\mathrm{Res}$	局部线性泛函	留数向量的加法; 全局约束 $\sum_p \mathrm{Res}_p(\omega) = 0$
$\mathrm{Per}$	双线性配对 $H_1 \times H^1_{\rm dR} \to \mathbb{C}$	配对的双线性性; 取值落在 $\mathcal{P}$
$\mathrm{Mon}$	$\pi_1$ 的表示	路径连接对应矩阵复合
$\mathrm{Stokes}$	按方向排序的算子族	沿方向圆的扇区过渡复合 (Stokes 群胚)
$G_\infty$	标量渐近不变量 (阶 $\rho$ 与型 $\sigma$)	无群运算; 阶与型上的偏序

关于末行. $G_\infty$ 的类型须明写, 因为记号中的 $G$ 会诱导群的联想. Paper 1 附录 B.5 所配的内容是整函数的阶与型加 Phragmén–Lindelöf 原理, 即标量渐近不变量, 不是群对象 (§ 2.5、§ 7.3). 本表如实登记为标量不变量.

由此产生一项类型上的不齐: 阶段 4 的两个分量中, $\mathrm{Stokes}$ 是矩阵族而 $G_\infty$ 是标量, 二者不构成群意义上的复合. § 7.2 的三层分析给出了处理方式——把该阶段的对象按层重新登记:

对象	类型	位置
$\mathrm{Slope}_\infty$	有理数族 (斜率过滤)	§ 7.2 第一层
指数环面 $\mathcal{T}$	代数环面	§ 7.2 第二层
$\mathrm{Stokes}$	按方向排序的算子族	§ 7.2 第三层
$G_{\mathrm{diff,loc}}$	代数群	三层的 Zariski 闭包 (§ 9.2 链 4)

指数环面的类型有一条精确刻画值得单独登记: 它是李代数包含指数因子 $Q$ 之全部系数的最小子环面. 故它由第二层 (系数) 决定, 不由第一层 (斜率) 决定. 这从微分 Galois 理论一侧独立印证了 § 7.2 的分层——该分层不是本文为组织方便而设的, 它在这里被另一套理论以自己的语言重述.

9.2 表二: 跨分量接口

接口	映射	陈述
1↔2	$d\log$	$\mathrm{Res}_p(df/f) = \mathrm{ord}_p(f)$
2↔3	小环路限制	$\oint_{\gamma_p}\omega = 2\pi i\,\mathrm{Res}_p(\omega)$
3↔4	(wild) Riemann–Hilbert	正则奇点联络 $\leftrightarrow$ 局部系统与 monodromy 数据; 不规则联络 $\leftrightarrow$ monodromy 加 Stokes 增强数据
4 内部	Ramis 稠密性	形式单值、指数环面与 Stokes 算子生成的子群在 $G_{\mathrm{diff,loc}}$ 中 Zariski 稠密

四条链各有一处须守的精确性.

链 1 · $d\log$. 对数微分的留数等于赋值的阶, 由此

$$\deg\,\mathrm{div}(f) = \sum_p \mathrm{ord}_p(f) = \sum_p \mathrm{Res}_p\!\left(\frac{df}{f}\right) = 0.$$

即主除子次数为零是全局留数定理对 $df/f$ 的一个特殊化. 该接口只经 $d\log$ 这一张映射成立; 不可笼统说所有除子定理与所有留数定理是同一守恒的两种记法 (§ 6.3).

链 2 · 小环路限制. 小环路周期是 $2\pi i$ 乘以留数. 方向不可逆: 不可由此式除以 $2\pi i$ 反推留数属于 $\mathcal{P}$, 因为 $1/\pi \in \mathcal{P}$ 未决 (§ 6.1、§ 6.4). 在代数数据条件下留数另因代数性属于 $\mathcal{P}$, 那是经 $\overline{\mathbb{Q}} \subset \mathcal{P}$ 的另一条理由.

链 3 · Riemann–Hilbert. 不规则侧仍然保留 monodromy; Stokes 是增加项, 不是替换项. 经典 RH 与 wild RH 的区别正在于 Betti 数据被 Stokes 信息增强.

且稠密性定理自身即呈现这一结构: Schlesinger 定理对正则奇点情形只用单值群 (单值群在微分 Galois 群中稠密); Ramis 定理对不规则情形增加指数环面与 Stokes 算子. 故 § 6.0 所立的"核心 $\hookrightarrow$ 增强"关系不是本文强加的读法, 而是这一对定理自身的关系.

链 4 · Ramis 稠密性, 及其两级结构. 须写 Zariski 稠密, 不可写"拓扑生成"——后者是参数化版本 (Kolchin 拓扑) 的说法, 非参数化的 Ramis 定理给出的是 Zariski 稠密性.

该链实际含两级, 且第一级是一条真正的合成律而非仅稠密性陈述:

• 形式层: 局部不规则 Riemann–Hilbert 对应在形式层给出范畴等价, 对应的群是半直积

$$G^{\rm formal} = \mathcal{T} \rtimes \overline{\langle \hat\gamma \rangle}^{\rm Zar},$$

即指数环面与形式单值 $\hat\gamma$ 所生成群之 Zariski 闭包的半直积. 这是阶段 4 内部两个对象之间的确定复合关系.

须注意闭包不可省: $\mathcal{T} \rtimes \mathbb{Z}$ 是抽象生成群, $\mathbb{Z}$ 不是代数群; $G^{\rm formal}$ 是其 Zariski 闭包. 本文在本链前两行刚立"须写 Zariski 稠密"的纪律, 此处依同一纪律书写.

• 非形式层: 过渡到收敛 Laurent 级数时, 普遍微分 Galois 群获得附加生成元, 它们依赖发散级数的重求和, 与 Stokes 现象相关. 取 Zariski 闭包得 $G_{\mathrm{diff,loc}}$.

故阶段 4 内部的合成是两级: 形式层的半直积, 加非形式层的 Stokes 生成元. 这一两级结构与 § 7.2 的形式层／解析层分界是同一条分界在群论语言中的呈现.

9.3 未还部分

明确登记本文未交付者.

未交付: 六类数据的统一范畴封装——即把 $\mathrm{Div}$、$\mathrm{Res}$、$\mathrm{Per}$、$\mathrm{Mon}$、$\mathrm{Stokes}$、$G_\infty$ 六者组织为单一范畴对象, 使 § 9.2 的全部接口成为该范畴内的态射.

障碍不是范畴论工具的缺失. Riemann–Hilbert 对应与其不规则 (wild) 扩展的完整形式化已经存在. 障碍在于六分量尚未被映到 Riemann–Hilbert 数据上; § 9.2 的四条链是该映射的局部片段, 不是全图. 尤其: 链 3 与链 4 已进入 RH 语言, 而链 1 与链 2 仍在经典复分析语言中, 二者之间的翻译本文未给出.

本文不把"给不出"诊断为发现. § 8.4 已阐述该封装具有元层级性质——它是关于本层缺陷装置如何复合的陈述, 而这类陈述按 Paper 1 § 5.5 的触发判词属规则层面. 但须区分两件事: "该封装具有元层级性质"是对该对象的描述, 与"本文给不出它"是两回事; 前者不构成后者的辩解. 按 § 12.4, 欠账登记为欠账.

登记: 未还挂账, 推至 $L_2 \to L_3$ 接口论文 (§ 13.1).

9.4 归属声明

两张表中的全部内容与四条链, 均为经典结果: 除子群与除子类群、留数定理、de Rham–Betti 配对、单值表示、Stokes 数据的扇区复合、对数微分的留数、小环路积分公式、Riemann–Hilbert 对应及其 wild 扩展、Ramis 稠密性定理. 本文引用而不证明这些结果.

本文所做的是组织: 按 § 7.2 的三层与 § 6.0 的核心-增强关系, 把六分量放到确定的位置上, 给出每一分量的类型, 并标出哪些接口已有、哪些尚无.

本文不主张发现了这些接口, 也不主张这一组织方式是唯一的. 组织本身属 SAE 登记, 其价值在于使未还的部分可见——在 Paper 1 的六元组清单形态下, 缺什么是看不出来的.

〔类型: 标准定理 (表与链的内容) + SAE 登记 (组织方式)〕

---

10. 侧支与替代闭合: Cayley–Dickson

10.1 判定

Paper 1 § 11.8 挂出一个开放问题: Cayley–Dickson 序列在 $L_2$ 处是否提供替代闭合, 以及充分性标准如何适用于它们. 本节作答.

判定: $\mathbb{H}$ 与 $\mathbb{O}$ 不是本文这条可交换复周期 trajectory 的等价并列实现.

判据是层内交换律阐述单位. 本文这条复化 trajectory 的层内交换律阐述单位是 $2\pi i$, 其交换律的具体形式是留数定理与辐角原理 (§ 6.3); 二者都要求可交换性. $\mathbb{H}$ 付掉的正是可交换性.

(本判据只涉及层内槽, 不涉及跃迁签名槽. 故它不受 Foundation v2 把数学 $L_1 \to L_2$ 的跃迁签名派为 $e$ 的影响, 见 § 11.5.3.)

按物理 Paper 1 所立的二层嵌套通则之一——子层不出新签名, 共用第一层签名——付掉该单位者不能算作同一层的并列实现.

该判据的力量来自 § 5.5. 若 $2\pi i$ 只是留数定理中的一个因子, 则"付掉可交换性即付掉它"只是一条关于某个公式的观察. 但 § 5.5 已阐述: $2\pi i$ 同时是 $\exp$ 的覆叠变换群的生成元, 与留数归一是同一条陈述的两种呈现. 故付掉可交换性所付掉的, 是整层覆叠结构在积分语言中的呈现, 而不只是某一条积分公式.

10.2 结论须收窄

本节的判定不能推得太远. 两处须明确.

其一, 纵向迭代论证撤回. 一种自然的论证是: Cayley–Dickson 是纵向迭代, $\mathbb{H} = \mathrm{CD}(\mathbb{C})$ 以 $\mathbb{C}$ 为输入, 故 $\mathbb{H}$ 本不是 $L_1 \to L_2$ 阈口上的兄弟. 该论证不成立——因为 $\mathbb{C} = \mathrm{CD}(\mathbb{R})$ 同样成立. 若纵向迭代即意味着"下一级", 则该论证推出的是"$\mathbb{H}$ 是 $L_3$", 而不是"$\mathbb{H}$ 是侧支". 它证得太多. 本文不用它.

其二, 四元数分析并非失去全部 Cauchy 型结构. slice-regular 理论在每个复切片

$$\mathbb{C}_J = \{x + Jy : x, y \in \mathbb{R}\}, \qquad J \in \mathbb{S}^2$$

上恢复复分析型的结构, 并有相应的 Cauchy 核与积分公式. 故变化不是"签名彻底消失", 而是单一的全局 $i$ 被一族虚单位 $J \in \mathbb{S}^2$ 取代.

因此本节所能得出的只是:

> $\mathbb{H}$ 不共享本文的单一 $2\pi i$ 签名, 亦不共享同一个六分量缺陷系统.

不得由此得出 $\mathbb{H}$ 被逐出整个 $L_2$. 后者需要的是关于全部 $L_2$ trajectory 空间的判断, 而本文只研究了其中一条. 按 Paper 2 的反本质化纪律, 用当前 trajectory 的签名去排除其他 trajectory, 与"复化只是 $L_2$ 的一条阐述"这一立场直接冲突.

〔类型: 不得作本篇结论——见 § 12.1〕

10.3 代价支与余项支

本节另立一条分类判据, 用以处理一个否则会出现的问题.

若 $\mathbb{H}$ 是"$L_2$ 之后"的某物, 而规则层面 (§ 8.4) 也是"$L_2$ 之后"的某物, 则本文有两个互相竞争的后继候选, 而 § 13.2 的预测只谈其中一个.

判据: 区分分支的来源.

• 代价支: 由付掉签名而得. $\mathbb{H}$、$\mathbb{O}$ 属此. 它们不是对本层余项的处理, 而是对本层某项结构约束的放弃所换得的.
• 余项支: 由处理本层余项而得. 指数方向 (§ 7.5) 与规则层面 (§ 8.4) 属此.

二者不同源, 故不竞争同一位置. $L_3$ 沿余项支而来; Cayley–Dickson 沿代价支而去. 二者不在同一条谱系上, 因而"两个后继候选"的问题不存在.

这条判据与 § 8.2 第一类余项的阐述相互支撑. § 8.2 阐述: $i$ 是被 $L_1$ 内部追问逼出的基底外余项, 故迫使跨层; 而 $\mathbb{C}$ 代数闭, 没有 $L_2$ 内部的追问逼出 $\mathbb{H}$. 用本节的语言: $L_2$ 没有被迫的代价支, 只有可施加的代价支.

〔类型: SAE 的 T3 分类提案, 非经典数学结论〕

10.4 对上游候选分类的修订

Paper 1 § 5.1 把 Cayley–Dickson 序列的各项列为 $L_2$ 阐述的候选. 本文把它们改判为侧向 trajectory 候选.

须说明这不只是偿还 Paper 1 § 11.8 的开放问题, 而是修订一项上游的候选分类. 前者是回答一个已被提出的问题, 后者是改变一个已被作出的归类. 本文按后者登记 (§ 11.1 第二项上游登记).

对 Paper 1 § 11.8 的直接回应: 替代闭合不支持同一个 $\omega_\rho^{(2)}$——签名不同, 六分量的类型随之不同. 但"不支持同一个"不等于"不支持任何缺陷系统": slice-regular 理论有其自身的结构, 其缺陷装置是什么, 本文未研究, 亦不预判.

---

11. 跨论文与跨系列关系

11.1 对 Paper 1 的四项上游处理

本文对 Paper 1 有四项上游处理: 三项精化／善意重构, 一项候选分类修订. 按本系列纪律, 精化明写而不默改; 且每一项须说明变的是什么、不变的是什么.

#	Paper 1 原阐述	本文处理	位置	类型
一	§ 5.2: $\pm\infty$ 的等同"由拓扑学强制"	一点紧化是亚纯与共形语境中被选择的规范闭合; 该选择的代价是方向分辨率, 于不规则边界处到期	§ 4.3、§ 8.1	精化
二	§ 5.1、§ 11.8: Cayley–Dickson 序列各项列为 $L_2$ 阐述候选	改判为侧向 trajectory 候选; 并回答 § 11.8: 替代闭合不支持同一个 $\omega_\rho^{(2)}$	§ 10.4	修订候选分类
三	§ 5.3: 欧拉公式"是闭合的实例, 而非 $L_2$ 的完整闭合"	准确内容: 它是跃迁闭合方程, 不是层闭合方程; 另作基底—模式叠加落点	§ 8.3	精化
四	附录 B.5: $G_\infty$ 为整函数的阶与型加 Phragmén–Lindelöf	Newton–Turrittin 数据为之提供候选比较接口, 尚未证明同一性	§ 7.3	善意重构

第一项: 等同本身不变——$\pm\infty$ 确实被合并为 $\hat\infty$; 变的是该等同的模态 (强制 → 语境中的选择) 与其代价的可见性.

第二项: 这不只是偿还一个已被提出的问题, 而是改变一个已被作出的归类. 按后者登记.

第三项: Paper 1 该句的判断不变——欧拉公式确实不是 $L_2$ 的完整闭合; 变的是给出了它是什么.

第四项须特别标注为善意重构. Paper 1 写下 $G_\infty$ 时所配的内容是标量渐近不变量; 说"记号预示群、本文补齐"是本文的重构, 不是 Paper 1 的原意. 本文保留 $G_\infty$ 记号与其原对象, 另立 $\mathrm{Slope}_\infty$ 与 $G_{\mathrm{diff,loc}}$ (§ 2.5), 不以新对象顶替旧记号.

此外, 本文兑现 Paper 1 § 11.1 所挂合成律债务的一部分 (§ 9), 未还部分明确登记为挂账 (§ 13.1); 并在 § 6.3 给出 Paper 1 § 4.5 所立 $L_2$ 交换律的具体形式 (留数定理与辐角原理).

11.2 与 Paper 2 的关系

Paper 2 立: $L_1$ 不是单一基础层级, 而是维度 × 路径的乘积; 并立反本质化纪律.

本文的 $L_2$ 同样是一条 trajectory, 不是唯一的 $L_2$. 该纪律在本文有三处具体执行:

• 题目不称"复化层 $L_2$"——复化是 $L_2$ 的一条阐述, 不是 $L_2$ 的定义;
• § 5.4 的登记后果不写作"结构性排除"——本文在当前 trajectory 内采用内部产物读, 跃迁读不再作本篇的主登记, 但本文不判断跃迁读在其他 trajectory 中的地位;
• § 10.2 的结论收窄——只得出 $\mathbb{H}$ 不共享本文这条 trajectory 的签名与缺陷系统, 不得出它被逐出整个 $L_2$.

关于 Paper 2 的三条闭合路径: 本文沿完整精确化路径. 有效 Cauchy 阐述模式属另一条路径上的模式, 其落点见 § 11.3.

11.3 与 Paper 3 的关系

格式承接: 四主步骤各含四内层步骤, 承 Paper 3 的阐述格式.

但主步骤 4 的登记类别不同——按 Paper 3 § 1.3 与 § 6.1 所立的三分类, 主步骤 4 是元层级闭合, 其 sub-articulation 是余项账本结构, 不遵循四步分形模式. 本文遵循该登记, 并以主步骤 4 横跨 § 6、§ 7、§ 8 三节的版面不对称承载之 (§ 2.1).

基底与阐述模式的区分: 承 Paper 3 § 2 并抬到 $L_2$ (§ 2.1). 本文另阐述该区分在两层之间发生了类型变化——承载体迁移 (§ 6.1).

余项账本三类: 承 Paper 3 § 6 并在 $L_2$ 重排 (§ 8.2). 其中第一类 (基底外) 在两层的地位不同: $i$ 由 $L_1$ 内部追问逼出, $\mathbb{H}$ 无对应来源.

sub-fractal 深度不对称: Paper 3 § 5.3 阐述 $L_1$ 的乘法路径较加法路径深一档; 本文 § 5.6 阐述 $L_2$ 重现同一不对称且格位相同, 但两层深处的实质内容不同.

通则③的压测: 物理 Paper 1 立"子步 d 的产物即母层余项". 本文 $L_2$ 满足之——主步骤 4 的最末阶段 (边界-扇区, § 7) 所产出的指数方向, 正是整层的闭合余项 (§ 7.5). 该满足是自动的, 未作任何安排.

通则①的使用: 子层不出新签名, 共用第一层签名. 本文在 § 10.1 使用之.

有效 Cauchy 阐述模式的落点

Paper 3 § 6.4 把有效 Cauchy 阐述模式登记为 post-$L_1$ 的解析完备化分支, 并明确不预设它究竟作为 $L_2$ 主轨内的 sub-articulation、作为 $L_{1\to 2}$ 桥、还是作为与复化主轨并行的 side-branch; 该判定留给后续层级论文. 本文是该后续层级论文.

本文所备的材料 (§ 8.2):

第一, 捕获范围的比较. 有效 Cauchy 以算法可逼近性为判据, 捕获可计算实数; 而 $\mathcal{P} \cap \mathbb{R}$ 严格含于可计算实数之中. 故它在实轴上捕获得严格多于 $L_2$ 的周期核心.

第二, 判据的类型. 三条判据——有限代数约束、代数积分表示、算法可逼近性——互不为对方的特例.

第三, 所响应的余项不同. 复化响应的是 $L_1$ 的基底外余项 $i$ (§ 3.2); 有效 Cauchy 响应的是 $L_1$ 的完备性余项——有限代数模式不阐述 Cauchy 序列的收敛.

本文的判定: 相对于本文这条复化 trajectory, 有效 Cauchy 阐述模式登记为响应另一项 $L_1$ 余项的并行 side-branch. 理由如下.

• 不是本文所研究的普通周期捕获模式内部的 sub-articulation: 若是, 其捕获范围应含于 $\mathcal{P} \cap \mathbb{R}$ 之内; 而它在实轴上捕获得严格多. 须注意此条只排除它落在该捕获模式内部, 不排除它落在某个别的 $L_2$ sub-articulation 内——本文反复声明 $\mathcal{P}$ 只是取值核心而不穷尽 $L_2$ (§ 2.6 其二), 故不得由"多于核心"推"不在任何 $L_2$ 之内".
• 不是本文所构造的复化跃迁路线的必要桥: 本文 § 3 至 § 5 所阐述的 $L_1 \to L_2$ 路线 (标而不构 $i$ → 二维化 → 覆叠) 不经过它. 此条只排除它是这条路线的必要桥, 不排除它是其他 $L_1 \to L_2$ trajectory 的桥.
• 是并行 side-branch: 按 § 10.3 的判据, 复化与有效 Cauchy 都是余项支——都由处理 $L_1$ 的某项余项而得——但所处理的余项不同. 二者是同层不同余项的两条回应, 故并行而不竞争.

据此, Paper 3 § 6.4 所留的开放问题在本文闭合.

该判定的范围须限定. 本文判定的是有效 Cauchy 与本文这条复化 trajectory 的关系, 不判定它在全部 $L_2$ trajectory 空间中的位置——与 § 10.2 对 Cayley–Dickson 所守的同一条限度. 且本判定属 SAE 登记, 非数学结论: 三条捕获范围的包含关系是标准结果, 而由该包含关系判定阐述模式之间的分支关系, 是本文在 SAE 框架内作出的识别.

〔类型: SAE 登记, 依据为 T2 事实 (捕获范围包含关系) 加 § 10.3 的分类判据〕

11.4 与 ZFCρ 线的关系

ZFCρ 线所提的问题是: $L_1$ 的单块 $\infty$ 不能分辨 fine ratio.

本文的处理分两半, 状态不同:

半	位置	状态
阶	§ 6.2	由赋值 $\mathrm{ord}_p$ 清. $\hat\infty$ 获得 $\mathbb{Z}$ 值分级; $\infty/\infty$ 的阶为整数减法
系数 (fine ratio)	§ 7.4	未清. 接口定位于 Newton 多边形的第二层 (指数因子的首项系数), 但未接通

两点须守:

其一, 与 ZFCρ 接口的是赋值, 不是除子——除子只是把各点的赋值打包 (§ 6.2). 说"除子给 $\infty$ 装刻度"是把打包动作误认为分级动作.

其二, 有接口不等于已接通. 本文未给出 fine ratio 与某个特征方程系数之间的显式映射、保持量或函子. 债务状态从"未接口"改登记为"有接口而未接通", 不得引作清账 (§ 2.6 其六、§ 12.1).

ZFCρ 线论文引用本文时, 应引本表的状态列, 不应引作该债务已解决.

11.5 与 SAE Physics Foundation v2 的关系

11.5.1 v2 的既有陈述

Foundation v2 是物理侧的系统性重述, 但其中含若干关于数学阶梯的陈述, 与本文直接相关:

• § 2.1 平行阶梯对照表: 数学 $L_1 \to L_2$ 为对偶化闭合, 闭合方程 $e^{i\pi}+1=0$, 余项 $i$ 与 $\pi$, 签名 $e$;
• § 7.3: 数学 $e$ 是标量层级闭合签名, 数学 $\pi$ 是标量层级闭合余项, 物理 $\hbar$ 是算符层级闭合签名;
• § 7.2: 数学 $L_2$ 关闭的是数域, 基底为纯数层; 物理 $L_2$ 关闭的是共轭读出, 多一层状态依赖;
• § 1.2: 两阶梯平行, 结构同型而内容独立; 通过跨阶梯本体对应互助——例如 $i$ 在两阶梯各自原生出现, 不是物理借用数学符号.

11.5.2 本文与 v2 一致的三处

余项 $i$ 与 $\pi$: 本文 § 8.3 与 v2 的派法一致. 本文另给出为何恰是这两项——$i$ 是跃迁的输入 (§ 3), $\pi$ 是本层的第一个结构产物 (§ 5.3 的 deck 格, § 6.4 的周期捕获). v2 未给出该理由.

$i$ 在两阶梯各自原生: 本文 § 3.2 阐述 $i$ 由 $L_1$ 自身的反向追问 (多项式求根) 逼出, 是 $L_1$ 的基底外余项. 这与 v2 "各自原生" 的立场一致, 且从数学侧提供了原生性的具体阐述.

数学 $L_2$ 的基底是纯数层: 与本文 § 2.1 一致. 但须区分基底与阐述模式——v2 所说的是基底 ($\mathbb{C}$ 关闭数域), 本文 § 6.1 的承载体迁移所说的是模式的捕获核心 ($\mathcal{P}$ 是取值像, 不是标量域). 二者不冲突; 并读时若不作此区分, 会误以为两篇对同一对象给出了不同类型的判定.

11.5.3 两个签名槽

v2 派给数学 $L_1 \to L_2$ 的签名是 $e$; Paper 1 § 4.5 派给 $L_2$ 的阐述交换单位是 $2\pi i$. 本文立: 二者是两个不同的槽, 不是同一槽的两个候选.

槽	内容	签署什么	位置
跃迁签名	$e$	跨越本身: 在 $e^{i\pi}+1=0$ 中绑定两项余项	§ 8.3
层内交换律阐述单位	$2\pi i$	本层两条阐述路线的可交换性: 留数定理与辐角原理	§ 6.3

一条结构后果值得登记: $e$ 既是跃迁签名, 又是本层闭合余项的数值见证 (§ 7.5). 即——签署这次跨越的那一项, 确定落在指数周期扩张之内, 而在本层普通周期核心中的成员状态至今未定.

(此处刻意不写作"这一层捕获不了的那一项": 后者在数值层面等价于断言 $e \notin \mathcal{P}$, 而该命题未决. 「确定在扩张内」是定理, 「不在核心内」是猜想; 结构张力由前者承担.)

该双重身份与 § 5.5 所阐述的 $2\pi i$ 双重身份同型而方向不同: $2\pi i$ 的双重身份是同一对象在层内两处 (deck 格与留数归一); $e$ 的双重身份是同一对象在跃迁与层内两处.

对 § 10 的影响: § 10 的 Cayley–Dickson 判据使用的是层内交换律单位 $2\pi i$, 不是跃迁签名 $e$. 该判据因此不受 v2 派签名的影响.

11.5.4 本文对 v2 数学行的重组

v2 § 4 在阐述辛结构化时使用"数学 $L_1 \to L_2$ 指数映射"作为 act type. 本文 § 5.3 与 § 5.4 阐述 $\exp$ 为 $L_2$ 的内部产物——主步骤 3 内层 4 所产出的覆叠映射——故把跨层行为改述为复化 (标而不构 $i$, § 3).

登记: Foundation v2 的主要工作在物理侧; 数学阶梯各行在 v2 中作平行对照而出现, 非专题阐述. 本文作为数学系列的 $L_2$ 专题论文, 在数学系列辖区内重组该行. 建议该重组在 Foundation v3 中被吸收.

本文取显式修订的模态, 不取兼容性解释. 即: Paper 4 修订 Foundation v2 数学行中 act type 的安放——把 $\exp$ 由跨层 act type 改判为 $L_2$ 内部产物. 该修订不影响 Foundation v2 的任何物理侧论证, 建议 Foundation v3 吸收.

之所以不取"v2 说的是粗分辨率下的闭合关系, 本文说的是生成住址, 二者兼容"这一解释: 该解释虽可使两文并存, 但它是本文代 v2 作出的重新诠释, 而 v2 原文并未作此区分. 按本系列纪律, 代上游重新诠释比显式修订更不透明.

11.5.5 词汇不统一, 且这是有意的

v2 § 6.3 给数学 $L_2$ 闭合派了一对词汇: 裸读出形式 ($e^{i\pi}+1=0$) 与不变形式 ($e^{ix} = \cos x + i\sin x$, 以及 Cauchy–Riemann 方程、谱定理等).

本文不采用该词汇对, 而使用数学系列自身的词汇 (落点、跃迁闭合方程、叠加落点).

理由出自 v2 自身: v2 § 6.4 明确把双重阐述 pattern 的普适性限定在物理量阶梯的阐述范围内, 不提前扩成所有可能阶梯的普适定理; 而 § 1.2 立两阶梯内容独立. 故数学阶梯按数学系列自身的组织方式阐述, 是遵循 v2 的限定而非背离它.

但须给出对照, 否则并读的读者会以为是两套互不相干的描述:

v2 (物理量阶梯词汇)	本文 (数学系列词汇)	是否同一对象
裸读出形式 $e^{i\pi}+1=0$	跃迁闭合方程 (§ 8.3)	是
不变形式 $e^{ix} = \cos x + i\sin x$	基底级经典构造与模式级产物的认同 (§ 5.5)	是
—	叠加落点 (§ 8.3)	本文另加的第三个谓词

一条收敛值得登记: 本文 § 5.5 由基底级经典构造与模式级产物 (内层 4 的覆叠映射在虚轴的限制) 相等, 推出 $\cos\theta + i\sin\theta = e^{i\theta}$ 的结构地位; v2 § 6.3 独立地把同一对象派为不变形式. 两条路线指向同一处. 该收敛不是重复, 是相互印证——本文由阐述过程的内部结构到达, v2 由跨阶梯的形式对照到达.

11.5.6 不焊的部分

$\hbar = h/2\pi$ 与 $[\hat x, \hat p] = i\hbar$ 同时含 $2\pi$ 与 $i$, 与数学 $L_2$ 的 $2\pi i$ 在形式上接近. 这是一处诱人的焊缝.

本文不焊.

v2 § 7.3 已给出为何 $\hbar$ 不是数学 $e$ 或 $\pi$ 的复制: 数学是标量层级闭合, 物理是算符层级闭合, 后者比前者多一层状态依赖. 本文接受该区分, 且不进一步焊接.

理由: 焊接需要跨阶梯的本体对应论证. v2 § 1.2 只对 $i$ 给出了这样的论证 ($i$ 作为闭合结构的普适特征在两阶梯各自原生); 对 $2\pi$ 未给出. 本文不代为给出.

〔登记为 T3 接口, § 12.1〕

11.5.7 数学 $L_2 \to L_3$

v2 § 2.1 把数学 $L_2 \to L_3$ 派为对角化, 签名 $\Omega_U$. 本文不处理该行 (§ 12.2 辖区).

须登记一项待检: 本文 § 13.2 的迁移预测与该派法是否相容, 属 Paper 5 ($L_3$) 的工作. 本文既不主张相容, 亦不主张不相容.

---

12. claim 三档、辖区、可证伪性与预测纪律

12.1 三档表

按物理 Paper 1 所立的三档 (T2 可锁 / T3 开放 / 不得作本篇结论), 本文主要陈述的档次如下.

陈述	档	依据 / 位置
$\mathsf{Int}_{\rm alg} \hookrightarrow \mathsf{Int}_{\exp}$	定义级 T2	普通周期与指数周期表达式类的定义; § 7.1
正则联络扩展至允许不规则奇点的联络	定理级 T2	Levelt–Turrittin 等形式分类; § 7.1
斜率全为零 $\Leftrightarrow$ 正则奇点	T2	标准斜率理论; § 7.2 一层
斜率是决定多项式的次数	T2	同上
增长的型依赖 $\lvert a_s\rvert$, 非斜率	T2	§ 7.2 二层
候选 Stokes 方向依赖指数因子之差 $q_\alpha - q_\beta$ 的首项系数与其辐角, 非斜率	T2	§ 7.2 二层
Stokes 乘子是形式分类之外的解析模量	T2	Hukuhara–Turrittin; § 7.2 三层
§ 9.2 四条链	T2	各条为经典结果; § 9.4 归属声明
$\overline{\mathbb{Q}} \subset \mathcal{P}$; $\mathcal{P}$ 可数; 存在可计算而非周期的实数	T2	标准结果与 Yoshinaga; § 8.2
$e \in \mathcal{P}^e$	T2	定义与计算; § 7.5
"正则 $\leftrightarrow \mathcal{P}$、不规则 $\leftrightarrow \mathcal{P}^e$" 的对齐	T2-条件项	exponential motives 一线研究级对应; § 7.1
$e \in \mathcal{P}$	T3	开放猜想; 作候选严格性见证
一般周期表达式等同性问题的可判定性	T3	开放; § 8.4
$G_\infty$ 与 $\mathrm{Slope}_\infty$ 的比较接口	T3	候选接口, 未证同一性; § 7.3
承载体迁移	SAE 登记	§ 6.1
$\mathcal{P}$ 为 $L_2$ 的主要捕获范围	SAE 登记	§ 6.4; 不宣称穷尽
主步骤 3 反向追问的不对称	SAE 登记 (由 T2 事实支撑)	§ 5.4
代价支／余项支判据	SAE 的 T3 分类提案	§ 10.3
$L_3$ 的迁移预测	T3 纲领性预测	§ 13.2
ZFCρ 的单块 $\infty$ 债务已清	不得作本篇结论	只清阶那一半; § 6.2、§ 7.4
$\mathbb{H}$ 被逐出整个 $L_2$	不得作本篇结论	§ 10.2
本文推导了复分析	不得作本篇结论	§ 2.6 其一
本文证明或反驳了 $e \notin \mathcal{P}$	不得作本篇结论	§ 2.6 其三
所有数学工具的并集不穷尽基底	不得作本篇结论	§ 8.2 辖区声明

12.2 辖区声明

本文管: $L_2$ 阐述模式的捕获范围与余项账本; 四主步骤的 sub-articulation 及其登记类别; $\omega_\rho^{(2)}$ 六分量的类型与已有接口; 上述内容与 Paper 1、Paper 2、Paper 3 及 Foundation v1 的关系.

本文不管: 周期数论具体结果的归属与优先权; 复分析各技术定理的归属; $\omega_\rho^{(2)}$ 的完整合成律 (§ 9.3、§ 13.1); $L_3$ 的 trajectory (§ 1.3); "数学工具"边界的划定 (§ 8.2); 数学安放坐标 (§ 13.3); 四元数分析自身的缺陷结构 (§ 10.4).

12.3 可证伪性: 双向登记

反事实一: 若 $e$ 被证明是周期 ($e \in \mathcal{P}$), 则须作结构性改写者, 只有 § 7.5 中 $e$ 的见证角色与 § 8.2 账本中的归类; 全文各处关于该命题"未知／开放"的事实状态说明须作机械同步更新 (摘要、§ 1.3、§ 2.6 其三、§ 12.1 三档表等). 承重结构不动: § 7.1 的两条扩张、§ 7.2 的三层、§ 5.4 的核心结构陈述、§ 9 的两张表与四条链.

此项成立的前提是本文的承重已从常数 $e$ 转到 $\exp$ 与表达式类扩张上 (§ 7.1). 若承重仍压在 $e \notin \mathcal{P}$ 上, 该反事实不成立——那时须改动的将包括全部余项阐述.

反事实二: 若"$e \notin \mathcal{P}$"被证明相对某形式系统 $F$ 独立, 则须更新者为该命题在 $F$ 中的证明论状态 (§ 2.4 三轴之第二轴). 该更新不自动决定一般周期表达式等同性问题的算法状态, 亦不改变数值 $e$ 的对象类型——独立性是命题的属性, 不是对象的属性 (§ 8.4 混层禁令其一). 故 $L_2/L_3$ 分栏不受影响.

为何须并列登记第二条. 一篇立有反免疫声明的论文, 若其可证伪性节只登记对自身无害的反事实, 则该声明本身成为装饰. 本条的起草经过见附录 A「版本与审稿来源记录」.

12.4 反免疫声明

承 Paper 1 附录 A.8.5: 不得把每一个候选例行重新分类为模式更新的机会——那会使模式免疫于反驳, 摧毁可证伪性.

本文有两处须执行该纪律.

其一, $e \notin \mathcal{P}$ 长期未被证伪, 只记为一致, 不记为支持. 该命题自 Kontsevich–Zagier 提出以来未被证伪, 这一事实与本文框架一致; 但不构成对框架的支持证据. 论证"若它是周期早该被找出表示"是似真性论证, 属本体边界一栏 (§ 7.5), 不进入承重.

其二, 合成律的欠账登记为欠账, 不转成发现. 完整合成律未被给出这一事实, 本身不构成对框架的任何支持——它是一笔欠账, 按 § 9.3 登记为未还挂账.

特别地, 该欠账不得因"该封装具元层级性质" (§ 8.4) 而被重述为一项识别: 对象的性质与本文是否给出它, 是两件事.

本文在起草过程中对该纪律有两处执行记录 (含一次自我改正), 见附录 A「版本与审稿来源记录」. 反免疫纪律的可信度取决于它在本文内部也被执行, 而非只在对外评价时被援引; 故该记录不省略, 只是不置于本节.

12.5 预测纪律

本系列出预测. 预测可能错. 被证伪时认.

此姿态与 § 12.4 的反免疫声明互为表里: 不免疫的代价是可能错; 而拒绝出预测本身也是一种免疫——一个从不作可检验陈述的框架, 同样无法被反驳.

Paper 1 已有先例: 该文明写交换律模式是"模式的一项非平凡跨层级预测", 并逐层列出检验结果.

> 〔v2 修订〕 v1 此处续写: "本文 § 13.2 的预测按同一方式登记: 给出预测、给出操作化的证伪条件、并指出已知的反例候选." 该句的前提已失效 —— § 13.2 的证伪条件已撤回 (其裁决性读法为空), 反例候选表的第一行亦已撤回. § 13.2 现登记为无操作化证伪条件的 T3 解释性提案. > > 另: Paper 1 的交换律预测本身也已在其 v2 中下调 —— 五层证据经两阶梯拆分后, 数学侧只剩三项且第三项待定, 且三项是构造集而非检验集. 故此处的"先例"须连同该下调一并读.

时序凭据. 本文先于 $L_3$ 论文公开发表. 该终签版本的版本级 DOI、存档页面及其公开记录, 构成第三方可核验的公开在先凭据: 它证明 § 13.2 在 $L_3$ 论文公开之前已经进入公共记录; 它不证明作者私人工作区中的起草顺序, 也不构成对该预测的独立检验.

引用与版本纪律: 应引该终签版本的版本级 DOI, 而非概念 DOI——后者的指向会随新版本移动. 后续若因 $L_3$ 论文而修改本文, 应发布为新版本, 不覆盖本版本; 本版本承载 § 13.2 的公开优先性.

此处刻意不以作者自填的日期承重: 自填日期只是自我声明, 无法向外部读者证明任何事; 若以之承重, 预测纪律就压在自述上, 与本文对 $e \notin \mathcal{P}$ 所守的承重纪律形成双重标准 (§ 7.1). 承重的是发表记录本身, 不是本文的任何陈述.

但两条限制须同时登记, 二者皆与时序无关, 不因先行发表而解除.

其一, 后续的 $L_3$ 论文不构成 § 13.2 的独立检验. 同一作者在同一框架内所写的续篇, 无论写出什么, 都不是对该预测的检验. 该预测的独立检验必须来自外部数学. 这一条不因降档而改变.

> 〔v2 修订〕 v1 此处续写: "按 § 13.2 给出的操作化条件, 真正能触发证伪的是对旧语言后果保守的对象扩张, 现登记的实例为保守的类理论扩张; 大基数扩展与强迫扩展仍是合法的压力测试对象, 但在当前证伪条件下不构成证伪路径." > > 该段整段撤回: (i) 保守对象扩张不可能裁决旧语言中原本不可判定的句子 (§ 13.2 (a)); (ii) 一旦把闭合目标精确化, 该条件对一切有效公理化扩张为空, 不分保守与否 (§ 13.2 (b)). 故"能触发者"与"不能触发者"的两栏划分本身不成立. 大基数与强迫扩展仍是合法的压力测试对象 —— 这一句保留, 但不再与"证伪路径"对举.

其二, 因而本文可主张"该预测在先", 不可主张"该预测已被检验". 二者差别不小: 前者由发表记录担保, 后者需要独立于本框架的检验者. 若后续 $L_3$ 论文与本预测相符, 本文不得据此声称预测已被证实.

> 〔v2 修订〕 v1 此处收尾: "故 § 13.2 的可检验性主要由其操作化的证伪条件承担; 发表记录只保证它是预测而非回填, 不保证它已被检验." > > 前一半的承担者已失效: 该条件已撤回, § 13.2 现无操作化证伪条件, 故其可检验性当前无承担者. 后一半仍成立: 发表记录只保证它是提案而非回填.

因而须补一条本节此前未规定的案例 〔增补〕.

v1 写"本系列出预测; 预测可能错; 被证伪时认". 但本次发生的不是证伪 —— 该预测未被任何反例推翻, 而是其证伪条件被发现为空. 该情形原节未规定. 拟补:

> 若某预测的证伪条件事后被证明为结构上不可满足, 则该条件失效, 该预测须降档为解释性提案或重建; 不得以"尚未被证伪"为由继续持有该预测.

理由: 一条只规定"错了要认"的纪律, 遇到"条件根本不可能触发"时是无声的 —— 而后者是更隐蔽的免疫形式. 故此补充是对本节的加强, 非削弱.

"结构上"三字有定理支撑: 由 § 13.2 (b), 一旦把闭合目标精确化为 $C_T^\vdash=C_{\mathbb{N}}^\vDash$, 该条件对一切有效公理化扩张为空. 这不是一条组合观察, 是全闭合障碍定理的直接后果.

修订后本节的执行状况: § 12.4 的反免疫纪律在本节的执行, 现由三件事承担 —— (i) 预测被降档而非被悄悄保留; (ii) 降档的理由与过程逐条登记 (§ 13.2 的 v2 修订); (iii) 上述新增案例把"条件为空"这一情形纳入纪律. 而不再由"§ 13.2 有一条操作化的证伪条件"承担.

---

13. 开放问题与后续 trajectory

13.1 $\omega_\rho^{(2)}$ 的完整合成律

状态: 未还挂账.

Paper 1 § 11.1 所挂的六分量合成律, 本文交付最小 typed composition 骨架 (§ 9.1、§ 9.2), 未交付六类数据的统一范畴封装. 障碍不是范畴论工具的缺失, 而是六分量尚未被完整映到 Riemann–Hilbert 数据上——§ 9.2 的链 3 与链 4 已进入该语言, 链 1 与链 2 仍在经典复分析语言中, 二者之间的翻译本文未给出 (§ 9.3).

推至: $L_2 \to L_3$ 接口论文. 该封装具元层级性质 (§ 8.4), 而按 Paper 1 § 11.4, 范畴论结构主义的充分接洽须在 $L_2$ 与 $L_3$ 两处处理.

13.2 $L_3$ 的迁移预测

预测: 在本文选定的完整精确化 trajectory 中, 反向追问的压力在对象域扩张耗尽之后, 转入表示与规则空间.

依据: § 5.4 的不对称 ($L_1$ 扩张标量基底; $L_2$ 因代数闭而转为覆叠) 与 § 8.4 的触发形式 (值的层面转为规则的层面). 两处显示同一方向: 压力在对象侧受阻时, 释放方式向表示与规则一侧转移.

证伪条件 〔v2 撤回, 详见本节末〕:

> 若 $L_3$ 的主要闭合动作能够仅通过向对象论语言加入新对象种类完成, 而无需改变等价、证明或可判定规则, 则本文所提"对象域至规则域的不对称迁移"被证伪.

v1 认为如此表述使"对象域扩张"与"压力被释放"两者皆有可操作的判定标准. 该判断不成立 —— 条件从未定义"$L_3$ 的主要闭合动作"指什么, 故它精细描述了扩张机制而未给出闭合目标. 详本节末的 v2 修订.

v1 曾按是否能触发证伪条件登记两栏反例候选. 该表的第一行现已撤回:

候选	v1 判定	v2 判定
保守的类理论扩张 (如 NBG)	能触发	不能触发 —— 见下文 (a)
非保守的类理论扩张 (如 MK)	否	否 (但理由须改, 见 (b))
大基数扩展	否	否 (但理由须改, 见 (b))
强迫公理 (MA、PFA 等)	否	否 (同上)

v1 由该表得出"故证伪路径只有一条, 且须限定到保守的那一支", 并称保守性是有判定标准的数学性质, 故该预测的可证伪性是具体可查的问题. 两条结论皆须撤回.

---

13.2 的 v2 修订

本节的修订源于 SAE 数学 Paper 5 (10.5281/zenodo.21538494) § 11.2. 分四项.

(a) 撤回"NBG 能触发证伪"的登记 〔改错〕

设 $S$ 在旧语言 $L$ 上保守扩张 $T$, $G$ 为 $L$-句且 $T\nvdash G$, $T\nvdash\neg G$. 若 $S\vdash G$, 由保守性得 $T\vdash G$, 矛盾; 对 $\neg G$ 同理. 故

> 保守的对象扩张不可能裁决旧语言中原本不可判定的句子.

v1 表格第一行的错在于: 它验证了 $S/T$ 保守且增加了类语言, 却未验证 $S$ 完成了 $L_3$ 的主要闭合. 二者不是一回事.

(b) 撤回"证伪路径只有一条", 且该路径为空的范围更大 〔改错〕

由 (a), 保守那一支为空. 而由 Paper 5 § 7.1 的推论 7.2, 该空可扩至一切有效扩张, 不分保守与否:

扩张类型	为何不触发
保守 (NBG 型)	裁决不了任何旧语言独立句 (由 (a))
非保守而仍有效 (大基数、强迫公理)	能裁决一些, 但永远达不到全闭合 (推论 7.2)

须注意 v1 对后两行的理由也不准: v1 说大基数"必然改变可证性故不满足无需改变证明规则", 这把改变可证后果与改变证明规则混为一谈 —— 二者分属下文 (d) 所列五轴中的 R3 与 R0.

但仍不可写成"原条件在所有读法下皆为空": 已证的是裁决性读法下为空; 表达性读法另论.

(c) 原证伪条件的地位 〔实质修订〕

原条件未定义闭合目标. 它精细描述了扩张机制 (加什么、不许动什么), 却从未说明"$L_3$ 的主要闭合动作"指什么.

Paper 5 把闭合目标精确化为句法后承闭包与标准算术语义闭包相等, 即 $C_T^\vdash=C_{\mathbb{N}}^\vDash$, 由此得到的不是一条新的证伪条件, 而是两条定理:

• 全闭合障碍: 不存在有效公理化的理论 $T$ 使 $C_T^\vdash=C_{\mathbb{N}}^\vDash$ (可靠性不是前提);
• 分片闭合阈值: 有效、$\Gamma$-可靠且 $\Gamma$-完备的理论存在 $\Rightarrow\mathrm{True}_\Gamma$ 是 c.e.; 阈值沿 $\Sigma$ 断在 $\Sigma_1\,|\,\Sigma_2$, 沿 $\Pi$ 自 $\Pi_1$ 即起, 无中间地带.

该定理不证明本节的迁移预测. 定理划的轴是有效／非有效, 迁移预测划的轴是对象／规则 —— 两根不同的轴. 事实上 $\mathrm{Th}(\mathbb{N})$ 达成全闭合而它并非有效规则系统, 也不必描述为"规则迁移" —— 它只是把完整语义真理直接放入公理集.

(d) 迁移句的地位 〔实质修订〕

档次不变, 仍为 T3 —— v1 已如此登记. 变的是 claim 的类型:

$$\text{T3 纲领性预测 (带操作化证伪条件)}\ \longrightarrow\ \text{T3 解释性提案 (无操作化证伪条件)}.$$

须说清发生了什么: 不可写"原预测后来被证明为定理" —— 那会掩盖对象的变化. 准确的登记是三件事的合取:

$$\text{撤回旧预测的操作化}\ +\ \text{新增两条定理}\ +\ \text{原迁移解释降为提案}.$$

另一处词义须拆开 〔澄清〕: 原条件中的"规则"至少含五轴 —— R0 逻辑演算 · R1 语言／签名 · R2 非逻辑公理 · R3 旧语言可证后果 · R4 算法结构. 原文措辞"等价、证明或可判定规则"读起来像 R0, 而两栏表实际取 R3 (以保守性作判据). 元语言表达力属 R1, 非 R2. 后续任何相关条件须写成条件向量, 不得再用裸词"规则".

修订后本节还剩什么: 迁移观察本身仍在 —— § 5.4 的不对称与 § 8.4 的触发形式都是已挣得的文本事实; 变的是本节不再声称该观察配有一条可操作的证伪条件.

档次: T3 解释性提案 (v1 作 T3 纲领性预测; 档次不变, claim 类型改, 见 (d)). 不是数学定理的预测——本文不预言任何具体定理的成立与否, 只提出压力释放方式的一种类型学读法.

本文不预设 $L_3$ 的 trajectory (§ 1.3). 上述提案是关于释放方式类型的, 不是关于 $L_3$ 内容的; 后者见 Paper 5.

关于时序: 本文先于 $L_3$ 论文公开发表, 该终签版本的版本级 DOI 与存档记录即本节公开在先的外部凭据 (不是写作在先——发表记录不能证明私人工作区中的起草顺序). 但本文不因后续论文与本预测相符而声称该预测已被检验——同一作者同一框架内的续篇不构成独立检验. 见 § 12.5.

v2 补: 降档之后, "在先"的对象须分开陈述 —— 观察在先 (压力似乎转入规则空间, 该观察确实先于 $L_3$ 论文公开), 而预测主张撤回 (其操作化不成立). 二者不分开会成为"保留优先权而撤回内容".

13.3 数学安放坐标

物理 Paper 1 立住址判据: 一个对象住在阐述它所需资源的最高根层. 该判据在数学侧的对应物——数学安放坐标——本文不做.

本文只在需要时局部使用了落点词汇 (§ 8.3 的叠加落点), 未建立系统的安放坐标. 全面的安放 (各常数、各定理、各构造分别住在何处) 留待合篇时决定.

13.4 其他接口

有效 Cauchy 阐述模式的落点: Paper 3 § 6.4 所留开放问题, 材料见 § 8.2, 判定见 § 11.3.

逻辑专题论文、审美数学专题论文、关系数学专题论文: 接口不变. 本文未涉及这三条线, 亦未改变 Paper 2 对它们所作的登记.

ZFCρ 线: 债务两半的状态见 § 6.2 与 § 7.4. 未清的一半 (系数层面的 fine ratio) 其接口定位于 Newton 多边形的第二层, 尚未接通.

Foundation 线: 本文承接并重组 Foundation v1 跨层账本表的 $L_1 \to L_2$ 行, 见 § 8.3 与 § 11.5.4.

---

附录 A · 版本与审稿来源记录

本文经四方内部审查 (Claude、ChatGPT、Gemini、Grok 各任一席), 大纲历 v1 → v2 → v2.1 redline 三轮, 其中 ChatGPT 席位对 v1 判 Major Revision.

此处所称 review 为作者组织的模型辅助内部结构审查, 不等同于期刊同行评审.

两处与本文 § 12.4 反免疫声明直接相关的改正, 在此登记.

其一. 大纲 v1 曾把"给不出完整合成律"诊断为 $L_2 \to L_3$ 触发的一个实例——即把一笔欠账转成一项识别. 该处理与本文对 $e \notin \mathcal{P}$ 所执行的同一纪律形成双重标准. 经审稿指出后改正; 现按 § 9.3 登记为未还挂账, 并按 § 9.1、§ 9.2 交付最小合成骨架作部分偿付.

其二. § 12.3 的可证伪性登记, 初稿只含良性反事实 (若 $e$ 被证明是周期). 一篇立有反免疫声明的论文, 若其可证伪性节只登记对自身无害的反事实, 该声明即成装饰. 且最初拟补的第二条反事实写作"余项本身成为 $L_3$ 对象"——该写法会破坏 § 8.4 所建立的值/规则分栏, 因为独立性是命题的属性, 不是对象的属性. 现按 § 12.3 作双向登记, 第二条采用修正后的形式.

另有三项承重决定在审稿过程中被降档或重组, 均已在正文相应位置登记: 捕获范围由"饱和于 $\mathcal{P}$"降为"正则周期取值核心" (§ 6.1); 余项的承重对象由常数 $e$ 改为指数映射 $\exp$ 与表达式类扩张 (§ 7.1、§ 7.5); Newton 多边形由"阶段 4 的共同发生器"改为"形式组织器", 且其与 Paper 1 附录 B.5 之 $G_\infty$ 的关系由"被安放"降为"候选比较接口, 尚未证明同一性" (§ 7.2、§ 7.3).

本附录承担的是来源可追踪功能, 不承担"论文因此可信"功能. § 12.4 反免疫声明的内部证据是正文自身: $e \notin \mathcal{P}$ 不承重 (§ 7.1)、合成律欠账未转成发现 (§ 9.3)、§ 12.3 登记了对本文不利的反事实. 本附录只补充版本链的可追溯性.

---

参考文献

周期与超越性

1. M. Kontsevich, D. Zagier, "Periods", in B. Engquist, W. Schmid (eds.), Mathematics Unlimited — 2001 and Beyond, Springer, Berlin, 2001, pp. 771–808. — 普通周期环 $\mathcal{P}$ 的定义; 加性、变量替换、Stokes 公式三条规则与周期猜想; 指数周期; $e$ 与 $\gamma$ 不是普通周期的期望.
2. M. Yoshinaga, "Periods and elementary real numbers", arXiv:0805.0349 [math.AG], 2008. — 周期可由初等有理 Cauchy 序列有效逼近; 构造出一个可计算而非周期的实数.
3. J. Ayoub, "Une version relative de la conjecture des périodes de Kontsevich–Zagier", Annals of Mathematics 181 (2015), 915–992. — Kontsevich–Zagier 猜想的相对/函数域版本.
4. J. Fresán, "Une introduction aux périodes", arXiv:2210.03407 [math.AG], 2022. — 周期理论综述; de Rham–Betti 配对.
5. J. Ax, "On Schanuel's conjectures", Annals of Mathematics 93 (1971), no. 2, 252–268. — Ax–Schanuel 定理 (Schanuel 猜想的微分/函数域类比).
6. Y. André, Une introduction aux motifs (motifs purs, motifs mixtes, périodes), Panoramas et Synthèses 17, Société Mathématique de France, 2004. — 周期与 motivic Galois 理论; Grothendieck 周期猜想.
7. J. Fresán, P. Jossen, Exponential motives, 手稿, 可于第一作者主页获取. — 指数周期的 motivic 框架.

不规则奇点、形式分类与 Stokes 现象

1. H. L. Turrittin, "Convergent solutions of ordinary linear homogeneous differential equations in the neighborhood of an irregular singular point", Acta Mathematica 93 (1955), 27–66.
2. A. H. M. Levelt, "Jordan decomposition for a class of singular differential operators", Arkiv för Matematik 13 (1975), no. 1–2, 1–27. DOI: 10.1007/BF02386195.

> 条目 8、9 与 M. Hukuhara 的早期工作合称 Hukuhara–Levelt–Turrittin 定理: 形式穿孔圆盘上的联络, 经适当的有限分歧后, 分解为只含单个指数因子的初等联络之直和. 本文 § 7.1 与 § 7.2 的定理级承重出自此. Hukuhara 原始论文 (Mémoires of the Faculty of Science, Kyushu University, 1940 年代初) 未经直接查阅; 该定理的陈述与归属见下条目 14.

1. W. Wasow, Asymptotic Expansions for Ordinary Differential Equations, Interscience, New York, 1965. — 不规则奇点处的渐近展开; 增长的阶与型; Stokes 扇区.
2. M. Loday-Richaud, "Stokes phenomenon, multisummability and differential Galois groups", Annales de l'Institut Fourier 44 (1994), no. 3, 849–906. — Stokes 现象、多重求和与微分 Galois 群的关系.
3. C. Sabbah, Introduction to Stokes Structures, Lecture Notes in Mathematics 2060, Springer, 2013. — Stokes 结构; 实定向吹起与方向圆上的局部系统.
4. K. S. Kedlaya, "Good formal structures for flat meromorphic connections, I–III", 2010–2015. — Hukuhara–Levelt–Turrittin 分解的高维推广; 不规则度与转折轨迹.

Riemann–Hilbert 与微分 Galois 理论

1. M. van der Put, M. F. Singer, Galois Theory of Linear Differential Equations, Grundlehren der mathematischen Wissenschaften 328, Springer, Berlin, 2003. — 微分 Galois 理论的标准参考; 正则情形的 Schlesinger 稠密性; 不规则情形的 Ramis 稠密性定理 (形式单值、指数环面与 Stokes 算子生成的子群在局部微分 Galois 群中 Zariski 稠密); 形式层的 $\mathcal{T} \rtimes \overline{\langle\hat\gamma\rangle}^{\rm Zar}$ 结构. 本文 § 9.2 链 4 出自此.
2. P. P. Boalch, "Geometry and braiding of Stokes data; Fission and wild character varieties", Annals of Mathematics 179 (2014), 301–365. arXiv:1111.6228 [math.AG]. — wild character variety; 不规则 Riemann–Hilbert 对应.
3. T. Mochizuki, Wild Harmonic Bundles and Wild Pure Twistor D-modules, Astérisque 340, Société Mathématique de France, 2011. arXiv:0803.1344 [math.DG]. — 不规则联络的高维理论.
4. N. M. Katz, "Nilpotent connections and the monodromy theorem: applications of a result of Turrittin", Publications Mathématiques de l'IHÉS 39 (1970), 175–232.

四元数分析

1. G. Gentili, D. C. Struppa, "A new theory of regular functions of a quaternionic variable", Advances in Mathematics 216 (2007), no. 1, 279–301. DOI: 10.1016/j.aim.2007.05.010. — slice regular 函数论的奠基论文: 沿 Cullen 的想法, 正则函数在每个复切片 $\mathbb{C}_J$ 上的限制是全纯的. 本文 § 10.2 前半出自此.

1. G. Gentili, C. Stoppato, D. C. Struppa, Regular functions of a quaternionic variable, Springer Monographs in Mathematics, Springer, Berlin–Heidelberg, 2013; 第二版 2022. — 该理论的系统展开, 含切片上的 Cauchy 核与积分公式. 本文 § 10.2 所引的 Cauchy 型结构出自此条, 不出自条目 18.

经典复分析

除子与除子类群、留数定理、辐角原理、de Rham–Betti 配对、单值表示、代数基本定理、一点紧化与 Riemann 球面. 这些结果的技术内容由 19 世纪至 20 世纪的数学工作确立, 载于任何标准教科书. 按 § 2.6 其一, 本文引用而不主张推导它们.

SAE 系列

1. 秦汉, 《SAE 数学奠基篇》(Paper 1). DOI: 10.5281/zenodo.20153791
2. 秦汉, 《从 $L_0$ 出发追问数学的不可避免性与多路径实现》(Paper 2). DOI: 10.5281/zenodo.20199082
3. 秦汉, Finite Algebraic Articulation of the Quantitative Dimension $L_1$ (Paper 3). DOI: 10.5281/zenodo.20266351
4. 秦汉, SAE Foundation v1. DOI: 10.5281/zenodo.19361951 — 跨层账本表; 本文承接并重组其 $L_1 \to L_2$ 行 (§ 8.3、§ 11.5.4).
5. 秦汉, SAE Physics Foundation v2. DOI: 10.5281/zenodo.20322460 — 平行阶梯对照表; 算符层级与标量层级闭合的区分 (§ 11.5).
6. 秦汉, 《SAE 物理 Paper 1》. DOI: 10.5281/zenodo.21227329 — 二层嵌套通则; 单点/叠加落点; 本体边界与定义边界; claim 三档.
7. 秦汉, 《Methodology 6: 相变窗口》. DOI: 10.5281/zenodo.19464507
8. 秦汉, 《Methodology 00: Via Rho》. DOI: 10.5281/zenodo.19657440