Skip to content

56. The Collatz first cause: the decree that fell

← 55. Collatz · Table of contents · 57. The well-foundedness canon →

Lean source: Engine/CollatzFirstCause.lean — now entirely green: the decree's post-mortem plus the surviving epistemics, with no quarantine import; the taint does not grow. The green front is Engine/CollatzTugOfWar.lean (chapter 55), where the refutation witness ropeLaw_universal_refuted was forged. The historical record of the removed field collatzBoundary — Engine/CausalClosureAxiom.lean §18. The P/NP epistemic complement — Engine/PNPFirstCause.lean. Notation: 🟢 — machine-proven (axioms stated explicitly); 🟡 — AXIOM-TAINTED, conditional on the repository's single axiom step00FirstCause; 🔴 — an open input.

Where we are

In chapter 55 Collatz was translated into the language of a tug of war and brought to an honest fork: the green mechanics (the window budget, descent under rope advantage, the HERO "law ⟹ halting") and one named 🔴 input — the universal law of rope dominance.

This chapter tells the story not of a static boundary but of its life cycle. We accepted a decree — and through the very tripwire (an intentional explosion detector — see the glossary) that we had planted in advance, we watched it fall. The order of events is this:

  1. the boundary is taken — the rope law becomes the fourth field of the single axiom;
  2. the decree discloses its honest price — only one arrow is proven, "the decree may overpay";
  3. the universal form of the law is machine-refuted (witness n = 27);
  4. the tripwire fires — keeping the field any longer would make the axiom inconsistent;
  5. the field is removed — at that moment the axiom was left with three boundaries (twins, Riemann, Navier–Stokes); later Riemann and Navier–Stokes were detached too (Option A), and the decree was left with one boundary — the twins.

There remained a second, deeper question as well: why does the problem resist solution so stubbornly — is there a rigorous reason why Collatz can be neither proven nor refuted from inside the system? This answer survives the decree's fall entirely: both internal decisions still cost a perpetual engine. To refute Collatz is literally to build a perpetual engine (a theorem with a genuine construction); to self-ground the law from inside is to cross one's own boundary, which self-destructs. The only key to the solution has remained external: verification.

Let us say it at once and plainly: the decree fell, and this is not a defeat of the programme but its best result. The discipline — the disclosed price and the armed tripwire — worked exactly as designed. Below is the whole story.

The story of the fourth boundary: how the decree was taken and removed

The universal law of rope dominance was accepted by decree — not as a separate axiom, but as a new field of the repository's single axiom step00FirstCause. At the moment of acceptance the structure looked like this:

structure Step00FirstCause : Prop where          -- historical state at the moment collatzBoundary was taken
  origin, firstFrame          : True                       -- markers of the event 0 → 1
  causalBoundary              : …                           -- the twin node (the only live boundary today)
  riemannBoundary             : RiemannManifestationLaw     -- Riemann (§10); detached later (Option A)
  nsBoundary                  : NsSolutionBalanceLaw        -- Navier–Stokes (§15); detached later (Option A)
  collatzBoundary             : ∀ n ≥ 1, RopeCountingLaw n  -- Collatz (§18) ← the fourth, soon removed

Why the boundary was taken

The heuristic sign was favourable — as with Riemann and Navier–Stokes, which is why the boundary was taken at all (unlike Mersenne §16, where the sign is inverted and the field was deferred). The harnesses (RESULTS_collatz.md) showed a contracting engine: a mean geometric per-step drift of 0.864 < 1, a balance halvings/triplings = 2.016 against the threshold log₂3 = 1.585, and the absorbing bottom 1 → 2 → 1.

And — the crucial point for the honesty of the moment — no forged refutation (a machine-built counterexample to the universal form of the law) was known at that hour. For RopeCountingLaw there was neither a green proof (that would have entailed Collatz) nor a witness of falsehood. In this respect the decree seemed to have exactly the same epistemic status as the twin node and the Riemann manifestation law: unknown in either direction, sign favourable — take it as a boundary.

The honest price, disclosed at once

As with Navier–Stokes (§15), the price was disclosed at the same moment as the decree. Only the arrow "boundary ⟹ conjecture" was proven; the converse was unknown. The counting law could turn out strictly stronger than the conjecture: the window budget is one-sided, and only the value law (the "condemned man's bridge") enjoyed an ⟺-form. The conclusion was recorded plainly: the decree may overpay.

The tripwires: detectors planted in advance

Together with the field, two 🟡 tripwires were laid into the theory — intentional explosion detectors (both were deleted together with the field and are therefore named here in italics, as history rather than as living code):

  • collatzQuarantine_inconsistent_if_law_refuted — a refutation of the law ⟹ False exactly here;
  • collatzQuarantine_inconsistent_if_nonconvergence_exhibited — a non-converging trajectory ⟹ False here.

Their meaning was one and the same: consistency of the extended theory ⟺ irrefutability of the conjecture. If the law should ever prove false, this construction was meant not to stay silent but to burn first — and point a finger.

The refutation and the removal

"May overpay" proved prophetic: the decree overpaid into falsehood. The mechanics is dissected in chapter 55; in brief — for the climbing trajectory n = 27, the road to one takes 41 engine moves against 29 rope pulls, and no window from the start gives the rope an advantage (the tail in the cycle 1 → 2 → 1 never recovers the deficit).

The heuristic measured the aggregate balance, while the law demanded an advantage in a window from every position — which climbing orbits violate. This is machine-recorded: not_ropeCountingLaw_27 🟢 and ropeLaw_universal_refuted 🟢 (kernel [propext, Quot.sound], witness n = 27).

The tripwire fired exactly as designed. collatzQuarantine_inconsistent_if_law_refuted showed: to keep the field any longer is to make the single axiom inconsistent. Therefore the field and the tripwires left the axiom together — the detector and the object it guarded were removed in one motion. Right after the removal Step00FirstCause looked like this (an intermediate historical state — Riemann and Navier–Stokes were detached later, Option A):

structure Step00FirstCause : Prop where          -- historical state, right after collatzBoundary was removed
  origin, firstFrame  : True
  causalBoundary      : Step00CausalClosureAxiom
  riemannBoundary     : RiemannManifestationLaw   -- detached later (Option A)
  nsBoundary          : NsSolutionBalanceLaw      -- collatzBoundary removed; nsBoundary detached later (Option A)

Today, after Option A, the single axiom is left with exactly one boundary — the twin node causalBoundary; riemannBoundary and nsBoundary are dead code in WITHDRAWN blocks.

Note (what exactly is proven). What was refuted is the universal form of the law — ∀ n ≥ 1, RopeCountingLaw n. This is not a refutation of the Collatz conjecture: the law was strictly stronger than it, and its fall says only that Collatz cannot be closed by such a decree. The conjecture itself remains open.

The decree's post-mortem is gathered in a single green summary, with three leaves: (1) the universal rope law is false (ropeLaw_universal_refuted) — a decree boundary on it is impossible; (2) the per-n law still entails halting (reaches_one_of_countingLaw, the conditional mechanics is alive); (3) a refutation of the conjecture itself still carries a perpetual engine (nonHalting_carries_perpetual_engine).

Theorem 56.1 (collatz_decree_postmortem). 🟢 Three statements at once: (1) the universal law is false, \(\neg\,\forall n\ge 1,\ \mathrm{RopeCountingLaw}(n)\); (2) under the law an orbit converges, \(\forall n\ge 1,\ \mathrm{RopeCountingLaw}(n)\Rightarrow \exists K,\ \mathrm{iter}\,K\,n = 1\); (3) every non-converging orbit carries a perpetual tail, \(\forall n,\ (\forall K,\ \mathrm{iter}\,K\,n\neq 1)\Rightarrow \exists j,\ \mathrm{NonDescendingOrbit}(\mathrm{iter}\,j\,n)\wedge \forall K,\ \mathrm{iter}\,K\,(\mathrm{iter}\,j\,n)\neq 1\).

The implication that "may have overpaid" survived as true and useless: "the universal law ⟹ convergence everywhere" holds, but its premise is now machine-false, so it can no longer be fed universally. All the content has moved into the per-n form reaches_one_of_countingLaw.

Theorem 56.2 (ropeLaw_would_imply_collatz). 🟢 \(\bigl(\forall n\ge 1,\ \mathrm{RopeCountingLaw}(n)\bigr) \Rightarrow \forall n\ge 1,\ \exists K,\ \mathrm{iter}\,K\,n = 1\). The implication holds, but by Theorem 56.1 its premise is machine-false.

The vacuum and its missing gap

Below are the cross-sections that the decree's fall did not touch: they were green and remain green.

At the bottom of the rope world lies the cycle 1 → 2 → 1 — an absorbing state with no way out. Let us call it the vacuum: the ground state of the dynamics, its lowest note. vacuum_has_no_gap 🟢 (even without Classical.choice): there is no leaving the vacuum, ever — iter d 1 ∈ {1, 2} for all d.

The conjecture asserts that the vacuum is unique and everything falls into it. To refute it is to exhibit a gap in the vacuum: a second bottom, another cycle, an orbit living forever without touching the ground state. What does such a gap cost?

The orbit minimum and the tail engine

The answer is obtained by one look at the minimum. exists_min_position 🟢 (the standard triple): every orbit has a position of global minimum of its value — the natural numbers are well-ordered.

The only non-constructivity is the choice itself of "is there a smaller one" (Classical.em does the work here), and we disclose this honestly: one cannot learn the position of the minimum without walking the orbit; but it exists unconditionally.

The tail of the orbit from the minimum never dips below its start — and that is exactly the perpetual engine of [chapter 55]: a non-descending orbit which (by nonDescendingOrbit_is_perpetual_engine) loses not a single window to the rope and never halts.

nonHalting_carries_perpetual_engine 🟢 (the central one). If an orbit never reaches 1, then its tail from the minimum position is a non-descending, never-halting orbit. The construction is genuine, not by contradiction: the tail is exhibited explicitly. Let us read it aloud: every Collatz counterexample carries a perpetual engine within it. A gap in the vacuum costs exactly a perpetual engine.

Conclusion. The Collatz conjecture does not resemble a statement about the impossibility of a perpetual engine — it is one, word for word, for the map T: an orbit reaches 1 ⟺ it has no perpetual tail.

The narrative: perpetual information extraction, which does not exist

Here is the spine of the chapter, and it explains ninety years of unsolvedness by a single physical "cannot". Note: all of this was proven without the decree and survives the decree's fall entirely.

Collatz is perpetual information extraction. A counterexample — an orbit living forever above its bottom and never falling into the vacuum — is a machine that draws endlessly: it loses not a single window to the rope, moves forever, never surrendering itself to the cycle (nonHalting_carries_perpetual_engine).

And "learning the cause from inside" — deriving the universal rope law by an internal proof — would mean extracting from the system an infinite fact about the whole number line without ever leaving it.

Infinite information extraction is impossible. Define an internal self-grounding of the law's foundation: a construction carrying the law itself AND a witness that it was derived from inside — crossing its own boundary.

Definition 56.3 (InternalisedCollatzGround). A structure with two fields: ground — the law itself, \(\forall n\ge 1,\ \mathrm{RopeCountingLaw}(n)\); beyondOwnHorizon — a witness of crossing the boundary, \(\neg\,\forall n\ge 1,\ \mathrm{RopeCountingLaw}(n)\). The abbreviation InternalKnowledgeOfCollatzCause identifies "internal knowledge of the cause" with this self-grounding.

As with the twins, the form is honestly tautological ("to prove P while crossing the boundary of P" collapses into \(P\wedge\neg P\)); substantively it is an identification: the attempt at internal grounding is already a burnt engine construction. Formally the self-grounding self-destructs by its very form — no_internalisedCollatzGround — and this yields the unknowability of the first cause.

Theorem 56.4 (collatzCause_unknowable). 🟢 (with no axioms at all) \(\neg\,\mathrm{InternalKnowledgeOfCollatzCause}\): internal knowledge of the Collatz first cause is impossible. The proof is direct: the two fields of Definition 56.3 contradict each other, \(\texttt{beyondOwnHorizon}\ \texttt{ground} : \mathrm{False}\).

And after the decree's fall — stronger still.

Theorem 56.5 (internalGround_impossible_a_fortiori). 🟢 \(\mathrm{InternalisedCollatzGround}\Rightarrow \mathrm{False}\), but now by content: the field ground is refuted directly (ropeLaw_universal_refuted), not through the self-destruction of the form. Unknowability now holds doubly — both as an impossible form (Theorem 56.4) and as false content.

This is the same law that forbids a perpetual engine on ℕ (no_infinite_descent) and that in the main programme takes the shape "knowledge from inside costs infinite energy" (the finite-knowledge barrier, Engine/FiniteKnowledgeBarrier, chapter 33; read via Landauer — knowledge is physical, and extracting the cause from inside would violate the same balance that forbids the engine).

That is why the engine is "perpetual" — as an object irremovable from inside. Since the conjecture is identified with "there is no perpetual engine", and internal knowledge of the cause is impossible, exactly the perpetual engine remains the thing that can be neither exhibited nor excluded by internal derivation: if a counterexample exists — it is genuinely perpetual; and that none exists cannot be proven from inside.

Hence — verification only.

Theorem 56.6 (unknowable_or_collatz_fails). 🟢 The substantive dichotomy (no ex falso in the statement): \(\neg\,\mathrm{InternalKnowledgeOfCollatzCause}\ \vee\ \exists n\ge 1,\ \forall K,\ \mathrm{iter}\,K\,n\neq 1\) — either the cause is unknowable, or Collatz is false. The left disjunct is proven (Theorem 56.4). The right one is discovered not by derivation but by verification — by exhibiting a concrete non-converging orbit. All of this is gathered in one green theorem that machine-encodes the narrative:

Theorem 56.7 (collatz_verification_not_derivation). 🟢 A conjunction of three links in a single statement: 1. infinite internal extraction of the rope law is impossible, \(\neg\,\mathrm{InternalKnowledgeOfCollatzCause}\) (Theorem 56.4); 2. therefore the only internal trace of a counterexample is a perpetual engine (nonHalting_carries_perpetual_engine); 3. hence the solution is either unknowable from inside, or accessible only by verification of an orbit (Theorem 56.6).

And verification finds-or-does-not-find exactly as far as we look. The harnesses (RESULTS_collatz.md) are precisely the "how far": n ∈ [2, 200000], a divergence guard at 100000 steps. In this window there are no counterexamples; but no finite window can prove the universal law — it can only fail to find a refutation within its own horizon. (And — as we now know — in the limit the universal form of the rope law is false after all: not because an orbit diverges, but because the law demanded more than is true. Verifying convergence and verifying the law are not the same thing.)

This is the same horizon of the finite observer as in chapter 33: a finite system sees only the pure classes of its own scale, and knowledge of the infinite requires cofinal (not finite) information. The answer exists; it can be found only by verification, and only up to the edge that the gaze can reach.

Summary: the solution is locked behind the engine

Theorem 56.8 (collatz_no_internal_decision_without_engine). 🟢 (green, axiom-free) A conjunction of three leaves: (1) to refute = to build a perpetual engine — \(\forall n,\ (\forall K,\ \mathrm{iter}\,K\,n\neq 1) \Rightarrow \exists j,\ \mathrm{NonDescendingOrbit}(\mathrm{iter}\,j\,n)\wedge\forall K,\ \mathrm{iter}\,K\,(\mathrm{iter}\,j\,n)\neq 1\) (a genuine construction); (2) to prove from inside = to self-ground — \(\mathrm{InternalisedCollatzGround}\Rightarrow\mathrm{False}\), it self-destructs; (3) the per-n rope law would entail per-n halting — \(\forall n\ge 1,\ \mathrm{RopeCountingLaw}(n) \Rightarrow\exists K,\ \mathrm{iter}\,K\,n = 1\). But the universal feed is gone: the universal law is refuted (Theorem 56.1), the decree door is closed by a forged witness.

Theorem 56.9 (collatz_open_status). 🟢 The final epistemic status of Collatz after the decree's fall (the mirror of pnp_locked_behind_engine_status, but now without the decree conjunct) — a conjunction of four: \(\neg\,\forall n\ge 1,\ \mathrm{RopeCountingLaw}(n)\) (the universal law is refuted, theorem) \(\wedge\) \(\neg\,\mathrm{InternalKnowledgeOfCollatzCause}\) (internal knowledge is impossible, theorem) \(\wedge\) \(\bigl(\forall n\ge 1,\ \mathrm{RopeCountingLaw}(n)\Rightarrow\exists K,\ \mathrm{iter}\,K\,n = 1\bigr)\) (the per-n law entails halting, conditional) \(\wedge\) \(\bigl(\forall n,\ (\forall K,\ \mathrm{iter}\,K\,n\neq 1) \Rightarrow\exists j,\ \mathrm{NonDescendingOrbit}(\mathrm{iter}\,j\,n)\wedge\forall K,\ \mathrm{iter}\,K\,(\mathrm{iter}\,j\,n)\neq 1\bigr)\) (a refutation of the conjecture would build a perpetual engine, theorem).

Note (what we do not claim). This is not Gödelian independence: we have not proven that Collatz is unprovable in Peano arithmetic. What is proven is something else, inside our system and rigorously: both paths of internal decision — refutation and self-grounding — run into one and the same forbidden object, and the decree third path is machine-closed (ropeLaw_universal_refuted). The Collatz conjecture remains open 🔴.

P/NP beyond the same horizon

This wall has a second slope, and it falls not into Collatz but into P/NP. The verifying machine there is not a metaphor: it is literally a pure computer on finite fuel with strict rules. In the repository it is assembled as RankedForwardGraph — the rules are its Step, the fuel is lexRank, the law of fuel consumption is RankFastTraversal — and launched from an arbitrary, "random" initial position (the 5-adic chain at A ≤ 4). This is exactly the "system with random initial conditions but strict rules" that was spoken of (chapter 39, the rank-payment machine).

To decide P/NP from inside means to pay with one finite key, injectively, for every certificate of an unbounded supply — FullRankCertificatePayment for UnboundedCertificateSupply (the supply is the infinite admissible family of certificates with which a deviation pays; see the glossary). It is the same perpetual-engine contradiction as with Collatz, translated word for word: "a non-falling tail = an engine" becomes "full payment of an unbounded supply = an engine".

Both perish against the same wall of well-foundedness — the pigeonhole: there is no injection of an infinite index into a finite key, hence no ℕ-descent.

This is recorded by the green theorems of Engine/PNPFirstCause: no_internalisedPNPGround, pnpCause_unknowable ("cannot be known from inside" — not a slogan but a theorem) and internalPNPDecision_carries_perpetual_engine, where the internal decision and the perpetual engine on ℕ are brought to a single wall through no_perpetual_engine_on_nat. The summaries are pnp_no_internal_decision_without_engine and pnp_locked_behind_engine_status.

The event horizon reads especially vividly here. The machine's finite fuel (its rank budget) is precisely its event horizon; a decision that needs to reach past it — to pay for an unbounded supply — is unattainable from inside. The answer is accessible only by verification and is found exactly as far as the finite-fuel gaze reaches. That is why P/NP becomes unprovable-from-inside in the very same sense in which the cause of Collatz is unknowable — "as far as the gaze reaches" here is not a figure of speech but the edge of the budget.

Let us say plainly what is not here. This is not a proof of classical P≠NP: the frame layer is plastic (allPFrame/constantsFrame), and nothing about the real P/NP follows from it. And this is not Gödel — neither an incompleteness theorem nor a fixed point, but a pigeonhole self-destruction.

And here is the difference that the decree's fall rewrote: previously P/NP had no decree boundary while Collatz had one. Now neither has one — both trilemmas, the mandatory tests of a boundary candidate, are closed machine-wise. For P/NP it is split (pnpLawUniversal_refuted, pnpLawExistential_green proved that no honest third field of step00FirstCause exists); for Collatz it has just been closed by the forged refutation ropeLaw_universal_refuted (the decree taken and removed).

Therefore this entire result is green, adds not a single axiom, and the repository's taint does not grow (it is 16, every declaration asserting the twins; the removal of the Collatz boundary did not change it). Everything holds at A ≤ 4.

Section takeaway. Thus the same prohibition closes, read now through a finite-fuel verifier: the machine that verifies any conjecture is itself the object it cannot surpass.

Philosophical digression: the decree that fell honestly

This chapter has a direct physical twin — the stability of the ground state. The vacuum of quantum theory is not "nothing" but the lowest energy state; all the stability of the world rests on the fact that one cannot sink below the vacuum. To find a "gap in the vacuum" — a state beneath the ground state — would bring everything down: from such a gap one could draw energy endlessly, and that is precisely a perpetual engine. Physics forbids it not by ordinance but by the very structure of the spectrum: there is a bottom, and it is unique.

But deeper than the physical analogy here lies a methodological one. We took a boundary by decree, disclosed its price, planted a tripwire — and the tripwire fired. A decree that fell honestly is worth more than a decree that stands unexamined.

The programme neither hid the overpayment nor forgot it: it recorded the fall as a theorem, not as oblivion. It was exactly the discipline one might have dismissed as pedantry — the disclosed asymmetry of the arrows, the armed detector collatzQuarantine_inconsistent_if_law_refuted, the demand for machine verification — that caught the falsehood. Honesty here is not an ornament but a working organ.

And here is where this places Collatz. It is no longer "a problem with an honest decree" but a problem with a closed trilemma — alongside Yang–Mills (chapter 40), where the universal candidate likewise fell to a forged refutation. In neither case is a decree possible: not because we failed to find a way to feed it, but because feeding it would mean feeding a falsehood, and the machine sees that.

What remains open is the heart itself: convergence. 1 → 2 → 1 is the vacuum of the rope world; a second bottom, a foreign cycle, a diverging orbit would cost exactly a perpetual engine. Whether there is a gap or not — this cannot be certified from inside: the attempt to walk the entire infinite line is the very forbidden engine. The one honest key has remained external — to verify exactly as far as the gaze can reach.

Place in the greater arc

Collatz completes its mask differently from how it began it. It entered as a mirror of the first cause (ch. 33), of Riemann (38) and of Mersenne (43): a green impossibility of the engine from inside plus an external boundary accepted by decree at an honestly disclosed price. And it exits with the same green impossibility, but without a boundary: the decree taken and removed, the trilemma closed by a forged refutation, as with Yang–Mills.

What sets Collatz apart, and what the decree's fall only sharpened, is that the engine here is literal (the tail of a non-converging orbit is perpetual motion on ℕ, not a surrogate), and that is why the narrative "verification, not derivation" reads most clearly on it: no_infinite_descent — one prohibition — turns out to be also a prohibition on infinite information extraction. Once more it brings us right up to the boundary of the knowable — and this time it shows that even the boundary itself, set too generously, honestly collapses under its own price.

← 55. Collatz · Table of contents · 57. The well-foundedness canon →