The Necessity of Twelve

The preceding essay asks what happens when unity acts upon itself, and reports that the resulting dynamics close at 12. This companion does not reopen the ontological question. It takes the closed form as given and asks a narrower one: why twelve, and what the arithmetic actually forces.

The original is entitled to its frame. This note is entitled to ignore it. What remains is a map on the integers, a short list of orbits, and a handful of statements that are either theorems or they are not.

Begin from the function

For n > 1 let Ω(n) be the number of prime factors of n counted with multiplicity, and let sopfr(n) be their sum. Define

and complete the definition at the boundary by A(−1) = 0, A(0) = 1, A(1) = 2. No other negative arguments are used.

This is the whole object. The names Kucwenga, FN, chamber, primitive unity are commentary. They may be returned to after the map has been made to confess.

Two elementary identities sit inside the formula and will be used constantly. If p is prime then Ω(p) = 1, sopfr(p) = p, so

If n = 2^k with k ≥ 1 then Ω = k, sopfr = 2k, so

The first says that primes step forward by one. The second says that powers of two step forward by their exponent. Both are strictly increasing. Everything interesting in the dynamics is about what happens after those steps, when the image is no longer prime and no longer a power of two.

What the extra term counts

sopfr is additive over the prime factors. The remainder 2^Ω − Ω is not a function of the primes themselves but of how many of them there are. It depends on n only through the length of its factorisation.

A counting story that would make the term inevitable is not supplied by the source essay, and is not invented here. What can be said without invention is the size:

• Ω = 1: 2^Ω − Ω = 1, A(n) = sopfr + 1
• Ω = 2: 2^Ω − Ω = 2, A(n) = sopfr + 2
• Ω = 3: 2^Ω − Ω = 5, A(n) = sopfr + 5
• Ω = 4: 2^Ω − Ω = 12, A(n) = sopfr + 12

The last line is already a rumour of the fixed point. It is only a rumour. 2^Ω − Ω equals 12 at Ω = 4, whereas the fixed point we will find has Ω = 3. The function does not spell twelve in its coefficients and then obediently sit there. Twelve has to be solved for.

The principal orbit

Start at the boundary and iterate.

Then 2, 3, 4 are forced by the two identities above: A(2) = 3, A(3) = 4, A(4) = A(2^2) = 4 + 2 = 6. Now 6 = 2 × 3 is the first argument that is neither prime nor a power of two.

Then 7 is prime, so A(7) = 8. Then 8 = 2^3, so A(8) = 8 + 3 = 11. Then 11 is prime, so A(11) = 12.

Five is not on this list. Neither is nine, nor ten. That is all the mystery the omissions have: this particular path did not go there. Their images, however, are on the list, or one step from it.

So 5 feeds the spine at 6; 9 feeds it at 8; 10 feeds 9, which feeds 8. The omitted points are preimages, not exiles. Any language that treats them as structurally refused has already left the orbit.

The interval is fourteen consecutive integers

The set {−1, 0, 1, …, 12} has fourteen elements because 12 − (−1) + 1 = 14. This is not a property of A. It is a property of counting. If the attractor had been 7, the same construction would have produced a “chamber” of nine integers and a different list of absences.

What is a property of A is that the orbit of −1 remains inside this interval and ends at its right-hand endpoint. That is a computation of eleven steps, not a cosmology. The interesting question is whether 12 is special for A in a way that 7 is not — whether it is a fixed point, whether it is the only one, whether other integers are compelled toward it. Those are questions the interval cannot answer. The function can.

Twelve is a fixed point

12 = 2^2 × 3, so Ω(12) = 3 and sopfr(12) = 7.

This is the only calculation in the subject that deserves to be called striking, and even it is only striking after one has asked the question. A number equal to the sum of its prime factors plus five (since 2^3 − 3 = 5) happens to be twelve. Equivalently:

For n = 12 the two sides are both 5. The rest of this section is the claim that no other n ≥ 2 balances the equation.

The unique fixed point

Write k = Ω(n). The fixed-point equation is n − sopfr(n) = 2^k − k. We run through small k. The pattern that emerges is that 2^k itself always undershoots the target gap, the next integers with the same k overshoot it, and only one composite in the middle lands exactly.

k = 1

Then n is prime, n − sopfr(n) = 0, and 2^k − k = 1. Never equal.

k = 2

Target gap 2. If n = p^2 then p^2 − 2p = 2, so p^2 − 2p − 2 = 0, which has no integer root. If n = p q with p < q then (p − 1)(q − 1) = 3. The only positive factorisation of 3 is 1 × 3, giving p = 2, q = 4, and 4 is not prime. No solutions.

k = 3

Target gap 5.

If n = p^3 then p^3 − 3p = 5. For p = 2 the left side is 2; for p = 3 it is 18.

If n = p^2 q with q ≠ p, the equation is p^2 q − 2p − q = 5. For p = 2 this becomes 4q − 4 − q = 5, hence 3q = 9, hence q = 3. Then n = 4 × 3 = 12, which works. For p = 3: 9q − 6 − q = 5, 8q = 11, not an integer. For p ≥ 5 the left side is already too large for small q and the q-coefficient is p^2 − 1 ≥ 24, which cannot yield q prime.

If n = p q r with distinct primes, the smallest candidate is 30, with gap 20, already larger than 5. Larger triples are worse.

So the only solution with k = 3 is 12.

k = 4

Target gap 12. The smallest n with Ω(n) = 4 is 16 = 2^4, whose gap is 16 − 8 = 8, too small. The next, 24 = 2^3 × 3, has gap 15, too large. Every later n with Ω = 4 is at least 24 and has gap at least 15. No solutions. (36, 40, 54, 56, 80, 81 all have still larger gaps.)

k ≥ 5

For each k the minimal n is 2^k, with gap 2^k − 2k. The target is 2^k − k. The power of two therefore undershoots by exactly k. The next integer of the same Ω is 2^{k−1} × 3, whose gap is already larger than the target: for k = 5, 32 undershoots (gap 22, target 27) and 48 overshoots (gap 37); there is no integer strictly between 32 and 48 with five prime factors counted with multiplicity, because 2^3 × 5 = 40 has only four. For larger k the jump from 2^k to 2^{k−1} × 3 grows as 2^{k−2} and the overshoot grows with it.

The cases k = 1, 2, 3, 4 are exhaustive and elementary. The case k ≥ 5 is the same comparison of 2^k with 2^{k−1} × 3, plus the observation that nothing of the right Ω lives in the gap between them. That is a proof, not a numerical survey — though a survey is easy, and agrees.

The non-decreasing points

A(n) ≥ n if and only if n − sopfr(n) ≤ 2^k − k: the gap of n does not exceed the gap that a fixed point would need.

• Every prime: gap 0, target 1. Strictly increasing, by 1.
• Every 2^k: gap 2^k − 2k, target 2^k − k. Strictly increasing, by k.
• n = 6: gap 1, target 2. A(6) = 7.
• n = 12: gap 5, target 5. Fixed.

Are there others? The same case analysis says no. For k = 2 the only n with gap at most 2 are 4 (already a power of two) and 6. For k = 3, gap at most 5: 8 (already a power of two) and 12; 18 has gap 10, 20 has gap 11, 27 has gap 18, 30 has gap 20. For k = 4, gap at most 12: only 16, a power of two (24 already has gap 15). For k = 5, gap at most 27: only 32. The pattern persists: the inequality holds for 2^k and fails for every larger n of the same Ω.

This is the original essay's fourth dynamical claim, now with a proof. It is also the reason the principal spine looks “mostly increasing.” The orbit of −1 happens to land on primes and powers of two until it reaches 6, and then on primes and powers of two again until it reaches 12. The omissions 5, 9, 10 are exactly the small composites in the interval that are not 6 or 12: they decrease, so a strictly climbing path that started below them will not visit them.

That is the architecture. It is real. It is also entirely accounted for by the two theorems above. No further property of five, nine or ten is required.

The basin

The original essay claims more: every n ≥ 2 reaches 12, and there are no cycles other than the fixed point.

Once uniqueness of the fixed point is known, “no cycles” and “global attraction” are nearly the same statement, because a cycle of length greater than 1 would be a second invariant set, and a wandering orbit that never hits 12 would have to grow without bound or cycle. Growth without bound is not plausible for long. For composite n with many factors, sopfr(n) + 2^{Ω(n)} − Ω(n) is much smaller than n: A(n)/n → 0 along factorials, along primorials, along n = p^k for odd p as k grows. The map is strongly contracting on highly composite arguments. The only expanding arguments are the non-decreasing ones just classified, and those expand slowly — primes by 1, powers of two by k, which is log n.

A prime p is sent to p + 1, which is even, hence composite for p > 2, hence typically contracted at the next step. A power of two is sent to 2^k + k, which for k ≥ 4 is even but not a power of two, and then tends to fall. The original spine is, in this light, the slowest route to 12, not a privileged one: it is the route that keeps finding expanding points.

This is a sketch, not the last page of a proof of global attraction. A complete argument would need a Lyapunov function, or a finite check of all n up to some bound past which A(n) < n and A(n) is already known to lie in the checked range. The second method is the obvious one. In the laboratory attached to this issue, every n from 2 through 12 000 falls to 12, and no other cycle appears. That is not yet “every integer.” It is enough to make the remaining claim a reasonable theorem rather than a hope.

What the generative criterion would require

The source essay's best sentence is not about twelve. It is: if the interpretation is true, let it show more.

Taken seriously, this sentence forbids the usual harvest of independent twelves. The dodecahedron, the clock, the chromatic scale, the apostles, the months — each is a twelve that was not produced by A. Listing them is the opposite of generation. Generation would look like one of the following.

• An independently motivated operator, written down for reasons that do not mention 12, turns out to be A, or to have the same attractor.
• A question in another field — a classification, a period, a obstruction — is answered by running A, or by using the expanding-point characterisation.
• A deformation of A (replace 2^Ω by 3^Ω, replace sopfr by sopf, restrict to odd n) produces a system whose difference from FN is itself structured, and that difference is predicted before it is computed.

Until one of these happens, Layer III has not begun. The arithmetic of this note is Layer I. Layer II, if it consists of saying that unity unfolds and then closes, fits the theorems but is not needed by them. Layers IV and V remain what the original said they were: optional, and not inheritors of the proof.

The laboratory on this site is a small instrument for the first kind of test. Pick a number that means something to you for reasons A does not know. Iterate. If you find a second attractor, the uniqueness theorem is false and this issue is withdrawn. If you find only 12, you have confirmed a computation, not a metaphysics.

Correspondences that do not count

It is difficult to write about 12 without the museum of twelves opening on its own. The museum should stay closed.

A correspondence, as opposed to a coincidence of cardinality, is a structure-preserving map. The vertices of an icosahedron are 12, but they come as a transitive action of A5, with golden-ratio coordinates, dual to the dodecahedron. Nothing in sopfr or Ω knows this. One may construct a map from the expanding points of A to some geometric 12; until the construction exists, the two twelves are as related as 12 and the hours of the clock, which is to say, not.

The same applies to the fourteen points of the interval. Fourteen is not 12. If a later paper should need fourteen, it should not be extracted from [−1, 12] ∩ ℤ, because that extraction is available to any attractor at 12 and therefore explains nothing about A.

The original essay is innocent of most of these moves. Its reviewers and successors will not be, unless they adopt its criterion in the strict sense: show more, or stop.

What the primitive permits

Return, briefly, to the original's language, now that the arithmetic has been made to speak.

If one insists on reading A as an action of unity on itself, the theorems restrict what that reading may say. Unity, so read, does not skip 5 as a matter of principle; it skips 5 because 5 is a decreasing point sitting beside a climbing orbit. Unity does not choose 12 as a sacred number; it lands on the unique solution of n − sopfr(n) = 2^{Ω(n)} − Ω(n). Unity does not close because closure is geometrically beautiful; it closes because after 12 the next value is 12, and because every expanding type eventually produces a composite that contracts.

Those restrictions are the gift of Layer I to every later layer. They make the metaphysics harder, which is the only thing that could make it better.

The original asked: what happens when the One acts upon itself? The arithmetic answers, without One:

Begin with the primitive. Apply only what the primitive permits. The consequences, in this case, are two short theorems and a basin that appears to be the whole of ℕ. That is the article. The rest is optional, and should remain so until it shows more.