The A166944 room is quantised by the event value: the slack is not a continuum
OpenA166944Research.a166944_room_quantumThe moving-horizon replay of the OEIS A166944 recurrence jumps from one event to the next, and each event leaves behind a quantity we call the room. The room was expected to be an arbitrary natural number above a minimum. This theorem says it is not: the room is quantised by the event's own value, and it may only land on the first lattice point or beyond the second.
Setup
Let be the Rowland-type recursion of OEIS A166944, its first differences and its offset from the identity. A moving-horizon step relates a state with to the next index at which the recursion's gcd is non-trivial:
where the step carries the guard .
Write for the event value and define the room left after the step as
The statement
For every moving-horizon step:
Equivalently: the parity of the hit index fixes a minimum room (, resp. after a shift of ), and the room can never land strictly between that minimum and the next quantum (, resp. ). The admissible values are exactly the two branches.
Why the gap is real
The mechanism is the quotient normal form. For an even hit one has
with odd and . Since is therefore even and at least , the room is always an even multiple of : the value is arithmetically unreachable, not merely unobserved. The even branch's quantum is , not — an earlier version of this statement claimed and omega refuted it, exhibiting the explicit profile , , with and . The odd branch is the same phenomenon shifted by : with even and , so the values are or at least .
Role
This is the sharp form of the slack estimate two_mul_increment_le_room_next, and it is what a summation over a replay needs: knowing the room cannot sit just above its minimum is exactly the information that a bound on the accumulated excess would have to consume. It is a local statement — see the note below on what it does not give.
What this does not do
The dichotomy is exact and it is still compatible with an arbitrarily large accumulated excess. On the degenerate even branch () the quotient is pinned to , but itself is free; the odd branch's case likewise leaves free. The local system therefore admits far above ; this is not a defect of the proof but a genuine limitation of the local structure, and it is recorded so the theorem is not over-claimed as progress on the global barrier.
Formalization Note Verbatim from A166944/RoomQuantum.lean at commit a3bc097, kernel-checked with no sorryAx (the proof depends only on propext and Quot.sound). The statement is the bundled two-parity form; moving_horizon_room_quantum_even and moving_horizon_room_quantum_odd are its two components. The companion divisibility results moving_horizon_event_dvd_room_even () and moving_horizon_event_dvd_room_odd () expose the same quotient normal form as a divisibility statement. Verified on all 230 steps of the fifteen canonical record replays below : zero violations.
import Definitions.Def_a166944_moving_horizon open A166944Research
namespace A166944Research
theorem a166944_room_quantum {s C r C' : Nat}
(hstep : IsMovingHorizonStep s C r C') :
(r % 2 = 0 → C' - r = 2 * d r ∨ 4 * d r ≤ C' - r) ∧
(r % 2 = 1 → C' - r + 4 = 3 * d r ∨ 5 * d r ≤ C' - r + 4) := by
sorry
end A166944Research