MT03 — Busy Beaver and Gödel Incompleteness

Field Value
Domain Mathematical Methods
System n-state, 2-symbol Turing machines
Group Sₙ × ℤ₂ (state permutations × symbol flip)
H^k tier
ISA Origami (β → ∞)
Status Validated
Opcodes ORBIT · LABEL · FLOP · BIND
Papers Paper 543

Physical system

The Busy Beaver function Σ(n) is the maximum number of 1s a halting n-state, 2-symbol Turing machine can write on an initially blank tape. Equivalently, S(n) is the maximum number of steps any halting n-state TM takes. These are:

n Σ(n) S(n)
1 1 1
2 4 6
3 6 21
4 13 107
5 4098 47,176,870
6 ≥ 10,910 ≥ 7.4×10³⁶,534

Σ(n) grows faster than any computable function. This is not a theorem about large numbers — it is a theorem about the boundary of computability. Σ(n) is well-defined (there are finitely many n-state TMs), but it cannot be computed by any Turing machine. For large n, the value of Σ(n) is independent of ZFC: to determine whether a specific TM halts may require axioms beyond any fixed formal system. The Busy Beaver function maps the boundary between the decidable and the undecidable onto a single, concretely defined integer sequence.


Target category

TM(n) — the category of n-state, 2-symbol Turing machine computations. Objects: tape configurations (bi-infinite sequences over {0,1} with a marked head position and state). Morphisms: single computation steps (state × symbol → new state × new symbol × head direction). The halting state H is a terminal object; non-halting TMs have no morphism to H.

Interpretation functor

F: C → TM(n) defined by:

Opcode F(opcode)
ORBIT Computation step: (state, tape) → (new state, new tape, head ±1); the fundamental TM transition
LABEL State assignment: the current state q ∈ {1,…,n,H}; the eigenvalue of the configuration at each step
FLOP Symbol write: 0 → 1 or 1 → 0 on the tape cell under the head; the bit-flip opcode as tape write
BIND Halting certificate: the H² class that distinguishes halting from non-halting TMs; the obstruction that cannot be detected by any finite ORBIT sequence alone

ISA programme

INIT:    LABEL[tape = 0^inf, state = q1, head = 0]   -- blank tape, start state
STEP:    ORBIT[(q, s) -> (q', s', d) | transition table]  -- one TM step
WRITE:   FLOP[s -> s' | per transition]              -- write new symbol
MOVE:    LABEL[head += d | d in {-1, +1}]            -- head movement
HALT?:   LABEL[state = H?]                           -- halting test
BIND:    BIND[halts? | undecidable for general n]    -- H2 obstruction
COUNT:   LABEL[Sigma(n) = max over halting TMs of ones-written]  -- Busy Beaver value

Computable output

  • Σ(n) for small n: Σ(1)=1, Σ(2)=4, Σ(3)=6, Σ(4)=13, Σ(5)=4098, Σ(6)≥10,910. These are validated by exhaustive enumeration of all n-state TMs and simulation of halting ones. They are the ORBIT outputs of the computation — integers counting the maximum 1s written.
  • Non-computability certificate: there is no Turing machine that computes Σ(n) for all n. The proof is the BIND obstruction: if Σ were computable, the halting problem would be decidable (simulate all TMs and check against Σ(n)). The halting problem’s undecidability is the H² cohomological obstruction in TM(n) — the morphism to H does not exist in the category, and no amount of ORBIT steps can detect its absence.
  • Independence from ZFC: for sufficiently large n, the statement “Σ(n) = k” for specific k is independent of ZFC. Scott Aaronson and Adam Yedidia (2016) showed a 7,918-state TM whose halting is independent of ZFC. Smaller examples likely exist. This is the ultimate BIND obstruction: the H² class is not just uncomputable, it is formally unprovable within any fixed axiom system.

Connection to the ISA framework

The Busy Beaver is an H² entry because halting is a BIND. The distinction between H⁰ (tropical, decidable), H¹ (Clifford, efficiently simulable), and H² (magic, not efficiently simulable) in quantum computing has a direct analogue in computability theory:

ISA tier Quantum analogue Computability analogue
H⁰ Origami Classical simulation (stabiliser) Decidable (recursive)
H¹ Forge Polynomial-time quantum RE (recursively enumerable)
H² Meld Universal quantum / magic Non-RE (undecidable)

The Busy Beaver sits at H²: the halting question is not in RE (you cannot enumerate halting TMs and confirm non-halting ones — there is no algorithm for the complement). The BIND opcode is the categorical morphism that would resolve the halting question if it existed; its non-existence in TM(n) is the Gödel incompleteness theorem in ISA language.

The β-plane reading: the Busy Beaver lives at β → 0 (Meld ISA) because:

  • At β → ∞ (Origami, tropical), computation is decidable — the tropical argmax selects the unique ground state in finite time
  • At β = β* (Forge, snap), computation may run for a very long time but halts
  • At β → 0 (Meld, Ambient), the computation has access to all paths simultaneously — it is the quantum superposition of all computation paths, and measuring the result corresponds to the halting certificate

The Busy Beaver function is the β → 0 limit of the computation landscape: the quantity you could compute if you could explore all paths simultaneously (at the Ambient) and then measure at β → ∞ (Origami). The obstruction is that this two-step process (Meld then Origami) cannot be implemented for general TMs because the Meld step (quantum superposition over all computation paths) is itself not computable — it requires an infinite-dimensional Hilbert space indexed by all computation histories.

Connection to Gödel: Gödel incompleteness is the statement that the BIND obstruction in TM(n) is not removable by adding more ORBIT steps (more axioms). Every consistent extension of ZFC has a Gödel sentence that is true but unprovable — a new H² class appears at each extension. This is the H^k tower failure at H²: there is no H³ that resolves all H² obstructions simultaneously, because a hypothetical “completeness oracle” would itself be subject to a higher-order incompleteness theorem.

The Fano connection: the appearance of 21 = 3×7 in the Apéry sequence (MT02) and the G₂ choreography conjecture (D02) may connect to the Busy Beaver: it is known that the TM corresponding to a specific mathematical conjecture (e.g., the Riemann Hypothesis) can be encoded in a TM with on the order of tens of states. The 5-state BB champion encodes aspects of Goldbach-type arguments. Whether the Fano plane (7 points, 7 lines) has a natural Busy Beaver encoding remains an open question.

Validation

  • Σ(1)–Σ(4): computed by Rado (1962) who introduced the function; independently verified many times.
  • Σ(5) = 4098: Lin & Rado (1965); confirmed by exhaustive enumeration.
  • Σ(6) ≥ 10,910: Marxen & Buntrock (1990); current record TM found.
  • Non-computability: Rado (1962), immediate from the halting problem.
  • ZFC independence: Yedidia & Aaronson (2016), arXiv:1605.04901.

Part of the ISA Zoo. Categorical foundations: Paper 591.