A 1996 knot-theory paper and the H² opcode

Plain-language explainer for doi:10.5281/zenodo.21278538 (#572)

The one-sentence version

Greg Kuperberg’s 1996 diagrammatic calculus for the exceptional Lie group G₂ — built to compute knot invariants — turns out to be the complete axiomatisation of the BIND opcode, the H² instruction that appears wherever a physical system is fundamentally non-associative.

What is the BIND opcode?

The ISA classifies computations by how complex their interactions are:

  • H⁰ (ORBIT): counting configurations — classical, tropical, associative and commutative
  • H¹ (TWIST): rotations and interference — quantum but still associative
  • H² (BIND): non-associative interactions — the rarest and hardest tier

BIND is the three-body vertex: it couples three objects at once in a way that depends on the order of coupling. Swap any two inputs and you can get a different answer. This non-associativity is the hallmark of octonions — the largest normed division algebra — and of the exceptional group G₂ = Aut(octonions).

A BIND operation appears, for example, when three iron atoms in FeMoco form a “Fano triple”: a triangular arrangement where the three magnetic orbitals are connected not by pairwise exchange (H¹) but by a genuine three-body interaction encoded in the Fano plane geometry.

What Kuperberg built in 1996

To compute invariants of knots whose strands are “coloured” by the 7-dimensional representation of G₂, Kuperberg needed a diagrammatic calculus — a set of local drawing rules that let you simplify any tangle until only a number remains.

He found five rules (call them R1–R5) for diagrams built from a single trivalent vertex V:

Rule Diagram Evaluation
R1 (circle) a closed loop = 7
R2 (bigon) two vertices nose-to-nose = 5 × (identity strand)
R3 (square) four vertices in a square decomposes into two simpler pieces
R4 (pentagon) five vertices in a ring = 0
R5 (isotopy) topological moves = same diagram

The key theorem: these five rules are complete. Any closed G₂ diagram, however complex, reduces to a number by applying R1–R5 finitely many times.

Why the trivalent vertex V is exactly BIND

The trivalent vertex in Kuperberg’s spider evaluates on basis vectors e_i, e_j, e_k of ℝ⁷ as:

V(eᵢ, eⱼ, eₖ) = φ(eᵢ, eⱼ, eₖ)

where φ is the Fano incidence function: it equals +1 if {i, j, k} form a positively-oriented Fano line, −1 for the opposite orientation, and 0 otherwise. This is identical to the ISA definition of BIND.

The 7-dimensional representation is the 7 imaginary octonion directions. G₂ is the automorphism group of the octonions. The trivalent vertex is the octonion trilinear form. BIND = Fano indicator = octonion trilinear form. All three descriptions are the same object.

What the spider gives us that we did not have

The ISA already defined the BIND opcode and knew it corresponded to G₂. What Kuperberg adds:

1. A precise normalisation (R2). BIND composed with its adjoint equals 5 times the identity. Before Kuperberg, the ISA had BIND defined up to a constant. R2 fixes the constant.

2. A 4-body rewrite rule (R3). A diagram with four BIND vertices decomposes into two simpler pieces with coefficients 5/7 and 3/7. This is the ISA analogue of arithmetic simplification: R3 lets you reduce multi-BIND programmes to normal form.

3. A minimum circuit theorem. The smallest non-trivial closed BIND programme uses exactly 3 vertices (not 1 or 2). Two BIND vertices only produce the bigon (R2), which is just a scalar — trivial at H². Three vertices form the “theta-graph,” evaluating to 5 × 7 = 35, which is genuinely H². This also explains why transition-metal spectroscopy is dominated by 4-body interactions: three BIND strands, each connecting a pair of electrons, give 3 × 2 − 3 = 3 shared electrons plus 3 free — a 4-body Coulomb integral, the Racah B-parameter.

4. A completeness theorem. Any closed BIND programme terminates with a definite integer answer. This is the ISA computation theorem for H²: BIND programmes always halt.

Where BIND appears in physics

Once you know the axioms, the BIND opcode is recognisable across many domains:

Domain BIND manifestation What R2–R4 compute
Transition-metal spectroscopy Racah B, C parameters 4-electron Coulomb integrals
Nuclear physics Tensor force S₁₂; seniority number Pairing-force matrix elements
FeMoco (nitrogen fixation) 7 Fe atoms as 7 spider strands Full G₂ catalytic programme
Knot theory G₂ link invariant Integer polynomial in q
Topological order Non-Abelian anyon fusion F-matrix elements

The BIND theorem — non-Abelian topological order ↔ BIND present — now has a complete proof via Kuperberg’s Theorem 6.1.

The β-deformation: continuous temperature

At the β-deformation parameter q = e^{iπβ}, the Kuperberg rules quantise: the integers 7, 5, 3 in R1–R2 become quantum dimensions [7]_β, [5]_β, [3]_β. Snap events — phase transitions in the BIND calculus — occur when one quantum dimension vanishes:

  • β = 1/3: the 3-dimensional sector collapses
  • β = 1/5: BIND self-pairs become degenerate (BIND can no longer close on itself)
  • β = 1/7: the full G₂ sector collapses; the Fano plane vanishes

These are the G₂ critical points of the β-deformed ISA. The β = 1/7 snap is the H² phase transition in the H^k complexity ladder.

A three-level hierarchy

The BIND opcode lives at three levels, each handled by a different framework:

Fano 731 crystal (3D, Paper #207)
      ↓  lossless 2D projection
2D Frog Diagram (Paper #281, 4-valent frogs)
      ↓  subdivide each frog into 2 trivalent vertices
Kuperberg G₂ spider (3-valent)
      ↓  apply R1–R5
Scalar ∈ ℤ[q, q⁻¹]

The 4-valent frogs of Paper #281 and the 3-valent Kuperberg vertices are not competing: one 4-valent frog equals two Kuperberg vertices joined by an internal edge. Paper #281 keeps the tetrahedron whole and compensates with two local rules (Excluded Volume Principle, Polarity Rule); Kuperberg splits the tetrahedron and needs neither rule. The Poison Frog (Associator Defect) of Paper #281 is exactly the non-zero R4 pentagon remainder made visible as a node in the diagram.

The four-storey building

Kuperberg’s spider is the top floor of a four-storey diagrammatic hierarchy:

Floor ISA opcode Calculus Key property
H² BIND G₂ spider (Kuperberg 1996) Non-associative
H¹ TWIST / FLIP / FLOP BMW algebra (1989) Non-commutative
H⁰’ FLIP only Hecke algebra (1964) Commutative, signed
H⁰ ORBIT / SPLIT / SPLAT Temperley-Lieb (1971) Commutative, positive

The Frog calculus is the full building plus the β-deformation elevator between floors. Kuperberg provides the axioms for the top floor; this paper adds the ISA interpretation, the physical applications, and the β-deformation path that connects all four floors.

See also: