Chemistry and quantum computing are the same computation

Plain-language explainer for doi:10.5281/zenodo.21345106 (#595)

The one-sentence version

The space of all 2-qubit gates is a tetrahedron; attaching the right homology theory to it (Bredon cohomology with isotropy coefficients) gives Euler characteristic 2 — a non-trivial topological invariant that distinguishes CNOT from iSWAP, and shows that the chemistry Grassmannian and the quantum-computing Weyl chamber are fibres of the same geometric object.

The tetrahedron of two-qubit gates

Every 2-qubit unitary U is locally equivalent (up to single-qubit rotations on each qubit) to a gate of the form:

\[U_W(c_1, c_2, c_3) = \exp\!\bigl(i(c_1\, XX + c_2\, YY + c_3\, ZZ)\bigr)\]

The three parameters (c₁, c₂, c₃) with π/4 ≥ c₁ ≥ c₂ ≥ c₃ ≥ 0 form the Weyl chamber — one point per equivalence class of 2-qubit gates. The Weyl chamber is a tetrahedron with four distinguished vertices:

Vertex Weyl coords Gate Entanglement H^k tier
v₀ (0, 0, 0) Identity none H⁰
v₁ (π/4, 0, 0) CNOT / CZ maximal H¹
v₂ (π/4, π/4, 0) iSWAP maximal H¹
v₃ (π/4, π/4, π/4) SWAP none (!) H⁰

The surprising fact: SWAP is at a vertex of the tetrahedron but produces no entanglement — it is a permutation gate. CNOT and iSWAP are both maximally entangling, but they sit on different faces of the tetrahedron and are not locally equivalent to each other.

Why standard homology misses the structure

The Weyl chamber, as a topological space, is just a solid tetrahedron — a 3-ball. Its standard homology is trivial: H₀ = ℤ, H₁ = H₂ = H₃ = 0. This tells you nothing about the gate structure.

The problem: the tetrahedron’s four vertices are not equivalent. The Identity vertex has a large centraliser (many rotations that commute with it — 6-dimensional in su(2)²). The iSWAP vertex has a small centraliser (1-dimensional). These symmetry differences are invisible to standard (constant-coefficient) homology.

The fix: Bredon cohomology with isotropy representation coefficients. Instead of attaching the same coefficient ring ℤ to every simplex, attach a coefficient L(v) = dim(centraliser of the gate at vertex v):

Vertex Gate Centraliser dim L(v)
v₀ Identity 6
v₁ CNOT 2
v₂ iSWAP 1
v₃ SWAP 3

The Bredon coboundary maps use these local coefficients. The result:

\[H^0_{\rm Bredon} = 1, \quad H^1_{\rm Bredon} = 0, \quad H^2_{\rm Bredon} = 1, \quad \chi_{\rm Bredon} = 2\]

The Euler characteristic jumps from 1 (standard) to 2 (Bredon). The H²_Bredon = 1 class is a genuine non-trivial topological invariant of the 2-qubit gate space — invisible to entanglement entropy alone.

CNOT vs iSWAP: same entanglement, different topology

Both CNOT and iSWAP are maximally entangling gates. Both have entanglement power E ≈ 0.35 (they produce the same amount of entanglement from a product input). Yet they are not equivalent — you cannot continuously deform CNOT into iSWAP via single-qubit rotations.

The Makhlin invariants (g₁, g₂) distinguish them:

Gate g₁ g₂ Entanglement power
Identity +1 +3 0
CNOT 0 +1 0.358
iSWAP 0 −1 0.351
SWAP −1 −3 0

CNOT has g₂ = +1; iSWAP has g₂ = −1. The sign of g₂ distinguishes them — even though their entanglement power is essentially identical. Makhlin g₂ is strictly finer than entanglement power as a gate classifier.

In ISA language: both CNOT and iSWAP are 1-BIND gates (each requires exactly one BIND opcode). Their different Weyl chamber position means they sit on different faces of the tetrahedron — they are topologically distinct within the H² stratum, even though both are H².

Three-qubit gates: the Bredon cascade

Extending to 3-qubit gates, the Weyl chamber becomes a 4-simplex (5 vertices). The Bredon Euler characteristic changes dramatically:

System Weyl dim χ_Bredon
2-qubit 3 2
3-qubit 4 0

The jump from 2 to 0 as you add one qubit is a non-trivial equivariant signature. It means:

  • 2-qubit gate space has one non-trivial Bredon H² class — the topological distinction between CNOT-type and iSWAP-type gates
  • 3-qubit gate space is equivariantly contractible — the Toffoli gate (the 3-qubit BIND) does NOT generate a new topological class; it is connected to Identity via equivariant deformation

This matters architecturally: the new topology (H² Bredon class) only appears in the 2-qubit sub-sector. The 3-to-7 qubit jump for quantum error correction is not arbitrary — it reflects this exact structure.

The 3-to-7 qubit architecture jump

The 3-qubit Weyl chamber has 4 Cartan parameters {ZZI, ZIZ, IZZ, ZZZ}. The Fano plane has 7 points. Why does useful quantum error correction require 7 qubits (Steane [[7,1,3]] code) rather than 3?

The Bredon theory gives a precise answer:

  • Gate space (3 qubits): 4-simplex, 4 Cartan parameters → classifies all 3-qubit gates
  • Stabiliser space (7 qubits): Fano plane PG(2,2), 7 points → [[7,1,3]] Steane code

The centraliser dimensions predict which stabiliser measurements are fragile:

  • High centraliser dim (e.g., Identity with dim=6): easy to measure, robust
  • Low centraliser dim (e.g., iSWAP with dim=1): fragile, needs more QEC overhead

The Steane code needs 7 physical qubits because it takes 4 gate parameters (dim of 3-qubit Weyl chamber) plus 3 (code distance) to achieve error correction. The Bredon cohomology bridges gate space and stabiliser space.

Chemistry and quantum computing: the same tetrahedron

The deepest result of this paper is that the chemistry Grassmannian (from Papers 570, 594) and the quantum-computing Weyl chamber are fibres of the same geometric object — the flag manifold Fl(1,2,…,n).

Side-by-side dictionary:

Concept Chemistry (Grassmannian) Quantum computing (Weyl chamber)
The space Gr(k,n) — orbital subspaces W ⊂ ℝ³ — gate equivalence classes
Invariant 1 NOON n ∈ [0,2] Makhlin g₂ ∈ {−3,−1,+1,+3,…}
Invariant 2 Grassmannian angle θ_G Entanglement power E
H⁰ tier Frozen orbital (n ≈ 0 or 2) Identity / SWAP (g₂ = ±3)
H¹ tier Covalent bond (n ≈ 1) CNOT / CZ (g₂ = +1)
H² tier Correlated / aromatic iSWAP / non-Clifford (g₂ = −1)
The snap NOON hits 1 (Schubert crossing) g₂ changes sign
Classical fails at H¹/H² boundary (DFT) H¹/H² boundary (Gottesman-Knill)
Algebraic layer H⁰ Cl(n) spinors Clifford algebra
Algebraic layer H¹ SU(k,n)/U(k) twistors Clifford + CNOT
Algebraic layer H² G₂ / octonions Magic states, T-gates

DFT fails at H¹/H²; Clifford simulation fails at H¹/H². They fail at the same boundary.

This is not a coincidence — it is the same topological obstruction (the Bredon H² class, χ_Bredon = 2) appearing in two different physical contexts. The ISA is the language that makes this visible.

The corrected NOON ↔ Weyl map

A key technical result: the correct map between natural orbital occupancies (NOONs) from chemistry and Weyl chamber coordinates from quantum computing is:

\[c_1 = \frac{\pi}{4}(1 - |n - 1|), \qquad c_2 = \frac{\pi}{4}\frac{\theta_G}{90°}\]

This map is symmetric around n = 1 (reflecting the particle-hole symmetry: a bond with NOON n is equivalent to one with NOON 2−n). Previous maps in the literature were asymmetric — this correction was confirmed by 5/5 alignment between chemistry benchmarks and Weyl chamber positions.

See also:

For the full technical treatment, see doi:10.5281/zenodo.21345106