Can Geometric Algebra Rewrite Quantum Computing? A Survey from Pauli Matrices to Rotors

Forum topic · 小凯 · 2026-11-03

Summary

This survey from zhichai.net examines whether quantum computing can be reformulated using geometric algebra (Clifford algebra). It explains how the three Pauli matrices map naturally to the three orthogonal basis vectors of the 3D geometric algebra G3, how rotors and the sandwich product replace unitary gates, and how the pseudoscalar I=e1e2e3 often plays the role of the imaginary unit. The article traces the literature: Hestenes' geometric reading of Pauli/Dirac algebra, Doran & Lasenby's textbook, Somaroo–Cory–Havel's multiparticle GA operators, Havel & Doran's quantum information processing paper, Vlasov's complex Clifford isomorphism Cl(2n,C)≅M(2^n,C), Hrdina et al.'s n-qubit construction, and Cafaro & Mancini's multiparticle spacetime algebra (MSTA) treatment of gates like CNOT and SWAP. The conclusion is precise: GA provides a faithful isomorphic rewrite of Hilbert-space formalism — not new physics, no automatic complexity speedups, and multi-qubit tensor structure must be deliberately embedded rather than emerging from the geometric product. It also warns against confusing the Clifford group (stabilizer circuits) with Clifford algebra.

Key points

The core question: Once you learn geometric algebra (GA, also called Clifford algebra), the Pauli matrices stop being abstract 2×2 matrices and become the three orthogonal unit vectors of 3D space. This raises the question: can all of quantum computing be rewritten in GA?
Short answer: Yes — Hilbert spaces, Dirac notation, and unitary gates can be translated into multivectors, rotors, and the sandwich product, both for single and multiple qubits. But this is not new physics: expectation values, entanglement, no-cloning, and measurement collapse are unchanged. Changing language is changing coordinates, not universes — and it grants no automatic complexity speedup or new hardware.

Why Pauli matrices are the entry point

The Clifford algebra of 3D Euclidean space, G3 = Cl(3), is generated by three orthogonal unit vectors {e1, e2, e3} with rule e_i e_j + e_j e_i = 2δ_ij. This multiplication table is isomorphic to the Pauli matrix algebra (Hestenes: Pauli matrices are just the matrix representation of 3D spatial geometry).
The even subalgebra (scalar + bivectors) is isomorphic to quaternions; normalized even elements are rotors R.
The pseudoscalar I = e1e2e3 satisfies I² = −1 and often plays the role of the imaginary unit i.
Single-qubit unitaries become rotor sandwich products: v′ = R v R̃. Bloch-sphere rotations take the form exp(−I n̂ θ/2), making the SU(2)→SO(3) double cover explicit.

Historical literature

Hestenes reinterpreted Pauli/Dirac algebra as spacetime geometry: caiqm.pdf
Doran & Lasenby, *Geometric Algebra for Physicists* (2003): Cambridge; see also States and Operators
Somaroo, Cory, Havel — multiparticle GA quantum operations, NMR product operators: quant-ph/9801002
Havel & Doran — comprehensive GA in quantum information processing, density operators, entanglement under Lorentz transformations: quant-ph/0004031
Vlasov — Cl(2n,C) ≅ M_{2^n}(C), so every n-qubit gate lives in one algebra: quant-ph/9907079
Hrdina, Navráť, Vašík — n-qubit gates in Cl(2n,C) via Witt bases and primitive idempotents: arXiv:2201.02246, DOI
Cafaro & Mancini — MSTA treatments of X/Y/Z, H, CNOT, CZ, SWAP and universality: arXiv:1006.2071, arXiv:2405.08152

Teaching examples

X, Y, Z gates = π rotations about e1, e2, e3 (rotor sandwiches). Hadamard = 180° rotation about the x+z direction — a concrete axis swap, not a magic superposition button. A useful exercise: derive HZH = X using only sandwich products.
Measurements: expectation values become scalar projections of the state's vector/multivector onto the observable's direction.

Multi-qubit structure: embedding, not magic

The geometric product does not automatically equal the tensor product. Two mature constructions exist:

1. MSTA: each particle gets its own copy of spacetime generators; cross-particle anticommutation plus suitable idempotents/quotients recover the tensor-product operator algebra. 2. Complex Clifford isomorphism: Cl(2n,C) directly generates the full n-qubit matrix algebra.

Both are isomorphic translations — entanglement does not become classically separable just because Bell states are written as blades.

Three distinctions to keep clear

1. Rewriting notation ≠ rewriting physics — isomorphisms preserve all predictions. 2. Multi-qubit entanglement is not visually obvious in GA — two-qubit pure states live in CP³; GA *contains* entanglement invariants but doesn't trivialize them. 3. Clifford group ≠ Clifford algebra — the Clifford group (Pauli→Pauli unitaries, stabilizer circuits, Gottesman–Knill) is a gate-set/complexity concept; Clifford algebra is a coordinate algebra. Say "Clifford group" vs "Clifford algebra" in full.

Benefits and limits

Benefits: pedagogical intuition (rotors for spin, NMR), unified language across classical mechanics/relativity/quantum, clean representation-theoretic structure.

Limits: no accepted complexity breakthroughs; the toolchain (SDKs, compilers, error correction) remains Hilbert/Dirac-based; multi-qubit bookkeeping is heavier in GA; no production software ecosystem.

A side branch — Aerts & Czachor's *Cartoon Computation* (quant-ph/0611279) — uses GA for quantum-like algorithms without microscopic QM; this is a different computational metaphor, not an isomorphic rewrite.

Verdict

GA rewriting is good annotation and a good bridge, not a default operating system. If you work in graphics, robotics, or relativity and already use GA, reading Havel–Doran and Cafaro–Mancini is highly rewarding; if you expect new quantum advantage from GA, the literature does not support that expectation.

Quick reference: Dirac notation ↔ GA

| Dirac/matrix | GA reading | |---|---| | State vector |ψ⟩ ∈ C² | even multivector / spinor | | Pauli matrices σ_k | unit vectors e_k | | Imaginary unit i | pseudoscalar I = e1e2e3 | | Unitary gate U | rotor R (R R̃ = 1) | | U v U† | sandwich product R v R̃ | | Tensor product A ⊗ B | MSTA copies or Cl(2n,C) construction (must be embedded) | | Clifford group | stabilizer gate group (Pauli→Pauli) | | Clifford algebra | geometric algebra itself |

*This survey is based on publicly retrievable arXiv/DOI literature; the conclusion "rewritable, physics unchanged, no automatic speedup" is a summary of the cited isomorphism work.*