Clifford Quantum Cellular Automata: Trivial group in 2D and Witt group in 3D

I. INTRODUCTION

• A QCA is a local-algebra automorphism whose image of each single-site operator is supported within a uniform finite-radius ball.The setting specializes to uniform lattices of prime-dimensional qudits.Source: §I · p. 2, A. Translation invariant Clifford QCAs · p. 2

A. Translation invariant Clifford QCAs

• TI Clifford QCAs map generalized Pauli matrices to Pauli products and commute with translations; translation-invariant layers of Clifford gates generate circuits.The lattice has finitely many prime-dimensional qudits at each site.Source: A. Translation invariant Clifford QCAs · pp. 2–3

B. Equivalent QCAs and the group in question

• Two QCAs are equivalent when their stabilized difference is a translation QCA followed by a weakly translation-invariant Clifford circuit, making their classes an abelian group.The paper treats Clifford circuits and translations as trivial.Source: B. Equivalent QCAs and the group in… · p. 3
• For D=2, α is a weakly TI Clifford circuit followed by a shift, so C(D=2,p)=0; for D=3, C(D,p) contains Z2 × Z2 and also contains Z4.Source: B. Equivalent QCAs and the group in… · p. 4

C. Overview of the proof

• The proof associates a TI Clifford QCA with a boundary algebra and a nondegenerate antihermitian form, with circuits and shifts accounting for lost information.The paper studies these forms over Laurent polynomial rings and identifies 2D triviality through the univariate case.Source: C. Overview of the proof · pp. 4–5

II. EXPONENTS OF CLIFFORD QCA

• For any lattice of prime-dimensional qudits, the group of all translation-invariant Clifford QCAs modulo Clifford circuits and shifts has an exponent of 2 or 4.Source: §II · p. 5

A. Translation invariant Clifford QCAs and polynomial symplectic matrix

• Translation invariance gives a compact symplectic-matrix representation of a Clifford QCA, and the determinant of any such matrix is c c−† for some invertible c.Tensor products of QCAs correspond to direct sums of their symplectic matrices.Source: A. Translation invariant Clifford QCAs… · pp. 5–6

• Suslin’s stability theorem turns invertible matrices of unit determinant and size at least 3 into elementary row operations realizable by circuits.Adding an ancilla per site or enlarging the unit cell can meet the size condition.Source: B. To Clifford circuits · p. 8
• For p ≡1 mod 4, α is equivalent to α^-1, so α^2 is trivial.The proof relates α to a matrix α′ and then identifies α′ with α^-1.Source: B. To Clifford circuits · p. 8
• For prime p, the paper states that Q ⊕ Q′ is a product of elementary row operations corresponding to TI Clifford circuit layers, so [Q′] is inverse to [Q].This generalizes the exponent-2 result for p=2; when p=2, Q′ can be set equal to Q.Source: B. To Clifford circuits · p. 9
• The exponent calculation is motivated by finite-field Witt groups, which are Z2×Z2 for p ≡1 mod 4 and Z4 for p ≡3 mod 4.The paper says the connection is closer in its D=3 examples.Source: B. To Clifford circuits · p. 10

• A QCA boundary is the finite-width region where the image of a half-space algebra becomes nontrivial near its boundary.The boundary approach has been used in earlier classification results [3] [5].Source: §III · p. 10
• The paper identifies boundary-algebra generators, defines an antihermitian form over the boundary translation group algebra, and constructs QCAs from invertible forms, in any spatial dimension.Source: §III · p. 10
• The correspondence among QCAs, boundary algebras, and invertible antihermitian forms loses precisely circuit and shift information.Because the forms live one dimension lower, the 2D case reduces to quadratic forms over a field of transcendence degree 1 above a finite field, which the paper finds trivial.Source: §III · pp. 10–11
• The boundary-algebra definition requires the boundary-direction variable to have exponents 0 or 1, achieved using shifts and a finite-index translation subgroup.The paper says the resulting algebras for arbitrary translation-subgroup and shift choices need comparison.Source: §III · p. 11

• The ranks of A and B are even and sum to 2q, and their kernels under the paired adjoint forms are the images of the other matrix.Their smallest nonvanishing determinantal ideals are both the unit ideal.Source: A. Commutation relations among… · p. 12
• The boundary-algebra Pauli operators modulo phases form a free module of even rank, whose antihermitian commutation form has determinant a nonzero square in Fp.The form depends on the choice of free basis only up to congruence.Source: A. Commutation relations among… · p. 14
• Reducing translations along directions other than the boundary axis can change the form, and invariance of its Witt class and congruence across axes remain open.Source: A. Commutation relations among… · p. 15

• For a TI Clifford QCA with boundary form Ξ, composing with a circuit and a shift can, after coarsening translation, yield a form congruent to diag(Ξ, λ^n).Source: B. Circuit invariance of induced… · p. 15
• If α is a TI Clifford circuit composed with shifts, its antihermitian form is congruent to λ^n under some smaller translation group.Source: B. Circuit invariance of induced… · p. 15
• The proof reduces the circuit case to a z-direction shift and expands (A+zB)(E0+zE1), separating standard λ blocks from the nonsingular part that supplies the induced form.Source: B. Circuit invariance of induced… · pp. 16–17

C. Witt group of antihermitian forms

• Stable equivalence identifies forms after congruence and addition of possibly different numbers of λ blocks.The displayed relation is E†(Ξ ⊕ λ^n)E = Ξ′ ⊕ λ^n′.Source: C. Witt group of antihermitian forms · p. 18
• The equivalence classes of finite-dimensional invertible antihermitian matrices form a group under orthogonal direct sum.The matrices are required to have zero constant terms on the diagonal; the paper establishes inverses even in characteristic 2.Source: C. Witt group of antihermitian forms · p. 18
• If the Witt group of a field has exponent n, the Witt group of invertible antihermitian matrices over a ring with a compatible involution also has exponent n.The proof represents the n-fold class by I_n ⊗ Ξ.Source: C. Witt group of antihermitian forms · p. 19
• The Witt group of antihermitian matrices is trivial over a field and, for finite fields, over F[x±].The one-variable result follows by showing forms of dimension greater than two are isotropic, so they can be reduced using hyperbolic planes.Source: C. Witt group of antihermitian forms · pp. 19–20
• The paper requires stable equivalence and does not establish that it implies equivalence without stabilization.It lacks an analogue of the Witt cancellation theorem.Source: C. Witt group of antihermitian forms · pp. 20–21

D. Boundary algebra determines QCA

• The boundary algebra contains all information about the bulk TI Clifford QCA modulo Clifford circuits and shifts.Blending two QCAs is defined by an interpolating finite-range QCA that agrees with each on its respective infinite subsystem.Source: D. Boundary algebra determines QCA · p. 21
• A TI Clifford QCA that blends into a shift between disjoint half-spaces is a weakly TI Clifford circuit followed by a shift.The blending QCA is required to be weakly translation invariant along directions orthogonal to the axis separating the half-spaces.Source: D. Boundary algebra determines QCA · p. 21
• If the induced antihermitian form is congruent to λ, the QCA is a weakly TI Clifford circuit followed by a shift.The proof constructs an interpolation that agrees with the QCA on one half-space and acts as a shift sufficiently far into the other.Source: D. Boundary algebra determines QCA · pp. 22–23
• The form induced by the inverse QCA is stably equivalent to the inverse of the original form.The proof uses that the tensor product of a QCA and its inverse is a weakly TI Clifford circuit.Source: D. Boundary algebra determines QCA · p. 23
• Every invertible antihermitian form over the stated Laurent polynomial ring can be realized as the boundary form of a TI Clifford QCA.The construction produces a QCA in one higher dimension.Source: D. Boundary algebra determines QCA · pp. 23–24
• The stable equivalence class of an invertible antihermitian form uniquely determines its TI Clifford QCA up to the paper’s QCA equivalence.The corollary also constructs a QCA inducing the given form.Source: D. Boundary algebra determines QCA · p. 24

E. Triviality of Clifford QCA in 2D

• Every 2D TI Clifford QCA is trivial because its induced invertible antihermitian form over a one-variable Laurent polynomial ring is congruent to λ^n.That congruence implies the QCA blends into a TI shift and is therefore trivial.Source: E. Triviality of Clifford QCA in 2D · p. 24
• The 2D-specific step is the ability to gap out the one-dimensional boundary; gapping a boundary is highly nontrivial in general.The paper identifies this as the special property used in the triviality proof.Source: E. Triviality of Clifford QCA in 2D · p. 24
• The proof leaves open whether triviality can be achieved while preserving full translation symmetry and the number of qudits per unit cell.The Cohn matrix has determinant 1 but is not a product of elementary row operations; the paper instead uses weak translation symmetry breaking.Source: E. Triviality of Clifford QCA in 2D · pp. 24–25

IV. CLIFFORD QCA IN 3D

• The 3D examples use scale-invariant surface topological order because weak translation symmetry breaking makes the boundary-form approach ambiguous.The section says the boundary-algebra theory cannot establish these examples’ nontriviality.Source: §IV · p. 25
• A TI Clifford QCA can be specified using a TI commuting Pauli Hamiltonian with a unique ground state on a sufficiently large torus.Rearranging its terms into a locally flippable separator yields a QCA, unique up to a TI Clifford circuit.Source: §IV · p. 25
• The example Hamiltonian acts on 2Z3 with odd-prime-dimensional qudits and has exactly two types of terms per site.Its stabilizer map is a 4-by-2 matrix over the three-variable Laurent polynomial ring.Source: §IV · p. 26
• In the symplectic matrix, each column specifies the Pauli image of a qudit operator, with monomials encoding Pauli factors and their lattice sites.The first column represents an operator with 10 nonidentity tensor factors.Source: §IV · p. 27
• The example’s induced antihermitian form has determinant 4f^2.Source: §IV · p. 27

A. Construction by condensation

• Each toric-code layer is viewed as two dyon theories, and each dyon is paired with its time-reversal conjugate in an adjacent layer for condensation.Source: A. Construction by condensation · pp. 27–28
• For p ≡ 1 mod 4, the toric code is suggested to admit two different decompositions; for p ≡ 3 mod 4, the two choices of f up to squares are f = ±1.Source: A. Construction by condensation · p. 28
• Commuting hopping terms for condensed boson pairs define a Pauli Hamiltonian with a nondegenerate ground state on any torus, created from a product state by a TI Clifford QCA.The paired dyons have zero topological spin; the proposed hopping operators require a modification to make them commute.Source: A. Construction by condensation · pp. 28–29
• The construction selects a particular spatial direction, and whether its final result has emergent isotropy is left open.Source: A. Construction by condensation · p. 29

B. Nontriviality

• An exposed boundary leaves an unpaired Θp,f, and the surface order is used to show α(p,f) cannot be a Clifford circuit followed by a shift.The proof constructs a surface Hamiltonian so that the half-space system satisfies local topological order.Source: B. Nontriviality · p. 29
• A hypothetical matching weakly TI Clifford circuit would yield a 2D commuting Pauli Hamiltonian whose required nontrivial boson contradicts Θp,f’s lack of non-vacuum bosons.Source: B. Nontriviality · pp. 29–30
• For p ≡ 1 mod 4, the examples generate Z2 × Z2, while for p ≡ 3 mod 4, αp,1 generates a Z4 subgroup.The p ≡ 3 mod 4 case follows because αp,1 ⊗ αp,1 is nontrivial and every group element has exponent 4.Source: B. Nontriviality · p. 30
• Nontriviality against arbitrary constant-depth gates is conjectured, relying on beliefs about chiral theories and anyon theories realizable by commuting Hamiltonians.Source: B. Nontriviality · p. 30
• For qudits of dimension p ≡ 1 mod 4, the toric code has two inequivalent decompositions: Θp,1 ⊕ Θp,−1 and Θp,g ⊕ Θp,−g, with Θp,1 not isomorphic to Θp,g.Source: B. Nontriviality · p. 30

C. Surface topological order

• The boundary Hamiltonian is checked for local topological order, a condition used to apply a result forcing a nontrivial boson in a suitable 2D commuting Pauli Hamiltonian.The proof uses a Gröbner-basis division algorithm to show that relevant Pauli operators are products of surface terms within a constant neighborhood.Source: C. Surface topological order · p. 33
• After locally created excitations are removed, pairs of flipped surface terms can be fused by hopping operators, so a single surface term represents each topological charge.Source: C. Surface topological order · p. 33
• Commutators of long string operators give θa = e^(2πim/p), with m = 1/(4f), identifying the boundary theory as Θp,1/(4f), congruent to Θp,f.The reported spin is independent of string length n ≥ 1.Source: C. Surface topological order · p. 34

V. DISCUSSION

• TI Clifford QCA are trivial modulo circuits and shifts in dimensions 0, 1, and 2, and every such QCA has fourth power trivial.Three-dimensional examples have true orders 2 and 4, so the exponents cannot be reduced.Source: §V · p. 34
• The number of nontrivial QCAs in three or higher dimensions remains unknown, and classifying C(D ≥ 3, p) is open.Source: §V · p. 34
• The proposed relation between boundary algebras and Witt groups is not fully settled because the effect of coarse-graining on the Witt group is not understood.The paper describes C(D, p) as a direct limit under these induced endomorphisms.Source: §V · p. 35

Appendix B.

• The authors describe the proof that their 3D examples are nontrivial as improvised; the boundary theory is physically motivated but not fully derived, and only the hopping operators are derived.They suggest a better classification invariant may come from the boundary algebra, including its locally generated maximal commutative subalgebra and spectrum.Source: Appendix B · p. 35
• The authors raise open questions about intrinsically defining boundary algebras and propose a possible Brauer-group framework, asking whether its 2D version matches the Witt group of braided fusion categories.They expect such a group to be at least as rich as the Witt group of finite-dimensional antihermitian forms.Source: Appendix B · p. 35

Appendix A: Every QCA in two dimensions is trivial

• For any QCA of range r on a compact two-dimensional manifold, composing with a shift of range at most cr yields an ancilla-free circuit of depth at most d(r).The constants depend on r, not the system size.Source: Appendix A · p. 36
• Alternating red and blue strips let the proof match the QCA to shifts strip by strip, using local algebra decompositions and Hall marriage to construct the red-strip QCA.The construction is extended from the torus to general two-manifolds by trivializing disk interiors, then one-skeleton segments, and finally neighborhoods of vertices.Source: Appendix A · pp. 37–39

Appendix B: Digression to automorphism groups of quadratic forms

• The quadratic-form digression encodes TI Clifford QCA using automorphisms over Laurent polynomial rings and defines C(D,p) through a direct-limit quotient; whether both embeddings are needed is unknown.The embeddings correspond to adding ancillas and breaking translation symmetry to a finite-index subgroup.Source: Appendix B · pp. 39–40

Appendix C: Equivalence of two QCA on qubits (p = 2)

• The coupled-layer construction and the earlier Walker–Wang-based qubit construction yield the same antihermitian form and are equivalent by Corollary III.16.The authors report this agreement as consistent with Conjecture V.1.Source: Appendix C · pp. 40–41

SUPPLEMENTARY MATERIAL

• The supplementary Mathematica notebook records calculations for 3D stabilizer Hamiltonians, disentangling QCAs and boundary forms, surface particle content, and a qubit QCA.The qubit example was first reported in [6].Source: SUPPLEMENTARY MATERIAL · p. 41