{"id":"444c5c60-e2c8-48fa-87d0-7a7d9a3ec059","arxiv_id":"2501.05349","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Fermionic cellular automata in one dimension are classified up to finite-depth circuits without ancillas, and a new 'forking' automaton is shown to be implementable but not expressible by qubit-style gates.","lead":"A mathematical study shows that one-dimensional fermionic cellular automata can be classified by their index without needing auxiliary 'ancilla' systems, and gives the first full list of nearest-neighbour fermionic automata. It also finds a genuinely fermionic automaton, the 'forking' automaton, that has no qubit counterpart, which matters for understanding what quantum circuits can and cannot simulate.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Ancilla removal on the infinite lattice is asserted via a 12-cell diagram; the finite-depth, homogeneous extension to Z is not proved, and the graded adaptation of Ref [41] is not established.","rationale":"After reading the full manuscript, the central claim to stress-test is Corollary 2: equal-index FCAs are F-equivalent without ancillas. Its proof depends on Proposition 2's ancilla removal. The reader's weakest_assumption already points here, and careful inspection supports that assessment. The 12-cell procedure in Figures 3-6 is illustrative: it does not define the infinite-lattice circuit, the layer structure, or the block tiling. It also does not address the Z2-graded complications (anticommuting odd operators) that distinguish fermionic from qudit systems. Without a rigorous graded analogue of the Freedman-Haah-Hastings bounded ancilla removal, the main result is conditional. I do not see a similar-scale gap in the classification of nearest-neighbour index-one FCAs: Proposition 3 and Theorem 3 come with explicit local rules and a concrete MS for the forking automaton (Corollary 5), so that part has independent support. Thus the most load-bearing concern is the ancilla-removal step, and it matches the reader's identification.","tokens_in":24671,"tokens_out":12287,"duration_ms":118537,"concrete_test":"Implement Proposition 2 as a formal circuit on a periodic chain of L cells for a non-trivial index-one FCA that is only stably M-implementable with one ancilla per cell (e.g., n=2 fermionic modes per site), and compute the circuit depth as a function of L; if depth grows with L, the infinite-lattice finite-depth claim fails. Alternatively, produce an explicit translation-invariant layer decomposition of the 12-cell block tiling and verify that every gate acts within a bounded partition and that the total number of layers is constant; failure to produce such a decomposition would confirm the proof gap.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Proposition 2, which is the sole bridge from stable equivalence to F-equivalence (Corollary 2), is proved by a finite 12-cell graphical construction. To conclude that an index-one FCA is F-implementable on the infinite lattice, one must show this construction tiles Z with a uniform, finite-depth circuit. The proof does not: it does not specify the block partition, the layer count, or how 'fully updated' cells used as simulated ancillae (e.g., cell 4 in Step 4) remain in the correct final state after extra gates are applied. The only external support cited is Ref [41], whose ancilla-removal theorem is proved for ungraded qudit QCAs; the paper asserts, but does not prove, that it carries over to Z2-graded CAR algebras where odd operators on disjoint sites anticommute and graded commutation replaces commutation in the support-algebra lemmas. If the sweeping depth grows with lattice size or the graded signs invalidate the borrowing argument, Corollary 2—the first main result—is unsupported. The nearest-neighbour classification (Theorem 4) is not affected by this concern, since it is backed by explicit MS implementations for the controlled-phase and forking cases.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper studies one-dimensional fermionic cellular automata (FCAs) over the Z2-graded CAR algebra. The authors first review the index theory of FCAs, in which two automata with equal index are known to be stably equivalent up to finite-depth fermionic circuits and the addition of inert ancillas. Their first result claims to remove the ancillas: equal-index FCAs are F-equivalent, i.e., connected by a finite-depth fermionic circuit acting only on physical cells (Corollary 2), via an ancilla-removal procedure (Proposition 2). The second part classifies all nearest-neighbour FCAs with one fermionic mode per site: any index-one FCA is either a local unitary, a controlled-phase type automaton (Proposition 3), or a new “forking automaton” (Theorem 3) that is explicitly implemented by a Margolus scheme (Corollary 5); non-unit index automata reduce to shifts or Majorana shifts composed with index-one FCAs. The classification demonstrates a fermionic circuit that cannot be written as qubit-style single-mode and controlled-phase gates.","tokens_in":24930,"tokens_out":10626,"duration_ms":95473,"significance":"The claimed strengthening of the index classification is conceptually important: it would remove the caveat that the fermionic classification requires ancillas, aligning the FCA theory with the ungraded QCA case. The nearest-neighbour classification is a concrete and useful result, and the forking automaton is a genuinely new object with an explicit and checkable Margolus implementation (Corollary 5). The paper is clearly structured and uses standard algebraic tools. However, the ancilla-removal result, which underpins the paper's headline claim, is not proved with the same rigor as the classification part; the latter is supported by explicit constructions and appears sound.","major_comments":[{"comment":"The proof of Proposition 2 is an informal finite-block construction (Figures 3-6) rather than a proof for the infinite lattice. It does not specify how the 12-cell block is tiled to cover all of Z, does not show that the resulting circuit has depth bounded by a constant independent of the block position, and does not prove that gates applied to cells that have already been “fully updated” (e.g., cell 4 in Step 4 and cell 3 in Step 3) leave those cells in their correct final state. Since Corollary 2 and the paper's first main result rest on this proposition, the claim that equal-index FCAs are F-equivalent without ancillas is not rigorously established as written. The text also invokes Ref. [41] for bounded ancilla removal, but that theorem is proved for ungraded qudit QCAs; the adaptation to the Z2-graded CAR algebra, where support-algebra lemmas involve graded commutators, is asserted rather than proved.","section":"IV, Proposition 2"},{"comment":"The final sentence of Corollary 2 claims that “upon suitable regrouping of cells one can recast F in an MS.” This does not follow from the cited Lemma 5, which concerns stable M-equivalence with ancillary copies (Eq. (25): T ⊖ I = M2∘M1∘(S⊖I)) rather than an ancilla-free MS relation K = M∘J. A separate argument that the ancilla-free FDFC from Proposition 2 has bounded depth and can be recast as a Margolus scheme after blocking is needed; as written, the “Moreover” clause is unsupported.","section":"Corollary 2"}],"minor_comments":[{"comment":"The section heading “AKNOWLEDGMENTS” is misspelled; it should be “ACKNOWLEDGMENTS.”","section":"VII"},{"comment":"In Eq. (22), the second layer is defined with unitaries M^{(2)}_{2x+1} acting over {y,y+1}, but the variable y is not defined in that expression; please clarify the action site.","section":"Definition 10"},{"comment":"In the proof of Lemma 10, point 2, “Theorem 8” should refer to “Lemma 8.”","section":"Lemma 10"},{"comment":"The sentence “Repeating the computation for I ⊖ Σ_{2i+1} we get X_{2i+i} ↦→ and Y_{2i+1} ↦→Y_{2i+1} while X_{2i+1} ↦→ −Y_{2i}” contains an incomplete image for X_{2i+i} (with a typo “2i+i”) and should be corrected.","section":"Appendix B 1"},{"comment":"In Eq. (42), the identity factors I3 and I1 are undefined; also in Eq. (44) the tensor factors are not labeled by lattice sites, which makes the controlled-phase expression unnecessarily hard to parse.","section":"Eq. (42)"},{"comment":"The displayed index values “1, 2±1/2, 2±1” should read 1, 2^{±1/2}, 2^{±1}; please ensure the exponents are typeset correctly.","section":"Section V"}],"recommendation":"major_revision","confidential_remarks":"The central issue is the rigor of Proposition 2; the classification part is sound and novel. The stress-test concern is accurate: the proof of the infinite-lattice, uniformly bounded ancilla removal is not supplied, and the reliance on Ref. [41] for the graded case is not justified. I would encourage the editor to request a rigorous treatment of this point before publication. The nearest-neighbour classification and the forking automaton are valuable and should not be delayed unnecessarily."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nYou should know this paper has one result that looks solid and one that is over-claimed. The solid result is the complete classification of nearest-neighbour FCAs over a single fermionic mode per site (Theorem 4): every index-one FCA is either controlled-phase type or the new forking automaton, and the forking automaton has an explicit Margolus circuit. That is a real contribution, checkable, and it identifies a genuinely fermionic primitive that qubit QCAs do not have.\n\nThe over-claimed result is the ancilla-free equivalence (Corollary 2). The bridge is Proposition 2, where the proof is a 12-cell diagrammatic sketch. It does not show how the construction extends to the infinite lattice with a uniform, finite-depth circuit, and it explicitly relies on Ref [41]'s ancilla-removal theorem for ungraded qudit systems without proving the graded-CAR version. The stress-test note is accurate: if the sweeping depth grows with system size or the graded signs break the borrowing argument, Corollary 2 is unsupported. This is the first main result in the paper, so the gap is load-bearing for that claim.\n\nThe rest of the index theory is imported from [15] and [39], and the classification logic is coherent. The forking automaton's circuit in Corollary 5 is a concrete computation; I checked the conjugation step and it works. The paper reads as honest work, not hand-waving in the conclusions, but the proof of Proposition 2 needs to be made rigorous.\n\nWho is this for? People working on QCA classification, Floquet phases, and fermionic simulation. They will want the classification and the forking automaton, and they will need to be careful before relying on the ancilla-free equivalence.\n\nMy recommendation: send it to peer review, but require either a complete proof of Proposition 2 on the infinite lattice, or a clear statement that the ancilla-free equivalence is conditional on a graded version of the ancilla-removal theorem. The classification part deserves publication on its own, and the gap is fixable.","headline":"Solid classification of nearest-neighbour fermionic automata, but the ancilla-free equivalence is built on a proof sketch.","tokens_in":25437,"tokens_out":3268,"would_cite":true,"duration_ms":30664,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"In one dimension, fermionic cellular automata with the same index are connected by a finite-depth fermionic circuit, with no additional ancillary systems.","keywords":["fermionic cellular automata","quantum cellular automata","index theory","finite-depth circuits","ancilla removal","Margolus partition scheme","forking automaton","Majorana shifts"],"falsifier":"Simulate the sweeping construction on a finite chain of length $N$ with periodic boundary conditions and record the minimal circuit depth and the number of cells used as temporary ancillas per update; if either grows without bound as $N$ increases, the ancilla-free equivalence theorem fails.","tokens_in":24516,"feed_emoji":"⚛️","tokens_out":11078,"duration_ms":93062,"temperature":0.7,"pith_summary":"This paper strengthens the classification of one-dimensional fermionic cellular automata. Earlier work showed that two automata with the same index—an invariant measuring the net left/right flow of information on the lattice—can be connected by a finite-depth circuit only if ancillary fermionic degrees of freedom are allowed. The paper shows that those ancillas can be removed: equal index now implies equivalence by a finite-depth fermionic circuit acting on the physical cells alone. It then completes the classification of nearest-neighbour automata with one fermionic mode per site, discovering a 'forking automaton' that is locally implementable but cannot be built from single-mode and controlled-phase gates composed with shifts, unlike every qubit cellular automaton. This matters because it shows fermionic locality sustains dynamics absent from qubit lattices while still admitting an index-based classification.","feed_headline":"No ancillas needed: 1D fermionic automata classified by index","feed_subtitle":"Equal index now means equal by a finite-depth circuit, and a 'forking' rule with no qubit analogue appears.","key_machinery":"The machinery is the fermionic index together with the support algebras of the cellular automaton. For a FCA $T$, the left and right support algebras $L_{2x}$ and $R_{2x+1}$ are the smallest graded subalgebras on which the evolved two-cell algebras are supported; the index is $\\mathrm{ind}[T]=\\sqrt{\\dim[L_{2x}]/\\dim[A_{2x}]}$, and it takes values $1$, $2^{\\pm1}$, and $2^{\\pm 1/2}$ for single-mode fermionic chains, with the half-integer powers of two coming from Majorana shifts. The paper's new step is an ancilla-removal procedure (Proposition 2): starting from the known fact that an index-one FCA is implemented by a Margolus partitioned scheme on the enlarged lattice with one ancilla per cell, it sweeps a 12-cell block across the lattice, using already-updated physical cells as temporary ancillas, to produce a finite-depth fermionic circuit on the physical system alone. The forking automaton is then isolated by a case analysis of the support algebras: if both supports are generated by odd anticommuting operators, the local rule must split the two odd generators between left and right neighbours. The classification is completed by combining these unit-index rules with shifts and Majorana shifts.","core_discovery":"The central claim is that the stable-equivalence classification of one-dimensional fermionic cellular automata can be made ancilla-free. Two FCAs $T$ and $S$ with the same index are $F$-equivalent: there exists a finite-depth fermionic circuit $F$ such that $T=F\\circ S$ (Corollary 2). For automata with one fermionic mode per site, every index-one FCA is $M$-implementable—realisable by a two-layer nearest-neighbour Margolus partition scheme—and the nearest-neighbour index-one FCAs are exhausted by the controlled-phase type of Proposition 3 and the forking automaton of Theorem 3. The forking automaton sends the two odd generators of a site to the two neighbouring sites, $T_0(\\eta)=X\\boxtimes I\\boxtimes I$ and $T_0(\\xi)=I\\boxtimes I\\boxtimes Y$, and cannot be expressed as single-mode and controlled-phase gates composed with shifts. This completes the classification: shifts and Majorana shifts cover all non-unit indices, and an index-one FCA is either a controlled-phase local unitary or a forking automaton.","pith_inferences":["The forking automaton could serve as a primitive for fermionic quantum information processing that has no qubit analogue; one testable extension is whether its two independent Majorana legs can be used for fermionic state transfer on a translation-invariant chain.","The ancilla-removal argument is presented for one dimension and one ancilla per cell; the same strengthening of stable equivalence might hold for lattices with boundaries or for higher-dimensional graded algebras, but the authors do not prove that here.","With more than one fermionic mode per site, the support algebras can be richer, so new index-one local rules beyond the controlled-phase and forking forms may appear.","The forking automaton could be probed numerically on finite chains: although it is locally implementable, its correlation or entanglement structure after one step may differ from controlled-phase automata, giving an observable fermionic signature."],"forward_implications":["Two fermionic cellular automata with the same index are equivalent by a finite-depth fermionic circuit without ancillas, matching the equivalence notion used for qubit cellular automata.","Every index-one FCA with one fermionic mode per site is implementable by a Margolus partitioned scheme, so index one is exactly the locally implementable class in this setting.","The nearest-neighbour index-one FCAs over $\\mathrm{Mat}(\\mathbb{C}^{1|1})$ are precisely the controlled-phase type and the forking automaton; composing with shifts and Majorana shifts exhausts all nearest-neighbour FCAs.","The forking automaton is a genuinely fermionic object: it is locally implementable but cannot be written as single-mode and controlled-phase gates composed with shifts, unlike every qubit cellular automaton.","Irrational index values $2^{\\pm 1/2}$ remain a fermionic phenomenon, associated with Majorana shifts that move odd fermionic degrees of freedom by half a cell per step."],"supporting_citations":[{"why":"Defines reversible quantum cellular automata and gives the qubit classification (single-mode and controlled-phase gates) that the fermionic results are compared against.","marker":"[4]"},{"why":"As cited in the paper, establishes the fermionic index theory and the stable-equivalence classification that this work strengthens by removing ancillas.","marker":"[15]"},{"why":"Proves that any cellular automaton can be implemented by a partitioned scheme with auxiliary systems, the starting point for the ancilla-removal argument.","marker":"[37]"},{"why":"Provides the index theory for one-dimensional qudit cellular automata and proves same-index equivalence for ungraded systems, the template extended to fermionic graded algebras.","marker":"[39]"},{"why":"Proves that ancilla removal can be done with a bounded number of ancillae in the ungraded case, the method adapted by Proposition 2.","marker":"[41]"},{"why":"Classified fermionic local rules with a technique that gives no structural insight; the present classification supersedes it by identifying the forking automaton.","marker":"[43]"}],"fun_headline_variants":["Ancilla-free classification of 1D fermionic automata","Forking automaton completes fermionic automata classification","Same index means same circuit for fermionic automata","No ancillas needed: fermionic automata fully classified","Fermionic automata: no ancillas, full classification"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The result rests on assuming that the trick of borrowing already-updated cells as temporary ancillas, shown on a 12-cell block, keeps working on the infinite chain with a bounded number of borrowed cells and a bounded circuit depth, and that the ancilla-removal theorem from the ungraded qudit setting transfers unchanged to $\\mathbb{Z}_2$-graded fermionic algebras.","fun_headline_variants_meta":{"raw":{"variants":["Ancilla-free classification of 1D fermionic automata","Forking automaton completes fermionic automata classification","Same index means same circuit for fermionic automata","No ancillas needed: fermionic automata fully classified","Fermionic automata: no ancillas, full classification"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00053,"raw_usage":{"total_tokens":2519,"prompt_tokens":873,"completion_tokens":1646,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":489,"completion_tokens_details":{"reasoning_tokens":1563}},"tokens_in":489,"tokens_out":1646,"duration_ms":10439,"temperature":1.0,"reasoning_tokens":1563,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T21:14:46.421682+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Simulate the sweeping construction on a finite chain of length $N$ with periodic boundary conditions and record the minimal circuit depth and the number of cells used as temporary ancillas per update; if either grows without bound as $N$ increases, the ancilla-free equivalence theorem fails.","supporting_citations":[{"cited_title":"g(T ) = 0 if T ∈ A0 and g(T ) = 1 if T ∈ A1, we have (T ⊠ I)(I ⊠ S) = ( −1)g(S)g(T )(I ⊠ S)(T ⊠ I), where I denotes the identity operator with g(I) = 0","cited_arxiv_id":null,"evidence_quote":"Defines reversible quantum cellular automata and gives the qubit classification (single-mode and controlled-phase gates) that the fermionic results are compared against."},{"cited_title":"Potter, and Ashvin Vishwanath","cited_arxiv_id":null,"evidence_quote":"As cited in the paper, establishes the fermionic index theory and the stable-equivalence classification that this work strengthens by removing ancillas."},{"cited_title":"Evenbly and G","cited_arxiv_id":null,"evidence_quote":"Proves that any cellular automaton can be implemented by a partitioned scheme with auxiliary systems, the starting point for the ancilla-removal argument."},{"cited_title":"In this paper, we can reformulate that notion for FCA as follows","cited_arxiv_id":null,"evidence_quote":"Provides the index theory for one-dimensional qudit cellular automata and proves same-index equivalence for ungraded systems, the template extended to fermionic graded algebras."},{"cited_title":"Unpaired majorana fermions in quantum wires","cited_arxiv_id":null,"evidence_quote":"Proves that ancilla removal can be done with a bounded number of ancillae in the ungraded case, the method adapted by Proposition 2."},{"cited_title":"Scalar fermionic ce llu- lar automata on finite cayley graphs","cited_arxiv_id":null,"evidence_quote":"Classified fermionic local rules with a technique that gives no structural insight; the present classification supersedes it by identifying the forking automaton."}],"review_version":1}