{"id":"8e47ba2b-cc09-40b4-bc27-ad0f72f9394a","arxiv_id":"2412.03955","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A UVCC state-preparation circuit that controls rotations on one qubit per vibrational mode instead of all qubits reduces CNOT counts while preserving the encoded physical action.","lead":"A new circuit construction for the Unitary Vibrational Coupled-Cluster ansatz cuts the number of entangling gates by exploiting the unused parts of the qubit Hilbert space in unary encoding. The authors report theoretical gate-count reductions up to 50% and demonstrate lower noise sensitivity on Quantinuum hardware for 6- and 8-qubit systems.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Equations 18-19 are inconsistent with the defined circuit: U_m|e_m> leaves control q1=0 (or contradicts Eq. 17), so the half-controlled RY cannot rotate |e_m> into |g_m>; the gate-count claim is unverified.","rationale":"The paper's central quantitative claim is the CNOT scaling in Tables I and II. That scaling depends entirely on Eq. 12 being an exact decomposition of the m-excitation unitary on the physical subspace. The supporting equations 15-19 are the only proof given, and under the paper's own notation they do not cohere. Starting from |e_m>, each S_j with q0=1 maps |10>_j to |01>_j (Eq. 17), and then CX(q0,q1) flips q1, so Eq. 19's |10>_0 cannot be the final state; if one tries to keep |10>_0, the control q1 is 0, so the central controlled-RY step is disabled. Either way the printed derivation fails. This is more specific than 'the proof is sketched': it is a concrete inconsistency. The reader's weakest assumption identified the same region of the argument but not this particular consequence. The hardware demonstration is encouraging evidence that some corrected version of the circuit works, and the TVD numbers point in the right direction, but without a corrected, machine-verified circuit the asymptotic gate-count claim should not be taken as established. A simulator-based unitary check is cheap and decisive, so the appropriate verdict remains conditional pending that verification.","tokens_in":10997,"tokens_out":15292,"duration_ms":132784,"concrete_test":"Implement Eqs. 12-14 exactly (m=2 and m=3) in a quantum simulator, choosing CNOT(a,b)=control a, target b and the standard 3-CNOT decomposition of the relative-phase Toffoli (or the explicit Figure 1 circuit). Compute the full unitary and check: (a) U_m|g_m> and U_m|e_m> match the printed Eqs. 18-19; (b) the composite U_m(θ) equals the desired block rotation (Eqs. 8-10) on all 4^m basis states, including all non-|g_m>/|e_m> physical states. If either check fails, the central gate-count claim in Section IV is not supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim (Section IV) is that Eq. 12 implements Eq. 6 with (8m-6) CNOTs. This requires that after the U_m of Eq. 13, |g_m> and |e_m> differ only in the target qubit q0 while all m controls q1,q3,...,q_{2m-1} are 1. The only proof of this is Eqs. 15-19, but as printed those equations are inconsistent with the preceding gate definitions under the standard convention CX(a,b)=control a, target b. Eq. 17 gives S_j|10>_j|1>_0 = |01>_j|1>_0, so for |e_m> the modes j=1..m-1 become |01> and q0 stays 1; the trailing CX(q0,q1) then flips q1 from 0 to 1, giving mode 0 as |11>, not the printed |10>_0 of Eq. 19. If instead the trailing CX is read with the opposite convention, Eq. 18 fails. Moreover Eq. 19 as printed leaves q1=0, so the controlled RY of Eq. 12 would not act on the |e_m> branch and no |g_m>-|e_m> rotation would occur. This is not an obscure phase issue: without a corrected and fully specified U_m, the claimed half-control reduction is unsupported, and the action of U_m^\\dagger on the |11> intermediate states needed to unwind other physical basis states is never given.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript proposes a new circuit decomposition for the m-fold excitation unitary U_m(θ) used in Unitary Vibrational Coupled-Cluster (UVCC) state preparation under unary (direct) encoding of vibrational modes. The key idea is to exploit the fact that only a small subspace of the qubit Hilbert space is physical, allowing the usual (2m−1)-controlled Y-rotation in the Givens-type decomposition to be replaced by an m-controlled Y-rotation. The authors claim that the intermediate unitary U_m can be built from CX gates and relative-phase Toffoli gates at a cost of (8m−6) CNOTs, yielding up to 50% reduction in entangling gates versus standard decompositions and an asymptotic 28% reduction versus ancilla-based methods such as Khattar-Gidney. They also report quantum-hardware experiments on Quantinuum H1-1 for 6- and 8-qubit UVCCSDT state preparation, with improved total variation distance compared with the Givens-rotation implementation.","tokens_in":11326,"tokens_out":33273,"duration_ms":305886,"significance":"If the central construction were correct, the paper would provide a simple and practically useful constant-factor reduction in two-qubit gate counts for UVCC state preparation in unary encoding, with a clear information-theoretic rationale. The improvement is analytic rather than fitted, and the authors are transparent about the fault-tolerant Toffoli overhead and about the method's restriction to vibrational, rather than electronic, unitary coupled cluster. The hardware demonstration, though small, is a concrete test of the proposed circuits. However, the printed derivation of the key unitary action is internally inconsistent, and the cost tables contain arithmetic errors. Since these issues are load-bearing, the claimed gate-count advantages and the hardware interpretation are not currently established.","major_comments":[{"comment":"The proof that U_m maps |g_m> and |e_m> to states differing only in q0 is not valid as printed. Under the standard convention CX(a,b) = control a, target b, Eq. (17) gives S_j|10>_j|1>_0 = |01>_j|1>_0; applied to |e_m>, the modes j=1,...,m-1 become |01> while q0 remains 1, after which the trailing CX(q0,q1) flips q1 to 1, producing |11>_0, not the |10>_0 printed in Eq. (19). If CX(q0,q1) is instead read with q0 as target, Eq. (19) is satisfied for mode 0, but modes j>=1 remain |01>, not |11>. Similarly, Eq. (18) cannot follow from Eq. (16) for |g_m>, because Eq. (16) leaves |01>_j unchanged when q0=0, so U_m|g_m> would remain |g_m>, not become all-|11>. In addition, the unitary action of the relative-phase Toffoli in Eq. (14) is never defined, so Eqs. (15)-(17) cannot be independently checked. Since the m-controlled RY in Eq. (12) acts correctly only if the odd control qubits are in |1> for both |g_m> and |e_m>, and only for those states, the central gate-count claim is unsupported until a complete and consistent truth-table derivation for the physical subspace is supplied.","section":"Section III, Eqs. (13)-(19)"},{"comment":"The reported CNOT counts are internally inconsistent. For the proposed method, U_m+U_m^dagger costs (8m-6) CNOTs with the stated 3-CNOT relative-phase Toffoli; adding an m-controlled rotation decomposed as 2m CNOTs, as the Table I caption says, gives 14 CNOTs for m=2 and 24 for m=3, whereas Table I lists 13 and 25. Table II's formula for the proposed method with A=2, B=3, namely 20m-32, gives 8 CNOTs for m=2, which is less than the (8m-6)=10 CNOTs needed for U_m and U_m^dagger alone, and the same formula is negative for m=1. The reduction implied by the Table II formulas is also m-dependent (50% at m=2, 36% at m=3, asymptotically 28%), which contradicts the unqualified '28% reduction' statement in Section IV. The exact counts in Tables I and II should be re-derived with an explicit, consistently applied decomposition rule, and the valid range of m should be stated.","section":"Section IV, Tables I and II"},{"comment":"The hardware comparison rests on single total-variation-distance values obtained from 220 and 512 shots, with no error bars or repeated runs. For the S-8 system the difference is 0.240 versus 0.320, but the sampling uncertainty in TVD at 512 shots is not negligible, so the claim of 'significantly higher fidelities' cannot be assessed from the data as presented. The authors should report shot-noise confidence intervals, repeated calibration runs, or a statistical test comparing the two circuits.","section":"Section V, hardware demonstration"}],"minor_comments":[{"comment":"The angle-state correspondence for system S-8 lists states such as |2,1,1> and |3,0,1>, although the text states that modes M0 and M1 have only two basis states each; this makes the experimental specification ambiguous and needs correction.","section":"Appendix A"},{"comment":"The order of the gates in the product in Eq. (13) should be stated explicitly (left-to-right versus right-to-left application); currently the circuit order must be inferred from the figure, which is not sufficient for reproducibility.","section":"Eq. (13) and Fig. 2"},{"comment":"The abstract's 'up to 28% reduction' is in tension with the larger small-m reductions implied by Table II; the wording should be qualified as an asymptotic statement.","section":"Section IV, abstract"},{"comment":"The convention for CX(a,b) should be defined once and used consistently; the paper appears to switch between control-target and target-control readings, which contributes to the inconsistency of Eqs. (15)-(19).","section":"Global notation"}],"recommendation":"major_revision","confidential_remarks":"The manuscript seems to be an early arXiv version with several typographical and arithmetic errors. The central idea is potentially salvageable, but the derivation and cost analysis must be corrected before the claims can be evaluated. I do not see evidence of intentional misrepresentation, but the current text cannot be accepted as is."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe interesting idea here is that unary encoding leaves most of Hilbert space unused, so a Givens-style decomposition of the m-excitation unitary might only need controls on every other qubit. The paper works out the gate-count consequences cleanly—(8m-6) CNOTs plus one m-controlled rotation versus the standard (2m-1) CNOTs plus a (2m-1)-controlled rotation—and Tables I and II are arithmetically consistent. The hardware TVD numbers on H1-1 support the qualitative point that fewer CNOTs helps fidelity. Credit where due: the gate-count reduction is analytic, not fitted, and the paper explicitly notes the method is less general than the Givens constructions.\n\nThe problem is that the circuit in Eqs. 13-14 does not do what the paper says it does. Take S_j as defined: RTOF on {q0,q_{2j}} targeting q_{2j+1}, then CX(q_{2j+1},q_{2j}). For input |01>_j|0>_0 (q0=0,q_{2j}=0,q_{2j+1}=1), the RTOF does nothing and the CX flips q_{2j} to 1, giving |11>_j|0>_0. Equation 16 says the output is |01>_j|0>_0. That is not a phase subtlety; the bit values differ. Same for the first case of Eq. 17. This is load-bearing: for |g_m>, each S_j acts with q0=0, so U_m maps |g_m> to a state with |11> on modes 1..m-1. After the controlled RY and U_m^dagger, the circuit leaks into states like |10>_0|01>_1 instead of staying in span{|g_m>,|e_m>}. A concrete m=2 example starting from |g> yields a superposition involving |0>_0|1>_1, which should be untouched. So Eq. 12 does not implement Eq. 6 for the circuit as written; the claimed half-control reduction is for a circuit that performs a different operation. The stress-test note flagged Eqs. 18-19 as inconsistent; the root cause is wider—the S_j transformations in Eqs. 15-17 do not follow from Eq. 14.\n\nThe idea may be salvageable with a corrected S_j, but that is not in the manuscript. As is, the central construction fails on its own terms. I would not cite it and would not bring it to a reading group. For peer review: the gate-count target is legitimate, but the formal foundation is broken; I would desk reject this version rather than send it out, though I would be curious to see a corrected construction.","headline":"The half-control construction is internally inconsistent: the S_j gate does not match the claimed transformations, so the gate-count savings are for a circuit that does not implement the intended excitation unitary.","tokens_in":11831,"tokens_out":33229,"would_cite":false,"duration_ms":236598,"reading_group":"no","serious_thinker":"no","would_accept_peer_review":false},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["03.67.Ac","03.67.Lx"],"model":"deepseek-v4-flash","headline":"The paper claims that the $m$-excitation UVCC unitary can be implemented with $(8m-6)$ CNOT gates plus one $m$-controlled Y-rotation, roughly halving the entangling-gate count by dropping controls that only act on unphysical states.","keywords":["Unitary Vibrational Coupled-Cluster","direct qubit mapping","unary encoding","CNOT gate reduction","multi-controlled rotations","relative-phase Toffoli gate","quantum state preparation","vibrational structure calculations"],"falsifier":"On a noiseless simulator, run the full $\\hat U_m(\\theta)$ circuit of Eq. 12 on every physical basis state for a small case such as $m=3$ and check that states other than $|g_3\\rangle$ and $|e_3\\rangle$ return to themselves exactly, including phases; any residual amplitude or phase on a spectator state, or any deviation from Eqs. 8-10 on the excitation pair, would falsify the half-control construction.","tokens_in":10813,"feed_emoji":"⚛️","tokens_out":12409,"duration_ms":107505,"temperature":0.7,"pith_summary":"The paper aims to show that the unitary vibrational coupled-cluster (UVCC) excitation operator, when vibrational modes are encoded in the direct (unary) qubit mapping, can be implemented with far fewer entangling gates than the standard constructions. The key move is to notice that half of the qubit controls in the Trotterized ansatz only ever act on states outside the physical subspace, so they can be dropped. The claimed result is concrete: an $m$-excitation unitary costing $(8m-6)$ CNOT gates plus one $m$-controlled Y-rotation, versus $2(2m-1)$ CNOTs plus a $(2m-1)$-controlled rotation for the usual Givens-rotation approach. That yields up to a 50% reduction in CNOT count against the standard baselines and a 28% reduction against the best ancilla-assisted decomposition. On trapped-ion hardware experiments with 6- and 8-qubit systems, the method reports lower total variation distance than the Givens baseline, backing the gate-count savings with measured fidelity gains.","feed_headline":"Half the controls, up to 50% fewer CNOT gates in chemistry circuits","feed_subtitle":"A new decomposition exploits unoccupied qubit states to shorten vibrational quantum state preparation.","key_machinery":"The load-bearing object is the $\\hat S_j$ gate defined in Eq. 14: a three-qubit gate made of one CNOT and one relative-phase Toffoli acting on the first qubit $q_0$ and the two qubits of mode $j$. It maps both physical occupation patterns $|01\\rangle_j$ and $|10\\rangle_j$ onto the marker state $|11\\rangle_j$ (Eqs. 15-17), so that the Y-rotation needs to be controlled by only one qubit per mode, $q_{2j+1}$, rather than by every qubit. The relative-phase Toffoli's 3-CNOT decomposition is what pins the CNOT count to $(8m-6)$, because each of the $m-1$ $\\hat S_j$ gates costs three CNOTs plus one CX.","core_discovery":"On the paper's own terms, the central discovery is that the unary encoding's forbidden $|11\\rangle$ states are not merely wasted space but a resource: because direct mapping uses only single-occupancy states of each mode, the excitation unitary only needs to act correctly on the physical subspace. The paper therefore replaces the standard decomposition, which controls the Y-rotation on all $2m-1$ relevant qubits, with Eq. 12, which controls on only every other qubit. Each mode's two allowed patterns $|01\\rangle$ and $|10\\rangle$ are mapped by the $\\hat S_j$ gate into a common $|11\\rangle$ pattern, up to phase, so the controlled rotation sees a single marker qubit. The paper claims this implements exactly the same $\\hat U_m(\\theta)$ on the physical subspace as Eqs. 8-10, with circuit cost $(8m-6)$ CNOTs plus one $m$-controlled rotation, and confirms numerically and experimentally that the shorter circuits prepare UVCCSDT states with lower total variation distance.","pith_inferences":["The same 'forbidden subspace' principle might be exploitable in other sparse encodings of bosonic or fermionic modes, though the paper notes the construction does not transfer straightforwardly to electronic structure.","Because the paper reports only total variation distance and not circuit depth, a natural next test is to scale to $m=4$ or larger and check whether the fidelity gap over the Givens baseline grows with the $(8m-6)$ gate-count advantage.","One could also insert the half-control excitation into quantum phase estimation trial-state preparation, not just VQE; the paper's argument applies to any use of the UVCC unitary."],"forward_implications":["For $m=3$ and $m=4$, the method's CNOT counts of 25 and 42 beat the Givens-rotation counts of 41 and 142 under the exponential $2n$-CNOT decomposition of controlled rotations.","For the UVCCSDT ansatz on 6 and 8 qubits, the paper reports total variation distances of 0.050 and 0.240 versus 0.136 and 0.320 for the Givens baseline, so the gate-count reduction carries through to measured fidelity.","Compared with the ancilla-assisted $C^nX$ decomposition used as a practical baseline, the method retains a 28% CNOT reduction, so the advantage is not simply an artifact of a weak baseline.","In fault-tolerant settings, halving the controls shifts cost into an $O(m)$ Toffoli overhead, so the final advantage depends on how $n$-controlled rotations are decomposed."],"supporting_citations":[{"why":"Establishes the direct qubit mapping for vibrational modes and the UVCC circuit context in which the redundancy argument applies.","marker":"[6]"},{"why":"Introduces the UVCC ansatz and its Trotterized excitation unitaries that the paper rewrites.","marker":"[33]"},{"why":"Provides the exponential decomposition baseline of $4m(2m-1)$ CNOTs and the $2n$-CNOT decomposition of $n$-controlled rotations used in Table I.","marker":"[42]"},{"why":"Gives the Givens-rotation implementation of the excitation unitary whose $2(2m-1)$ CNOTs and $(2m-1)$-controlled rotation is the main baseline.","marker":"[43,44]"},{"why":"Supplies the single- and double-excitation Givens-rotation circuits used as the hardware comparison baseline.","marker":"[40]"},{"why":"Supplies the 3-CNOT relative-phase Toffoli decomposition used inside each $\\hat S_j$ gate to keep the count at $(8m-6)$.","marker":"[45]"},{"why":"Provides the ancilla-assisted $C^nX$ decomposition with $A=2, B=3$ against which the paper reports a 28% CNOT reduction.","marker":"[41]"}],"fun_headline_variants":["Redundant qubit states slash control qubits in UVCC circuits","Exploiting unary encoding redundancies to halve CNOT counts in chemistry","Hilbert-space redundancy cuts UVCC entangling gates by up to 50%","Qubit-space redundancy enables half the control qubits in UVCC state prep"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the single-mode transformation rules in Eqs. 15-17 force the full circuit to act as identity on every allowed multi-mode state outside the two-state excitation subspace, even though the paper does not derive that behavior for arbitrary allowed inputs.","fun_headline_variants_meta":{"raw":{"variants":["Redundant qubit states slash control qubits in UVCC circuits","Exploiting unary encoding redundancies to halve CNOT counts in chemistry","Hilbert-space redundancy cuts UVCC entangling gates by up to 50%","Qubit-space redundancy enables half the control qubits in UVCC state prep"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001107,"raw_usage":{"total_tokens":4618,"prompt_tokens":956,"completion_tokens":3662,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":572,"completion_tokens_details":{"reasoning_tokens":3577}},"tokens_in":572,"tokens_out":3662,"duration_ms":26292,"temperature":1.0,"reasoning_tokens":3577,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T21:55:32.620616+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"On a noiseless simulator, run the full $\\hat U_m(\\theta)$ circuit of Eq. 12 on every physical basis state for a small case such as $m=3$ and check that states other than $|g_3\\rangle$ and $|e_3\\rangle$ return to themselves exactly, including phases; any residual amplitude or phase on a spectator state, or any deviation from Eqs. 8-10 on the excitation pair, would falsify the half-control construction.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes the direct qubit mapping for vibrational modes and the UVCC circuit context in which the redundancy argument applies."},{"cited_title":"McArdle, A","cited_arxiv_id":null,"evidence_quote":"Introduces the UVCC ansatz and its Trotterized excitation unitaries that the paper rewrites."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the exponential decomposition baseline of $4m(2m-1)$ CNOTs and the $2n$-CNOT decomposition of $n$-controlled rotations used in Table I."},{"cited_title":"Maslov, On the advantages of using relative phase toffolis with an application to multiple control toffoli optimization, Physical Review A 93 (2015)","cited_arxiv_id":null,"evidence_quote":"Supplies the 3-CNOT relative-phase Toffoli decomposition used inside each $\\hat S_j$ gate to keep the count at $(8m-6)$."}],"review_version":1}