{"id":"d0870cc4-4d1f-434f-bdc7-5843d3851022","arxiv_id":"2608.11042","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The minimal ancilla dimension for optimal quantum channel estimation equals the minimum rank of a maximizer of a concave variational functional.","lead":"The paper characterizes how large an entangled ancilla must be to reach the best possible precision in quantum channel parameter estimation. The main result reduces this to a rank-constrained optimization and gives examples where no ancilla is needed and where the full ancilla dimension is required.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 3(i) is proved only under an unstated attainability condition; as stated, the fixed-measure-and-prepare zero-gap claim is not established.","rationale":"The reader's weakest-assumption analysis correctly identifies the exact place where Theorem 3(i) overreaches: the proof silently adds an attainability condition (Eq. 35) that is neither stated nor derived from the fixed measure-and-prepare hypothesis. This is the most load-bearing concern because the abstract and conclusions advertise fixed measure-and-prepare channels as a sufficient class with k*=1, and an uncorrected reader could take an unsupported claim as one of the paper's main results. The central contribution, Theorem 2 and the k-ancilla variational formula, does not depend on Theorem 3(i); the rank characterization is supported by a self-contained argument that I found no concrete flaw in. The conditional bound of Theorem 4 is explicitly conditional and not central. I therefore do not change the reader's conditional verdict: the paper needs revision of Theorem 3(i) (adding the missing assumption or proving the theorem under the stated one), but the main variational and rank-based conclusions can stand.","tokens_in":12890,"tokens_out":31333,"duration_ms":297003,"concrete_test":"Independently re-derive Theorem 3(i) from the stated fixed-POVM condition (33) without invoking Eq. (35). If the inequality J(ξ_{i,θ}) ≤ J_1 cannot be justified from (33) alone, the proof is incomplete as stated. As a numerical probe, take the qubit family M_1=diag(2/3,1/3), M_2=diag(1/3,2/3), ξ_{1,θ}=|ψ_θ⟩⟨ψ_θ|, ξ_{2,θ}=|ψ_{θ+π}⟩⟨ψ_{θ+π}| with |ψ_θ⟩=cos(θ/2)|0⟩+sin(θ/2)|1⟩, and compute J_1=max_ρ J(Φ_θ(ρ)) and J_2=max_σ̃ J((id⊗Φ_θ)(σ̃)) at θ=0. This POVM is full rank, so condition (35) fails; if J_2>J_1 the theorem statement is false, while J_2≤J_1 still leaves the proof gap unresolved.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The proof of Theorem 3(i) in Sec. V adds, after Eq. (35), the condition that for every i there is a θ-independent ρ_i with Φ_θ(ρ_i)=ξ_{i,θ}. This condition is not stated in the theorem and is not implied by the fixed POVM decomposition (33): the states ξ_{i,θ} in a measure-and-prepare representation need not be channel outputs for any fixed input. The convexity argument yields J(ω_θ) ≤ Σ_i p_i J(ξ_{i,θ}) with θ-independent p_i; to conclude this is ≤ J_1 one must know each J(ξ_{i,θ}) ≤ max_ρ J(Φ_θ(ρ)), and the extra attainability condition is exactly what supplies that. Without it, a prepared state can in principle carry more Fisher information than any unassisted output of Φ_θ, so the zero-gap conclusion does not follow. The abstract and conclusions repeat the stronger statement that a fixed measure-and-prepare representation alone suffices for k*=1. Thus one of the paper's two advertised zero-gap criteria is overstated. The central rank characterization of Theorem 2 and the variational formula of Theorem 1 appear sound; this concern is specific to Theorem 3(i) and its repetition in the narrative.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the minimal ancilla dimension needed to attain the maximum SLD Fisher information in one-parameter quantum channel estimation. It defines a k-ancilla Fisher information J_k, proves a variational formula J_k = max_{\\sigma \\in S_k(H)} g(\\sigma) based on the Fujiwara–Imai functional, and derives a rank characterization k^* = rank(\\sigma^*) where \\sigma^* is a minimum-rank global maximizer of the concave functional g. It then gives sufficient conditions for k^* = 1, namely a fixed measure-and-prepare representation and a horizontality condition, and derives a conditional lower bound on incremental gains J_{k+1} - J_k. The results are illustrated on unitary channels, qubit depolarizing, and amplitude damping channels.","tokens_in":13093,"tokens_out":16602,"duration_ms":147344,"significance":"The paper offers a clean and potentially useful reformulation of the ancilla-resource problem: the optimal ancilla dimension is determined entirely by the rank structure of maximizers of a concave functional. Theorem 1 and Theorem 2 appear logically sound and are built on the established Fujiwara–Imai variational formula. The examples are instructive and the writing is generally clear. The main defect is that one of the two advertised zero-gap criteria, Theorem 3(i), is not proven as stated, because the proof introduces an unstated attainability condition. Since the abstract and conclusions repeat the stronger claim, this is a load-bearing issue that must be fixed before publication.","major_comments":[{"comment":"The theorem statement claims that a fixed measure-and-prepare representation (a θ-independent POVM {M_i} and states ξ_{i,θ}) suffices for k^* = 1. However, the proof adds the additional assumption that for every i there exists a θ-independent state ρ_i with Φ_θ(ρ_i) = ξ_{i,θ}. This condition is not stated in the theorem and is not implied by the fixed-POVM representation alone: the states ξ_{i,θ} in a measure-and-prepare representation need not belong to the image of Φ_θ. The step J(ξ_{i,θ}) = J(Φ_θ(ρ_i)) ≤ J_1 at Eq. (40) relies exactly on this added attainability assumption. Without it, the convexity argument only gives J(ω_θ) ≤ Σ_i p_i J(ξ_{i,θ}), which can exceed J_1 because a prepared state can in principle carry more Fisher information than any unassisted channel output. Consequently, the theorem as stated is not established, and the abstract and conclusions repeat the stronger claim that a fixed measure-and-prepare representation alone suffices for k^* = 1. Please either add the attainability condition to the statement, prove that it follows from the fixed-POVM representation, or provide a counterexample and adjust the narrative accordingly.","section":"Sec. V, Theorem 3(i) and proof after Eq. (35)"}],"minor_comments":[{"comment":"The phrase 'input states of rank at most k' is imprecise: in Eq. (14) the input is a state on H_k ⊗ H whose rank is not constrained to be at most k. The rank constraint applies to the reduced state σ = Tr_1 |ψ⟩⟨ψ| after the reduction to pure inputs. Please clarify this wording.","section":"Abstract"},{"comment":"In the sentence 'if ψ⟩= Pk i= 1√λi |ei⟩ ⊗ |fi⟩, is the Schmidt decomposition', a backslash before “psi” is missing and the equation is typeset incorrectly.","section":"Sec. III, proof of Theorem 1, step (iii)"},{"comment":"The notation 'p_i, τ_i' in the display after Eq. (37) is confusing; the comma appears to be a typographical artifact. Please use the standard notation p_i τ_i ⊗ ξ_{i,θ}.","section":"Sec. V, Eq. (37)"},{"comment":"The paragraph distinguishing condition (i) from entanglement-breaking channels is useful, but it does not address the additional attainability assumption in the proof. Please reconcile this discussion with the revised theorem statement.","section":"Sec. V, discussion after Theorem 3(i)"}],"recommendation":"major_revision","confidential_remarks":"The central rank characterization (Theorem 2) and the variational formula (Theorem 1) appear sound and are the main contribution. The obstacle to acceptance is the unproven status of Theorem 3(i) as stated; if the theorem is false in its current form, that would be a serious flaw, but the issue seems fixable by explicitly adding the attainability condition to the theorem statement, which would still leave a valid sufficient condition. The authors should also revisit the abstract and conclusions to ensure they do not overstate the fixed-measure-and-prepare criterion."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The core of this paper is worth reading. The k-ancilla Fisher information J_k and the variational formula J_k = max over rank-k states of the concave functional g (Theorem 1) cleanly extend Fujiwara and Imai to rank-constrained probes. Theorem 2, identifying the minimal ancilla dimension k* with the rank of a minimum-rank maximizer of g, is a neat and correct observation, and the concavity argument is solid. The examples (unitary, depolarizing, amplitude damping) are illustrative and the conclusions for them are plausible. If you work on channel estimation or quantum metrology resource theory, the rank characterization is a useful tool. The soft spot is exactly where the stress-test points. Theorem 3(i) states that a fixed measure-and-prepare representation suffices for k* = 1, but the proof introduces, after Eq. (35), the extra assumption that each prepared state xi_{i,theta} is itself realizable as Phi_theta(rho_i) for a theta-independent input rho_i. That assumption is not in the theorem statement and does not follow from the fixed POVM alone. Without it, the convexity bound leaves J(xi_{i,theta}) uncontrolled, so the zero-gap conclusion does not follow. The abstract and conclusions repeat the stronger statement, and Appendix A does not repair it; the appendix justifies the theta-independent decomposition of the output, but not the attainability of the xi_{i,theta}. This is a genuine defect, but it is localized: Theorems 1, 2, 3(ii), and the conditional Theorem 4 are fine. There is also a minor typo in step (iii) of Theorem 1's proof, but that is cosmetic. Who is this for: researchers in quantum channel estimation, convex optimization approaches to Fisher information, and entanglement as a resource in metrology. A serious referee should be assigned, with the clear request to fix Theorem 3(i) by either adding the attainability condition to the statement or proving the result under the stated assumptions. The paper is honest in its framing and builds on cited work rather than re-deriving it; the main characterization is new and useful. With that theorem corrected, this would be a solid contribution.","headline":"A clean rank-based characterization of ancilla dimension, with one zero-gap theorem that is proved only under an unstated attainability assumption.","tokens_in":680,"tokens_out":1147,"would_cite":true,"duration_ms":24671,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81P50","81P47","90C25"],"pacs":["03.67.-a","06.20.-f"],"model":"deepseek-v4-flash","headline":"The minimal ancilla dimension needed to saturate the quantum channel Fisher information is exactly the minimum rank among the maximizers of a concave functional over input states.","keywords":["quantum channel estimation","ancilla dimension","entanglement depth","Fisher information","symmetric logarithmic derivative","rank-constrained optimization","measure-and-prepare channel","quantum metrology"],"falsifier":"For the qubit depolarizing family $\\Phi_\\theta(\\rho)=(1-\\theta)\\rho+\\frac{\\theta}{2}\\mathbb{1}$ at $\\theta=0.1$, compute $\\max_{\\operatorname{rank}\\sigma\\le 1}g(\\sigma)$ and $g(\\mathbb{1}/2)$; the paper predicts the first is strictly smaller and equals $J_1<J_2=J_d$, so agreement would falsify the rank formula. More generally, any channel family whose minimum-rank maximizer has rank $r$ should show a strict jump in $J_k$ at $k=r$; absence of that jump would falsify Theorem 2.","tokens_in":12654,"feed_emoji":"🎯","tokens_out":8876,"duration_ms":75557,"temperature":0.7,"pith_summary":"The paper asks how much ancillary entanglement a probe needs to estimate a parameter encoded in a quantum channel at the ultimate precision allowed by quantum mechanics. Its central answer is that the minimal ancilla dimension $k^*$ equals the minimum rank of any maximizer of a concave functional $g(\\sigma)$ over input density matrices: $k^* = \\operatorname{rank}\\sigma^*$, where $\\sigma^*$ is a lowest-rank global maximizer of $g$. This reduces an optimization over bipartite probe states to a rank-constrained convex problem, and it yields a monotone sequence $J_1 \\le J_2 \\le \\cdots \\le J_d$ of achievable Fisher informations parametrized by ancilla dimension. The paper also gives sufficient conditions—fixed measure-and-prepare channel families and \"horizontal\" generator curves—under which no ancilla is needed, and it illustrates channels where the full ancilla dimension is required (qubit depolarizing) and where none is (unitary and amplitude-damping channels).","feed_headline":"Smallest ancilla dimension equals rank of optimal probe state","feed_subtitle":"If the optimal probe state is pure, a one-dimensional ancilla suffices; if it is full-rank, the full ancilla is required.","key_machinery":"The load-bearing identity is the variational formula $J_k(\\Phi_\\theta)=\\max_{\\sigma\\in S_k(\\mathcal{H})} g(\\sigma)$, where $g(\\sigma)=\\min_{X\\in u(q)} f_{\\theta_0}(\\sigma,X)$ with $f_{\\theta_0}(\\sigma,X)=\\sum_k \\operatorname{Tr}[\\sigma \\dot B_k^\\dagger \\dot B_k]-\\operatorname{Tr}[iC(\\sigma)X]+\\operatorname{Tr}[XG(\\sigma)X]$, and $G(\\sigma)_{k\\ell}=\\operatorname{Tr}[\\sigma B_k^\\dagger B_\\ell]$, $C(\\sigma)_{\\ell k}=\\operatorname{Tr}[\\sigma(\\dot B_k^\\dagger B_\\ell - B_k^\\dagger \\dot B_\\ell)]$ for a reference Kraus generator $B(\\theta)$. The minimization over $X$ solves a continuous Lyapunov equation with a unique Hermitian solution when $G(\\sigma)$ is positive definite. The function $g$ is concave, so its maximizer set is convex and compact; Theorem 2 uses that structure to identify $k^*$ as the minimum rank among maximizers. The rank constraint $S_k(\\mathcal{H})$ inherits all information about the ancilla because every pure state on $\\mathcal{H}_k\\otimes\\mathcal{H}$ reduces to a state of rank at most $k$, and every such state has a purification.","core_discovery":"On the paper's own terms, the discovery is a rank formula for entanglement depth in quantum channel estimation. For a one-parameter family $\\Phi_\\theta$, define the $k$-ancilla Fisher information $J_k(\\Phi_\\theta)$ as the best SLD Fisher information attainable with a $k$-dimensional ancilla. Theorem 1 states $J_k = \\max_{\\operatorname{rank}\\sigma \\le k} g(\\sigma)$, where $g$ is obtained by minimizing the channel-generator cost $f(\\sigma,X)$ over Hermitian generator corrections $X$; $g$ is concave and continuous, so its maximizer set $\\mathcal{M}$ is nonempty, convex, and compact. Theorem 2 then shows $k^* = \\min_{\\sigma\\in\\mathcal{M}} \\operatorname{rank}\\sigma$. In words: the minimum ancilla dimension sufficient to saturate the fully extended Fisher information is exactly the smallest rank among optimal reduced probe states, so entanglement depth is a rank property of a concave variational problem.","pith_inferences":["Editorial extension: Theorem 2 suggests a practical numerical recipe—evaluate $g$ via the Lyapunov solution and increase the allowed rank until the maximum stops changing; the first saturating rank is $k^*$. The paper does not spell out this algorithm.","Editorial extension: The proof distinction between fixed measure-and-prepare and merely entanglement-breaking families points to a concrete test: a family of entanglement-breaking channels whose Holevo POVM varies with $\\theta$ may show $J_d>J_1$, with the classical Fisher term of the varying probabilities being the obstruction.","Editorial extension: Because rank is a discontinuous function of the state, small perturbations of a channel family can make $k^*$ jump sharply even when $J_d$ changes little, so minimal-ancilla metrology could be fragile to model error.","Editorial extension: In the multiparameter case, the rank characterization would likely make the required ancilla dimension depend on the chosen estimation direction, since the matrix-valued Fisher information has no single scalar optimum; the paper lists multiparameter extension as open."],"forward_implications":["For any one-parameter channel family, the whole sequence $J_1\\le J_2\\le\\cdots\\le J_d$ is controlled by the single concave function $g$, so finding the optimal ancilla dimension becomes a finite-dimensional rank-constrained convex optimization problem.","The optimal ancilla dimension is determined by purity of the optimizer: if a minimum-rank maximizer of $g$ is pure then $k^*=1$ and entanglement buys nothing; if all maximizers are full-rank then the full ancilla $k^*=d$ is required.","Channel families admitting a fixed measure-and-prepare representation with the same POVM for all $\\theta$, and families with horizontal generator curves, have zero gap $J_d=J_1$ under the stated attainability condition.","The incremental bound $J_{k+1}-J_k \\ge t(g(\\tau)-J_k)$ quantifies the marginal value of an extra ancilla dimension when successive optimal states are convex mixtures.","The worked examples give concrete resource answers: unitary channels need $k^*=1$, qubit depolarizing needs $k^*=2$, and amplitude damping needs $k^*=1$."],"supporting_citations":[{"why":"Supplies the variational formula for the fully extended channel Fisher information that Theorem 1 adapts to $k$-dimensional ancillas, along with the reference-generator technique used throughout.","marker":"[8]"},{"why":"Defines the quantum channel identification problem that frames the paper's objective of optimizing input probes for channel parameter estimation.","marker":"[6]"},{"why":"Provides the measure-and-prepare (Holevo) representation that Theorem 3(i) uses as the sufficient condition for $k^*=1$.","marker":"[22]"},{"why":"Supplies the extended convexity inequality for the SLD Fisher information, which the appendix uses to show why $\\theta$-dependent Holevo decompositions do not automatically give the zero gap.","marker":"[23]"}],"fun_headline_variants":["Optimal probe rank sets needed ancilla dimension","Ancilla size equals rank of optimal probe state","Quantum estimation: ancilla depth = probe rank","Probe rank dictates minimal ancilla in channel estimation","Smallest ancilla is probe rank for optimal estimation"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that, for a channel family written in fixed measure-and-prepare form, each prepared state $\\xi_{i,\\theta}$ is itself the output $\\Phi_\\theta(\\rho_i)$ of a single $\\theta$-independent input state $\\rho_i$; this attainability condition is added in the proof of Theorem 3(i) after Eq. (35) and is not part of the theorem's stated assumptions. If it fails, the convexity argument for $J_d\\le J_1$ does not go through.","fun_headline_variants_meta":{"raw":{"variants":["Optimal probe rank sets needed ancilla dimension","Ancilla size equals rank of optimal probe state","Quantum estimation: ancilla depth = probe rank","Probe rank dictates minimal ancilla in channel estimation","Smallest ancilla is probe rank for optimal estimation"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00053,"raw_usage":{"total_tokens":2548,"prompt_tokens":937,"completion_tokens":1611,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":553,"completion_tokens_details":{"reasoning_tokens":1538}},"tokens_in":553,"tokens_out":1611,"duration_ms":10409,"temperature":1.0,"reasoning_tokens":1538,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T11:28:17.262490+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For the qubit depolarizing family $\\Phi_\\theta(\\rho)=(1-\\theta)\\rho+\\frac{\\theta}{2}\\mathbb{1}$ at $\\theta=0.1$, compute $\\max_{\\operatorname{rank}\\sigma\\le 1}g(\\sigma)$ and $g(\\mathbb{1}/2)$; the paper predicts the first is strictly smaller and equals $J_1<J_2=J_d$, so agreement would falsify the rank formula. More generally, any channel family whose minimum-rank maximizer has rank $r$ should show a strict jump in $J_k$ at $k=r$; absence of that jump would falsify Theorem 2.","supporting_citations":[{"cited_title":"A fibre bundle over manifolds of quantum channels and its application to quantum statistics,","cited_arxiv_id":null,"evidence_quote":"Supplies the variational formula for the fully extended channel Fisher information that Theorem 1 adapts to $k$-dimensional ancillas, along with the reference-generator technique used throughout."},{"cited_title":"Quantum channel identification problem,","cited_arxiv_id":null,"evidence_quote":"Defines the quantum channel identification problem that frames the paper's objective of optimizing input probes for channel parameter estimation."},{"cited_title":"Quantum coding theorems,","cited_arxiv_id":null,"evidence_quote":"Provides the measure-and-prepare (Holevo) representation that Theorem 3(i) uses as the sufficient condition for $k^*=1$."},{"cited_title":"Extended convexity of quan- tum fisher information in quantum metrology,","cited_arxiv_id":null,"evidence_quote":"Supplies the extended convexity inequality for the SLD Fisher information, which the appendix uses to show why $\\theta$-dependent Holevo decompositions do not automatically give the zero gap."}],"review_version":1}