{"id":"c3060f7b-a717-4a4f-81d3-c7400b46c259","arxiv_id":"2501.12891","paper_version":1,"verdict":"REJECT","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The deimaginarity cost of a quantum state equals its regularized relative entropy of imaginarity.","lead":"This paper defines the deimaginarity cost of a quantum state as the minimum randomness needed to make many copies of the state real by random unitaries. It claims this cost equals the regularized relative entropy of imaginarity, giving that quantity an operational meaning.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Direct part's typical subspace is generally not real: for rho=|+i>, the averaged output under a real-orthogonal twirl preserving H_{n,delta} is rho itself, so the construction cannot produce the claimed real state.","rationale":"The reader's REJECT verdict is supported, but the most decisive defect is not merely the unproved twirl ensemble. Even granting such an ensemble, the direct part's output is I_{n,delta}/D_{n,delta}, which is real only if H_{n,delta} is invariant under transposition. The standard typical subspace of an arbitrary rho is spanned by its eigenvectors; for rho=|+i> these are complex, so H_{n,delta} is not real. The proof's claim that E(W) is real is therefore false in exactly the cases where imaginarity is nonzero. The missing normalization in T and the additivity assertion are secondary and repairable; the typical-subspace reality issue is internal and load-bearing. Without a real invariant subspace or a symmetrized construction over H_{n,delta} plus its conjugate, the randomized real-orthogonal unitaries cannot both twirl on H_{n,delta} and produce a real output. The theorem may still be true by a different argument, but the manuscript's proof does not establish it.","tokens_in":9209,"tokens_out":31331,"duration_ms":339134,"concrete_test":"Instantiate the direct construction with rho = |+i><+i| and n=2 with delta small. Compute Pi = |v><v| with v = |+i>^{otimes n}. Verify that every real orthogonal U with U Pi U^dagger = Pi satisfies U rho^{otimes n} U^dagger = rho^{otimes n}, so the ensemble average E(W) equals rho^{otimes n}. Then compute the trace distance from rho^{otimes n} to the nearest real state; for n=2 it is 1, and it does not vanish as n grows. This settles whether the claimed E(W) is real and whether the direct part can output a free state.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The direct part assumes an ensemble of covariant-free (real orthogonal) unitaries such that for every theta supported on the delta-typical subspace H_{n,delta}, the average U theta U^dagger equals I_{n,delta}/D_{n,delta}. But H_{n,delta} is built from eigenvectors of rho in its eigenbasis, which are generally complex in the fixed reference basis. Real orthogonal unitaries that preserve H_{n,delta} also preserve its complex conjugate; when H_{n,delta} is not invariant under transposition, the operator I_{n,delta}/D_{n,delta} is not a real matrix. The proof then asserts 'E(W) is real' after Eq. (A.19), but E(W) = (mu/D) I_{n,delta}, which is real only if the typical subspace is real. Example: rho = |+i><+i| and n=1 gives H_{1,delta} = span{|+i>}, D=1; every real orthogonal U preserving this line acts by a phase on |+i>, so U rho U^dagger = rho and the average is rho, not a real state. The same obstruction persists for rho^{otimes n} for every n. Thus the achievability construction (A.16)-(A.19) outputs the maximally mixed state on a complex typical subspace, not a free state. This is a concrete gap in the proof of C_d(rho) <= I_r^infty(rho), independent of whether a twirl ensemble exists.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces a task called deimaginarity, in which n copies of a quantum state ρ are mapped close to a real state by applying a randomly chosen covariant-free (real orthogonal) unitary, and the cost C_d(ρ) is the minimum asymptotic rate of randomness needed. The main result, Theorem 2, states that C_d(ρ) equals the regularized relative entropy of imaginarity I_r^∞(ρ) = lim_{n→∞} (1/n) I_r(ρ^⊗n). The proof is given in Appendix A: a converse part uses entropy inequalities and Fannes' inequality, and a direct part uses typical subspaces, an assumed twirl ensemble, and an operator Chernoff bound.","tokens_in":9459,"tokens_out":23308,"duration_ms":230814,"significance":"If Theorem 2 were established, it would provide a clean operational interpretation of the regularized relative entropy of imaginarity as a randomness cost, analogous to results for coherence and entanglement erasure. The converse part follows a standard template and contains the right entropy estimates modulo a normalization slip. The paper also correctly identifies the free operations and the covariance condition. However, the direct part relies on an unproved and generally unavailable twirl assumption and contains a false assertion that the constructed channel output is real; this invalidates the proof of achievability. The false claim that the relative entropy of imaginarity is additive is an additional factual error, although the theorem itself uses the regularized quantity.","major_comments":[{"comment":"The direct part assumes the existence of an ensemble {p(dU), U} of covariant-free unitaries such that for every state θ on the typical subspace H_{n,δ}, ∫ p(dU) U θ U† = I_{n,δ}/D_{n,δ}. This is a strong twirl (2-design) property on a subspace that is generally complex, and no proof or construction is given. The sentence 'Assume ...' is not an argument. Since the operator Chernoff step and the final trace-norm estimate depend on this ensemble, the achievability claim C_d(ρ) ≤ I_r^∞(ρ) is unsupported without a proof of existence or an alternative construction.","section":"§VI A 2, Eq. (A.16)"},{"comment":"The assertion 'E(W) is real' is false. From Eq. (A.16), E(W) = (μ_{n,δ}/D_{n,δ}) Π_{n,δ}, where Π_{n,δ} is the projector onto the typical subspace. The projector is a real matrix only if H_{n,δ} is invariant under transposition, which is not generally the case. For ρ = |+i⟩⟨+i| and n=1, H_{1,δ}=span{|+i⟩}, D=1, and every real orthogonal U preserving this line acts by a phase on |+i⟩; the averaged output is then ρ itself, not a real state. Hence the final state (1/μ_{n,δ}) E(W) in the construction is not free, and the direct part does not produce the claimed real state.","section":"§VI A 2, after Eq. (A.19)"},{"comment":"Even if a twirl ensemble existed, the construction targets the maximally mixed state on H_{n,δ}. For non-real typical subspaces this state is not a real density matrix. The example ρ=|+i⟩⟨+i| shows the obstruction persists for all n, since the typical subspace is spanned by |+i⟩^{⊗n}, and the real-orthogonal stabilizer acts by phases. The proof therefore cannot be repaired by a minor adjustment of parameters; the achievability strategy itself must be replaced.","section":"§VI A 2, Eq. (A.16)–(A.19)"},{"comment":"The displayed inequality S(O_n(ρ⊗n)) ≥ (1/2^{nR}) Σ_k S(Θ(O_k ρ⊗n O_k†)) is not a consequence of concavity. For ρ=|+i⟩⟨+i|, n=1, m=1, O=I, the left side is S(ρ)=0 while the right side is S(Θ(ρ))=1. The intended argument should combine Eq. (A.7) with concavity applied to Θ(O_n(ρ⊗n)); this yields an additional entropy term −nη(2ε)log d. The final bound survives after that correction, but the chain as written is mathematically false.","section":"§VI A 1, Eq. (A.8)"},{"comment":"The isometry T is defined as T = (1/2^{nR}) Σ_{k=1}^{2^{nR}} |k⟩_E ⊗ O_k. With this normalization T†T = 2^{-nR} I, so T is not an isometry and the state |ψ_n⟩ is not normalized. The correct coefficient is 1/√(2^{nR}). This is a local error, but it affects the entropy estimates in Eq. (A.6).","section":"§VI A 1, Eq. (A.6)"},{"comment":"The statement 'The REI is additive' is false as written. For ρ=|+i⟩⟨+i|, I_r(ρ)=S(I/2)-S(ρ)=1, while I_r(ρ⊗2)=S(Re(ρ⊗2))-S(ρ⊗2)=1 because Re(ρ⊗2) has eigenvalues 1/2,1/2. Thus I_r(ρ⊗2) ≠ 2I_r(ρ). This false claim is not used in the proof of Theorem 2 (which uses the regularized quantity), but it should be corrected or removed.","section":"§II B"}],"minor_comments":[{"comment":"The title contains a typo: 'Deimginarity' should be 'Deimaginarity'.","section":"Title"},{"comment":"The text uses 'ortthogonal' and 'O_kΘ(·)O_k†' where the adjoint should be the transpose for real orthogonal operators; these are notational inconsistencies.","section":"§III, Definition 1"},{"comment":"The lower bound on λ_{n,δ} should be stated as an inequality for the minimum nonzero eigenvalue; the current text says 'the minimum nonzero eigenvalue ... is bounded by' without specifying direction, and the denominator D should be explicitly the dimension of H_{n,δ}.","section":"§VI A 2, Eq. (A.18)"},{"comment":"The notation N = 2^{I_r(ρ⊗n)+3nδ} should be understood as an integer ceiling; as written it is not an integer in general.","section":"§VI A 2"},{"comment":"The η(2ε) term should be η(2ε)log d multiplied by n, since the Fannes inequality is applied on the n-copy Hilbert space of dimension d^n; the displayed formula is consistent with this, but the text should make the dimension explicit.","section":"§VI A 1, Eq. (A.11)"}],"recommendation":"reject","confidential_remarks":"The paper is very short and the main proof is in a sketchy appendix. The direct part is not a minor gap: the asserted reality of the constructed output is false for the paper's own construction, and the twirl ensemble is assumed without proof. Even though the theorem may be true, the submitted proof does not establish it. I would encourage the authors to rework the achievability argument and to correct the additivity claim before resubmission."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The claimed result—that deimaginarity cost equals the regularized relative entropy of imaginarity—is a natural operational interpretation, and if a correct proof existed it would be a useful addition to the imaginarity literature. But the manuscript as written does not establish it. The errors are load-bearing, not cosmetic.\n\nWhat's good: the task is a sensible adaptation of the Groisman-Popescu-Winter / Wakakuwa erasure template, and the single-copy example (equal mixture of I and sigma_z) is a nice illustration. The converse part follows the standard converse structure, and the paper is clearly organized.\n\nThe problems are in the proof. First, the claim that the relative entropy of imaginarity is additive is false. For rho = |+i><+i|, I_r(rho) = 1, but I_r(rho^{otimes n}) = 1 for all n, so the regularized value is 0, not n. Second, the concavity step in (A.8)-(A.10) does not go through. Concavity gives S(On(rho)) >= average S(Ok rho Ok^T), but you cannot insert Theta into the argument without an additional step; Lemma 6 only relates S(Theta(O rho O^T)) to S(Theta(rho)), not to S(O rho O^T). Third, the isometry T is missing a 1/sqrt(2^{nR}) normalization, which shifts the entropy bound. Fourth, the direct part assumes a twirl ensemble on the delta-typical subspace with no proof. The stress-test note is right: for rho = |+i>, the typical subspace is not invariant under transposition, so a real-orthogonal twirl preserving it cannot average to the maximally mixed state on that subspace; the construction outputs rho itself, not a real state. That is a concrete counterexample to the achievability argument.\n\nThese are not minor typos. Both halves of the theorem are unsupported as written. The underlying idea is worth pursuing, but the paper needs a rewritten proof before it can be taken seriously.\n\nWho is this for? Researchers in quantum resource theories, but as a draft, not a citable result. I would not send this to peer review in its current form; I'd encourage the author to fix the proof and resubmit. A reading group might find the typical-subspace issue instructive.\n\nRecommendation: reject in current form.","headline":"Plausible result, but the proof as written has load-bearing errors; reject for now.","tokens_in":10046,"tokens_out":5636,"would_cite":false,"duration_ms":52726,"reading_group":"maybe","serious_thinker":"no","would_accept_peer_review":false},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81P45","81P68"],"pacs":["03.65.Ud","03.67.Mn"],"model":"deepseek-v4-flash","headline":"The cost to erase imaginarity from a quantum state equals a regularized relative entropy.","keywords":["deimaginarity","imaginarity resource theory","relative entropy of imaginarity","quantum randomness cost","real quantum states","resource erasure","quantum resource theory"],"falsifier":"Take the qubit maximally imaginarity state $|\\phi\\rangle=(|0\\rangle+i|1\\rangle)/\\sqrt{2}$, whose regularized relative entropy of imaginarity is 1 bit per copy. For n=2 and n=3, search explicitly over ensembles of real orthogonal unitaries on $(\\mathbb{C}^2)^{\\otimes n}$ that satisfy the twirl property on the δ-typical subspace; if for any n such an ensemble provably does not exist, the achievability proof breaks. Alternatively, compute the deimaginarity cost for small n by direct optimization over finite ensembles and check whether the sequence converges to $I_r(\\rho)$; finding a rate below 1 bit per copy at any n would contradict the theorem.","tokens_in":8938,"feed_emoji":"⚛️","tokens_out":10675,"duration_ms":85351,"temperature":0.7,"pith_summary":"The paper addresses a task it calls deimaginarity: using random real orthogonal (covariant-free) unitary operations to transform a quantum state into a real state, and asks for the minimum rate of randomness needed in the limit of infinitely many copies with vanishing error. It proves that this minimum rate, the deimaginarity cost $C_d(\\rho)$, equals the regularized relative entropy of imaginarity $I_r^{\\infty}(\\rho)=\\lim_{n\\to\\infty}\\frac{1}{n}I_r(\\rho^{\\otimes n})$. Because the relative entropy of imaginarity had no known operational meaning, this gives the measure an interpretation: it is the asymptotic randomness cost of erasing imaginarity. The result places imaginarity among the resources whose erasure cost is characterized by a regularized entropy.","feed_headline":"Deimaginarity cost equals an entropy rate","feed_subtitle":"The paper proves the minimum randomness to make a state real is the regularized relative entropy of imaginarity.","key_machinery":"The central object is the relative entropy of imaginarity, $I_r(\\rho)=\\min_{\\sigma\\in F} D(\\rho\\|\\sigma)=S((\\rho+\\rho^T)/2)-S(\\rho)$, and its regularized version $I_r^{\\infty}(\\rho)=\\lim_{n\\to\\infty}\\frac{1}{n}I_r(\\rho^{\\otimes n})$. Covariant-free unitaries are real orthogonal operators $O$ that commute with the real-part map $\\Theta(\\cdot)=((\\cdot)+(\\cdot)^T)/2$, so they cannot create imaginarity from real states. The proof machinery consists of $\\delta$-typical subspaces of $\\rho^{\\otimes n}$, the gentle measurement lemma, the Fannes inequality, and the operator Chernoff bound. The load-bearing step in the achievability direction is an unproved assumption: that for each $n$ and $\\delta$ there exists an ensemble $\\{p(dU),U\\}$ of covariant-free real orthogonal unitaries on the typical subspace satisfying $\\int p(dU)\\,U\\theta U^\\dagger = I_{n,\\delta}/D_{n,\\delta}$ for every state $\\theta$ on that subspace, and that these unitaries extend to real orthogonal unitaries on the full space $H^{\\otimes n}$. This ensemble is what converts the projected state into the maximally mixed state on the typical subspace, from which a real state is reached.","core_discovery":"The central claim is Theorem 2: for every finite-dimensional state $\\rho$, the deimaginarity cost equals the regularized relative entropy of imaginarity, $C_d(\\rho)=I_r^{\\infty}(\\rho)$. The converse part shows that any protocol achieving rate $R$ must have $R\\geq I_r^{\\infty}$, using the subadditivity of von Neumann entropy, the Fannes inequality, and the fact that covariant-free unitaries leave the real-part entropy $S((\\rho+\\rho^T)/2)$ invariant. The direct part constructs an achievability protocol at rate $I_r^{\\infty}$ by projecting onto a $\\delta$-typical subspace, applying an assumed ensemble of real orthogonal unitaries that twirls the subspace to the maximally mixed state, and using the operator Chernoff bound to replace the continuous ensemble by a finite one. The paper presents this as the first operational interpretation of the relative entropy of imaginarity.","pith_inferences":["The unproved twirl-ensemble assumption is the point most likely to need repair; a failure of existence for even one state would turn the achievability statement into an open problem rather than a theorem.","A natural next step is to check whether the same formula holds when the free operations are restricted to covariant-free channels rather than unitaries, which would generalize the result to a noisy setting.","The equality suggests a quantitative duality between imaginarity and randomness: the amount of imaginarity a state carries is exactly the random bits needed to wash it out, a relation that could be tested experimentally by realizing the optimal ensembles for small states."],"forward_implications":["The relative entropy of imaginarity now has an operational meaning: it is the asymptotic randomness cost of converting a state into a real state using random real orthogonal unitaries.","For states where the relative entropy of imaginarity is additive, the deimaginarity cost equals the single-copy relative entropy of imaginarity, so the resource can be erased at that rate per copy.","Deimaginarity joins correlation, coherence, and asymmetry as resources whose optimal erasure cost is a regularized entropy, extending the erasure-cost family to imaginarity.","The result provides a target for constructive protocols: any protocol that deimaginarizes at a rate below the regularized relative entropy of imaginarity is impossible, and the optimal rate is known even before a protocol is found."],"supporting_citations":[{"why":"defines the relative entropy of imaginarity and its closed form as S(Re ρ)−S(ρ); this is the quantity whose regularized limit is claimed to equal the deimaginarity cost.","marker":"[28]"},{"why":"sets up the resource theory of imaginarity with real states as the free states, the framework in which deimaginarity is a resource-erasure task.","marker":"[17]"},{"why":"supplies the erasure-cost method used for the direct part, adapted here to the imaginarity setting.","marker":"[29]"},{"why":"provides the symmetrizing-cost technique that the achievability proof follows.","marker":"[32]"},{"why":"Fannes inequality, used in the converse part to bound the entropy difference between adjacent states.","marker":"[38]"},{"why":"defines δ-typical subspaces and their entropy bounds, used to restrict the direct construction to the typical subspace.","marker":"[39]"},{"why":"gentle measurement lemma, used to approximate ρ^{⊗n} by its projected version on the typical subspace.","marker":"[40]"},{"why":"operator Chernoff bound, used to show that a finite ensemble of unitaries approximates the continuous twirl average.","marker":"[41]"}],"fun_headline_variants":["Randomness to erase quantum imaginarity is an entropy rate","Cost to make a quantum state real: entropy of imaginarity","Minimum randomness for real states equals relative entropy rate","Deimaginarity cost proven: regularized entropy of imaginarity"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof's achievability part assumes, without proof, that for every state and every number of copies there is a set of real orthogonal unitaries on the typical subspace that sends every state on that subspace to the maximally mixed state on average, and that these unitaries can be extended to the full n-copy space; if this fails, the construction that reaches the claimed rate does not exist.","fun_headline_variants_meta":{"raw":{"variants":["Randomness to erase quantum imaginarity is an entropy rate","Cost to make a quantum state real: entropy of imaginarity","Minimum randomness for real states equals relative entropy rate","Deimaginarity cost proven: regularized entropy of imaginarity"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000768,"raw_usage":{"total_tokens":3332,"prompt_tokens":805,"completion_tokens":2527,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":421,"completion_tokens_details":{"reasoning_tokens":2456}},"tokens_in":421,"tokens_out":2527,"duration_ms":18341,"temperature":1.0,"reasoning_tokens":2456,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T16:44:52.415510+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take the qubit maximally imaginarity state $|\\phi\\rangle=(|0\\rangle+i|1\\rangle)/\\sqrt{2}$, whose regularized relative entropy of imaginarity is 1 bit per copy. For n=2 and n=3, search explicitly over ensembles of real orthogonal unitaries on $(\\mathbb{C}^2)^{\\otimes n}$ that satisfy the twirl property on the δ-typical subspace; if for any n such an ensemble provably does not exist, the achievability proof breaks. Alternatively, compute the deimaginarity cost for small n by direct optimization over finite ensembles and check whether the sequence converges to $I_r(\\rho)$; finding a rate below 1 bit per copy at any n would contradict the theorem.","supporting_citations":[{"cited_title":"Quantification of resource theory of imaginarity,","cited_arxiv_id":null,"evidence_quote":"defines the relative entropy of imaginarity and its closed form as S(Re ρ)−S(ρ); this is the quantity whose regularized limit is claimed to equal the deimaginarity cost."},{"cited_title":"Quantifying the imaginarity of quantum mechanics,","cited_arxiv_id":null,"evidence_quote":"sets up the resource theory of imaginarity with real states as the free states, the framework in which deimaginarity is a resource-erasure task."},{"cited_title":"Quantum, classical, and total amount of correlations in a quantum state,","cited_arxiv_id":null,"evidence_quote":"supplies the erasure-cost method used for the direct part, adapted here to the imaginarity setting."},{"cited_title":"Symmetrizing cost of quantum states,","cited_arxiv_id":null,"evidence_quote":"provides the symmetrizing-cost technique that the achievability proof follows."},{"cited_title":"A continuity property of the entropy density for spin lattice systems,","cited_arxiv_id":null,"evidence_quote":"Fannes inequality, used in the converse part to bound the entropy difference between adjacent states."},{"cited_title":"Thomas and A","cited_arxiv_id":null,"evidence_quote":"defines δ-typical subspaces and their entropy bounds, used to restrict the direct construction to the typical subspace."},{"cited_title":"Coding theorem and strong converse for quantum channels,","cited_arxiv_id":null,"evidence_quote":"gentle measurement lemma, used to approximate ρ^{⊗n} by its projected version on the typical subspace."},{"cited_title":"Strong converse for iden- tification via quantum channels,","cited_arxiv_id":null,"evidence_quote":"operator Chernoff bound, used to show that a finite ensemble of unitaries approximates the continuous twirl average."}],"review_version":1}