{"id":"d50d898d-1619-41f1-916f-bd0b36135ca4","arxiv_id":"2505.12187","paper_version":1,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The L2 convergence rate of a lifted quantum Markov semigroup is at most the square root of the spectral gap of its collapsed dynamics, and matching lower bounds hold under explicit structural assumptions.","lead":"This paper extends the 'lifting' trick for speeding up random walks to quantum Markov processes, showing the possible speed-up is at most quadratic. It also gives a general mathematical framework and constructs examples that achieve the optimal speed-up.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 4.11's lower bound drops the K1=sqrt(J0) factor from its own proof, so optimality for depolarizing and Schur multipliers is not established.","rationale":"The reader's weakest assumption was the case-by-case verification of Assumption 3 for the lower bound. My concern is narrower and more specific: a concrete algebraic gap in the proof of Theorem 4.11, which is the basis for the abstract's claim of optimal lifts for depolarizing semigroups and Schur multipliers. The proof of Eq. (4.65) explicitly obtains K1 = sqrt(J0), and the subsequent deduction of the lower bound in Theorem 4.11 silently omits this factor. Since J0 is dimension-dependent in the advertised examples, the condition K1 + K2 lambda_O^{-1/2} = O(1) used to define optimality in (2.42) is not verified by the argument presented. This does not invalidate the upper bound theorem or the structural lower-bound framework, but it does undercut a central claimed application. A corrected proof showing K1 = O(1) due to cancellations, or a revised theorem with J0 in the denominator, would resolve the issue. I therefore recommend a conditional accept rather than an unqualified accept.","tokens_in":52785,"tokens_out":12088,"duration_ms":119732,"concrete_test":"For the depolarizing semigroup L = id - EN on B(C^d) with d = 4, 8, 16 and EN the conditional expectation onto the diagonal algebra, construct the lift of Section 4.3.2 with sigma_B proportional to 1 and kappa = 0. Numerically or symbolically compute the exact constant K1 in (2.35), K1 = sup_{Y in ker(LO)^perp} ||(id-ES)L_A^dagger(-L_S)^{-1}L_A Y||_{2,sigma} / ||LO(Y)||_{2,sigma}, and also compute the maximal convergence rate nu(L_{gamma_max}) by exact diagonalization of L_gamma. If K1 grows like sqrt(d), or nu decays like 1/sqrt(d), then Theorem 4.11's optimality claim is false for depolarizing semigroups. If K1 remains O(1) despite the proof's crude estimate, then a sharper argument must be added to justify the theorem as stated.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"In Section 4.3.2, the proof of Theorem 4.11 derives K1 = sqrt(J0) and K2 = sqrt(J0 max{0,-kappa}) immediately after Eq. (4.65). Substituting these into the paper's own optimality condition (2.42), K1 + K2 lambda_O^{-1/2} = O(1), requires J0(1 + sqrt(max{0,-kappa} lambda_O^{-1})) = O(1). Yet the statement of Theorem 4.11 gives the lower bound Omega(sqrt(lambda_O)/(1 + sqrt(max{0,-kappa} lambda_O^{-1}))), which is valid only if K1 is O(1), i.e., if J0 is bounded independently of the system. For the examples highlighted in the abstract, J0 grows with dimension: for the depolarizing generator L = EN - id on B(C^d) with N the diagonal algebra, J0 = d^2 - 1, and for the Schur multipliers in Example 4.2, J0 = d. Therefore the claimed optimality does not follow from the proof as written. The universal upper bound O(sqrt(lambda_O)) and the conditional lower bound of Theorem 2.17 may still be correct, but the load-bearing application claim that these examples admit optimal lifts is unsupported unless a refined bound showing K1 = O(1) is supplied.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a theory of second-order non-reversible lifts for detailed-balanced quantum Markov semigroups, in analogy with the classical lifting framework of Eberle and Lörler. The lifted generator is written as L_gamma = L_A + gamma L_S, and the overdamped limit is identified with an effective generator L_O. The main theoretical results are an upper bound on the convergence rate in terms of the square root of the spectral gap of L_O, obtained via the singular-value gap, and a lower bound obtained via a flow Poincaré inequality. The paper also presents an abstract Hilbert-space lifting framework and applies it to reversible diffusions, finite Markov chains on a cycle, depolarizing semigroups, Schur multipliers, and group von Neumann algebras, claiming optimal Theta(sqrt(lambda_O)) lifts in several cases.","tokens_in":53014,"tokens_out":12881,"duration_ms":136666,"significance":"If the main theorems are correct, the paper gives a unified quantitative framework for hypocoercive quantum lifts and connects it to the classical lifting literature. The singular-value-gap upper bound and the flow-Poincaré lower bound are derived from first principles, with no free parameters fitted to examples, and the main proofs are largely self-contained. The paper is also transparent about the conditional nature of the lower bound: Assumption 3 is verified only case by case, and the authors explicitly note the absence of general sufficient conditions. These are genuine strengths. However, the application section contains a load-bearing gap: the claimed optimality for depolarizing semigroups and Schur multipliers does not follow from the proof as written, because the constant K_1 obtained in the proof scales with the dimension. In addition, the universal upper bound on the convergence rate is stated with a prefactor C that is not controlled uniformly in gamma, so the abstract's phrasing overstates what Theorem 2.16 establishes.","major_comments":[{"comment":"The proof of Theorem 4.11 derives K_1 = sqrt(J_0) and K_2 = sqrt(J_0 max{0,-kappa}) immediately after Eq. (4.65). Substituting these constants into the optimality condition (2.42) requires J_0(1 + sqrt(max{0,-kappa} lambda_O^{-1})) = O(1). However, in the highlighted examples J_0 grows with dimension: for the depolarizing generator L = E_N - id on B(C^d), J_0 is of order d^2, and for the Schur multipliers in Example 4.2, J_0 is of order d. The statement of Theorem 4.11, which gives Omega(sqrt(lambda_O)/(1 + sqrt(max{0,-kappa} lambda_O^{-1}))), is therefore not a consequence of the proof unless a refined argument shows K_1 = O(1). Since this theorem is the basis for the claimed optimal lifts of depolarizing semigroups and Schur multipliers, the application claim in the abstract is unsupported as written. The theorem should either be reproved with a construction that avoids the J_0 factor, or restated with the J_0-dependent bound and the optimality claims appropriately weakened.","section":"Section 4.3.2, Eq. (4.65) and Theorem 4.11"},{"comment":"The abstract and the discussion after Theorem 2.16 claim that the L2 convergence rate of the lifted semigroup is bounded above by O(sqrt(lambda_O)) for every gamma, independent of gamma. The proof, however, uses Lemma 3.4, which yields nu <= (1 + log C) s(L), where C is the prefactor in the exponential decay estimate. The prefactor C is not shown to be bounded uniformly in gamma; for hypocoercive semigroups C generally depends on the parameters of the generator and can grow as gamma varies. Thus Theorem 2.16 as stated is a relation between an admissible pair (C, nu) and the singular-value gap, not a uniform upper bound on the sharp convergence rate. The relaxation-time lower bound trel >= 1/(2 sqrt(fsm^{-1} lambda_O)) is valid and does support the impossibility of more than a square-root speed-up in mixing time, but the claim about the convergence rate itself needs to be qualified, or a uniform bound on the optimal C must be supplied.","section":"Theorem 2.16 and Lemma 3.4"}],"minor_comments":[{"comment":"There is a typo in the phrase 'fintie Markov chains'; it should read 'finite Markov chains'.","section":"Introduction, Section 1"},{"comment":"The proof of Lemma 2.9 is omitted with a reference to the authors' earlier work [LL24, Lemma 2.5]. Since this lemma is used in the proof of the central upper bound, either a full proof or a precise statement of the cited lemma should be included for the paper to be self-contained.","section":"Lemma 2.9"},{"comment":"The proof of Lemma 4.7 is omitted with the comment that it is almost identical to [VW23, Theorem 2.5]. Given that this lemma underpins the continued verification of Condition D', the authors should provide a detailed proof or at least a complete derivation of the Kraus form claimed in the lemma.","section":"Lemma 4.7"},{"comment":"The normalization factor in the definition of Z_j^A is written ambiguously as 1/sqrt(mu_{2j-1} + mu_{2j}); the parentheses should be explicit so that the formula reads 1/sqrt(mu_{2j-1} + mu_{2j}), matching the norm computation immediately after Eq. (4.65).","section":"Eq. (4.60)"},{"comment":"The notation J_0 is used both for a set of indices (J_{O,0} = {1 <= j <= J_0}) and for its cardinality; this makes the counting arguments in the proof of Theorem 4.11 more difficult to follow than necessary.","section":"Section 4.3.2, notation"}],"recommendation":"major_revision","confidential_remarks":"The central theoretical framework appears sound, and the issues identified are local rather than global: they concern the precision of the universal upper-bound claim and the validity of the optimality assertions in the applications. These are fixable within the scope of the manuscript, so I recommend major revision rather than rejection. The authors' reliance on their own prior work is appropriate and does not constitute circularity."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nHere's my read on Li–Lu, arXiv:2505.12187. The core is good, but the paper's showcase examples of optimal quantum lifts have a real gap.\n\nWhat's new and working: The paper lifts the Eberle–Lörler second-order lifting construction to quantum Markov semigroups and, more abstractly, to any symmetric contraction C0-semigroup on a Hilbert space. The upper bound ν(L_γ) = O(√λ_O) via the singular value gap is clean and general. The lower bound via the flow Poincaré inequality is a serious piece of work, and the framework does recover the classical diffusion results and give a genuinely optimal quantum lift for the symmetric random walk on a chain (Section 4.2 verifies the matrix inequality with O(1) constants).\n\nThe soft spot is Theorem 4.11. In the proof, after Eq. (4.65), the authors get K1 = sqrt(J0) and K2 = sqrt(J0 max{0,−κ}). Plugging these into their own lower bound (2.41) yields ν = Ω(√λ_O / (1 + sqrt(J0)(1 + sqrt(max{0,−κ} λ_O^{-1})))). The theorem statement instead gives Ω(√λ_O / (1 + sqrt(max{0,−κ} λ_O^{-1}))), effectively dropping the sqrt(J0). For the two headline applications, J0 grows with dimension: for the depolarizing semigroup J0 is on the order of d^2, and for the Schur multipliers it is at least d. So the optimality condition K1 + K2 λ_O^{-1/2} = O(1) is not met, and the claimed Θ(√λ_O) rate for these examples does not follow from the proof as written. The upper bound and the conditional lower-bound framework are untouched; the problem is specifically the application claim in the abstract.\n\nMinor issues: Lemma 2.9 and Lemma 4.7 are deferred to earlier work, and Assumption 3 is verified case-by-case with no general sufficient conditions. The authors say so explicitly, so it's a limitation, not a hidden one.\n\nBottom line: this deserves a serious referee. The general framework is novel and likely correct; Theorem 4.11 needs fixing, either by a sharper estimate showing K1 = O(1) (I suspect that's not possible for these examples) or by weakening the optimality claim. I'd send it to review with a request for revision. I'd cite the upper-bound framework, but not the claim of optimal quantum lifts until the J0 issue is resolved.","headline":"Solid general lifting framework for quantum Markov semigroups, but Theorem 4.11 overclaims optimality by dropping a dimension-dependent J0 factor; the depolarizing and Schur optimal-lift examples are not proven.","tokens_in":53546,"tokens_out":9411,"would_cite":true,"duration_ms":88795,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60J27","60J60","81S22"],"pacs":[],"model":"deepseek-v4-flash","headline":"Lifted quantum Markov processes converge at most a square-root faster than their collapsed dynamics.","keywords":["quantum Markov semigroup","lifting","hypocoercivity","mixing time","detailed balance","space-time Poincaré inequality","singular value gap","Lindblad dynamics"],"falsifier":"Take a small finite-dimensional detailed-balanced quantum Markov semigroup, construct a second-order lift satisfying Conditions A-D, diagonalize L_gamma directly, and compute the L2 convergence rate as gamma varies; if the maximum rate exceeds the upper-bound constant (1+log C) $\\sqrt$($fsm^{{-1}}$ lambda_O), the upper bound is false, and if a generator with growing K1+K2 $lambda_O^{{-1/2}}$ shows an optimal-rate scaling slower than $\\sqrt$(lambda_O), the matching lower bound is false.","tokens_in":52553,"feed_emoji":"⚛️","tokens_out":6005,"duration_ms":54308,"temperature":0.7,"pith_summary":"This paper extends the classical idea of lifting—adding auxiliary variables to a reversible Markov process to break detailed balance—to quantum Markov semigroups. It claims that any second-order lift of a detailed-balanced quantum Markov process can reduce the L2 relaxation time by at most a square root: the convergence rate is bounded by O(sqrt(lambda_O)), where lambda_O is the spectral gap of the original (overdamped) dynamics. It also proves a matching lower bound at an optimally chosen lifting strength, under technical structural conditions, and identifies cases where the square-root speed-up is actually attained. The significance is that a simple, physically motivated operation—adding a coherent Hamiltonian term to a dissipative generator—has a provable speed limit, and that this limit is generically reachable.","feed_headline":"A square-root ceiling governs lifted quantum mixing","feed_subtitle":"New rate bounds show non-reversible lifts can turn diffusive mixing into ballistic, but no faster than a square root.","key_machinery":"The load-bearing object is the second-order lift: adding a fast symmetric dissipation gamma L_S and a Hamiltonian-type coherent term L_A to a detailed-balanced generator, so the slow dynamics emerges only after rescaling time by gamma. The effective collapsed generator L_O is found by eliminating the fast subspace through a second-order perturbation expansion, which is the reverse of the overdamped limit. The upper bound follows from comparing the singular value gap of L_gamma with the spectral gap of L_O. The lower bound uses a flow Poincaré inequality—a time-augmented coercivity estimate for the degenerate (hypocoercive) generator—whose constants are controlled by the lifting structure and by technical inequalities (C0)-(C2).","core_discovery":"The central claim is a pair of theorems about generators of the form L_gamma = L_A + gamma L_S, with L_S self-adjoint and L_A anti-self-adjoint in the KMS inner product, where gamma is a damping parameter. When exp(tL_gamma) is a second-order lift of a detailed-balanced quantum Markov semigroup exp(tL_O), its L2 convergence rate satisfies nu(L_gamma)=O($\\sqrt$(lambda_O)) for every gamma, with lambda_O the spectral gap of the collapsed generator L_O=-(L_A E_S)^*(-L_S)^{-1}L_A E_S. At the optimal gamma, the rate satisfies nu = $\\Omega$($\\sqrt$(lambda_O)/(1+K1+K2 $lambda_O^{{-1/2}}$)); if K1+K2 $lambda_O^{{-1/2}}$=O(1), the lift is optimal and nu=Theta($\\sqrt$(lambda_O)). The paper further formulates an abstract Hilbert-space version that applies to any symmetric contraction C0-semigroup, unifying classical diffusions and quantum dynamics, and constructs optimal lifts for a symmetric random walk, the depolarizing semigroup, Schur multipliers, and quantum Markov semigroups on group von Neumann algebras.","pith_inferences":["The upper bound is robust: it uses only the lifting identities and finite dimensionality, so any lift satisfying those identities—even one not arising from an overdamped limit—inherits the square-root ceiling.","The lower bound is the fragile part: since inequalities (C0)-(C2) are verified case-by-case and no general sufficient conditions are given, the matching square-root speed likely holds for a narrower class than the upper bound; characterising that class is a natural next step.","A testable prediction is that generators with strong negative intertwining curvature, where -kappa lambda_O^{-1} is large, should exhibit an optimal rate degraded away from sqrt(lambda_O); measuring the rate for such a generator would directly probe the constants K1 and K2."],"forward_implications":["For any detailed-balanced quantum Markov process, a second-order non-reversible lift can cut the L2 relaxation time by no more than a square-root factor; mixing speed at best changes from diffusive to ballistic.","At the optimally tuned lifting strength gamma_max, the lower bound matches the upper bound whenever the structural constants satisfy K1 + K2 lambda_O^{-1/2} = O(1), so the square-root speed-up is achieved, not just possible.","The abstract framework applies to any symmetric contraction C0-semigroup, so the same upper and lower rate bounds hold for classical reversible diffusions and their lifts, such as underdamped Langevin dynamics and randomized Hamiltonian Monte Carlo.","Concrete optimal lifts exist for the symmetric random walk on a chain, the depolarizing semigroup, Schur multipliers, and quantum Markov semigroups on group von Neumann algebras, giving explicit examples where the bound is saturated."],"supporting_citations":[{"why":"Introduced second-order non-reversible lifts for reversible diffusion processes and the upper-bound argument this paper generalizes.","marker":"[EL24a]"},{"why":"Established the square-root speed-up ceiling and first-order lifts for finite Markov chains.","marker":"[CLP99]"},{"why":"Provided the variational hypocoercivity framework with space-time Poincaré inequalities used for the lower bound.","marker":"[AAMN24]"},{"why":"Extended lifts and flow Poincaré inequalities to piecewise-deterministic processes and boundary problems, serving as the template for Section 3.","marker":"[EGH+25]"},{"why":"Proved quantum space-time Poincaré inequalities for Lindblad dynamics, the direct predecessor for the quantum lower bound.","marker":"[LL24]"},{"why":"Introduced the κ-intertwining condition used to bound K1 and K2 in the optimal-lift examples.","marker":"[CM17]"},{"why":"Supplied the space-time divergence lemma that controls the flow Poincaré constants.","marker":"[EL24b]"},{"why":"Connected detailed balance to spectral-gap relaxation times for quantum Markov semigroups, grounding the L2 notion of mixing.","marker":"[TKR+10]"}],"fun_headline_variants":["Lifting quantum Markov processes: speedup capped by square root","Quantum lifts: diffusive to ballistic, but only sqrt gap gain","Square-root limit on accelerated quantum mixing via lifting","Lifting speeds quantum mixing, bound set by spectral gap sqrt","Quantum Markov lifts: max ballistic speedup is sqrt spectral gap"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The lower-bound theorem rests on the assumption that the lifted generator satisfies a list of structural inequalities (C0)-(C2) with finite constants, verified only example by example; if those inequalities fail, the claimed $\\Omega$($\\sqrt$(lambda_O)) rate need not hold.","fun_headline_variants_meta":{"raw":{"variants":["Lifting quantum Markov processes: speedup capped by square root","Quantum lifts: diffusive to ballistic, but only sqrt gap gain","Square-root limit on accelerated quantum mixing via lifting","Lifting speeds quantum mixing, bound set by spectral gap sqrt","Quantum Markov lifts: max ballistic speedup is sqrt spectral gap"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000147,"raw_usage":{"total_tokens":1217,"prompt_tokens":1011,"completion_tokens":206,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":627,"completion_tokens_details":{"reasoning_tokens":122}},"tokens_in":627,"tokens_out":206,"duration_ms":2998,"temperature":1.0,"reasoning_tokens":122,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T20:38:53.148137+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a small finite-dimensional detailed-balanced quantum Markov semigroup, construct a second-order lift satisfying Conditions A-D, diagonalize L_gamma directly, and compute the L2 convergence rate as gamma varies; if the maximum rate exceeds the upper-bound constant (1+log C) $\\sqrt$($fsm^{{-1}}$ lambda_O), the upper bound is false, and if a generator with growing K1+K2 $lambda_O^{{-1/2}}$ shows an optimal-rate scaling slower than $\\sqrt$(lambda_O), the matching lower bound is false.","supporting_citations":[],"review_version":1}