{"id":"deed9add-6e9e-4a39-87c4-7793e5faf632","arxiv_id":"2411.18529","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A quantum symmetry is robust against a perturbation exactly when it commutes with the limit spectral subprojections of the perturbed Hamiltonian; against all perturbations, only functions of the unperturbed Hamiltonian survive.","lead":"This paper classifies which quantum symmetries survive small Hamiltonian perturbations for all times, showing that the survivors are exactly the operators commuting with certain projections induced by the perturbation. The result extends a finite-dimensional theorem to infinite-dimensional systems, with consequences for quantum simulation and thermalization.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 3.1(ii) overstates Kato: eigenvalue branches need not be distinct when a perturbation preserves degeneracy (e.g., V=0); Theorem 3.3's fragility proof uses sup_t |e^{it(h_m-h_n)}-1|=2, which vanishes for degenerate blocks, so the characterization is unproven for such perturbations.","rationale":"The reader's weakest assumption identifies exactly the load-bearing gap: Theorem 3.1(ii) overstates Kato by claiming analytic eigenvalue branches with h_n≠h_m everywhere, which is false when a perturbation preserves a degeneracy. The fragility proof in Theorem 3.3 depends on sup_t |e^{it(h_m-h_n)}-1|=2, which requires a nonzero eigenvalue difference for ε≠0. If degeneracies persist, that sup is 0, the argument breaks, and the stated criterion [S,P_n(0)]=0 is actually too strong, as V=0 demonstrates with a basis-mixing symmetry that is robust but fails the condition for rank-one limiting projections. This is the central theorem on which Theorems 4.1-4.3 rest, so the proof is incomplete as written. However, the issue is repairable by grouping eigenvalue branches with coincident limits and defining P_n(0) as the spectral projections of V restricted to each degenerate eigenspace, in line with standard degenerate perturbation theory. I see no other concern of comparable weight: the reduction in Theorem 4.3 from H-bounded to bounded perturbations is straightforward because bounded perturbations are a subset, and the infinite-dimensional constructions in Lemma 3.1 and the unitary U(ε) appear defensible. Thus the verdict should remain CONDITIONAL, unchanged from the reader.","tokens_in":17429,"tokens_out":12410,"duration_ms":106034,"concrete_test":"Take a finite-dimensional H with a degenerate eigenspace, e.g., H = diag(0,0,1) on C^3, and set V=0. Let S be a symmetry that mixes the two degenerate basis vectors but commutes with H (so S∈{H}′). Apply Theorem 3.2(i) with the analytic labeling used in the proof: for a generic eigenbasis the individual limiting projections P_n(0) are rank-one, and [S,P_n(0)]≠0, so the theorem would declare S fragile; but with V=0 the perturbed evolution is unchanged, so S is plainly V-robust. Then check the corrected grouping criterion: the spectral projection of V on the degenerate eigenspace is the whole 2D projection, and S commutes with it, recovering V-robustness.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The proof of Theorem 3.2(i) invokes Theorem 3.1(ii), which asserts that the perturbed eigenvalues h_n(ε) of H+εV can be analytically labeled with h_n(ε)≠h_m(ε) for all n≠m and ε∈(-1,1). This is not a consequence of compact resolvent. If V commutes with H and acts nontrivially on a degenerate eigenspace, the degeneracy persists for all ε, so any analytic labeling has h_n=h_m on that block; if V=0, all eigenvalues are constant and any repeated eigenvalue gives coincident branches. Theorem 3.3's 'only if' direction proves fragility by choosing ψ_n∈Pn(0)H and ψ_m∈Pm(0)H with ⟨ψ_m|Sψ_n⟩≠0 and bounding ‖A(t,ε)ψ_n‖ ≥ |e^{it(h_m(ε)-h_n(ε))}-1| |⟨ψ_m|Sψ_n⟩|. The supremum over t of |e^{itΔ}-1| is 2 only when Δ≠0; for a degenerate block Δ=0 the bound vanishes and the argument gives no fragility. In fact, symmetries mixing states within a persistently degenerate block are V-robust, so the condition [S,P_n(0)]=0 for each individual branch is too strong; the theorem as stated fails for V=0 (or any V not splitting all degeneracies). A correct statement must group branches whose eigenvalue difference vanishes on a punctured neighborhood and require commutation with the corresponding spectral subprojections, replacing Theorem 3.1(ii) with a precise grouping statement from Kato-Rellich theory. This affects the proof of the central characterization, though the result is likely repairable.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper defines P-robust symmetries of a Hamiltonian H as those whose Heisenberg evolution under H+εV stays close to the unperturbed evolution uniformly in time as ε→0. Assuming H has compact resolvent, the authors prove (Theorem 3.2) that a symmetry S is V-robust exactly when it commutes with the limiting eigenprojections P_n(0) of the perturbed family H+εV. This is then used to characterize robustness against sets of perturbations and to show that completely robust symmetries are precisely the bicommutant {H}'' (bounded Borel functions of H), extending a finite-dimensional result from Ref. [9] to unbounded Hamiltonians. The paper also analyzes symmetry-restricted perturbations, shows that the wandering range can decay as |ε|^γ with arbitrarily small γ in infinite dimensions, and constructs adiabatic invariants from robust symmetries.","tokens_in":17855,"tokens_out":12225,"duration_ms":120736,"significance":"The algebraic characterization is clean and, once the degeneracy issue described below is fixed, would provide a substantial infinite-dimensional generalization of the finite-dimensional theorem. The paper is methodologically sound in its use of Kato perturbation theory and von Neumann algebras, contains no fitted parameters, and includes a concrete example (Example 5.1) that gives explicit wandering-range scaling. The main theorems are falsifiable and would be useful for classifying conserved quantities under perturbations. However, the central characterization currently has a load-bearing gap for perturbations that do not split all degeneracies, so the result is not yet established as stated.","major_comments":[{"comment":"The assertion that the analytic eigenvalue branches h_n(ε) of H(ε)=H+εV satisfy h_n≠h_m for all n≠m and all ε∈(-1,1) overstates what Kato-Rellich theory provides. When H has a degenerate eigenvalue and V splits that degeneracy, different branches have h_n(0)=h_m(0); when V preserves the degeneracy for all ε (for example V=0 on a degenerate eigenspace, or V commuting with H and acting as a scalar on that eigenspace), the branches coincide identically and no such labeling exists. Since the proof of Theorem 3.2 invokes Theorem 3.1(ii) to verify assumption (ii) of Theorem 3.3, the applicability of Theorem 3.2 to arbitrary H-bounded V is not justified.","section":"Section 3, Theorem 3.1(ii)"},{"comment":"The 'only if' direction of Theorem 3.3 uses the identity sup_t |e^{it(h_m(ε)-h_n(ε))}-1|=2, which requires h_m(ε)≠h_n(ε) for all ε in a punctured neighborhood of 0. For a persistently degenerate block, where h_m≡h_n on the whole interval I, this supremum is 0 and the fragility argument gives no lower bound. A symmetry that mixes states inside such a block is V-robust even though it fails to commute with an arbitrary analytic labeling of the subprojections P_n(0). Thus the condition [S,P_n(0)]=0 in Theorem 3.2 is too strong, and the characterization is unproved for perturbations that do not split all degeneracies. The theorem should group branches whose eigenvalue difference vanishes identically and require commutation with the corresponding spectral subprojections; with that grouping the proof appears repairable.","section":"Section 3, Theorem 3.3 and proof of Theorem 3.2"}],"minor_comments":[{"comment":"In the last displayed estimate before 'lim inf', the vector φ_m should be ψ_m, matching the notation introduced earlier in the proof.","section":"Proof of Theorem 3.3"},{"comment":"The statement c_α/‖ψ_α‖ → 1/√2 as α↓1 is plausible but not immediately evident; a one-line derivation would improve readability.","section":"Section 5.1, Eq. (62)"},{"comment":"It would be helpful to state explicitly that for perturbations that do not split a degeneracy, the limiting projections P_n(0) are not canonically defined by H and V, and that the correct object is the collection of spectral projections of V restricted to each degenerate eigenspace of H.","section":"Section 3, discussion after Theorem 3.2"}],"recommendation":"major_revision","confidential_remarks":"The stress-test concern about persistent degeneracies is legitimate and lands on the main theorem. The proposed repair by grouping branches with identically vanishing eigenvalue differences is natural and likely preserves the paper's principal results, so I view this as a major revision rather than a rejection. The paper fits the journal's scope and contains a valuable extension of the finite-dimensional result, so it is worth a careful revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Main point: this paper gives the infinite-dimensional version of the finite-dim robustness classification: for compact-resolvent H, the completely robust symmetries are exactly the bounded functions of H, and against a single perturbation they are the operators commuting with the perturbed spectral subprojections. Those results are genuinely new relative to [9,12], and so is the example showing the wandering range can be O(epsilon^gamma) for arbitrarily small gamma.\n\nWhat's good: the proof structure is clear. The eternal block-diagonal approximation (Lemma 3.1) and the splitting of the evolution into robust and fragile pieces (Lemma 3.2) are elegant, and the algebra in Section 4 (symmetry-restricted perturbations, bicommutant results) is natural and well argued. The harmonic-oscillator example is worked out carefully.\n\nWhere the soft spots are: the stress-test note is right. Theorem 3.1(ii) says the perturbed eigenvalues can be labeled analytically with h_n != h_m for all n != m and all epsilon. That is not what Kato's theorem guarantees when the perturbation does not split a degeneracy. V=0 is the trivial case, and any V that is a scalar on a degenerate eigenspace keeps the branches coincident for all epsilon. The proof of Theorem 3.3 uses sup_t |e^{it(h_m-h_n)}-1| = 2, which is zero on such branches, so the fragility argument misses those cases. This is a real gap in the written proof, but it is repairable: group the branches that stay degenerate and state the characterization in terms of the grouped subprojections. The main Theorems 3.2 and 4.3 likely survive unchanged once this is made precise.\n\nAnother minor point: Theorem 4.3's reduction from H-bounded to bounded perturbations is not fully justified; it deserves a sentence or two.\n\nVerdict: this is a paper worth refereeing. The core idea and the infinite-dim extension are solid, the example is sharp, and the issues are presentation-level. A careful revision of Section 3 should fix the gap. I'd bring it to a reading group and would cite it once the labeling issue is cleaned up.","headline":"A solid infinite-dimensional extension of the finite-dim robustness classification, with a repairable technical gap about eigenvalue branches that stay degenerate.","tokens_in":18348,"tokens_out":6947,"would_cite":true,"duration_ms":62144,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81Q15","47A55","46L10"],"pacs":[],"model":"deepseek-v4-flash","headline":"A quantum symmetry survives a perturbation exactly when it commutes with the limiting projections the perturbation induces on the Hamiltonian's energy levels, and the only symmetries that survive every perturbation are bounded functions…","keywords":["quantum symmetries","conserved quantities","Hamiltonian perturbations","robustness","commutant and bicommutant","von Neumann algebras","analytic perturbation theory","quantum adiabatic invariants"],"falsifier":"Diagonalize $H = 0$ on $\\mathbb{C}^2$ with $V = 0$: $H$ has compact resolvent and a single doubly degenerate level, and $H + \\varepsilon V = 0$ for every $\\varepsilon$, so no analytic labeling with $h_1(\\varepsilon) \\neq h_2(\\varepsilon)$ exists. This direct example shows that the distinct-level premise is not a consequence of compact resolvent, and it forces one to check whether the characterization survives when the limiting projections are grouped by the spectrum of $V$ on each degenerate eigenspace.","tokens_in":17240,"feed_emoji":"⚛️","tokens_out":16184,"duration_ms":129680,"temperature":0.7,"pith_summary":"This paper asks a practical question about quantum models: when the Hamiltonian is slightly perturbed, which of its conserved quantities remain almost conserved at all later times, not just for short times? The authors call a symmetry robust when its perturbed evolution stays close to its unperturbed value uniformly in time as the perturbation strength goes to zero, and fragile when some state eventually drifts away. Their central result is an exact algebraic answer: against a fixed perturbation, the robust symmetries are precisely the bounded operators that commute with the family of limiting projections that the perturbation induces on the Hamiltonian's degenerate energy levels. Against all admissible perturbations, the robust symmetries reduce to the bounded functions of the Hamiltonian itself. This matters because only such robust quantities can be trusted to describe the long-time behavior of a model whose Hamiltonian is known only approximately.","feed_headline":"Only functions of the Hamiltonian survive every perturbation","feed_subtitle":"A new algebraic theorem tells exactly which quantum symmetries stay conserved at all times under Hamiltonian noise.","key_machinery":"The carrying device is analytic perturbation theory for compact-resolvent Hamiltonians. It yields analytic eigenvalue branches $h_n(\\varepsilon)$ and eigenprojections $P_n(\\varepsilon)$ of $H + \\varepsilon V$, the limiting subprojections $P_n(0) = \\lim_{\\varepsilon \\to 0} P_n(\\varepsilon)$ that refine the unperturbed spectral projections, and a strongly continuous unitary family $U(\\varepsilon)$ with $P_n(\\varepsilon) = U(\\varepsilon) P_n(0) U(\\varepsilon)^\\dagger$. From these the paper forms the eternal block-diagonal approximation $\\tilde H(\\varepsilon) = \\sum_n h_n(\\varepsilon) P_n(0)$, whose unitary group stays uniformly close in time to the true perturbed group, and splits the perturbed evolution of a symmetry into a robust component that vanishes uniformly and a fragile component $A(t,\\varepsilon) = e^{itH(\\varepsilon)}[S, e^{-it\\tilde H(\\varepsilon)}]$. Fragility is exposed by the phase factor $e^{it(h_m(\\varepsilon)-h_n(\\varepsilon))} - 1$, whose supremum over $t$ is $2$ whenever the corresponding levels are distinct, while the bicommutant theorem identifies $\\{H\\}''$ with the algebra of bounded Borel functions of $H$.","core_discovery":"On its own terms, the paper establishes Theorem 3.2: for a self-adjoint Hamiltonian $H$ with compact resolvent and an $H$-bounded symmetric perturbation $V$ (one whose size is controlled by $H$, so that $H + \\varepsilon V$ stays self-adjoint), a symmetry $S$ (a bounded operator commuting with $H$) is $V$-robust if and only if $[S, P_n(0)] = 0$ for every limiting eigenprojection $P_n(0) = \\lim_{\\varepsilon \\to 0} P_n(\\varepsilon)$ of $H + \\varepsilon V$. Because the $P_n(0)$ are subprojections of the unperturbed eigenprojections, the $V$-robust symmetries form the von Neumann algebra $\\{P_n(0)\\}'$ lying between the bicommutant $\\{H\\}''$ and the commutant $\\{H\\}'$. For a whole set $\\mathcal{P}$ of perturbations, $\\mathcal{P}$-robustness is the intersection of these commutants over $V \\in \\mathcal{P}$; when perturbations are restricted to preserve a family of protected symmetries $\\mathcal{J}$, Theorem 4.2 gives exactly the bicommutant $(\\{H\\} \\cup \\mathcal{J})''$; and when perturbations are unrestricted, Theorem 4.3 gives $R(H) = \\{H\\}''$, the bounded Borel functions of the Hamiltonian, extending the known finite-dimensional result to unbounded compact-resolvent Hamiltonians. The paper also constructs an explicit harmonic-oscillator example in which a completely robust symmetry has a wandering range that decays only as $O(|\\varepsilon|^{\\alpha - 1})$ with $\\alpha$ arbitrarily close to $1$, so in infinite dimensions the convergence need not be uniform in the state and can be arbitrarily slow.","pith_inferences":["The algebraic characterization likely does not need compact resolvent: Theorem 3.3 in the paper is already stated for continuous pure-point deformations without it, so any perturbation family whose limiting subprojections are well defined should obey the same commutant criterion.","The slow-wandering example is a quantum analogue of a known diffusion phenomenon of classical Hamiltonian mechanics, where conserved quantities drift anomalously slowly under weak perturbations, so quantitative stability estimates are needed before robust symmetries are used for long-time predictions.","Because completely robust observables are exactly spectral functions of $H$, the result gives a sharp error-robustness criterion for quantum simulation: only functions of the effective Hamiltonian have long-time expectation values insensitive to arbitrary small control errors.","When perturbations are required to preserve a symmetry $J$, the theorem predicts that robust conserved quantities are block-diagonal in the common eigenspaces of $H$ and $J$; this could be tested numerically on lattice models with degenerate bands by computing the limiting projections directly."],"forward_implications":["Every symmetry of the form $f(H)$ with $f$ a bounded Borel function is completely robust: it survives every admissible perturbation uniformly in time, so spectral functions of the bare Hamiltonian are the only exactly protected conserved quantities.","A symmetry that fails to commute with some limiting projection $P_n(0)$ is fragile: no matter how small the perturbation strength, some state will drift away from its unperturbed value at long times, so the deviation does not vanish uniformly.","If physical law restricts perturbations to preserve a protected symmetry $J$, the robust symmetries are exactly the algebra generated by the common spectral projections of $H$ and $J$; perturbing within the commutant of $J$ cannot break those conserved quantities.","In infinite-dimensional systems the finite-dimensional linear bound on the wandering range fails: even completely robust symmetries can deviate as $O(|\\varepsilon|^\\gamma)$ with $\\gamma$ arbitrarily small, so robustness does not by itself give a practical rate of convergence.","Every robust symmetry can be continuously deformed into a symmetry of the perturbed Hamiltonian, a quantum adiabatic invariant, so the symmetry is bent rather than broken by the perturbation."],"supporting_citations":[{"why":"Supplies the analytic perturbation theorem used for the spectral behavior of $H + \\varepsilon V$, the limiting subprojections $P_n(0)$, and the unitary intertwiner $U(\\varepsilon)$.","marker":"[23]"},{"why":"Proved in finite dimension that completely robust symmetries are functions of the Hamiltonian, the result Theorem 4.3 extends to unbounded compact-resolvent operators.","marker":"[9]"},{"why":"A companion finite-dimensional analysis of quantum symmetries under perturbations, cited with [9] as the finite-dimensional characterization being generalized.","marker":"[12]"},{"why":"Related construction of equilibrium states as adiabatic invariants with respect to all perturbations, used to connect the bicommutant result to thermal equilibrium states.","marker":"[13]"},{"why":"Earlier small-time robustness bound with which the paper contrasts its large-time uniform-in-time robustness notion.","marker":"[10]"}],"fun_headline_variants":["Perturbation-proof symmetries are just functions of H","Algebraic theorem classifies all robust quantum symmetries","Only bounded Borel functions of H survive every noise","Robust quantum symmetries: exactly the functions of H"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof relies on labeling the perturbed energy levels so that they remain distinct for small nonzero perturbation strength; if a perturbation does not split a degenerate level, such as the zero perturbation, that labeling is impossible and the oscillating-phase argument that detects fragility has no phase to oscillate.","fun_headline_variants_meta":{"raw":{"variants":["Perturbation-proof symmetries are just functions of H","Algebraic theorem classifies all robust quantum symmetries","Only bounded Borel functions of H survive every noise","Robust quantum symmetries: exactly the functions of H"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000293,"raw_usage":{"total_tokens":1710,"prompt_tokens":948,"completion_tokens":762,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":564,"completion_tokens_details":{"reasoning_tokens":696}},"tokens_in":564,"tokens_out":762,"duration_ms":7253,"temperature":1.0,"reasoning_tokens":696,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T11:08:29.997694+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Diagonalize $H = 0$ on $\\mathbb{C}^2$ with $V = 0$: $H$ has compact resolvent and a single doubly degenerate level, and $H + \\varepsilon V = 0$ for every $\\varepsilon$, so no analytic labeling with $h_1(\\varepsilon) \\neq h_2(\\varepsilon)$ exists. This direct example shows that the distinct-level premise is not a consequence of compact resolvent, and it forces one to check whether the characterization survives when the limiting projections are grouped by the spectrum of $V$ on each degenerate eigenspace.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the analytic perturbation theorem used for the spectral behavior of $H + \\varepsilon V$, the limiting subprojections $P_n(0)$, and the unitary intertwiner $U(\\varepsilon)$."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Proved in finite dimension that completely robust symmetries are functions of the Hamiltonian, the result Theorem 4.3 extends to unbounded compact-resolvent operators."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Related construction of equilibrium states as adiabatic invariants with respect to all perturbations, used to connect the bicommutant result to thermal equilibrium states."}],"review_version":1}