{"id":"abfd3449-9679-4b85-952d-a853670b5f8e","arxiv_id":"2504.16827","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The Hardy-Littlewood maximal operator, adjoint Riesz transforms, and Poisson/heat semigroup approximations are shown to be bounded on the Schrödinger-associated CMO_L space.","lead":"This paper proves three endpoint boundedness results for singular integral operators on the vanishing mean oscillation space CMO_L associated with a Schrödinger operator L=-Δ+V. A generalist might read it because it connects maximal functions, Riesz transforms, and semigroup approximation in a non-classical harmonic analysis setting, and it recovers classical CMO results as a special case.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The proof of Theorem 1.2, Step III, drops the τ-integral after substituting an estimate with a (1+|τ|^{1/2}R_y)^m factor; with the stated bounds that integral diverges, so the continuity of T1 and hence Theorem 1.2 is not established as written.","rationale":"The reader's weakest assumption is exactly the Step III τ-integral issue. My stress-test confirms and sharpens it: the problem is not merely an unjustified interchange of limits, but an outright divergence of the τ-integral if the stated estimates are used. I surveyed the surrounding argument for other weak spots. Theorem 1.1 uses the classical boundedness of M on BMO and a standard L1-to-L2 reduction via John-Nirenberg; no fatal gap appears there. Theorem 1.3 relies on the Poisson kernel estimate (5.8) and the tent-space characterization from [18]; the trace arguments are plausible and the concern is not load-bearing. Lemma 4.1 inherits the status of Theorem 1.2, so the τ-integral step is the single most important unresolved point. The concern is technical rather than foundational: the endpoint claim may still be true and repairable using more delicate Schrödinger kernel estimates, but as written the proof of Theorem 1.2 does not go through. Since the reader already conditioned acceptance on clarifying exactly this step, I do not recommend changing the verdict.","tokens_in":30007,"tokens_out":15621,"duration_ms":146851,"concrete_test":"Verify the key inequality in Theorem 1.2 Step III against the full τ-integral. Concretely: (1) in the model case V=0, where Γ is explicit, check whether the displayed bound can be derived by substituting the manuscript's gradient estimate into ∫_R (-iτ)^{-1/2}(1+|τ|^{1/2}R_y)^m dτ; this integral diverges, so identify the missing cancellation or decay. (2) Determine whether Shen's [17] estimates provide a genuinely τ-decaying bound, e.g. |∇_yΓ(y,x_1,τ)-∇_yΓ(y,x_0,τ)| ≲ |x_1-x_0|^δ |τ|^{-1/2-η} for some η>0, uniformly in the relevant annuli. If no such bound is available from [17] or proved in the manuscript, Theorem 1.2 remains incomplete.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 4, Step III (the displayed estimate after 'Furthermore'): the kernel difference R_j(y,x_1)-R_j(y,x_0) is written as -1/(2π)∫_R (-iτ)^{-1/2}[∇_yΓ(y,x_1,τ)-∇_yΓ(y,x_0,τ)]dτ. The preceding estimates give only |∇_yΓ(y,x_1,τ)-∇_yΓ(y,x_0,τ)| ≲ (1+|τ|^{1/2}R_y)^m (|x_1-x_0|/r_0)^δ (···). Substituting this into the τ-integral would require bounding ∫_R |τ|^{-1/2}(1+|τ|^{1/2}R_y)^m dτ, which diverges at infinity for every m ≥ 0. No principal-value or cancellation mechanism is stated, and no τ-decay estimate for the gradient difference is proved. The displayed inequality jumps from the integral representation to a finite right-hand side without this step. Since T1 is the nonlocal part of R_j^*(φ), this gap is what carries the continuity of R_j^*(φ); without it (4.6) and (4.4) do not follow. Thus Theorem 1.2 and the measure representation Lemma 4.1 rest on an unproved technical step. The concern does not directly affect Theorems 1.1 or 1.3, but it is central to the paper's Part II.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies endpoint boundedness for Schrödinger operators L = -Δ + V with nonnegative V in the reverse Hölder class RH_q, q ≥ n/2, acting on the space CMO_L(R^n), the closure of C∞_c in BMO_L. Theorem 1.1 states that the Hardy–Littlewood maximal operator is bounded on BMO_L and maps CMO_L into itself. Theorem 1.2 states that each adjoint Riesz transform (∂_j L^{-1/2})^* maps C_0(R^n) into CMO_L. Theorem 1.3 states that the Poisson and heat semigroups associated to L give an approximation to the identity in BMO_L norm for functions in CMO_L, with uniform convergence for smooth compactly supported functions. The paper also derives a Riesz-type representation for continuous linear functionals on CMO_L in terms of finite Borel measures whose Riesz transforms are again measures.","tokens_in":30322,"tokens_out":13020,"duration_ms":116793,"significance":"If the results are correct, they fill a natural gap in the endpoint theory for CMO_L: the maximal operator, Riesz transforms at the C_0-to-CMO endpoint, and semigroup approximation are all relevant to applications in harmonic analysis and PDE. The paper is clearly written and builds on established machinery from Shen [17] and Song–Wu [18]; the proof of Theorem 1.1 is a careful adaptation of classical arguments, and Theorem 1.3 is a plausible extension of known results. The main novel contribution, Theorem 1.2, is the most technically demanding but contains an unproved convergence step that is load-bearing. The Riesz representation lemma, which depends on Theorem 1.2, is also affected. The paper is honest about limitations and open questions, and the overall strategy is credible, but the proof of Theorem 1.2 is not complete as written.","major_comments":[{"comment":"The proof of (4.6) writes the difference |R_j(y,x_1)-R_j(y,x_0)| as the absolute value of an integral over τ of (-iτ)^{-1/2} [∇_yΓ(y,x_1,τ)-∇_yΓ(y,x_0,τ)] dτ and then bounds this by the finite right-hand side. However, the preceding estimate for the gradient difference contains a factor (1+|τ|^{1/2} R_y)^m; substituting that estimate into the τ-integral would require bound on ∫_R |τ|^{-1/2} (1+|τ|^{1/2} R_y)^m dτ, which diverges at infinity for every m ≥ 0. No principal value, cancellation mechanism, or τ-decay estimate for the gradient difference is given. Since this step is exactly what proves the continuity of T_1(φ), Theorem 1.2, and hence Lemma 4.1, rest on an unproved technical point.","section":"Section 4, Step III (display after 'Furthermore')"},{"comment":"The identity expressing R_j(y,x_1)-R_j(y,x_0) as a τ-integral of (-iτ)^{-1/2}∇_yΓ is stated with the word 'Furthermore' but no proof or citation. This identity is the starting point of the estimate for T_1, so its validity and the precise sense in which the integral converges (in particular, whether it is an ordinary improper integral or a principal value) need to be established. If the identity is taken from Shen [17] or another reference, the source should be cited explicitly and the convergence condition checked.","section":"Section 4, Step III (identity for the Riesz kernel difference)"}],"minor_comments":[{"comment":"The text contains typos: 'there eixsts a cube Q whose sigdelength is 2 r_B' should read 'there exists a cube Q whose sidelength is 2 r_B'; and after (3.6c), 'cP dentoes' should be 'c_P denotes'.","section":"Section 3, Step II"},{"comment":"The definition of ΔE(x) is written incorrectly: the expression (E_x ∩ E_{x_0}) \\ (E_x ∩ E_{x_0}) is always empty; the intended object is the symmetric difference E_x Δ E_{x_0}. This should be corrected to make the convergence argument for E(x_1) meaningful.","section":"Section 4, Step III"},{"comment":"In the verification of η_2(F_s), the text writes 'η_2(F_2)=0' where the index should be s; also in (5.2) the constant is written as 'CC' with a double C.","section":"Section 5, Lemma 5.1 proof"},{"comment":"The notation 'R^*_j(C_0(R))' should be 'R^*_j(C_0(R^n))'; the current notation suggests a one-dimensional domain.","section":"Remark 3.1"},{"comment":"The norm expression in the display after (4.2) is typeset awkwardly with nested norms of the maximal function; this should be clarified for readability.","section":"Section 4, Step I"}],"recommendation":"major_revision","confidential_remarks":"The main issue is confined to Theorem 1.2, specifically the convergence of the τ-integral in Step III. The rest of the paper (Theorems 1.1 and 1.3) appears sound. The gap is technical but essential; if the authors can supply a valid τ-decay estimate for the gradient difference or otherwise justify the passage from the integral representation to the finite bound, the paper would be suitable for publication. Given the paper's reliance on existing machinery from [17] and [18], this is likely repairable within the scope of a revision. The presentation is generally careful but contains several typos and one incorrect set-difference expression."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is a genuine new contribution to the CMO_L program. The three theorems are new in the Schrodinger setting, and the paper correctly notes that the classical CMO maximal-operator boundedness follows as a corollary. The overall plan is sensible: use the Song-Wu vanishing characterizations, Shen's auxiliary function, maximal estimates, and tent-space/semigroup tools. Theorems 1.1 and 1.3 read as plausible, and the proof of Theorem 1.3 via tent spaces plus the Kato-Trotter comparison is a clean route. I would not desk-reject this.\n\nBut the stress-test note is right, and it is not a nit. In Section 4, Step III of the proof of Theorem 1.2, the kernel difference R_j(y,x1)-R_j(y,x0) is written as a tau-integral of (-i tau)^{-1/2} grad_y Gamma(y,x,tau). The preceding estimate gives only a pointwise bound on the gradient difference with a factor (1+|tau|^{1/2} R_y)^m. Substituting that bound into the tau-integral would require bounding the integral of |tau|^{-1/2}(1+|tau|^{1/2})^m d tau, which diverges at infinity. The paper simply drops the tau-dependent factor and writes a finite right-hand side. No cancellation, principal value, or tau-decay for the gradient difference is shown. This step is the core of the continuity proof for T1, hence for R_j^*(phi) in C0 and for Theorem 1.2. Lemma 4.1 leans on Theorem 1.2, so that consequence is also not established as written.\n\nSecond, in Theorem 1.1 Step III, the authors replace L2 oscillation with L1 oscillation and say a John-Nirenberg type inequality takes care of it. That is a real but softer gap: the local inequality one needs is not stated, and the printed step is hand-wavy. I expect it can be repaired with a standard local John-Nirenberg estimate, but a referee should ask for details.\n\nWhat the paper does well: no invented entities, no fitted parameters; the methods are honest modifications of Bennett-DeVore-Sharpley, Shen, and Song-Wu; the exposition is mostly clear; and the citation pattern is fine. The gap in Theorem 1.2 does not directly damage Theorems 1.1 or 1.3.\n\nBottom line: send it to a serious referee, but be explicit that Part II needs a written proof of the missing tau estimate, or a different argument for continuity of R_j^*. If Theorem 1.2 cannot be fixed, the paper still has real value through Theorems 1.1 and 1.3. I would not cite Theorem 1.2 as it stands.","headline":"Theorem 1.2 has a real, load-bearing gap in the tau-integral step; Theorems 1.1 and 1.3 look solid, and the paper deserves a serious referee.","tokens_in":749,"tokens_out":880,"would_cite":true,"duration_ms":94793,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["42B20","42B25","42B35"],"pacs":[],"model":"deepseek-v4-flash","headline":"For Schrödinger operators with reverse-Hölder potentials, the paper establishes endpoint boundedness on the associated CMO space: the maximal operator is bounded, adjoint Riesz transforms send C0 into it, and semigroup approximations…","keywords":["maximal operator","Riesz transforms","CMO space","Schrödinger operators","reverse Hölder class","vanishing mean oscillation","BMO_L","semigroup approximation"],"falsifier":"Keep the τ-dependent factor in the displayed inequality bounding |R_j(y,x_1) - R_j(y,x_0)|: direct application of the triangle inequality would require integrating |τ|^{-1/2}(1 + |τ|^{1/2}R_y)^m over τ ∈ R, which diverges for every m ≥ 0. A reader can settle the claim by finding a cancellation or decay estimate that makes this integral converge, or by exhibiting a potential V ∈ RH_q for which the corresponding kernel difference fails the stated bound.","tokens_in":29791,"feed_emoji":"📐","tokens_out":8461,"duration_ms":81352,"temperature":0.7,"pith_summary":"The paper aims to prove endpoint boundedness for the standard singular integral operators on CMO_L(R^n), the space of functions of vanishing mean oscillation associated to a Schrödinger operator L = -Δ + V with nonnegative potential V in a reverse Hölder class. It claims three results: the Hardy–Littlewood maximal operator preserves CMO_L; each adjoint Riesz transform (∂_j $L^{{-1/2}}$)^* is bounded from continuous functions vanishing at infinity into CMO_L; and the Poisson and heat semigroups $e^{{-t√L}}$ and $e^{{-tL}}$ converge to the identity on CMO_L in BMO_L norm as t→0. If true, these results give endpoint boundedness and an approximation identity for the nonclassical CMO_L space, and they recover the classical CMO analogues by taking V=0.","feed_headline":"Schrödinger CMO space gains maximal and Riesz endpoint bounds","feed_subtitle":"Hardy–Littlewood maximality, adjoint Riesz transforms, and semigroup limits all behave well on the nonclassical CMO_L space.","key_machinery":"The load-bearing object is the critical-radius function ρ(x) introduced in the paper's main reference [17]; the BMO_L norm averages over balls of radius r_B < ρ(x_B) with the mean subtracted and over balls of radius r_B ≥ ρ(x_B) without the mean. Around this, the arguments use three structural tools: the characterization of CMO_L by vanishing of the mean-oscillation functionals γ_1–γ_5 from [18], the tent-space characterization of CMO_L via the Poisson semigroup also from [18], and Shen's kernel estimates comparing the Riesz kernel of L with the classical Riesz kernel. For the Riesz adjoints, the proof writes the kernel through an integral in τ of ∇_yΓ(y,x,τ), where Γ is the fundamental solution of -Δ + (V + iτ), and uses Morrey-type Hölder estimates for ∇_yΓ. For the semigroup approximation, the key quantitative input is a Kato–Trotter comparison showing the heat kernel of L differs from that of -Δ by a factor (√t/(√t+ρ(x)))^δ with δ>0.","core_discovery":"The central discovery is that endpoint singular-integral theory, classically formulated for the Laplacian, survives for Schrödinger operators once the critical-length function ρ(x) = sup{r>0 : $r^{{-(n-2)}}$ ∫_{B(x,r)} V ≤ 1} is used to split local and nonlocal estimates. With V ∈ RH_q for q ≥ n/2, the paper proves that M is bounded on BMO_L and maps CMO_L into itself, that each R_j^* maps C_0(R^n) into CMO_L, and that $e^{{-t√L}}$f and $e^{{-tL}}$f converge to f in BMO_L norm for every f ∈ CMO_L, with uniform convergence for compactly supported smooth f. As a consequence, the paper derives a Riesz-type representation for the dual of CMO_L: every continuous linear functional is integration against a finite Borel measure whose L-Riesz transforms are also finite Borel measures; specializing to V=0 recovers the classical duality (CMO)^* = $H^{1}$.","pith_inferences":["The representation lemma invites an L-analogue of the F. and M. Riesz theorem: show that the representing measure is absolutely continuous with density in H^1_L. The paper leaves this open; a natural test is whether the Poisson–Stieltjes integrals satisfy subharmonicity of |F_L|^p for p ≤ 1.","Theorem 1.1's mechanism suggests a transfer principle: any operator bounded on BMO_L whose estimates respect the ρ(x) splitting should preserve CMO_L once the three vanishing conditions γ_1, γ_3, γ_5 are checked; fractional maximal operators are a direct test case.","The strict inclusions CMO_L ⊊ CMO ⊊ VMO mean the endpoint results for L are genuinely sharper than the classical ones; comparing whether M maps CMO into VMO could identify which of the classical vanishing conditions is lost versus gained."],"forward_implications":["The maximal operator becomes a usable tool on CMO_L: since M preserves the space, density and truncation arguments that require applying M to CMO functions are valid in the nonclassical setting.","The semigroup approximation theorem gives a canonical mollifier compatible with L: to prove a statement for all CMO_L functions it suffices to prove it for e^{-t√L}f or e^{-tL}f and then pass t→0, avoiding convolution kernels that generally leave CMO_L.","The duality representation upgrades the abstract predual information: every bounded linear functional on CMO_L is integration against a finite measure, and the L-Riesz transforms of that measure are finite measures; with V=0 this recovers the classical identification (CMO)^* = H^1.","For the classical CMO with V=0, Theorem 1.1 settles the boundedness of M on CMO(R^n) for the functions on which M f is finite, while the example f(x)=ln ln |x| shows the obstruction to finiteness is real.","The results complement the known L^p bounds for R_j^* with the endpoint p=∞ statement, giving a new route to boundary-value problems for harmonic functions of L with CMO data."],"supporting_citations":[{"why":"Supplies the auxiliary critical-radius function ρ, the doubling property, and the kernel estimates comparing R_j with the classical Riesz kernel; load-bearing for Theorems 1.1 and 1.2.","marker":"[17]"},{"why":"Characterizes CMO_L via mean oscillation and tent spaces, and shows CMO_L is the BMO_L-closure of C_0; used throughout all three theorems and in Lemma 4.1.","marker":"[18]"},{"why":"Defines BMO_L, proves the John–Nirenberg type inequality, and establishes boundedness of the Poisson maximal operator on BMO_L; foundational for Theorem 1.1 and Lemma 5.1.","marker":"[9]"},{"why":"The Bennett–DeVore–Sharpley theorem on boundedness of the Hardy–Littlewood maximal operator on classical BMO is the argument being refined in Theorem 1.1.","marker":"[1]"},{"why":"Provides the representation of BMO_L functions in terms of R_j^* applied to L^∞ functions, used in Remark 3.1 and around Theorem 1.2.","marker":"[21]"},{"why":"Supplies the Kato–Trotter comparison estimate for the heat kernel of L versus that of -Δ, used in the proof of Theorem 1.3.","marker":"[4]"},{"why":"Gives the boundedness of R_j from H^1_L to L^1, used in the duality argument of Lemma 4.1.","marker":"[10]"}],"fun_headline_variants":["Maximal and Riesz endpoint bounds on Schrödinger CMO","CMO_L tamed: maximal, Riesz, and semigroup limits","Schrödinger CMO gains endpoint singular integral bounds","Riesz transforms map C0 into Schrödinger CMO space","Dual of CMO_L identified via Riesz measures"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof of the Riesz-transform continuity assumes that an integral over the auxiliary evolution parameter τ remains convergent after a τ-dependent factor is discarded; this interchange is not shown in the paper, and without it Theorem 1.2 is not established.","fun_headline_variants_meta":{"raw":{"variants":["Maximal and Riesz endpoint bounds on Schrödinger CMO","CMO_L tamed: maximal, Riesz, and semigroup limits","Schrödinger CMO gains endpoint singular integral bounds","Riesz transforms map C0 into Schrödinger CMO space","Dual of CMO_L identified via Riesz measures"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000353,"raw_usage":{"total_tokens":2016,"prompt_tokens":1132,"completion_tokens":884,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":748,"completion_tokens_details":{"reasoning_tokens":798}},"tokens_in":748,"tokens_out":884,"duration_ms":8289,"temperature":1.0,"reasoning_tokens":798,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T10:54:51.544369+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Keep the τ-dependent factor in the displayed inequality bounding |R_j(y,x_1) - R_j(y,x_0)|: direct application of the triangle inequality would require integrating |τ|^{-1/2}(1 + |τ|^{1/2}R_y)^m over τ ∈ R, which diverges for every m ≥ 0. A reader can settle the claim by finding a cancellation or decay estimate that makes this integral converge, or by exhibiting a potential V ∈ RH_q for which the corresponding kernel difference fails the stated bound.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the auxiliary critical-radius function ρ, the doubling property, and the kernel estimates comparing R_j with the classical Riesz kernel; load-bearing for Theorems 1.1 and 1.2."},{"cited_title":"Song and L.C","cited_arxiv_id":null,"evidence_quote":"Characterizes CMO_L via mean oscillation and tent spaces, and shows CMO_L is the BMO_L-closure of C_0; used throughout all three theorems and in Lemma 4.1."},{"cited_title":"Dziuba´ nski, G","cited_arxiv_id":null,"evidence_quote":"Defines BMO_L, proves the John–Nirenberg type inequality, and establishes boundedness of the Poisson maximal operator on BMO_L; foundational for Theorem 1.1 and Lemma 5.1."},{"cited_title":"Bennett, R.A","cited_arxiv_id":null,"evidence_quote":"The Bennett–DeVore–Sharpley theorem on boundedness of the Hardy–Littlewood maximal operator on classical BMO is the argument being refined in Theorem 1.1."},{"cited_title":"Wu and L.X","cited_arxiv_id":null,"evidence_quote":"Provides the representation of BMO_L functions in terms of R_j^* applied to L^∞ functions, used in Remark 3.1 and around Theorem 1.2."},{"cited_title":"Bui, X.T","cited_arxiv_id":null,"evidence_quote":"Supplies the Kato–Trotter comparison estimate for the heat kernel of L versus that of -Δ, used in the proof of Theorem 1.3."},{"cited_title":"Dziuba´ nski and J","cited_arxiv_id":null,"evidence_quote":"Gives the boundedness of R_j from H^1_L to L^1, used in the duality argument of Lemma 4.1."}],"review_version":1}