{"id":"998fcc6d-35a2-4a30-8b30-48caa42d3b3e","arxiv_id":"1908.04028","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The BMO to BLO norm of the natural and classical dyadic maximal operators equals 1 for all n, via an explicit Bellman function on α-trees.","lead":"This paper computes the sharp Bellman function for the dyadic maximal operator acting from BMO into BLO, and proves the operator norm is 1 in every dimension. It also supplies an explicit norm-optimizing sequence, settling a question open since Bennett's 1982 paper.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 3.4's pasting construction does not preserve the BMO_d(Q0) bound: a dyadic subcube mixing both types can have variance >1, so the proof of local concavity fails.","rationale":"The reader's CONDITIONAL verdict focused on the α-concavity verification (Lemma 2.11). I agree that Section 4 is technical, but the more basic structural gap is Lemma 3.4. The upper-bound half may well be correct: modulo Lemma 2.11, Lemma 2.8 gives B_n≤A_n. The lower-bound half, however, is not merely a long computation; the only proof offered for local concavity of B_n uses a pasting operation that can leave the admissible class F_x. The concrete counterexample with constants shows the asserted membership is false for a standard binary partition. Since Lemma 3.4 is the base for Lemmas 3.5, 3.6, and the proof of Lemma 3.1, the sharpness of Theorem 1.1 and the equality in Theorem 2.13 currently rest on an invalid lemma. This does not show the theorem is false, and the norm-optimizing sequence in §3.1 is independent, but the general lower bound is unsupported. A repair might be possible by using a stopping-time partition with uniform conditional proportions and proving the BMO bound on all dyadic subcubes, or by replacing the pasting argument with a dynamic-programming induction. The verdict remains CONDITIONAL: the manuscript should not be accepted until Lemma 3.4 is repaired or replaced.","tokens_in":31379,"tokens_out":33819,"duration_ms":341664,"concrete_test":"Re-run Lemma 3.4's construction for n=1, α=1/2, γ=1/5 with φ−=−5/2χ_{Q0} and φ+=0, pasted according to the greedy binary partition realizing γ (left half −, then recursively 2/5 in the right half). Compute the variance on the dyadic cube R=(1/2,1); if it equals 3/2>1, the construction fails and a correct proof of local concavity must supply a BMO estimate on mixed dyadic cubes or restrict to dyadic-rational γ with an approximation argument.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 3's lower-bound argument rests on Lemma 3.4, which asserts B_n is locally concave on Ω*. The proof takes optimizing sequences for x− and x+, splits Q0 into two sets of measures 1−γ and γ, writes each set as a union of dyadic subcubes, pastes rescaled copies, and asserts the result lies in F_{(1−γ)x−+γx+}. No check is made on dyadic subcubes that contain pieces of both types. This is not a minor omission: for a general dyadic partition, such mixed subcubes exist, and the variance there can exceed 1 even when the global variance is 1. For n=1, γ=1/5, take φ−=−5/2 on Q0 and φ+=0 (both have BMO 0 and L=0), and realize γ by the greedy dyadic partition (left half −, then recursively target 2/5 in the right half). On the dyadic cube R=(1/2,1) the proportion of + pieces is 2/5, so ⟨φ²⟩_R−⟨φ⟩_R²=(2/5)(3/5)(5/2)²=3/2>1, while the variance on Q0 is 1. Thus the pasted function is not in F_x, and the proof of Lemma 3.4 as written is invalid. Lemma 3.4 is used in Lemmas 3.5, 3.6 and 3.1 to prove B_n≥B, hence the sharpness half of Theorem 1.1 and Theorem 2.13 are not established by the given argument.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the action of the natural dyadic maximal operator N (and the classical maximal operator M) from dyadic BMO into dyadic BLO, on general α-trees. The authors construct an explicit Bellman function for the corresponding extremal problem and prove an upper bound: for a function φ with BMO norm at most 1, ⟨Nφ⟩_Q ≤ L + Φ_n(t)‖φ‖_{BMO}, where L = inf_Q Nφ and t = L − ⟨φ⟩_Q. They show that Φ_n(k(2^{n/2}−2^{−n/2})) = 2^{−nk}, which implies that the BMO→BLO norm of N (and M) is at most 1. They also claim sharpness of the constants and provide an explicit norm-optimizing sequence.","tokens_in":31720,"tokens_out":10593,"duration_ms":102268,"significance":"If the claims are fully established, the paper would provide the first exact BMO→BLO norm for a maximal operator, equal to 1 in all dimensions, together with an explicit Bellman function and an explicit optimizer. The upper-bound portion is carefully developed: the α-concavity of the candidate B is verified in Section 4, and the Bellman induction in Section 2 is well motivated. The construction of an explicit norm-optimizing sequence in Section 3.1 is also a valuable contribution. However, the lower-bound/sharpness argument currently contains a gap that must be repaired before the main conclusions can be accepted.","major_comments":[{"comment":"The proof of Lemma 3.4 asserts that the pasted function φ_j belongs to F_{(1−γ)x−+γx+}; this is not generally true, because dyadic subcubes that contain pieces of both types can have variance exceeding 1. For example, in dimension n=1, take γ=1/5, φ−=−5/2 on Q0, φ+=0, and partition Q0 as in the stress-test: the left half is of type −, and in the right half a greedy dyadic partition is used to realize measure 2/5 of type +. On the dyadic subcube R=(1/2,1), the proportion of + pieces is 2/5, giving ⟨φ²⟩_R−⟨φ⟩_R² = (2/5)(3/5)(5/2)² = 3/2 > 1. Thus φ_j ∉ F_{(1−γ)x−+γx+} in general. Since Lemma 3.4 is used in the proofs of Lemmas 3.5, 3.6, and 3.1, the lower bound B_n ≥ B and hence Theorem 2.13 and the sharpness claims of Theorem 1.1 that depend on it are not established by the given argument.","section":"Section 3, Lemma 3.4"},{"comment":"Lemma 3.7, which establishes the properties of the norm-optimizing sequence {ψ_j}, is stated with its proof left as an exercise. This lemma is used immediately afterwards to conclude that the sequence {φ_j} is optimizing and that ‖M‖_{BMO→BLO} ≥ 1. For a central claim of sharpness, the proof should be included in the manuscript or a reference with a complete proof should be provided; delegating a load-bearing computational verification to the reader is not acceptable.","section":"Section 3.1, Lemma 3.7"}],"minor_comments":[{"comment":"The monotonicity and convexity of b on [0,∞) are left as an exercise; these properties are asserted in the proof of Theorem 1.4 and stated in Theorem 1.1 for Φ_n. Please include the verification or a citation.","section":"Section 2.3, after (2.20)–(2.21)"},{"comment":"There is a typo: 'both really heavily on' should be 'both rely heavily on'.","section":"Section 3, Lemma 3.4"},{"comment":"The proof of Lemma 2.1 is concise but the notation 'maximal subset' could be clarified; this is a presentation issue only.","section":"Section 2.1, Lemma 2.1"}],"recommendation":"major_revision","confidential_remarks":"The paper contains a substantial and interesting contribution, but the gap in Lemma 3.4 is serious and affects the sharpness half of the main theorem. I recommend major revision. The authors should address the counterexample and either repair Lemma 3.4 or replace it with a valid argument. The norm-optimizing sequence in Section 3.1 is a promising alternative but its proof must be included."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The headline: this paper claims the first sharp BMO-to-BLO norm for a maximal operator, constant 1 in every dimension, and backs it with an explicit Bellman function and an explicit optimizer. The upper-bound half is impressive and, as far as I can tell, correct. The lower-bound half has a gap in the proof of Lemma 3.4 that looks like a genuine flaw, not a minor omission.\n\nWhat is genuinely new: the explicit Bellman candidate in (2.11)–(2.18), the decay function Φ_n, the α-tree generalization in Theorem 1.4, and the norm-optimizing sequence in Section 3.1. The α-concavity verification in Section 4 is long and technical, but it is a real proof, not a hand-wave, and the upper bound follows from it cleanly. The paper does not fit parameters; the candidate is found heuristically and then verified independently. Credit where it is due.\n\nNow the soft spot. Lemma 3.4 asserts that the Bellman function B is locally concave on Ω*, and the proof pastes optimizing sequences along a dyadic partition. The claim that the pasted function lies in the admissible class F_x is not justified. On a dyadic subcube that contains pieces of both types—which will exist for any partition that is not aligned with the two functions—the variance can exceed 1 even when the global variance is 1. Concrete example, n=1, γ=1/5: take φ− = -5/2 and φ+ = 0 on Q0, both BMO 0, and use the greedy dyadic partition realizing γ. On R=(1/2,1), the proportion of the + piece is 2/5, so the variance there is (2/5)(3/5)(5/2)^2 = 3/2 > 1. So the pasted function is not in F_x, and Lemma 3.4 as written is invalid. Since Lemma 3.4 feeds into Lemmas 3.5, 3.6, and 3.1, the full Bellman equality (Theorem 2.13) and the sharpness of the decay function Φ_n for all t are not established by the given argument. The separate norm-optimizing sequence in Section 3.1 does prove the operator norm is at least 1, so the main norm result may survive, but the paper currently overclaims what is proved.\n\nAlso noted: monotonicity and convexity of b are left as exercises. That is minor, but it should be supplied or referenced.\n\nWho is this for: harmonic analysts and Bellman-function people. The paper deserves a serious referee, but the authors need to repair Lemma 3.4 or restrict the sharpness claims. I would send it to peer review, with a clear request for a corrected lower-bound argument.","headline":"A serious paper with a real upper-bound result and a load-bearing gap in the sharpness proof: Lemma 3.4's pasting argument does not preserve the BMO bound.","tokens_in":32277,"tokens_out":4784,"would_cite":true,"duration_ms":49661,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["42A05","42B35","49K20"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that the BMO-to-BLO norm of the dyadic maximal operator is 1 in every dimension, via an explicit Bellman function and an optimizing sequence.","keywords":["BMO","BLO","α-trees","maximal functions","explicit Bellman function","sharp constants","dyadic maximal operator","natural maximal operator"],"falsifier":"Evaluate the $\\alpha$-concavity inequality (2.7) numerically for the candidate $B$ on a dense grid of pairs $(x^-,x^+)$ and weights $\\beta\\in[\\alpha,1/2]$, taking $\\alpha=2^{-n}$ for small $n$ and refining near the boundaries between the domains $\\Omega_k$ and near $\\Gamma_1$. The inequality should be tight at the known equality configurations, for instance $\\beta=\\alpha$ with $x^-$ on $\\Gamma_0$ and $x^+$ on $\\Gamma_1$; a negative margin anywhere would disprove Lemma 2.11 and with it the upper bound in Theorem 1.1. Separately, compute $\\langle N\\varphi_j\\rangle_Q$ for the paper's explicit sequence and compare its limit with $L+\\Phi_n(t)$ to test the sharpness claim.","tokens_in":78,"feed_emoji":"📏","tokens_out":14287,"duration_ms":254125,"temperature":0.7,"pith_summary":"The paper solves the sharp norm problem for the dyadic maximal operator acting from BMO into the smaller space BLO, a problem for which no operator norm had been computed before. It proves that for the natural dyadic maximal operator $N$, every $\\mathrm{BMO}_d$ function satisfies $\\|N\\varphi\\|_{\\mathrm{BLO}_d} \\le \\|\\varphi\\|_{\\mathrm{BMO}_d}$, and that the constant 1 cannot be improved, in any dimension; the same sharp constant holds for the classical dyadic maximal operator $M$. The route is an explicit Bellman function on a parabolic domain, built from a quasi-periodic $\\alpha$-concave function adapted to $\\alpha$-trees. The same machinery yields a more detailed inequality with a decay function $\\Phi_n$ that is exponential at a natural ladder of points. A reader should care because exact norm constants for maximal operators are rare, and the Bellman construction here is explicit enough to produce both the extremal sequence and the dimension-free bound.","feed_headline":"Sharp BMO-to-BLO norm of dyadic maximal operator equals 1","feed_subtitle":"Explicit Bellman function and extremal sequence pin the constant at 1 in all dimensions.","key_machinery":"The machinery is the Bellman function $\\mathcal{B}_n(x,L)=\\sup\\{\\langle N\\varphi\\rangle_Q : \\varphi\\in E_{x,L,Q}\\}$, together with a family of auxiliary functions $A(\\,\\cdot\\,;L)$ built from a single candidate $B$ on the parabolic domain $\\Omega=\\{(x_1,x_2): x_1^2\\le x_2\\le x_1^2+1\\}$. The candidate $B$ is defined piecewise on subdomains $\\Omega_k$: it is linear on a foliation of line segments (the 'extremals'), and in the left part of the domain it repeats under the parabolic shift $T_a(x_1,x_2)=(x_1-a,x_2-2ax_1+a^2)$ with a scaling factor $\\alpha^k$, giving the quasi-periodic structure. Lemma 2.8 (Bellman induction) converts $\\alpha$-concavity of $A(\\,\\cdot\\,;L)$ into the upper bound on $\\langle N_T\\varphi\\rangle_K$, Lemma 2.12 packages $B$ into the required family, and the converse inequality $\\mathcal{B}_n\\ge\\mathcal{A}_n$ is proved from local concavity and boundary comparisons on the upper parabola $\\Gamma_1$.","core_discovery":"The core discovery is an exact bound for the natural dyadic maximal operator $N\\varphi(x)=\\sup_{J\\ni x}\\langle\\varphi\\rangle_J$: for any dyadic cube $Q$, with $L=\\inf_Q N\\varphi$ and $t=L-\\langle\\varphi\\rangle_Q$, one has $\\langle N\\varphi\\rangle_Q \\le L + \\Phi_n(t)\\|\\varphi\\|_{\\mathrm{BMO}_d(Q)}$, where $\\Phi_n$ is decreasing and convex and satisfies $\\Phi_n(k(2^{n/2}-2^{-n/2}))=2^{-nk}$ for nonnegative integers $k$. Setting $t=0$ and taking suprema gives $\\|N\\varphi\\|_{\\mathrm{BLO}_d}\\le\\|\\varphi\\|_{\\mathrm{BMO}_d}$, and the same inequality holds for the classical dyadic maximal operator $M$; both inequalities are sharp, so the norm is exactly 1 in every dimension. The proof identifies the Bellman function $\\mathcal{B}_n$ of the extremal problem with an explicit candidate $\\mathcal{A}_n$, and establishes the identification by proving the candidate is $\\alpha$-concave and by matching it on the boundary of the domain.","pith_inferences":["- The paper leaves open whether the same quasi-periodic Bellman candidate gives the exact norm on non-atomic $\\alpha$-trees with a guaranteed child of measure $\\alpha\\mu(K)$; the paper notes such trees admit optimizing sequences, so the missing piece is a matching upper-bound verification.","- Because the authors record an equivalence between $N:\\mathrm{BMO}\\to\\mathrm{BLO}$ and $M:\\mathcal{A}_\\infty\\to\\mathcal{A}_1$, the explicit candidate here may serve as a template for the sharp $\\mathcal{A}_\\infty\\to\\mathcal{A}_1$ constant, though sharpness does not transfer automatically.","- The norm-optimizing sequence is a one-dimensional construction extended by tensorization; one could test numerically whether genuinely $n$-dimensional rearrangements change the rate of approach to 1 for finite $n$, which would indicate whether the dimension-free constant is part of a stronger finite-dimensional phenomenon."],"forward_implications":["- For every $\\varphi\\in\\mathrm{BMO}_d(\\mathbb{R}^n)$, both the natural and the classical dyadic maximal operators satisfy $\\|N\\varphi\\|_{\\mathrm{BLO}_d}\\le\\|\\varphi\\|_{\\mathrm{BMO}_d}$, and the constant 1 is sharp for both.","- The refined inequality (1.3) holds with the explicit function $\\Phi_n$; at $t=k(2^{n/2}-2^{-n/2})$ the upper bound decays as $2^{-nk}$, so large gaps between the infimum and the cube average force the maximal-function average close to $L$.","- For any $\\alpha$-tree, the same Bellman induction yields $\\langle N_T\\varphi\\rangle_K \\le L+F_\\alpha(t)\\|\\varphi\\|_{\\mathrm{BMO}(T(K))}$ with $F_\\alpha(k(\\alpha^{-1/2}-\\alpha^{1/2}))=\\alpha^k$, and $N_T:\\mathrm{BMO}(T)\\to\\mathrm{BLO}(T)$ has norm at most 1.","- The Bellman identity $\\mathcal{B}_n=\\mathcal{A}_n$ solves the extremal problem for every admissible triple of average, square average, and external maximal level, not just for the norm constant.","- A simpler majorant $\\mathcal{A}_0$ already yields the norm constant 1 directly, while the full candidate is needed to capture the sharp decay of $\\Phi_n$."],"supporting_citations":[{"why":"Establishes that the Hardy–Littlewood maximal function maps BMO to itself, the starting point for the BMO-to-BLO question.","marker":"[3]"},{"why":"Proves the maximal function maps BMO into BLO but only with non-sharp estimates, the gap this paper closes.","marker":"[2]"},{"why":"Introduces BLO and characterizes it through A1 weights, fixing the target space of the theorem.","marker":"[4]"},{"why":"Supplies the Bellman-function framework for dyadic maximal operators and the convention of pinning the external maximal level L.","marker":"[10]"},{"why":"Introduces α-trees and provides the α-concavity lemmas, including the sufficiency criteria used in Section 4, on which the upper bound relies.","marker":"[17]"},{"why":"Gives the Monge–Ampère/PDE approach used to derive the Bellman candidate from boundary conditions.","marker":"[16]"},{"why":"Computes Bellman functions for dyadic-like maximal operators on L^p, whose boundary-function procedure Section 5 adapts.","marker":"[6]"},{"why":"Develops Bellman functions for extremal problems in BMO, providing the test-function setup behind the definition of B_n.","marker":"[5]"}],"fun_headline_variants":["BMO-to-BLO norm of dyadic maximal operator is exactly 1","Sharp norm: dyadic maximal operator maps BMO to BLO with constant 1","Exact Bellman function yields maximal operator norm 1 from BMO to BLO","Dyadic maximal operator: BMO-to-BLO norm equals 1 in every dimension","Dimension-free norm 1 for dyadic maximal operator BMO to BLO"],"cache_read_input_tokens":34304,"weakest_assumption_plain":"The upper bound depends on the claim that the specially constructed auxiliary function satisfies the required concavity inequality at every point of its domain; if that single condition fails, the norm-1 conclusion is not established.","fun_headline_variants_meta":{"raw":{"variants":["BMO-to-BLO norm of dyadic maximal operator is exactly 1","Sharp norm: dyadic maximal operator maps BMO to BLO with constant 1","Exact Bellman function yields maximal operator norm 1 from BMO to BLO","Dyadic maximal operator: BMO-to-BLO norm equals 1 in every dimension","Dimension-free norm 1 for dyadic maximal operator BMO to BLO"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00027,"raw_usage":{"total_tokens":1635,"prompt_tokens":966,"completion_tokens":669,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":582,"completion_tokens_details":{"reasoning_tokens":562}},"tokens_in":582,"tokens_out":669,"duration_ms":6291,"temperature":1.0,"reasoning_tokens":562,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:53:32.836755+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Evaluate the $\\alpha$-concavity inequality (2.7) numerically for the candidate $B$ on a dense grid of pairs $(x^-,x^+)$ and weights $\\beta\\in[\\alpha,1/2]$, taking $\\alpha=2^{-n}$ for small $n$ and refining near the boundaries between the domains $\\Omega_k$ and near $\\Gamma_1$. The inequality should be tight at the known equality configurations, for instance $\\beta=\\alpha$ with $x^-$ on $\\Gamma_0$ and $x^+$ on $\\Gamma_1$; a negative margin anywhere would disprove Lemma 2.11 and with it the upper bound in Theorem 1.1. Separately, compute $\\langle N\\varphi_j\\rangle_Q$ for the paper's explicit sequence and compare its limit with $L+\\Phi_n(t)$ to test the sharpness claim.","supporting_citations":[{"cited_title":"Bennett, R","cited_arxiv_id":null,"evidence_quote":"Establishes that the Hardy–Littlewood maximal function maps BMO to itself, the starting point for the BMO-to-BLO question."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Proves the maximal function maps BMO into BLO but only with non-sharp estimates, the gap this paper closes."},{"cited_title":"Coifman, R","cited_arxiv_id":null,"evidence_quote":"Introduces BLO and characterizes it through A1 weights, fixing the target space of the theorem."},{"cited_title":"Nazarov, S","cited_arxiv_id":null,"evidence_quote":"Supplies the Bellman-function framework for dyadic maximal operators and the convention of pinning the external maximal level L."},{"cited_title":"Slavin, V","cited_arxiv_id":null,"evidence_quote":"Introduces α-trees and provides the α-concavity lemmas, including the sufficiency criteria used in Section 4, on which the upper bound relies."},{"cited_title":"Slavin, A","cited_arxiv_id":null,"evidence_quote":"Gives the Monge–Ampère/PDE approach used to derive the Bellman candidate from boundary conditions."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Computes Bellman functions for dyadic-like maximal operators on L^p, whose boundary-function procedure Section 5 adapts."},{"cited_title":"Ivanishvili, N","cited_arxiv_id":null,"evidence_quote":"Develops Bellman functions for extremal problems in BMO, providing the test-function setup behind the definition of B_n."}],"review_version":1}