{"id":"003be400-b9f2-4fe6-9b08-d86bfd4f9d65","arxiv_id":"1908.07830","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"The intrinsic area of a self-similar growth-fragmentation has a C-infinity density that decays like r^{-1-omega_+/omega_-}, and this density permits tilting the process to condition on area A=r.","lead":"This paper proves that the intrinsic area of a self-similar growth-fragmentation has a smooth, everywhere positive density with a power-law tail, and then uses that density to condition the process on having a given area. This gives a rigorous way to speak of growth-fragmentations with fixed 'area', a step toward area-conditioned random surfaces such as the Brownian disk.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 3.3(i) for 2<p<ω+/ω− is not proved: the displayed BDG inequality used for p≤2 is reversed for p>2, so the O(x) moment bound backing Theorem 1.2(i) remains a gap.","rationale":"The reader's verdict is already CONDITIONAL, and the reader explicitly flags the omitted proof of Lemma 3.3(i) for 2<p<ω+/ω−. My read agrees with that concern and sharpens it: the sentence preceding the omission is not merely incomplete but, taken literally, relies on an inequality that reverses for p>2. The promised iterative compensation cannot be a straightforward repetition of the p≤2 argument. This gap is load-bearing because Lemma 3.3(i) is used at a strictly larger exponent in the proof of Lemma 3.3(ii), and Lemma 3.3(ii) is the remainder term in Lemma 3.4(ii), from which Theorem 1.2(i) follows. The no-positive-jump assumption is a genuine scope restriction rather than an internal inconsistency; the paper works within that assumption, so I do not object on that ground. If the moment estimate can be supplied with a careful BDG/compensator argument, the central claim should hold; until then the current conditional verdict is appropriate. My agreement is partial because the reader's named 'weakest_assumption' points to no-positive-jumps, whereas the verdict-driving issue in my reading is the unproved moment bound inside Lemma 3.3(i).","tokens_in":22279,"tokens_out":10834,"duration_ms":107549,"concrete_test":"Independently re-derive Lemma 3.3(i) for a fixed p∈(2,ω+/ω−), say p=3, using the Lévy–Itô decomposition and the compensator of [N(c)], without invoking (Σ a_i^2)^{p/2}≤Σ a_i^p. A valid derivation must establish E[[N(c)]^{p/2}(t+(x))]=O(x) as x→0+ and state any extra conditions on the Lévy measure of η+. A concrete case worth checking first is η+ a spectrally negative stable process of infinite activity with parameters satisfying κ(ω±)=0; if the O(x) bound fails or requires assumptions not present in the paper, then Lemma 3.3(i) is not established and the proof of Theorem 1.2(i) remains conditional.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing gap is in the proof of Lemma 3.3(i), which supplies E((A+(x))^p)=O(x) for every 1≤p<ω+/ω−. The text proves p≤2 via (Σ a_i^2)^{p/2} ≤ Σ a_i^p, which is valid exactly for p≤2, and then says the case p∈(2,4] is 'mostly similar', with an iterative argument 'left to the reader'. For p>2 the displayed inequality goes the other way: (Σ a_i^2)^{p/2} ≥ Σ a_i^p, since ℓ^p ⊂ ℓ^2, so the direct jump-sum bound cannot be iterated as stated. The promised compensation of [N(c)] to a martingale is not carried out; at minimum one must control E[[N(c)]^{p/2}(t+(x))] through the predictable compensator and handle the unbounded jumps of η+. This matters because Lemma 3.3(ii) invokes (i) at an exponent p′ strictly between p and ω+/ω−, Lemma 3.4(i) uses convergence of A+(x) to 0, and Lemma 3.4(ii) feeds directly into the density tail in Theorem 1.2(i). Without a correct proof of Lemma 3.3(i) for all p<ω+/ω−, the main theorem is incomplete. The no-positive-jump assumption is explicit and is not an internal flaw; the issue is the missing moment estimate inside the claimed proof.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies self-similar growth-fragmentations whose driving self-similar Markov process has no positive jumps and satisfies Cramér's condition, and investigates the law of their intrinsic area A, which is the terminal value of the intrinsic martingale. The main results are: Theorem 1.1, asserting that under P1 the law of A is absolutely continuous with a C∞ density a; Theorem 1.2, giving the sharp tail expansion a(r) ∼ (c ω+/ω−) r^{−1−ω+/ω−} as r→∞ together with positivity of a on (0,∞); and Section 4, where the density is used to construct regular conditional laws P1(·|A=r) by probability tilting, to prove a disintegration formula, to show that conditioning on A=r with r→∞ converges to tilting by the martingale M+(n), and to extend the construction to the canonical measure N0^- via pseudo-excursion measures. The proofs use the smoothing transform, a random affine equation derived by stopping the Eve trajectory at first passage of the associated spectrally negative Lévy process, path decompositions at the overall supremum, and the published global tail estimate for A.","tokens_in":22620,"tokens_out":32097,"duration_ms":293989,"significance":"If the results are correct, they give a local version of the Kesten–Grincevičius–Goldie theorem in a branching setting, and they provide a rigorous construction of growth-fragmentations conditioned on their intrinsic area, with explicit asymptotic descriptions. The paper is well organized and the overall architecture is convincing: the density is obtained through the smoothing transform, the asymptotic constant is identified with the global tail constant rather than introduced as a free parameter, the conditioning construction is explicit as a density tilt, and the authors are careful to state the scope of the no-positive-jump assumption. The reliance on the imported tail estimate (1) and the size-bias identity from [3] is transparent, so there is no circularity. However, one load-bearing proof in Section 3.3 is incomplete as written, which prevents me from recommending acceptance without further work.","major_comments":[{"comment":"The displayed estimate for p≤2 uses (Σ a_i^2)^{p/2} ≤ Σ a_i^p, which is valid exactly for p≤2; for p>2 the inequality is reversed, so the sentence “The case p∈(2,4] is mostly similar” does not follow from the preceding computation. The missing piece is an L^p bound on [N^{(c)}]^{p/2}(t_+(x)) for 2<p<ω+/ω−, together with control of the unbounded jumps of η+. This is load-bearing because Lemma 3.3(ii) invokes (i) at an exponent p′ strictly between p and ω+/ω−, and Lemma 3.4(ii) feeds directly into the density tail in Theorem 1.2(i). Please provide a complete proof, or a precise quotation of a lemma (for instance Lemma 2.3 in [3]) that covers the range p>2.","section":"Section 3.3, proof of Lemma 3.3(i)"},{"comment":"The paper applies Liu’s Theorem 2.1, stated for smoothing transforms with finitely many terms, and asserts that the arguments work for the infinite series in (12). This extension should be justified: the characteristic function involves an infinite product, and a truncation or domination argument is needed to pass from finite approximations. Since Theorem 1.1 supplies the density used in Lemma 3.1 and hence in the proof of Theorem 1.2, this is a substantive point, even though it is likely fixable by adding a short argument or a precise reference.","section":"Section 3.2, proof of Theorem 1.1"}],"minor_comments":[{"comment":"After dividing by x and letting x→0+, the text says “we get (i)”; this should read “we get (ii)”.","section":"Proof of Lemma 3.4(ii)"},{"comment":"The definition of a−(r) has “r ∈ R”; it should be “r > 0”.","section":"Lemma 3.1"},{"comment":"The displayed formula for the law of e^{xω−}A− appears to have the exponential factor in the wrong place; please check the change of variables and verify that the subsequent integral bounds correspond to the corrected expression.","section":"Proof of Lemma 3.3(ii)"},{"comment":"The Fatou step after conditioning on the sequence (γ_i^{ω−}) is very terse. Since the conclusion is a−(r)>0 (equivalently a(r)>0), please spell out the lower bound for the conditional density and the passage to the unconditional density.","section":"Proof of Theorem 1.2(ii)"},{"comment":"The phrase “in particular, except on a nowhere dense subset” is not a consequence of “except on a set with zero Lebesgue measure”; the intended statement is that the exceptional set has dense complement, which is sufficient for the continuity argument.","section":"Proof of Theorem 4.1"}],"recommendation":"major_revision","confidential_remarks":"The paper is transparent about the imported results from [3] and about the scope of the no-positive-jump assumption. The main unresolved issue is internal to the proof of Lemma 3.3 and is likely fixable within the manuscript's scope; I see no circularity or fit problem."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: the results are likely right and the machinery is genuinely useful, but there is a real hole in the proof of Lemma 3.3(i) for 2<p<ω+/ω−. The paper should go to a serious referee, with the expectation of a major revision.\n\nWhat's new: the smooth density for A (Theorem 1.1), the sharp local tail in Theorem 1.2(i), and positivity in 1.2(ii). The tilting martingale (a(B(n),r)) and the conditional limits (Corollary 4.5, Lemma 4.6, Proposition 4.7) are new and well-motivated by the planar map connection. The imported inputs from [3] are published theorems, not fitted constants, so circularity is not a concern.\n\nWhere it gets shaky: the proof of Lemma 3.3(i). For p≤2 the displayed inequality (Σ a_i^2)^{p/2} ≤ Σ a_i^p is correct and the BDG argument works. For p>2 that inequality is backwards. The text says the case p∈(2,4] is 'mostly similar' and leaves the details, and that is exactly where the proof breaks. You can't iterate the same step because the naive jump-sum bound goes the wrong way. At minimum you need to compensate [N(c)] to a martingale and control its p/2 moment through the predictable compensator, handling the unbounded jumps of η+; that is not written. Since Lemma 3.3(i) feeds directly into Lemma 3.4 and thus into the density tail in Theorem 1.2(i), the main theorem is incomplete as written.\n\nOther notes: the positivity proof in Theorem 1.2(ii) has a terse Fatou/conditioning step, but that looks repairable. The no-positive-jump assumption is explicit and not an internal flaw. Section 3.4 is a discussion, not a proof, so it shouldn't carry weight.\n\nBottom line: this is a serious paper by people who know the area, with a testable and substantial claim. The gap is real but probably fixable; it deserves peer review and a major revision. If you work on smoothing transforms or growth-fragmentations, cite it once you've verified the fix.","headline":"Strong and useful paper on conditioning self-similar growth-fragmentations by intrinsic area, but the proof of the key moment bound (Lemma 3.3(i)) has a genuine gap for p>2 that needs fixing.","tokens_in":23106,"tokens_out":2381,"would_cite":true,"duration_ms":21659,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60G18","60J80"],"pacs":[],"model":"deepseek-v4-flash","headline":"A self-similar growth-fragmentation can be conditioned on its intrinsic area, and the conditional law has a smooth positive density with a power-law tail.","keywords":["self-similar growth-fragmentation","intrinsic area","smoothing transform","random affine equation","spectrally negative Levy process","Cramer condition","conditional distribution","power-law tail"],"falsifier":"Simulate the intrinsic area A of a self-similar growth-fragmentation satisfying Cramer's condition and estimate the density a(r) at large r: Theorem 1.2 predicts a(r) > 0 for every r and a(r) ~ (c omega_+/omega_-) $r^{{-1-omega_+/omega_-}}$. Finding any r with a(r) = 0, or a large-r exponent different from 1 + omega_+/omega_-, would disprove the main claim; applying the same test to a positive-jump process would settle whether the extension discussed in Section 3.4 holds.","tokens_in":22096,"feed_emoji":"📈","tokens_out":7314,"duration_ms":158855,"temperature":0.7,"pith_summary":"This paper asks what it means to condition a self-similar growth-fragmentation — a branching process of cell masses — on the value of its intrinsic area, a random variable that arises as the terminal value of an additive martingale. The authors prove that the intrinsic area has a smooth density a(r), that a(r) is positive for every r > 0, and that a(r) decays as a power law with exponent 1 + omega_+/omega_- fixed by the two roots of the cumulant under Cramer's condition. They then use this density to build, by probability tilting, a regular version of the conditional law given A = r. This matters because such growth-fragmentations appear in random planar geometry, where the intrinsic area plays the role of the area of a random surface, so conditioning on A = r is a concrete way to fix the area of such a surface.","feed_headline":"Growth-fragmentations can be conditioned on their intrinsic area","feed_subtitle":"A martingale tilt builds a conditional law for every area r, and large areas match conditioning on indefinite growth.","key_machinery":"The intrinsic area A is the terminal value of the additive martingale M_-(n) = sum over generation-n birth masses raised to the power omega_-, where omega_- < omega_+ are the two roots of the cumulant kappa(q) = 0 under Cramer's condition. The load-bearing identities are the smoothing transform A = sum gamma_i A_i, the size-bias relation Q^-_1(A ∈ dr) = r a(r) dr, and the random affine equation A^- = A^+(x) + $e^{{x omega_-}}$ A^- obtained from the first-passage decomposition of a spectrally negative Levy process at level x. Letting x tend to 0 and using the path decomposition at the overall supremum converts the tail of A^- into the tail of its density a^-(r) = r a(r), which yields the exponent 1 + omega_+/omega_-. Positivity of a(r) then makes the density a(x,r) of weighted sums of independent intrinsic areas available as a martingale in the branching random walk, and tilting by a(B(n),r) constructs the conditional law.","core_discovery":"The central result is that the intrinsic area A of a self-similar growth-fragmentation has a C^∞ density a on (0,∞) with a(r)>0 for all r, and with the sharp asymptotic a(r) ~ (c omega_+/omega_-) $r^{{-1-omega_+/omega_-}}$ as r tends to infinity, where c is the same constant as in the tail P1(A>r) ~ c $r^{{-omega_+/omega_-}}$. The proof passes through the size-biased variable A^- with density r a(r), and through a random affine equation obtained by stopping the trajectory of the ancestor cell at its first passage above a level; the absence of positive jumps makes that first passage continuous, and letting the level go to zero converts tail information into local density information. From strict positivity of the density, the paper shows that (a(B(n),r)) is a martingale and that tilting by it defines a probability measure P1(· | A = r) which is a regular disintegration of P1 given the intrinsic area. It further shows that as r → ∞ this conditional law converges to the law obtained by tilting with the M+ martingale, so conditioning on a huge area is asymptotically equivalent to conditioning the growth-fragmentation on indefinite growth, and it constructs a canonical version starting from initial mass 0 with tail N^-_0(A>r) = c $r^{{-omega_Delta/omega_-}}$.","pith_inferences":["Beyond the paper, the density tail for infinite weight sequences should be testable numerically: Corollary 3.7 gives a lower bound, and the paper's conjecture in Section 3.5 says equality should hold, which would make the tail depend only on the sum of omega_+-moments of the birth masses.","Beyond the paper, the large-area limit in Corollary 4.5 suggests that random surfaces built from these growth-fragmentations should look, at large area, like surfaces conditioned on indefinite growth; comparing area-conditioned and unconditioned observables in simulations of the branching random walk could confirm this.","Beyond the paper, the canonical-measure construction is carried out for alpha < 0; adapting Lemma 4.6 to alpha >= 0 or to the boundary case omega_Delta = 0 would require a different normalization and is left open."],"forward_implications":["A regular conditional law P_x(· | A = r) exists for every positive initial mass x and every r > 0, and it disintegrates P_x through the intrinsic area.","For large areas, P_1(· | A = r) converges to the tilt of P_1 by the M_+(n) martingale, meaning that conditioning on a very large intrinsic area is asymptotically the same as conditioning the growth-fragmentation on indefinite growth.","Under the canonical measure N^-_0, conditioning on A = r remains well-defined for growth-fragmentations started from initial mass 0, and the area tail is exactly N^-_0(A>r) = c r^{-omega_Delta/omega_-}.","The density of a weighted sum of independent intrinsic areas is continuous in the weight sequence, and for finitely supported weights the density tail is c (omega_+/omega_-) (sum x_j^{omega_+}) r^{-1-omega_+/omega_-}.","The tilting construction itself uses only positivity of the density a, so it extends to the positive-jump growth-fragmentations discussed in Section 3.4 of the paper."],"supporting_citations":[{"why":"Defines the intrinsic area, supplies the tail estimate (1), the smoothing transform (12), and the spinal-decomposition identity used in Lemma 3.1.","marker":"[3]"},{"why":"Provides the smoothing-transform criterion for a smooth density that is used to prove Theorem 1.1.","marker":"[18]"},{"why":"Supplies the general tail estimate for additive martingales behind the tail behavior (1) imported through [3].","marker":"[17]"},{"why":"Gives the positivity P(A_i < b) > 0 used in the proof that a(r) > 0 for every r > 0.","marker":"[6]"},{"why":"Records the Kesten–Grincevičius–Goldie theorem and perpetuity equations that underlie the random affine equation analysis in Section 3.4.","marker":"[7]"},{"why":"Provides background on spectrally negative Levy processes, their first-passage subordinators, and the Lamperti transform used throughout Section 3.","marker":"[14]"},{"why":"Gives the path decomposition at the overall supremum of a spectrally positive Levy process used in Lemma 3.4.","marker":"[1]"},{"why":"Supplies the pseudo-excursion measures for self-similar Markov processes from which the canonical measure N^-_0 is constructed.","marker":"[24]"}],"fun_headline_variants":["Conditioning growth-fragmentations by their intrinsic area","Tilting martingales yields area-conditioned growth-fragmentations","For huge areas, conditioning matches indefinite growth","Smooth density for intrinsic area in growth-fragmentations","Area conditioning: from martingale tilt to large-area limit"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole density-tail proof assumes the cell-mass process only ever jumps downward; if upward jumps are allowed, the key decomposition at the moment the mass first exceeds a level breaks down.","fun_headline_variants_meta":{"raw":{"variants":["Conditioning growth-fragmentations by their intrinsic area","Tilting martingales yields area-conditioned growth-fragmentations","For huge areas, conditioning matches indefinite growth","Smooth density for intrinsic area in growth-fragmentations","Area conditioning: from martingale tilt to large-area limit"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000328,"raw_usage":{"total_tokens":1885,"prompt_tokens":1053,"completion_tokens":832,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":669,"completion_tokens_details":{"reasoning_tokens":752}},"tokens_in":669,"tokens_out":832,"duration_ms":425705,"temperature":1.0,"reasoning_tokens":752,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T11:55:37.774821+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Simulate the intrinsic area A of a self-similar growth-fragmentation satisfying Cramer's condition and estimate the density a(r) at large r: Theorem 1.2 predicts a(r) > 0 for every r and a(r) ~ (c omega_+/omega_-) $r^{{-1-omega_+/omega_-}}$. Finding any r with a(r) = 0, or a large-r exponent different from 1 + omega_+/omega_-, would disprove the main claim; applying the same test to a positive-jump process would settle whether the extension discussed in Section 3.4 holds.","supporting_citations":[{"cited_title":"Bertoin, T","cited_arxiv_id":null,"evidence_quote":"Defines the intrinsic area, supplies the tail estimate (1), the smoothing transform (12), and the spinal-decomposition identity used in Lemma 3.1."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the smoothing-transform criterion for a smooth density that is used to prove Theorem 1.1."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the general tail estimate for additive martingales behind the tail behavior (1) imported through [3]."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the positivity P(A_i < b) > 0 used in the proof that a(r) > 0 for every r > 0."},{"cited_title":"Buraczewski, E","cited_arxiv_id":null,"evidence_quote":"Records the Kesten–Grincevičius–Goldie theorem and perpetuity equations that underlie the random affine equation analysis in Section 3.4."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides background on spectrally negative Levy processes, their first-passage subordinators, and the Lamperti transform used throughout Section 3."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the path decomposition at the overall supremum of a spectrally positive Levy process used in Lemma 3.4."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the pseudo-excursion measures for self-similar Markov processes from which the canonical measure N^-_0 is constructed."}],"review_version":1}