{"id":"b423babb-f061-4f2a-aa07-6d98224580f7","arxiv_id":"2412.15988","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Limiting heights of generic translates of complete intersections in toric varieties equal mixed integrals of roof and Ronkin functions, settling Gualdi's conjecture.","lead":"The paper proves a formula for the limiting height of generic translates of complete intersections in toric varieties, confirming a conjecture of Roberto Gualdi. It also builds a topological space for schemes over globally valued fields on which fiber heights vary continuously, a new tool for studying heights in families.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 4.2's non-Archimedean Fubini step is unproved: the equality of the height of V with the Ronkin-divisor intersection product assumes boundary-null and product-measure statements on analytic cycles that are not established (Section 4, 'We sketch an argument').","rationale":"The reader's weakest_assumption is exactly the non-Archimedean Fubini principle in Theorem 4.2; I agree that this is the load-bearing point. Lemma 4.1 uses Theorem 3.1 and equidistribution to express the limit as a height on V; Theorem 4.2 converts that height into an intersection with Ronkin divisors; Lemma 4.3 then evaluates the latter combinatorially. If Theorem 4.2's Fubini claim fails, there is no bridge between the GVF limit and the mixed integrals, so Gualdi's conjecture is not proved. The authors themselves flag the step as a sketch and cite [Sto21] without checking the product-measure statement for the specific cycles and metrics used here. The boundary-nullset assertion is particularly delicate: for toric compactifications, Chambert-Loir measures often charge boundary strata, and the restriction of these measures to a complete intersection is not automatically supported in the open torus. The proposed test isolates this issue: for the two hypersurfaces in P^2, the complete intersection is a finite set whose points can be translated by small points, and the boundary in X is a toric divisor; explicit computation at p=3 would show whether the open-torus iterated integral reproduces the true intersection number. I do not see a contradiction with the result, and if the Fubini step can be proved the argument likely goes through, so the verdict should remain CONDITIONAL. The paper does provide substantial independent scaffolding (e.g., Theorem 2.21 from [BPS14] and the known [GS23] example), but the central application depends on this unproved analytic step, so a conditional verdict is appropriate.","tokens_in":30018,"tokens_out":23469,"duration_ms":210705,"concrete_test":"Test the Fubini step in the minimal non-trivial case: K=Q_p, T=P^2, n=2, m=2, f_1=x_1+x_2+1, f_2=x_1+1, D_0=O(1) with the canonical metric at a non-Archimedean place (say p=3). Compute the limit height of the finite sets ζ_{1,j}V_1∩ζ_{2,j}V_2 directly and compare with the mixed-integral formula of Theorem 4.5. Then compute both sides of Theorem 4.2 explicitly: the left-hand intersection number deg(R_1 R_2 π^*D_0 | X) via toric combinatorics (Theorem 2.21), and the right-hand height of \\tilde{V}; finally, evaluate the Fubini-sketched iterated integral over the open torus and check whether the boundary term is zero. If the boundary term is non-zero, the Fubini step is false.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In the proof of Theorem 4.2, the authors reduce integrals over div(g_1)∩...∩div(g_{r-1})∩X to integrals over the open torus chart (T×∏_{i<r} T)×T×∏_{i>r} T, asserting 'a Zariski closed subset with empty interior is a nullset with respect to a measure associated to differential forms' and then invoking an adaptation of [Sto21, Prop 3.4.21] with 'the general case follows by approximation.' This is not a proof. For Chambert-Loir measures attached to semipositive toric metrics on a toric compactification, the toric boundary can carry positive mass; the restriction of such measures to the complete-intersection cycle is a residual measure whose support can meet the boundary (e.g., when some g_i is a monomial, or when the intersection contains toric strata). The subsequent 'inner integral' computation on torus fibers is exactly what makes the left-hand side equal to the right-hand side. If the Fubini factorization or the boundary vanishing fails, the identity deg(R_1...R_m D...|X)=deg(D...|\\tilde{V}) is unsupported, and Theorem 4.5 collapses because Lemma 4.1 and Lemma 4.3 both rely on this equality.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a theory of globally valued field (GVF) analytifications for finite type schemes and proves a continuity theorem for arithmetic intersection numbers in flat projective families over a GVF base. The main abstract result is Theorem 3.1/3.23: for a flat projective morphism X→S of relative dimension d and globally integrable line bundles L0,...,Ld on X/S, the fiber height s↦deg(L0,...,Ld|Xs) is continuous on S_GVF. The proof proceeds by expressing the intersection number as a limit of heights of resultants over the base, with a uniform error term (Remark 3.16). This continuity result is then applied to prove Gualdi's conjecture on limit heights of complete intersections in toric varieties (Theorem 4.5): for Laurent polynomials f1,...,fm over a number field and semipositive toric Zhang divisors D0,...,D_{n-m} on a proper toric variety T, the limit of the heights of ζ_{1,j}V1∩...∩ζ_{m,j}Vm along a generic small sequence equals a sum of mixed integrals of the roof functions of the Di and the Legendre transforms of the Ronkin functions of the fi. The proof combines the continuity theorem with a polarised GVF structure on the function field of the translation torus T^m, a resultant/Ronkin divisor computation (Theorem 4.2), and convex-geometric projection formulas (Lemmas 2.26 and 2.27).","tokens_in":30255,"tokens_out":6178,"duration_ms":58799,"significance":"If the missing technical arguments are supplied, this is a substantial contribution. It provides a general and apparently new continuity theorem for heights in flat families over GVF bases, with an explicit and quantitative resultant route rather than a purely abstract compactness argument. The application to Gualdi's conjecture is a concrete, checkable result: it gives the limit height as a sum of mixed integrals, generalizing the earlier special cases of Gualdi-Sombra. The paper also contains a self-contained Appendix A with an elementary Mahler-measure estimate needed for the resultant comparison. The GVF framework is imported from companion papers by the same authors, but the central uniform-convergence argument for resultant heights and the toric application appear original. However, two load-bearing steps are only sketched: the non-Archimedean Fubini/vanishing argument in Theorem 4.2 and the density of arithmetically ample divisors in semipositive Zhang divisors used in Theorem 3.24. These need to be proved or precisely referenced before the main theorems can be considered established.","major_comments":[{"comment":"The identity deg(R1...Rm·π_h^*D0...π_h^*D_{n-m}·π_1^*O(1)^n...π_m^*O(1)^n|X) = deg(π_h^*D0...π_h^*D_{n-m}·π_1^*O(1)^n...π_m^*O(1)^n|V-tilde) is not proved. The paragraph beginning 'We need to be slightly careful when applying Fubini's theorem' explicitly says 'We sketch an argument' and concludes 'The general case follows by approximation.' The reduction to the open torus chart assumes that 'a Zariski closed subset with empty interior is a nullset with respect to a measure associated to differential forms,' and then invokes an adaptation of [Sto21, Proposition 3.4.21]. For Chambert-Loir measures attached to semipositive toric metrics on a toric compactification, this null-set statement is not automatic: the toric boundary can carry positive mass, and the residual measure obtained on div(g1)∩...∩div(g_{r-1}) can meet toric strata, for example when one of the gi is a monomial. This equality is load-bearing: Lemma 4.1 and Lemma 4.3 both feed into Theorem 4.5 through it, so Theorem 4.5 collapses if the Fubini factorization or the boundary vanishing fails. A complete proof of these analytic facts, or a reference covering exactly this situation, is required.","section":"Section 4, proof of Theorem 4.2"},{"comment":"The statement 'Arithmetically ample divisors in turn are dense in semipositive Zhang divisors allowing us to finish the proof' is asserted without proof or reference. This density is not a formality; it is an arithmetic Demailly-type approximation statement, closely related to work of Charles and of Qu-Yin (see [QY23]). It is exactly the step that upgrades continuity from lattice line bundles to integrable Zhang divisors, and it is used in Corollary 3.2 and in the application in Section 4 (Lemma 4.1 says the construction 'makes sense by Theorem 3.24'). The proof given only addresses approximation by arithmetically ample divisors; the semipositive case is not supplied. Please provide a proof of the density assertion or a precise reference with a verification that its hypotheses are satisfied in the present setting.","section":"Section 3.4, proof of Theorem 3.24"},{"comment":"The reduction 'We apply the projection formula to restrict to the case, where the NP(fi) define divisors on T' is not spelled out. Theorem 4.4 assumes NP(fi) define divisors on T, while Theorem 4.5 does not; the reduction must be checked on both sides of the limit identity, including the behavior of the translated hypersurfaces ζ_{i,j}Vi and of the Ronkin divisors under the toric modification. Please state the projection formula used and explain why the mixed-integral expression is unchanged under this reduction.","section":"Section 4, proof of Theorem 4.5"}],"minor_comments":[{"comment":"The letter T is used both for the proper toric variety and for its open torus T=G^n; this is very confusing in statements such as 'T = G^n ⊂ T'. Please introduce a separate notation for the torus, for instance T0 or G^n.","section":"Throughout, Theorem 1.1 and Section 4"},{"comment":"The notation ω(a):=−log|a|_ω and then h(x1,...,xn):=∫ −min_i(ω(x_i)) dν(ω) overloads ω for both a place and the associated valuation; please write val_ω(a) or |a|_ω consistently.","section":"Definition 2.2"},{"comment":"The map V→T^m is said to be flat over a dense Zariski open U⊆T^m and Theorem 3.1 is applied to V/U, but the limit point η_can of the generic small sequence is asserted to lie in U_GVF. Please justify this, or state explicitly that the construction of the limit point accounts for the flat locus.","section":"Section 4, before Lemma 4.1"},{"comment":"The notation for the Ronkin divisors changes between the theorem and the lemma: in Theorem 4.2 the Ri are associated to gi=fi(z1w_{1,i}^{-1},...,znw_{n,i}^{-1}) on a toric blow-up of T×∏ P^n, while Lemma 4.3 assumes 'Ri denote the Ronkin line bundle associated to fi and suppose it is already defined on T'. Please clarify the relationship between these two sets of divisors and the toric blow-up X.","section":"Theorem 4.2 and Lemma 4.3"},{"comment":"The statement of Proposition A.1 is correct in substance, but the notation m_{S_{m1}}×...×S_{mn}(P) is easy to misread as a product of measures rather than an integral over a product of spheres; a short sentence defining the measure would improve readability.","section":"Appendix A, Proposition A.1"}],"recommendation":"major_revision","confidential_remarks":"The manuscript relies heavily on [Ben+24] and [Sza23], which are coauthored by two of the present authors and by the third author respectively. This is not circular: the central resultant-based continuity argument and the toric application are new. However, the editor may wish to verify that the cited GVF foundations are publicly available and that the novelty of the present paper is clearly separated from that foundational material. The paper fits the journal's scope, but the two sketched arguments identified in the major comments are essential; I would recommend acceptance only after those points are resolved."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: this paper takes a real shot at Gualdi's conjecture for arbitrary m and, to my reading, hits it modulo two explicitly flagged technical gaps. The main new idea — prove a continuity/definability statement for heights in flat families over globally valued fields, then apply it to a family of translates of complete intersections and use Yuan's equidistribution — is original and well suited to the problem. The earlier special cases (P^2 with f1=f2=x1+x2+1, then m=2) used local logarithmic equidistribution; this paper instead goes through uniform convergence of resultant heights and a polarised GVF structure, which is a different and genuinely new route.\n\nThe resultants section is the strongest part: the intersection product on lattice line bundles is computed as a limit of heights of resultants with an explicit error bound (Remark 3.16), and the continuity of the fibre height then falls out in a clean epsilon argument. The appendix on mixed Mahler measures is self-contained and useful. The authors also honestly flag where they are sketching: Theorem 3.24 asserts without proof that arithmetically ample divisors are dense in semipositive Zhang divisors, and the non-Archimedean Fubini step in Theorem 4.2 is introduced with 'we sketch an argument' and 'the general case follows by approximation.' Both are load-bearing for the final theorem. I do not see an outright contradiction, and the specific worry that toric boundary strata carry positive Chambert-Loir mass does not land for the toric metrics used here — their supports sit in the open torus — but the iterative slicing argument on the complete-intersection cycle does need a real proof, not a sketch. The self-citation point is a red herring: the GVF framework comes from papers by two of the authors, but the central resultant argument is independent and the reliance is legitimate.\n\nMy overall take: this is a substantial advance, with a plausible and mostly explicit proof, but the two sketched steps mean the paper is not yet a finished reference. I'd want a referee to push hard on Theorem 4.2 before accepting the main theorem as proved. Who is it for: Arakelov geometers and model theorists working on heights, and toric geometers. It deserves serious peer review, and in the meantime I would cite the continuity theorem and the Gualdi result with a caveat.","headline":"Proves the full Gualdi conjecture via a new GVF-analytification continuity theorem; the strategy is original and mostly solid, but two load-bearing steps are sketched rather than proved.","tokens_in":30853,"tokens_out":6398,"would_cite":true,"duration_ms":57137,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14G40","14M25","52B20","03C66"],"pacs":[],"model":"deepseek-v4-flash","headline":"Heights of cycles vary continuously in flat families, yielding an exact toric limit-height formula.","keywords":["toric variety","Arakelov geometry","globally valued field","height of cycles","complete intersection","Ronkin function","mixed integral","limit height"],"falsifier":"Compute both sides of Theorem 4.5 independently in a new configuration, for example $K=\\mathbb{Q}$, $T=\\mathbb{P}^2$, $m=2$, $f_1=f_2=x_1+x_2+1$ with the same toric divisors used in the paper; the known value $2\\zeta(3)/(3\\zeta(2))$ provides a sharp target, so any mismatch would refute the theorem. Alternatively, test the Fubini step in Theorem 4.2 directly: over a non-Archimedean field, evaluate the double integral of $\\log|g\\,s|$ against $\\pi^*c_1(\\mathcal{O}(1))^n$ on the vanishing locus of $g$ and check whether the inner fiber integral vanishes identically; a nonzero boundary term would break the equality.","tokens_in":29778,"feed_emoji":"📐","tokens_out":14931,"duration_ms":117878,"temperature":0.7,"pith_summary":"The paper establishes that arithmetic heights of cycles vary continuously in flat projective families, once the base is viewed through its 'globally valued field' (GVF) analytification—a space whose points carry height functions satisfying a product formula rather than just valuations. From this continuity result it derives an exact limit formula: if Laurent polynomials define hypersurfaces in a proper toric variety, then the height of the intersection of their generically translated copies, in the limit of a generic small torus point, equals a sum over places of mixed integrals of the roof functions of the ambient divisors and the Legendre transforms of the Ronkin functions of the polynomials. This proves the conjecture posed in [Gua18a]. The payoff is that a transcendental-looking limit of heights becomes an explicit convex-geometric quantity, computable from Newton polytopes and concave functions; in one worked case it equals $2\\zeta(3)/(3\\zeta(2))$.","feed_headline":"Exact limit formula for heights of translated toric intersections","feed_subtitle":"Continuity of heights over globally valued fields turns the limit into a mixed-integral formula.","key_machinery":"The load-bearing device is the GVF analytification of a finite-type $K$-scheme $S$: points are pairs of a scheme point and a height function on its residue field extending the height of $K$, topologized so that every tuple of regular functions has continuous height. Over this space the paper defines globally integrable line bundles by uniform approximation from lattice line bundles pulled back from projective spaces with Weil or Fubini–Study metrics. The continuity theorem for $\\widehat{\\deg}$ on fibers is proved by expressing the intersection number through heights of resultants, which are manifestly continuous in the GVF topology. In the toric application the computation is carried by three further objects: Ronkin divisors, whose roof functions convert the height of a hypersurface into a height on the toric variety; the mixed integral, which polarizes integrals of concave functions just as mixed volume polarizes volume; and a non-Archimedean Fubini principle that lets integrals over the product torus be evaluated fiber by fiber, with the mutual vanishing of the Ronkin function and $\\log|f|$ on each fiber doing the required cancellation.","core_discovery":"On the paper's own terms, the central discovery is Theorem 4.5: for Laurent polynomials $f_1,\\dots,f_m$ in $n$ variables over a number field $K$, a proper toric variety $T$ with torus $\\mathbb{G}_m^n\\subset T$, and semipositive toric Zhang divisors $D_0,\\dots,D_{n-m}$, the limit $$\\lim_{j\\to\\infty}\\widehat{\\deg}(D_0,\\dots,D_{n-m}|_{\\zeta_{1,j}V_1\\cap\\cdots\\cap \\zeta_{m,j}V_m})=\\sum_{v\\in M_K} n_v \\operatorname{MI}(\\theta_{0,v},\\dots,\\theta_{n-m,v},\\rho_1^\\vee,\\dots,\\rho_m^\\vee)$$ holds for every generic small sequence $(\\zeta_{1,j},\\dots,\\zeta_{m,j})$ in the torus, i.e. for torus points whose Weil height tends to zero and which are chosen generically. Here $V_i$ is the hypersurface of $f_i$, $\\theta_{i,v}$ are the local roof functions of the toric divisors, and $\\rho_i^\\vee$ are Legendre transforms of the Ronkin functions. The proof combines a general continuity theorem for fiber heights in flat projective families over a GVF with a toric comparison between intersections on the family and intersections with Ronkin divisors; the core of that comparison is a non-Archimedean Fubini step.","pith_inferences":["A fully formal proof of the non-Archimedean Fubini step would make the same average-intersection method available for other families equipped with a torus fibration and product measures, not only complete intersections in toric varieties.","The continuity theorem suggests that the GVF analytification itself is a natural home for height-convergence statements: rather than fixing one polarisation and applying an equidistribution theorem, one could study convergence of all fiber heights directly on this space.","The mixed-integral right-hand side is an explicit computational target: for new polynomials and toric divisors, one can produce numerical predictions for generic small sequences and check them by direct height computation, which tests both the theorem and the Fubini step."],"forward_implications":["For a flat projective family over a number field, if a sequence of base points has convergent small-point heights, then all fiber intersection heights converge, with the limit read off from the GVF analytification and identified with an arithmetic intersection number by Proposition 3.26.","The height limit for translated complete intersections is independent of the chosen generic small sequence and depends only on Newton polytopes and concave roof functions of the polynomials and divisors.","Arithmetic heights of these intersections become explicit convex-geometric quantities; in the $m=2$, $T=\\mathbb{P}^2$, $f_1=f_2=x_1+x_2+1$ case the limit equals $2\\zeta(3)/(3\\zeta(2))$.","Because every integrable Zhang line bundle on a projective variety over a number field is globally integrable, the continuity theorem applies to all semipositive toric divisors appearing in the conjecture."],"supporting_citations":[{"why":"States the conjecture (Conjecture 6.4.4) that the generic-small limit of heights of translated complete intersections equals a mixed integral of roof and Ronkin functions; proving it is the paper's goal.","marker":"[Gua18a]"},{"why":"Introduces Ronkin metrics for Laurent polynomials and the identity replacing the height of a hypersurface by an intersection with its Ronkin divisor, which the proof applies fiberwise.","marker":"[Gua18b]"},{"why":"Provides the arithmetic geometry of toric varieties: roof functions of toric Zhang divisors, mixed integrals, and the formula expressing intersection numbers as sums of mixed integrals over places.","marker":"[BPS14]"},{"why":"Supplies the adelic-curve intersection product, the resultant-height computation used to prove continuity, and Proposition 4.5.1 used to identify intersections over a polarised GVF with arithmetic intersections on a model.","marker":"[CM21]"},{"why":"The equidistribution theorem is the source of the generic small sequences and the polarised GVF structure on the function field, letting the continuity theorem apply to the toric limit.","marker":"[Yua08]"},{"why":"The non-Archimedean Fubini principle adapted in the proof of Theorem 4.2; the paper sketches the adaptation and says the general case follows by approximation.","marker":"[Sto21]"},{"why":"Underpins the theory of semipositive and integrable Zhang divisors and the arithmetic-ampleness approximation used to show every integrable Zhang line bundle is globally integrable.","marker":"[Zha95a]"},{"why":"Defines adelic curves, semipositivity, integrability, and the upper-integral distance between metric families, which the definition of global line bundles over a GVF uses.","marker":"[CM20]"}],"fun_headline_variants":["Height continuity gives exact toric limit formula","Gualdi conjecture proved via height continuity","Limit heights of toric intersections become integrals","Flat families yield continuous heights, Gualdi resolved"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is the non-Archimedean Fubini principle used in Theorem 4.2: integrals of the forms defining the height of the intersection can be evaluated iteratively over the torus fibers with product measures and no boundary contributions, a step the paper sketches rather than proves in full.","fun_headline_variants_meta":{"raw":{"variants":["Height continuity gives exact toric limit formula","Gualdi conjecture proved via height continuity","Limit heights of toric intersections become integrals","Flat families yield continuous heights, Gualdi resolved"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000255,"raw_usage":{"total_tokens":1560,"prompt_tokens":923,"completion_tokens":637,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":539,"completion_tokens_details":{"reasoning_tokens":580}},"tokens_in":539,"tokens_out":637,"duration_ms":6120,"temperature":1.0,"reasoning_tokens":580,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T10:54:04.491942+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute both sides of Theorem 4.5 independently in a new configuration, for example $K=\\mathbb{Q}$, $T=\\mathbb{P}^2$, $m=2$, $f_1=f_2=x_1+x_2+1$ with the same toric divisors used in the paper; the known value $2\\zeta(3)/(3\\zeta(2))$ provides a sharp target, so any mismatch would refute the theorem. Alternatively, test the Fubini step in Theorem 4.2 directly: over a non-Archimedean field, evaluate the double integral of $\\log|g\\,s|$ against $\\pi^*c_1(\\mathcal{O}(1))^n$ on the vanishing locus of $g$ and check whether the inner fiber integral vanishes identically; a nonzero boundary term would break the equality.","supporting_citations":[],"review_version":1}