REVIEW 2 major objections 6 minor
Log Canonical Models and Positive Geometries
T0 review · 2 major / 6 minor · reviewed 2026-07-31 · grok-4.5
Pith's one-line read Canonical forms of positive geometries supply explicit coordinates and equations for log canonical compactifications of a large class of open varieties.
desk verdict Clean conditional theorem that turns canonical forms into log-canonical coordinates, plus real equations and code for classical families; soft spots are already scoped as open generation questions. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
Theorem 3.2: the Brown–Dupont canonical-form map from relative homology of a genus-zero pair onto the space of logarithmic forms identifies, under degree-one generation of the log canonical ring, the Proj of that ring with the image of the map by those forms; positive-geometry residues then give concrete coordinates.
What would settle it
Exhibit a variety meeting the geometric hypotheses of Theorem 3.2 whose log canonical ring is not generated in degree one, or show that the real-region canonical forms span a proper subspace of the logarithmic forms while degree-one generation still holds—either breaks the claim that the image is the log canonical model.
Extended reading notes
Core claim
If an open variety U has a log canonical compactification and a compactification (X, Y) of genus zero whose log canonical ring is generated in degree one, then the rational map given by a basis of logarithmic top forms is the log canonical embedding of U. When (X, Y) is moreover a positive arrangement with real combinatorial rank equal to combinatorial rank, the canonical forms of the real regions alone realize that embedding, so the equations of the model can be read off by implicitization.
Load-bearing premise
The identification needs the log canonical ring to be generated in degree one, a ring-theoretic condition the paper assumes or cites rather than proves for every new example.
Editorial extensions
If this is right
- Log canonical models of essential connected hyperplane arrangement complements are cut out scheme-theoretically by quadrics and are recoverable from adjoint polynomials of bounded regions.
- The eighth Veronese of a smooth cubic surface is the log canonical embedding of the surface minus its 27 lines, with an explicit ideal of 5586 quadrics.
- The homogeneous ideals of the log canonical model of Y(3,6) and of the canonical-form closure of X(3,6) have 9605 and 6616 minimal quadratic generators respectively.
- Parke–Taylor varieties are the special case of this construction for the moduli space M_{0,n}.
- Whenever genus zero, degree-one generation, and equality of real and combinatorial rank can be checked, the same recipe produces coordinates and equations.
Reading between the lines
- If degree-one generation holds for X(3,6) and X(3,7), the already-computed canonical-form spans would settle the equations of their log canonical models.
- Sehr-schön tropical compactifications offer a combinatorial route to combinatorial rank, so the method may scale to other configuration spaces once generation is known.
- Equality of real and combinatorial rank is likely the practical bottleneck for nonlinear examples beyond del Pezzo surfaces; higher Y(3,n) is a natural test.
- The degeneration examples suggest that flat limits of these embeddings may record how log canonical models jump when arrangements specialize.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper shows that for open varieties U admitting a log canonical compactification, if a compactification (X,Y) has genus zero and the log canonical ring is generated in degree one, then the rational map given by a basis of logarithmic n-forms is the log canonical embedding; when moreover (X,Y) is a positive arrangement with equal real and combinatorial rank, the canonical forms of the real regions realize that embedding (Theorem 3.2). The argument rests on Brown–Dupont’s canonical-form map and invariance under modifications, together with the definition of Proj of the log canonical ring. The authors recover the Hacking–Keel–Tevelev models of hyperplane-arrangement complements (with an algorithm and explicit equations), compute the eighth Veronese embedding of a cubic surface with its 27 lines removed, give quadratic generators for the images associated to X(3,6) and Y(3,6), and identify Parke–Taylor varieties with the log canonical models of M_{0,n}. Code for the computations is provided.
Significance. The work supplies a uniform, computationally effective coordinate system for log canonical models in a class of examples that includes several classical moduli and arrangement spaces. Theorem 3.2 cleanly unifies prior observations (hyperplane arrangements, Parke–Taylor/M_{0,n}, cubic surfaces) under the Brown–Dupont framework and yields new explicit equations (e.g., 5586 quadrics for the cubic surface, 6616/9605 quadrics for the X(3,6)/Y(3,6) images). The accompanying code and the scheme-theoretic/quadratic claims for arrangement complements are concrete contributions that other researchers can reuse. The conditional nature of the main theorem is appropriately scoped: generation in degree one is cited from the literature where known and flagged as open (Conjecture 5.5) where not.
major comments (2)
- [Proposition 4.1] Proposition 4.1 asserts that the HKT compactification S is defined by quadrics as a scheme. The proof states that Ge is cut out by quadrics and that the matroid toric variety T is cut out by quadrics “up to saturation” (citing [29, Thm 3]), then concludes the claim for the scheme-theoretic intersection. Saturation can change the scheme structure along the irrelevant locus; a short argument is needed that the saturation step does not affect the intersection with Ge inside the Plücker space (or that the saturated ideal remains generated by quadrics after intersecting). As written, the scheme-theoretic statement is not fully justified.
- [§5.2, Abstract, Introduction] In §5.2 the authors construct X_cf(3,6) via canonical forms, prove it is four-dimensional and birational to X(3,6), and obtain 6616 quadratic generators, but correctly note that identification with the log canonical model still requires degree-one generation of the log canonical ring (left open). The abstract and introduction list “the moduli space of marked cubic del Pezzo surfaces” and configuration spaces among the applications of the theory without always distinguishing the cases where Theorem 3.2 applies unconditionally (Y(3,6) via [18,38]) from those where an extra ring-theoretic hypothesis remains (X(3,6)). A brief clarifying sentence in the introduction would prevent over-reading the X(3,6) computation as a completed log-canonical embedding.
minor comments (6)
- [Affiliations] Affiliation line for the second author: “Max Planck Instiute for Physics” → “Institute”.
- [Example 2.6] Example 2.6 and Figure 1: the non-SNC arrangement is helpful; a one-line reminder that the four-dimensional space of logarithmic forms is recovered after blow-up would make the comparison with H^0(P^2, Ω^2(5H)) fully self-contained.
- [Algorithm 1] Algorithm 1, step 7: “Implicitize the variety parametrized by this basis” — specify whether a Gröbner-basis or numerical/interpolation method is intended, since the Julia code is the actual reference implementation.
- [Theorem 3.2, proof] In the proof of Theorem 3.2, the sentence “Since R(U) is generated in degree one, there is a surjective map Sym^•(Ω^n_log(U)) → R(U)” is correct but could note explicitly that this uses the identification Ω^n_log(U) ≅ H^0(X, K_X+Y) already established for SNC (and then for general) compactifications.
- [References] References [21] and [38] are arXiv preprints by overlapping author sets; ensuring the published or final versions are cited when available would help readers.
- [Throughout] Typographical: “Poincar´ e” and similar accented characters appear inconsistently encoded in a few places (e.g., Definition 2.7); normalize to a single LaTeX accent convention.
Circularity Check
No significant circularity: Theorem 3.2 is a conditional identification via external foundations, not a self-referential derivation.
-
self citation load bearing
[Section 5.3, Theorem 5.6 and its proof]
"The Parke–Taylor forms are the canonical forms of the regions of (M_{0,n}, D) [3, Sec. 5]. The log canonical ring is Koszul and thus generated in degree one by [25, Theorem 1.2]. We have g(M_{0,n}, D) = 0 by [5, Sec. 6.8]. Moreover, (M_{0,n}, D) is a modification of a real hyperplane arrangement, thus cr_R(M_{0,n}, D) = cr(M_{0,n}, D). Thus Theorem 3.2 applies."
Theorem 5.6 identifies PT_n with the log canonical compactification of M_{0,n} by invoking the authors' own prior paper [21] (which already claimed that PT_n is Keel–Tevelev's embedding) together with Theorem 3.2. The step is only mildly circular: the generation and genus facts are external ([25], [5]), and [21] is used mainly to name the forms; the logical content is the application of the new conditional theorem rather than a closed self-referential loop.
full rationale
The central claim (Theorem 3.2) assumes existence of the log canonical compactification and degree-one generation of the log canonical ring, then concludes that the map by logarithmic/canonical forms is the log canonical embedding when genus is zero (and that real-region forms realize it when cr_R = cr). The short proof in Section 3 uses the Brown–Dupont isomorphism Ω^n_log(U) ≅ H^0(X, K_X + Y) and the definition of Proj of the log canonical ring; both are external. Applications supply the generation hypothesis from independent prior theorems (Hacking–Keel–Tevelev, Eur–Fink–Larson, Keel–Tevelev on M_{0,n}, projective normality of cubic surfaces). Self-citations ([21] Parke–Taylor, [38] cubic surfaces) supply computational input and examples whose independence is supported by released code and by matching known combinatorial ranks (e.g. 126 for Trop Gr(3,6)). For X(3,6) the paper explicitly leaves degree-one generation open (Conjecture 5.5) and does not claim the identification unconditionally. No step reduces a prediction to a fitted input or renames the target by construction.
Assumptions & free parameters
assumptions (5)
- domain assumption Brown–Dupont: genus-zero pairs admit a surjective canonical-form map from relative homology to logarithmic top forms, invariant under modifications.
- standard math Log canonical compactification exists and is unique when it exists (Prop. 2.2); equals Proj of the log canonical ring when the ring is finitely generated.
- domain assumption For the example classes (connected essential hyperplane arrangements; smooth cubic with real lines; M_{0,n}; Y(3,6)), existence of the log canonical compactification and (where used) very ampleness or degree-one generation are taken from prior literature.
- domain assumption Real regions of the treated positive arrangements are positive geometries in the sense of Arkani-Hamed–Bai–Lam (with smooth complement of algebraic boundary).
- domain assumption Log canonical ring generated in degree one (hypothesis of Thm 3.2(1)); verified by citation for arrangements and M_{0,n}, left open for X(3,6).
Cite this review
Pith. "Pith review of Log Canonical Models and Positive Geometries." pith.science (2026). https://pith.science/paper/U7WBFJ7U
@misc{pith2026260728368,
author = {Pith},
title = {Pith review of: Log Canonical Models and Positive Geometries},
year = {2026},
howpublished = {\url{https://pith.science/paper/U7WBFJ7U}},
note = {Machine review of arXiv:2607.28368}
}
read the original abstract
Constructing log canonical compactifications of open varieties is a central problem in birational geometry. Finding a natural coordinate system and obtaining the equations of these models is difficult in general. We show that for a large class of varieties explicit coordinates for the log canonical model are provided by canonical forms of positive geometries, and use this to compute the equations of these models. Our theory applies, for instance, to complements of hyperplane arrangements, cubic surfaces with lines removed, and the moduli space of marked cubic del Pezzo surfaces.
Figures
Reviewed July 31, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.