REVIEW 3 major objections 3 minor 9 references
A note on specializing semi-orthogonal decompositions
T0 review · 3 major / 3 minor · reviewed 2026-08-10 · deepseek-v4-flash
Pith's one-line read There is a smooth projective family of rational surfaces over a discrete valuation ring whose generic-fiber semi-orthogonal decomposition cannot be extended to the whole family, giving a negative answer to a question from [BOR20].
desk verdict Likely-correct negative answer to BOR20's question, but the key K0 comparison in Proposition 3.5 is not proven as written. 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
A semi-orthogonal decomposition is an ordered list of full subcategories whose morphisms vanish in one direction and which together generate the derived category. The machinery is the moduli space $\mathrm{SOD}_f \to B$ of semi-orthogonal decompositions in a family, the relative topological K-theory functor $K^{\mathrm{top}}_0(-/B)$, and the notion of a phantom: a nonzero admissible subcategory whose Grothendieck group is zero. The proof positions a phantom from [Kra23] on every general fiber, extends it étale-locally, and uses the local-system property of topological K-theory to force a phantom on the chosen central fiber, where [BK24] says none can exist.
What would settle it
Look at one general fiber of the constructed family and compute the comparison map $K_0(P_b) \to K^{\mathrm{top}}_0(P_b/\mathbb{C})$ for the phantom component; if it is not an isomorphism, the proof's propagation step is not valid, and the claimed contradiction is unsupported.
Extended reading notes
Core claim
Theorem 1.1 asserts the existence of a smooth projective family of rational surfaces $f: X \to \mathrm{Spec}(R)$ over a discrete valuation ring $R$ and a $K$-linear semi-orthogonal decomposition $\langle A_K, B_K\rangle$ of $D^b\mathrm{Coh}(X_K)$ that is not the base change of any $R$-linear semi-orthogonal decomposition of $D^b\mathrm{Coh}(X_R)$. This gives a negative answer to Question 8.8 of [BOR20] about the existence part of the valuative criterion for properness. The proof assumes a specialization exists, pulls it back to the closed fiber, shows the relevant component is a phantom, and derives a contradiction from a choice of closed fiber that cannot admit a phantom.
Load-bearing premise
The argument needs the K-theory vanishing of the phantom to propagate from a general fiber to the central fiber; this relies on a proposition that assumes the fiber has a full exceptional collection, while the fibers carrying the phantom do not have one, and no variant for this case is supplied.
Editorial extensions
If this is right
- The moduli space of semi-orthogonal decompositions of a family is not proper when the base is a curve, since the existence part of the valuative criterion for properness fails.
- Question 8.8 in [BOR20] is answered negatively: a decomposition over the generic point cannot always be completed to a decomposition over the whole family.
- Phantom components provide an obstruction to specializing semi-orthogonal decompositions, detected through K-theoretic invariants.
- The failure already occurs for families of rational surfaces, so it is not a feature of exotic or high-dimensional varieties.
Reading between the lines
- A natural next step, not taken in the paper, is to look for an explicit cohomological or deformation-theoretic obstruction class that detects when a generic decomposition cannot specialize; this example shows such an obstruction exists asymptotically but does not name it.
- If the K-theory comparison used for phantom components were replaced by another invariant that is deformation-invariant and local in families, the same strategy might produce counterexamples in other settings, such as the K3-category version the author leaves open.
- The role of the central fiber is only to be phantom-free; families whose special fiber is a del Pezzo surface might show specialization always holds, drawing a sharper boundary around the failure.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper claims to disprove the existence part of the valuative criterion for semi-orthogonal decompositions, answering negatively [BOR20, Question 8.8]. The proposed construction is a smooth projective family of rational surfaces over a discrete valuation ring whose generic fiber is a blow-up of the projective plane at ten general points, carrying the Krah phantom as a semi-orthogonal component. The proof shows that this generic decomposition cannot extend to an R-linear decomposition because the central fiber would then carry a phantom, contradicting the Borisov–Kemboi non-existence theorem for blow-ups at points on a smooth cubic. The argument relies on the BOR20 moduli theory of semi-orthogonal decompositions and on Perry's relative topological K-theory.
Significance. If the proof can be made rigorous, the result is a significant and clean counterexample in the deformation theory of semi-orthogonal decompositions, resolving a concrete open question. The paper is commendably short and uses deep external results appropriately, including Krah's phantom and Perry's relative K-theory, and it does not rely on any fitted parameters or ad hoc normalizations. However, as written, the proof contains a load-bearing gap in the step that shows the central fiber component is a phantom; this step is essential for the contradiction. The existence of the family itself is also asserted rather than constructed. With the main gap repaired, the result would be of clear interest to the derived-category and moduli communities.
major comments (3)
- [§3, construction of f] The proof states: "We now apply Proposition 3.1 to C and Bo respectively to obtain K0(P_k)=0." Proposition 3.1 requires a fiber X_b with a full exceptional collection. The fibers on which the Krah phantom lives are blow-ups of P^2 at ten general points; by Krah's Theorem 3.4, the phantom is the right orthogonal of an exceptional collection of length 13. If that exceptional collection were full, its right orthogonal would be the zero category, contradicting the existence of a nonzero phantom. Thus no such fiber satisfies the hypothesis. The central fiber is only known to have no phantom (Theorem 3.6), not to have a full exceptional collection. Since K0(P_k)=0 is the key step that makes P_k a phantom and produces the contradiction, this gap is load-bearing. The manuscript needs a separate lemma, for instance a weak form of Proposition 3.1: if over an open U the component B_U is the right orthogonal of a relative exceptional collection and K0(B_b)=K_top^0(B_b)=0 for b in U, then K0(P_c)=K0(P_b) for c in an étale neighborhood of b. This variant must be stated and proved, including the topological K-theory comparison for the phantom component, before the argument can proceed.
- [§3, construction of f] The proof begins "we choose a smooth projective morphism f : X → B ≅ P^1 such that..." but no construction or reference is given showing the existence of a family whose general fibers are blow-ups of P^2 at ten general points and whose special fiber is a blow-up at ten points lying on a smooth cubic. Since Theorem 1.1 is an existence statement, this is part of the required data. A short argument using a rational curve in the Hilbert scheme Hilb^10(P^2) passing through the two specified configurations would fill the gap, but as written the existence of this family is an unproved assumption.
- [§3, Proposition 3.1] In the commutative diagram argument, the proof claims that because the left and right arrows in the right diagram are injective, the top arrow K0(A_b)→K_top^0(A_b) is an isomorphism. Injectivity of the vertical maps together with an isomorphism on the bottom row does not imply an isomorphism on the top row unless the rows are compatibly split. The desired conclusion does follow from Theorem 2.2(3), which gives explicit direct-sum decompositions of both K0 and K_top^0, but the proof does not state this. The text should be rewritten to use those splittings explicitly; as written, the inference is invalid.
minor comments (3)
- [Throughout] The manuscript contains several typographical errors in the extracted text, such as "semi-orth ogonal", "d o not", "givin g", and "Ricolfi"; these should be corrected in the source file.
- [Notation] The symbol B is used both for the base scheme and for a component of the semi-orthogonal decomposition (e.g., ⟨A, B⟩), and later B_o is an open subset and B′ is a component. This is confusing; renaming the component, for instance to ⟨A, C⟩, would improve readability.
- [§3, proof of Proposition 3.5] The statement "Since C and SOD_f are both étale over B, we know that C → SOD_f is also étale" is not a standard cancellation property of étale morphisms and requires either a proof or a reference. If the statement is false in general, the argument should be rephrased to avoid needing it.
Circularity Check
No circular reasoning found: the construction is assembled from external theorems (BOR20, Perry, Krah, BK24) and contains no fitted parameters or self-citations.
full rationale
The paper derives Theorem 1.1 by combining BOR20's moduli-space theorem and étale extension result, Perry's relative topological K-theory, Krah's phantom on blow-ups of P^2 at ten general points, and BK24's non-existence of phantoms. None of these are author-defined normalizations or fitted parameters, and the paper contains no self-citations. The central claim—that a K-linear semi-orthogonal decomposition cannot be specialized over a DVR—is not defined in terms of the result it proves: the decomposition to be specialized is pulled back from the generic point over an étale neighborhood of a general fiber, and the obstruction is established by a contradiction argument using K-theory and uniqueness of local deformations. The only substantive concern is Proposition 3.5's invocation of Proposition 3.1: the stated hypothesis of a fiber with a full exceptional collection is not manifestly satisfied by the Krah general fibers, which carry a nonzero phantom and therefore cannot admit a full exceptional collection. If this is a genuine gap, it is a failure of the proof's inference rather than circularity: the conclusion K0(P_k)=0 does not coincide by construction with any input, and the needed weak variant is unstated rather than assumed. There is no equation in which the claimed result is equivalent to its premises, no fitted parameter is renamed as a prediction, and no load-bearing argument reduces to a self-citation. Accordingly no circular step is identified.
Assumptions & free parameters
assumptions (5)
- standard math BOR20 Theorem A: SOD_f is an algebraic space etale over B representing two-component linear semi-orthogonal decompositions, with pullback and etale extension properties.
- standard math Perry's Theorem 2.2: relative topological K-theory is a local system and splits along semi-orthogonal decompositions.
- standard math Krah's Theorem 3.4: a phantom exists on the blow-up of P^2 at ten general points.
- standard math BK24 Theorem 3.6: no phantom exists on the blow-up of P^2 at points in very general position on a smooth cubic.
- ad hoc to paper Existence of a family f: X -> P^1 with the prescribed general fibers (10 general points) and prescribed special fiber (points on a smooth cubic) and a compatible Krah phantom over an open subset.
Cite this review
Pith. "Pith review of A note on specializing semi-orthogonal decompositions." pith.science (2026). https://pith.science/paper/G32QH2O4
@misc{pith2026250110679,
author = {Pith},
title = {Pith review of: A note on specializing semi-orthogonal decompositions},
year = {2026},
howpublished = {\url{https://pith.science/paper/G32QH2O4}},
note = {Machine review of arXiv:2501.10679}
}
read the original abstract
We prove that families of semi-orthogonal decompositions do not satisfy the existence part of the valuative criterion for properness, giving a negative answer to a question posed by Belmans, Okawa, and Ricolfi.
Reference graph
Works this paper leans on
-
[1]
Lev Borisov and Kimoi Kemboi, Non-existence of phantoms on some non-generic blowups of the projective plane, arxiv:2405.01683
-
[2]
Anthony Blanc, Topological K -theory of complex noncommutative spaces, Compos. Math. 152 (2016), no. 3, 489–555
work page 2016
-
[3]
Ricolfi, Moduli spaces of semiorthogonal decompositions in families, arXiv:2002.03303
Pieter Belmans, Shinnosuke Okawa, and Andrea T. Ricolfi, Moduli spaces of semiorthogonal decompositions in families, arXiv:2002.03303
arXiv 2002
-
[4]
Sergey Gorchinskiy and Dmitri Orlov, Geometric phantom categories, Publ. Math. Inst. Hautes Études Sci. 117 (2013), 329–349
work page 2013
-
[5]
Krah, A phantom on a rational surface, A phantom on a rational surface, Invent
J. Krah, A phantom on a rational surface, A phantom on a rational surface, Invent. Math. 235 (2024), no. 3, 1009–1018
work page 2024
-
[6]
Alexander Kuznetsov, Base change for semiorthogonal decompositions, Compos. Math. 147 (2011), no. 3, 852–876
work page 2011
-
[7]
Tasos Moulinos, Derived Azumaya algebras and twisted K -theory, Adv. Math. 351 (2019), 761–803
work page 2019
-
[8]
Alexander Perry, The integral Hodge conjecture for two-dimensional Calabi-Yau categories, Compos. Math. 158 (2022), no. 2, 287–333
work page 2022
Show all 9 references
-
[9]
Dmitrii Pirozhkov, Admissible subcategories of del Pezzo surfaces, Adv. Math. 424 (2023), Paper No. 109046, 62 pp
2023
Reviewed August 10, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.