REVIEW 4 major objections 4 minor 9 references
On Extensions of Partial Isomorphisms
T0 review · 4 major / 4 minor · reviewed 2026-08-14 · deepseek-v4-flash
Pith's one-line read Every countable structure extends to an ultraextensive one.
desk verdict A mostly solid paper with real new results; one proof step in Theorem 4.6 is unjustified as written, but repairable. 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
The load-bearing construction is the canonical HL-extension of a finite structure $C$. One partitions $C$ into maximal pieces of elements realizing the same unary predicates, fixes a representative $a_i$ in each piece, and forms the free group $F(P(C))$ generated by all partial isomorphisms of $C$. For each $i$, the stabilizer $H_i$ of $a_i$ under the partial action is a subgroup of $F(P(C))$, and the disjoint union of coset spaces $F(P(C))/H_i$ carries a natural $\mathcal{L}$-structure on which every generator $p$ acts by left multiplication; this gives the minimal HL-extension $(\Gamma,\Phi)$. Replacing each $H_i$ by $N_iH_i$ for finite-index normal subgroups $N_i$ yields finite quotients $\Gamma_{\vec N}$, and the paper shows that every finite minimal HL-extension is a homomorphic image of some such quotient. The passage to infinite ultraextensive structures is carried by a coherence condition on HL-extensions: extensions of nested finite substructures must agree on overlapping partial isomorphisms and embed the corresponding automorphism groups. The coherence theorem imported from the literature guarantees such coherent finite extensions exist, and the paper telescopes them into a countable union to obtain $U$.
What would settle it
Produce a finite $T$-free structure $C_2$, a substructure $C_1$, and a finite minimal HL-extension $(D_1,\varphi_1)$ of $C_1$ for which no finite $T$-free HL-extension of $C_2$ coherent with $(D_1,\varphi_1)$ exists, where every member of $T$ is a Gaifman clique. Theorem 4.3 says this can never happen; such an example would sink Theorem 4.3 and the subsequent ultraextensive existence results.
Extended reading notes
Core claim
The central claim is that ultrahomogeneity can be strengthened to ultraextensiveness without leaving the countable realm: for every countable $\mathcal{L}$-structure $C$ there is a countable ultraextensive $\mathcal{L}$-structure $U$ containing $C$, and if $T$ is a finite set of finite $\mathcal{L}$-structures each of which is a Gaifman clique, then the extension can be chosen $T$-free whenever $C$ is $T$-free. The proof builds $U$ as a union of finite structures in which every finite minimal HL-extension of a smaller substructure is coherently extended to a minimal HL-extension of every larger substructure, so that the partial isomorphisms at every level are recorded by genuine automorphisms in a compatible way. A second claim, proved by a free-group construction, is that every finite $T$-free, minimal HL-extension of a finite structure is a homomorphic image of one of the canonical extensions $\Gamma_{\vec N}=\bigsqcup_i F(P(C))/(N_iH_i)$ obtained by taking quotients by finite-index normal subgroups. A third claim is that the HL-property of a group—a strengthening of the property that finite products of finitely generated subgroups are closed in the profinite topology—is equivalent to a finite approximation property for actions on structures and is closed under finite free products. The paper's derived corollary is that $\mathrm{Aut}(U)$ has a dense locally finite subgroup for every countable ultraextensive $U$.
Load-bearing premise
The construction of ultraextensive extensions rests on an imported coherence theorem for finite structures: every finite structure avoiding a finite list of forbidden configurations has a finite extension, still avoiding them, in which automorphisms are assigned to all partial isomorphisms so that compositions match. If that theorem failed for even one forbidden family, the paper's main existence theorem would not follow.
Editorial extensions
If this is right
- Every countable relational structure can be embedded in a countable ultraextensive structure, so ultrahomogeneous structures with coherent extension behavior exist in abundance.
- When the forbidden family consists of Gaifman cliques, the ultraextensive extension can be chosen to avoid the family, so classes such as $K_n$-free graphs have ultrahomogeneous limits that are ultraextensive.
- The automorphism group of every countable ultraextensive structure has a dense locally finite subgroup, providing many examples of symmetry groups with this strong property.
- All finite minimal HL-extensions of a finite structure arise as homomorphic images of the canonical quotients $\Gamma_{\vec N}$, reducing the classification of minimal extensions to finite-index subgroups of free groups.
- The HL-property of groups is closed under finite free products, extending the known free-product closure of the weaker profinite-topology property.
Reading between the lines
- The coherent-extension construction suggests a route to building automorphism groups with prescribed dense locally finite subgroups: one could try to choose the finite stages so that the limiting subgroup is a specified direct limit of finite groups, and ask whether coherence forces the whole automorphism group to be determined by that subgroup.
- The paper's counterexample shows the Gaifman-clique condition is necessary for the coherent-extension theorem; one could test whether a weaker hypothesis, such as allowing only finitely many non-clique obstructions, still yields a modified ultraextensive existence theorem.
- Because the HL-property is characterized by a finite-approximation condition on actions, the same characterization might be formulated for other kinds of structures, such as metric or topological structures, where the free-product closure would then give new examples of groups with the property.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper studies HL-extensions, a framework for extending partial isomorphisms of finite relational structures to automorphisms of larger structures. In Section 2 it gives a description of all finite minimal T-free HL-extensions of a finite T-free structure as homomorphic images of canonical quotient structures built from finite-index normal subgroups of a free group on the partial isomorphisms. In Section 3 it introduces a group-theoretic HL-property, characterizes it in terms of a finite approximability condition for structures (Theorem 3.3), and claims that the HL-property is closed under free products (Corollary 3.8). In Section 4 it defines coherent HL-extensions and ultraextensive structures, proves that every countable structure has a countable ultraextensive extension (with a T-free version when T consists of Gaifman cliques), and concludes that every countable ultraextensive structure has an automorphism group with a dense locally finite subgroup. The paper also contains a counterexample showing that the Gaifman-clique assumption is necessary for the coherent extension theorem.
Significance. If the results are correct, the paper makes several meaningful contributions: the explicit description of minimal HL-extensions in Theorem 2.4; the equivalence between the HL-property and a structural approximation property in Theorem 3.3; the preservation of the HL-property under free products, extending Coulbois's corresponding result for the RZ-property; and the construction of ultraextensive structures with dense locally finite automorphism groups, including Henson graphs as a corollary. The proofs are often detailed and the paper is careful to point out the role of the Siniora–Solecki coherence theorem and to provide a counterexample for sharpness. However, several proof gaps need to be addressed before the claims are fully established.
major comments (4)
- [Theorem 4.6, proof, pp. 19–20] The proof concludes that \bigcup_{i=1}^\infty \operatorname{Aut}(D_i) is a dense locally finite subgroup of \operatorname{Aut}(U). This is not justified as written: each \operatorname{Aut}(D_i) is the automorphism group of a finite substructure of U, not a subgroup of \operatorname{Aut}(U), and the coherence condition in Definition 4.1(iii) only provides embeddings of the subgroups K_i = \langle \varphi_i(P_{C_i}) \rangle into K_{i+1}, not of the full groups \operatorname{Aut}(D_i). Arbitrary automorphisms of D_i need not extend to automorphisms of D_{i+1}. The argument can be repaired by taking the directed system K_1 \hookrightarrow K_2 \hookrightarrow \cdots given by the unique embeddings \kappa_i; each K_i is finite because D_i is finite, so the direct limit is locally finite, and density follows because every finite partial isomorphism of U lies in some P_{C_i} and is extended by an element of K_i. As written, the proof of Theorem 1.6 does not establish the claimed result.
- [Theorem 4.4, proof, pp. 18–19] In the induction step the proof says 'Apply Theorem 4.3 to obtain a finite T-free, minimal HL-extension (D_n,\varphi_n) of C_n that is coherent with (D_{n-1},\varphi_{n-1})'. But Theorem 4.3 as stated only provides a finite T-free HL-extension, with no guarantee of minimality. Minimality is needed because Definition 4.2(iii) is stated only for minimal extensions. This gap is repairable: given any coherent HL-extension (D,\varphi) of C_n that extends C_{n-1}, one can pass to the minimal substructure generated by C_n under \varphi(P_{C_n}); this substructure contains D_{n-1} and the restricted map still satisfies the coherence conditions. The manuscript should either supply this argument or state a strengthened version of Theorem 4.3.
- [Corollary 3.8, proof, p. 16] The proof applies the HL-property of G_1 and G_2 by invoking Theorem 3.3(iii) to obtain finite structures D'_1 and D'_2 with F_k-embeddings from C'. The hypotheses of Theorem 3.3(iii) require a T-free L-structure E_k on which G_k acts faithfully by isomorphisms and transitively on each unary part S_i^{E_k}. The proof does not identify such input structures. From the given assumption that G_1 * G_2 acts transitively on each S_i^D, it does not follow that the factor G_k is transitive on each S_i^D; for example, the free product of two copies of \mathbb{Z}/2 generated by transpositions (1\,2) and (2\,3) acts transitively on {1,2,3}, while neither factor does. A separate argument is needed to justify this application, for instance by constructing appropriate input structures for each factor or by proving a transitivity-free variant of the approximation property.
- [Proposition 3.6, p. 15] Proposition 3.6 is stated for an arbitrary finite T-free L-structure C and claims the existence of a finite T-free HL-extension (D,\varphi) with the stated embedding property for all substructures E. However, the cited coherence result of Siniora and Solecki (and the combinatorial proof in [5]) requires that every structure in T is a Gaifman clique. Without this hypothesis the statement is not supported and can fail, as the counterexample after Theorem 4.3 indicates. The proposition should include the Gaifman-clique assumption; this does not affect Corollary 3.8, where the assumption is present.
minor comments (4)
- [Theorem 4.6, proof, p. 19] The proof should explicitly justify the existence of the increasing sequence (D_i,\varphi_i) by induction using Definition 4.2(iii) with each (D_i,\varphi_i) chosen minimal; as written, the proof does not state that the chosen HL-extensions are minimal, which is required for the quoted property of ultraextensiveness.
- [Section 3.2, proof of (C2), p. 10] In the explanation of condition (C2), the displayed equality '(p_1(a_{i_1}),\ldots,p_m(a_{i_m})) = (gp_1(a_{i_1}),\ldots,gp_m(a_{i_m}))' appears to contain a typo; from the coset equality (p_1H_{i_1},\ldots,p_mH_{i_m})=(gq_1H_{i_1},\ldots,gq_mH_{i_m}) one obtains (p_1(a_{i_1}),\ldots,p_m(a_{i_m})) = (gq_1(a_{i_1}),\ldots,gq_m(a_{i_m})).
- [Section 3.4 and references] The name 'Siniora–Solecki' is misspelled as 'Sinora–Solecki' in the paragraph before Proposition 3.6 and in the proof of Corollary 3.8; the reference list uses the correct spelling.
- [Theorem 4.5, proof, p. 19] The proof of Theorem 4.5 is very abbreviated, saying only that a similar argument to Theorem 4.4 works with modifications. Since the differences (using the defining property of ultraextensiveness instead of Theorem 4.3, and unions instead of free amalgamations) are straightforward but not entirely trivial, a few more details would improve the presentation.
Circularity Check
No significant circularity; the main results are derived from external theorems (Herwig-Lascar, Siniora-Solecki) and independent constructions, not from the conclusions being proved.
full rationale
The paper's derivation chain is self-contained with respect to the question of circularity. Theorem 1.4 and Theorem 4.4 are proved by an explicit induction using Theorem 4.3, whose proof invokes the external coherence result of Siniora-Solecki [9] (also proved independently by Hubicka-Konecny-Nesetril [5]), together with the free amalgamation property for T-free structures when T consists of Gaifman cliques. These are genuine external inputs, not restatements of the paper's conclusions. The HL-extension concept is imported from Herwig-Lascar with a precise definition, and the canonical extension construction in Section 2 is built from the free group on the set of partial isomorphisms; the characterization in Theorem 2.4 is proved by defining normal subgroups from the given extension, not by assuming the target result. The only self-reference is the remark that results in Sections 2 and 4 are 'analogous to previous work by the authors [2] on similar concepts in the context of metric spaces.' This analogy is not used as a proof input, so it is not load-bearing. No equation is defined in terms of the quantity it is meant to derive, and no fitted parameter is renamed as a prediction. The identified gap in the proof of Theorem 4.6, concerning whether the union of the Aut(D_i)'s is a subgroup, is a correctness or exposition issue, not a circularity issue, since the intended directed system of coherent subgroups is supplied by Definition 4.1(iii) rather than by circularly assuming the conclusion. Accordingly, the circularity score is 0.
Assumptions & free parameters
assumptions (4)
- standard math Herwig-Lascar theorem (Theorem 1.3): every finite T-free L-structure has a finite T-free HL-extension.
- standard math Siniora-Solecki strong coherence theorem: for every finite T-free structure C there is a finite T-free extension D and a map phi from P(C) to Aut(D) with p contained in phi(p) and phi(p) composed with phi(q) equals phi(r) whenever p composed with q equals r.
- standard math Siniora-Solecki free amalgamation characterization: the class of finite T-free structures has free amalgamation iff every structure in T is a Gaifman clique.
- standard math The paper assumes ZFC and standard facts about free groups, profinite topology, and Fraisse limits.
Cite this review
Pith. "Pith review of On Extensions of Partial Isomorphisms." pith.science (2026). https://pith.science/paper/767D5GUD
@misc{pith2026190802965,
author = {Pith},
title = {Pith review of: On Extensions of Partial Isomorphisms},
year = {2026},
howpublished = {\url{https://pith.science/paper/767D5GUD}},
note = {Machine review of arXiv:1908.02965}
}
abstract
In this paper we study a notion of HL-extension (HL standing for Herwig--Lascar) for a structure in a finite relational language $\mathcal{L}$. We give a description of all finite minimal HL-extensions of a given finite $\mathcal{L}$-structure. In addition, we study a group-theoretic property considered by Herwig--Lascar and show that it is closed under taking free products. We also introduce notions of coherent extensions and ultraextensive $\mathcal{L}$-structures and show that every countable $\mathcal{L}$-structure can be extended to a countable ultraextensive structure. Finally, it follows from our results that the automorphism group of any countable ultraextensive $\mathcal{L}$-structure has a dense locally finite subgroup.
Reference graph
Works this paper leans on
-
[2]
On extensions of partial isometries
M. Etedadialiabadi, S. Gao, On extensions of partial isometries , preprint. Available at arXiv:1903.09723
work page Pith review arXiv 1903
-
[5]
Jan Hubiˇ cka, Matˇ ej Koneˇ cn` y, and Jaroslav Neˇ setˇ ril, A combinatorial proof of the extension property for partial isometries , Comment. Math. Univ. Carolinae 60 (2019), no. 1, 39–47
work page 2019
-
[1]
Coulbois, Free product, profinite topology and finitely generated subg roups, Internat
T. Coulbois, Free product, profinite topology and finitely generated subg roups, Internat. J. Algebra Comput. 11 (2001), no. 2, 171–184
work page 2001
- [3]
-
[4]
Hrushovski, Extending partial isomorphisms of graphs , Combinatorica 12 (1992), no
E. Hrushovski, Extending partial isomorphisms of graphs , Combinatorica 12 (1992), no. 4, 411–416
work page 1992
- [6]
-
[7]
Rosendal, Finitely approximable groups and actions part I: the Ribes– Zalesskii property, J
C. Rosendal, Finitely approximable groups and actions part I: the Ribes– Zalesskii property, J. Symbolic Logic 76 (2011), no. 4, 1297–1306
work page 2011
-
[8]
Solecki, Extending partial isometries , Israel J
S. Solecki, Extending partial isometries , Israel J. Math. 150 (2005), 315–331
work page 2005
Show all 9 references
-
[9]
Siniora, S
D. Siniora, S. Solecki, Coherent extension of partial automorphisms, free amalgam ation, and automorphism groups , J. Symbolic Logic, to appear. Available at arXiv:1705.018 88. Department of Mathematics, University of North Texas, 1155 Uni on Circle #311430, Denton, TX 76203, ...
Reviewed August 14, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.