REVIEW 2 major objections 2 minor 4 cited by
An edge-spectral Erd\H{o}s-Stone-Simonovits theorem and its stability
T0 review · 2 major / 2 minor · reviewed 2026-08-05 · deepseek-v4-flash
Pith's one-line read This paper proves an edge-spectral analogue of the classical forbidden-subgraph extremal theorem, showing the squared spectral radius of any F-free graph with m edges is at most (1 − 1/r + o(1))2m, and establishes the matching stability res
desk verdict New edge-spectral ESS bound looks plausible and important, but the stated stability theorem has a disconnected counterexample 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
The central object is the spectral radius λ(G), and the paper's key quantity is the normalized squared spectral radius λ²(G)/2m. The claim is that, asymptotically, this quantity is bounded above by 1 − 1/r for any graph that forbids a fixed F with chromatic number r+1 — the same constant that bounds edge density in the classical extremal theorem. The stability result uses closeness to Turán graphs as the structural description, and the common-neighbor application follows from the spectral bound.
What would settle it
A concrete refutation would be a forbidden graph F with chromatic number r+1 and a sequence of F-free graphs G with m edges such that λ²(G)/2m tends to a limit strictly greater than 1 − 1/r. Equivalently, for some ε > 0, infinitely many F-free graphs with λ²(G) > (1 − 1/r + ε)2m would break the theorem.
Extended reading notes
Core claim
On the paper's own terms, the central result is the inequality λ²(G) ≤ (1 − 1/r + o(1))2m for every F-free graph G with m edges, whenever the chromatic number of F is r+1 with r ≥ 2. The paper derives this as a unified statement that contains both the classical edge-density extremal theorem and its vertex-spectral analogue, and it goes beyond existence to stability: if λ²(G) is within o(m) of the extremal value, then G differs from the appropriate complete multipartite extremal graph by only o(m) edge edits. The proof is not visible in the abstract, but the claimed consequences are stated cleanly and include the improved common-neighbor bound for graphs with λ(G) > √m.
Load-bearing premise
The argument depends on showing that any graph that pushes the spectral bound must already look almost like an r-partite Turán graph; if that structural reduction fails for some forbidden F, the constant 1 − 1/r may not hold.
Editorial extensions
If this is right
- If true, the theorem unifies the classical edge-density bound and the vertex-spectral bound under a single edge-spectral inequality.
- It confirms the open conjecture, resolving a proposed line of spectral extremal results.
- The stability statement gives a quantitative structural description of F-free graphs near the spectral threshold: they are almost r-partite.
- The common-neighbor corollary says that spectral radius above √m guarantees a dense pair of vertices, improving the earlier bound and matching the random construction.
Reading between the lines
- One implicit consequence: the bound suggests that the squared spectral radius can serve as a weighted edge count for extremal problems, so other edge-density phenomena (such as supersaturation) may admit spectral analogues with the same constants.
- The stability result may transfer to algorithmic settings, where spectral measurements could certify near-Turán structure without examining all edges.
- A testable extension would be to replace the spectral radius with other symmetric functions of the adjacency eigenvalues; the proof method may indicate which spectral statistics inherit the same extremal constant.
- The common-neighbor bound is likely sharp only in the asymptotic sense; understanding the small-m regime could require additional eigenvalue constraints.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The manuscript (arXiv:2508.15271, abstract only) claims an edge-spectral analogue of the Erdős–Stone–Simonovits theorem: for any graph F with chromatic number r+1 ≥ 3, every F-free graph G with m edges satisfies λ²(G) ≤ (1 − 1/r + o(1))2m, confirming a conjecture of Li, Liu and Feng. The abstract also claims an accompanying edge-spectral stability theorem: if λ²(G) is asymptotically equal to the maximum (1 − 1/r)2m, then G is o(m)-close, in edit distance, to a complete bipartite graph for r=2 and to an r-partite Turán graph for r≥3. Finally, the paper claims an application improving a result of Zhai, Lin and Shu: if λ(G) > √m, then two vertices share at least (1/2)√m − O(1) common neighbors, and this is best possible.
Significance. If the main spectral bound is correct, it is a useful unified result connecting the classical edge-count extremal theorem with spectral extremal results, and the parameter-free asymptotic constant is appealing. The manuscript also promises an extension of the Erdős–Simonovits stability theorem to the spectral edge-setting, which would be a significant contribution. However, because the full text is not available, the proof of the main bound cannot be verified from the abstract; and the stability claim as stated appears to be false, as shown by a concrete counterexample. Thus the significance of the paper depends on a substantial revision of the stability statement and a careful statement of hypotheses.
major comments (2)
- [Abstract, stability theorem (r=2 case)] The stability statement as written is false. Let G = K_{a-1,a} ∪ K_{1,a} with m = a² edges. G is bipartite, hence K₃-free (here r=2). Its spectral radius is √(a(a−1)), so λ² = a²−a = m−√m = (1/2 − o(1))2m, satisfying the near-extremal hypothesis. But no complete bipartite graph H on the same vertex set has e(G△H)=o(m). Indeed, any such H must contain almost all edges of the K_{a-1,a} component, forcing its two parts to contain all vertices of that component on opposite sides. The edges of the K_{1,a} component then force at least one of the two possible orientations, and in either case H adds at least a²−O(a) edges not in G (or misses equally many), so the edit distance is Ω(m), not o(m). Therefore the stability theorem needs an additional hypothesis (e.g., connectivity or a condition that almost all edges lie in one component) or a different notion of closeness; otherwise it is incorrec
- [Abstract, stability theorem (r≥3 case)] The same disjoint-union obstruction applies to r≥3. Take a near-extremal F-free graph on m₀ edges, e.g., a complete r-partite Turán graph, and add a small disjoint F-free component with o(m₀) edges. The spectral radius of the disjoint union is the maximum of the component spectral radii, which is still asymptotically (1−1/r)2m for the total m, since the small component contributes negligible spectral radius. But the resulting graph is not o(m)-close to an r-partite Turán graph on the same vertex set: placing the added vertices into parts of a complete r-partite graph creates Ω(m) edges to the existing parts, and deleting the small component incurs only o(m) edge difference but also requires ignoring a positive fraction? Actually the edit distance to a spanning complete r-partite graph is Ω(m), because any assignment of the small component's vertices to the r parts introduces all cross ed
minor comments (2)
- [Abstract, stability hypothesis] The notation λ²(G) = (1 − 1/r − o(1))2m is ambiguous. The intended meaning is presumably that λ²(G) is at least (1 − 1/r)2m − o(m), i.e., the value is asymptotically concentrated near the extremal bound from below. Please state this as a one-sided inequality with a quantifier (e.g., for every ε>0, λ²(G) ≥ (1 − 1/r − ε)2m for sufficiently large m) to avoid confusion.
- [Abstract, application] The phrase 'at least 1/2√m − O(1) common neighbors' should be clarified: the O(1) constant may depend on the graph? Also the 'random construction' witnessing sharpness is not described; a reference or precise model would help the reader assess the optimality claim.
Circularity Check
No circularity evident from the abstract; the main bound is a parameter-free asymptotic theorem.
full rationale
This is an abstract-only review, so the derivation chain cannot be walked in detail. The main claim (λ²(G) ≤ (1−1/r+o(1))2m for F-free G with χ(F)=r+1) is a universal asymptotic bound with no fitted parameters and no visible reduction to its own input. The stability statement and the common-neighbor application are presented as derived consequences. The only self-referential element is the phrase 'confirms a conjecture proposed by Li, Liu and Feng,' where two of the present authors share surnames with the conjecturers; however, the conjecture is cited as motivation rather than used as a load-bearing proof step, and the theorem is asserted to be proved, not assumed. No equation or construction is given that would let one exhibit a fitted parameter renamed as a prediction or a definition circularly defining the target. The skeptic's counterexample to stability is a substantive mathematical objection about correctness, not a circularity argument, and would require the full proof to assess; it does not establish that any claim reduces to its inputs. Therefore the appropriate finding is no significant circularity.
Assumptions & free parameters
assumptions (3)
- domain assumption Graphs are finite, simple, undirected; F-free means no subgraph isomorphic to F.
- standard math Spectral radius lambda(G) is the largest eigenvalue of the adjacency matrix.
- domain assumption Known extremal results, such as the Erdős–Stone–Simonovits theorem and Nikiforov's vertex-spectral result, are used as background or are extended.
Cite this review
Pith. "Pith review of An edge-spectral Erd\H{o}s-Stone-Simonovits theorem and its stability." pith.science (2026). https://pith.science/paper/BHJICNSV
@misc{pith2026250815271,
author = {Pith},
title = {Pith review of: An edge-spectral Erd\Hos-Stone-Simonovits theorem and its stability},
year = {2026},
howpublished = {\url{https://pith.science/paper/BHJICNSV}},
note = {Machine review of arXiv:2508.15271}
}
abstract
We study the extremal problem that relates the spectral radius $\lambda (G)$ of an $F$-free graph $G$ with its number of edges. Firstly, we prove that for any graph $F$ with chromatic number $\chi (F)=r+1\ge 3$, if $G$ is an $F$-free graph on $m$ edges, then $\lambda^2(G)\le {(1-\frac{1}{r} + o(1))2m}$. This provides a unified extension of both the Erd\H{o}s--Stone--Simonovits theorem and its vertex-spectral version due to Nikiforov, and confirms a conjecture proposed by Li, Liu and Feng. We also establish the corresponding edge-spectral stability, showing that if $G$ is an $F$-free graph on $m$ edges with $\lambda^2(G)=(1- \frac{1}{r} - o(1))2m$, then $G$ differs from a complete bipartite graph by $o(m)$ edges when $r=2$, and $G$ differs from an $r$-partite Tur\'{a}n graph by $o(m)$ edges when $r\ge 3$. This extends the classical Erd\H{o}s--Simonovits stability theorem. As an application of our method, we improve a result of Zhai, Lin and Shu by showing that if $\lambda (G)>\sqrt{m}$, then there exist two vertices in $G$ that have at least $\frac{1}{2}\sqrt{m} - O(1)$ common neighbors. This bound is the best possible as witnessed by a random construction.
Forward citations
Cited by 4 Pith papers
-
A sharp fixed-size spectral bound for $kK_3$-free graphs
For all fixed k and all sufficiently large m, every m-edge graph with no k vertex-disjoint triangles has adjacency spectral radius at most (k-1)+sqrt(m-k(k-1)), with equality only for the join of K_{2k-1} with an inde...
-
Supersaturation in Nosal graphs: Triangles and books
Every m-edge graph with spectral radius λ has at least m(λ−√m) triangles, and every Nosal graph contains a book of size > ¼√m and at least (1/8−o(1))m kites.
-
An edge-spectral supersaturation of Mubayi's theorem for color-critical graphs
For color-critical F with χ(F)=r+1≥4, λ²(G)≥2(1-1/r)m+q with 0<q≤δ_F√m forces at least (B_F-o(1))q m^{(f-2)/2} copies of F, and B_F is best possible.
-
On a spectral booksize problem fo non bipartite graphs
Every sufficiently large m-edge non-bipartite graph without isolated vertices satisfying rho(G)^2 >= m-1+2/(rho(G)-1) is either an exceptional graph S+_{m,s} or contains a book of size at least (1/4-o(1)) sqrt(m), and...
Reviewed August 5, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.