{"id":"9a7309eb-9af5-4205-a1cf-6d94ece07520","arxiv_id":"2608.04367","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The equality graphs for the energy-independence inequality are exactly disjoint unions of isolated vertices, balanced complete multipartite graphs, and the new family H_r(a,b).","lead":"Every connected component of a graph whose energy equals twice its independence number is an isolated vertex, a balanced complete multipartite graph, or a graph built from two such multipartite graphs by joining corresponding parts. This settles the equality case of a recently proved inequality in spectral graph theory.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The equality characterization is internally coherent, but its proof is contingent on the unverified external inequality (5) from [8]; if that lemma fails, the equality propagation and rigidity arguments collapse.","rationale":"The paper's central claim, Theorem 1.3, is an exhaustive equality characterization. I read Sections 2-9 in good faith and found the internal argument coherent: equality propagation to anti-neighborhoods, edge rigidity via the F_uv chain, the P-class/cell partition, the fractional matching argument, and the identification of the two connected families all follow once the external inequalities are granted. The spectral verification of the extremal families is sound, including the computation for H_r(a,b). The weakest point is exactly the one the reader identified: inequality (5) and the associated Schur-complement claims are imported from the preprint [8] and are not proved here. Because Proposition 2.1 and Lemma 3.1 are the first two load-bearing steps, a flaw in those imported claims would invalidate the proof, even if the statement itself might still be true. I do not see an error in the paper's own combinatorial or spectral arguments, and the external dependency is a standard, though non-negligible, risk. Hence I agree with the reader's assessment: accept with moderate confidence, pending independent confirmation of [8].","tokens_in":11468,"tokens_out":34278,"duration_ms":330507,"concrete_test":"Reconstruct the proof of [8, Lemma 2.2 and Claims 2.1/2.2] from the PSD decompositions in (2)-(4), checking specifically that the chain Σ_v E(G_v) ≤ 2Σ_v tr(P_v) ≤ n E(G) − 4m is valid and that the second difference equals Σ_{uv∈E(G)} F_uv with F_uv ≥ 0; if the chain cannot be completed, (5) is unsupported. As a secondary check, enumerate all graphs on at most nine vertices and verify both (5) and the Theorem 1.3 classification numerically, since a single counterexample would falsify the claimed characterization.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The single most load-bearing premise is the imported inequality (5): 4|E(G)| + Σ_v E(G_v) ≤ |V(G)| E(G), together with the Schur-complement claims in [8, Claims 2.1/2.2]. Proposition 2.1 uses (5) to force equality in the summed anti-neighborhood lower bound, which is the first step of the entire proof. Lemma 3.1 then requires not only (5) but also the assertion that the gap in the second inequality of the chain is exactly Σ_{uv∈E(G)} F_uv, with every F_uv ≥ 0. Neither component is proved in the present paper. If either is incorrect or is not applicable in the equality case, Proposition 2.1 and Lemma 3.1 fail, and with them the P-class/cell construction, the weighted bipartite matching argument, and the classification in Theorem 1.3. This is a correctness risk rather than an internal inconsistency: the combinatorial development from rigidity onward is coherent, and the listed families are directly verified to attain equality by the spectra in Lemmas 7.1 and 7.2. The concern is that the analytic foundation is outsourced to an unpublished preprint, so the theorem is established only conditionally on [8].","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper determines all finite simple graphs attaining equality in the recent Kumar–Pragada inequality E(G) ≥ 2(n − α(G)). The main result (Theorem 1.3) states that equality holds exactly when every connected component is an isolated vertex, a balanced complete multipartite graph K_{a,a,...,a}, or a member of the new family H_r(a,b) with r ≥ 3, obtained from two balanced complete r-partite graphs by completely joining corresponding parts (and adding no other cross-edges). The proof propagates equality to anti-neighborhoods, uses positive-semidefinite rigidity of the absolute adjacency matrix to derive a P-class/cell structure, encodes cell adjacencies in a weighted bipartite incidence graph, and uses weighted matching duality to show that at most two P-classes occur. The paper closes by verifying the spectra of the extremal families.","tokens_in":11642,"tokens_out":13265,"duration_ms":130560,"significance":"The equality case of the Graffiti/energy–independence conjecture is a natural and valuable contribution. If the supporting inequality from Kumar–Pragada is accepted, the classification is clean and the internal development from the P-class/cell construction onward is coherent. The spectral verification of the H_r(a,b) family is explicit and correct. The main weakness is that the initial equality propagation and edge rigidity depend on unproved technical facts from an unpublished preprint [8]; the paper is therefore conditional unless those facts are supplied or a published version is cited.","major_comments":[{"comment":"The proof of Theorem 1.3 is conditional on the external inequality (5) of Kumar and Pragada and on the accompanying Schur-complement/F_uv chain imported from [8, Claims 2.1/2.2]. Proposition 2.1 uses (5) to force equality in the summed anti-neighborhood bound, and Lemma 3.1 uses the unproved assertions that F_uv ≥ 0 and that the gap in the second inequality is exactly Σ_{uv∈E(G)} F_uv. Because [8] is an unpublished preprint, the characterization is not self-contained at a load-bearing point. Please either prove these supporting facts in the paper (for example, in an appendix) or cite a published/refereed version; at minimum, state the imported claims explicitly.","section":"§2–§3, Eq. (5), Lemma 3.1"},{"comment":"The displayed lower bound for F_uv is algebraically incorrect. From F_uv = ((x−1)^2−z^2)/x + ((y−1)^2−z^2)/y = (x+y)/(xy)(xy+1−z^2)−4, the AM–GM bound x+y ≥ 2√(xy) gives F_uv ≥ (2/√(xy))((√(xy)−1)^2−z^2), not 2√(xy)((√(xy)−1)^2−z^2). The claimed conclusion (F_uv=0 implies x=y and (√(xy)−1)^2=z^2) still follows from the corrected bound, so this is repairable, but the current formula is false (e.g., x=y=100, z=0).","section":"§3, Lemma 3.1"}],"minor_comments":[{"comment":"The example of the cell incidence graph is understandable in words; if the figure is omitted in the published version, the caption should be adjusted accordingly.","section":"§5.3, Figure 1"},{"comment":"The phrase 'contain triangles inside each of their two complete multipartite halves' is slightly imprecise; the triangles lie inside the X-side and the Y-side of H_r(a,b), not in the two P-classes as 'halves' in a geometric sense.","section":"§9, Corollary 9.2"},{"comment":"Reference [8] is an arXiv preprint with a 2026 date; if a refereed or published version appears, it should be cited in its place so readers can verify the imported inequality.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The proof is internally strong once the Kumar–Pragada inequality and the associated F_uv chain are accepted. If the editor knows that [8] has been accepted for publication, I would be comfortable downgrading to minor revision; otherwise, the manuscript should be revised to include or state the supporting lemmas. The algebra slip in Lemma 3.1 is easily fixed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe headline: this paper gives the first complete list of equality graphs for the energy–independence inequality E(G) ≥ 2(n−α(G)), and the list is very plausible. It is worth a close read if you work in spectral graph theory, though the proof leans on an external inequality that is still unpublished.\n\nWhat is genuinely new: the H_r(a,b) family, and the proof that equality holds exactly for isolated vertices, balanced complete multipartite graphs, and that family. Earlier literature had only examples (K2□Kr and balanced complete multipartite graphs) plus partial bounds. The paper introduces a clean P-class/cell decomposition and reduces the cell adjacency structure to a weighted matching LP. That LP step is the most elegant part: a uniform fractional matching is shown to be optimal, and dual complementarity forces each P-class to have exactly R cells, all of the same size. The converse direction is handled properly—both families are verified by direct spectral computation, and the spectrum of H_r(a,b) is computed correctly.\n\nThe soft spot is the one flagged in the stress-test. Everything after Proposition 2.1 depends on inequality (5), 4|E(G)| + Σ E(G_v) ≤ n E(G), plus the F_uv ≥ 0 claim, both taken from Kumar–Pragada's preprint [8]. Neither is proved here. That is not a citation sin—the base inequality is the same authors' theorem—but it makes this paper's theorem conditional on an unpublished proof. A referee should verify [8, Lemma 2.2 and Claims 2.1/2.2] before signing off. Once (5) is granted, the combinatorial development is coherent; I did not find an internal gap.\n\nMinor concerns: the notation is heavy but the structure is clear, and the literature review is honest about the partial results. The paper is a within-subfield contribution, not a breakthrough, but it closes a natural open question.\n\nWho is this for: spectral graph theorists working on graph energy and extremal graph problems. It deserves a serious referee, not a desk reject. If [8] checks out, this should be accepted.\n\nRecommendation: send to peer review, and ask the referee to verify the external inequalities—or have the author append a proof of (5).","headline":"Complete and likely correct equality classification, but the proof rests on an unverified inequality from another preprint.","tokens_in":12182,"tokens_out":2469,"would_cite":true,"duration_ms":57148,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["05C50","05C35"],"pacs":[],"model":"deepseek-v4-flash","headline":"Every graph that attains the energy–independence lower bound decomposes into isolated vertices, balanced complete multipartite graphs, or $H_r(a,b)$ with $r\\ge 3$.","keywords":["graph energy","independence number","vertex cover number","extremal graphs","complete multipartite graphs","spectral graph theory","fractional matching","well-covered graphs"],"falsifier":"Exact spectral enumeration of all connected graphs of order at most 10: compute $\\mathcal E(G)$ and $\\alpha(G)$ for each, and any graph with $\\mathcal E(G)=2(n-\\alpha(G))$ that is not a disjoint union of the three listed families would refute Theorem 1.3.","tokens_in":11217,"feed_emoji":"⚡","tokens_out":11124,"duration_ms":106912,"temperature":0.7,"pith_summary":"The paper determines exactly which graphs make the lower-bound inequality between adjacency energy $\\mathcal E(G)$ and the independence number $\\alpha(G)$, namely $\\mathcal E(G)\\ge 2(n-\\alpha(G))$, an equality. It claims that every connected equality graph is either a single vertex, a balanced complete multipartite graph with at least two parts, or a graph $H_r(a,b)$ with $r\\ge 3$ built by taking two balanced complete multipartite graphs with $r$ parts and completely joining corresponding parts; disconnected equality graphs are disjoint unions of these. The proof shows that equality is rigid enough to force every anti-neighborhood to be an equality graph and every edge to satisfy a tight numerical relation in the absolute adjacency matrix, then compresses the remaining structure into at most two algebraic classes. The payoff is a complete, checkable list of the extremal graphs, closing the equality problem for a spectral bound proposed in the 1980s.","feed_headline":"Three component families exhaust the energy lower bound","feed_subtitle":"The energy bound is tight only for isolated vertices, balanced complete multipartite graphs, and $H_r(a,b)$ blow-ups.","key_machinery":"The machinery centers on the spectral decomposition $A=P-Q$ and the absolute adjacency matrix $B=P+Q=|A|$, the spectral-calculus absolute value of the adjacency matrix. Equality in the earlier proof's chain makes a local quantity $F_{uv}$ vanish on every edge, giving Lemma 3.1: for every edge $uv$, $B_{uu}=B_{vv}$ and $|B_{uv}|=B_{uu}-1$. The Gram representations $P_{xy}=\\langle p_x,p_y\\rangle$ and $Q_{xy}=\\langle q_x,q_y\\rangle$ then split edges into two types, either $p_u=p_v$ or $q_u=-q_v$. Vertices with equal $p$-vectors form P-classes, subdivided into cells with equal $q$-vectors; cell adjacencies are encoded by a weighted bipartite incidence graph whose matchings are exactly the independent sets. A uniform fractional matching and LP duality show that every P-class has exactly $R=n/\\alpha$ cells of equal size, and the zero-sum identity $q_1+\\cdots+q_R=0$ leaves only one or two P-classes, which are precisely the balanced complete multipartite and $H_R(a,b)$ cases.","core_discovery":"The central claim, Theorem 1.3, is that $\\mathcal E(G)=2(n-\\alpha(G))$ holds if and only if every connected component of $G$ is an isolated vertex, a balanced complete multipartite graph $K_{a,a,\\ldots,a}$ with at least two parts, or a graph $H_r(a,b)$ with $r\\ge 3$ and $a,b\\ge 1$. The graph $H_r(a,b)$ is formed from two balanced complete multipartite graphs, one with $r$ parts of size $a$ and one with $r$ parts of size $b$, by completely joining the $i$-th part of the first to the $i$-th part of the second for each $i$. The proof establishes necessity by showing that equality propagates to every anti-neighborhood, so the graph is well-covered, and then using the Gram vectors of the positive and negative spectral parts of the adjacency matrix to force a cell structure with at most two classes.","pith_inferences":["A stability version is a natural extension not claimed in the paper: graphs with energy within $\\epsilon$ of the bound should be close in edit distance to the listed families, since the rigidity lemmas are quantitative.","The integrality constraint $n/\\alpha(G)\\in\\mathbb{Z}$ gives a cheap computational filter: any graph with $n/\\alpha$ not an integer can be excluded from equality testing without computing its spectrum.","The same LP and Gram-vector machinery may characterize equality in other spectral lower bounds, such as the matching-number energy bound, where a different equality list is known but the method of proof here is transferable."],"forward_implications":["For any connected equality graph with $\\alpha(G) < n/2$, the ratio $n/\\alpha(G)$ is an integer.","A bipartite graph satisfies the equality if and only if every nontrivial component is a balanced complete bipartite graph.","The previously known examples $K_2\\square K_r$ are exactly the case $H_r(1,1)$ inside the new family, so the earlier examples are covered and no further ones exist.","A disconnected graph attains equality exactly when each connected component does, so the connected classification is the whole classification."],"supporting_citations":[{"why":"Proves the lower bound $\\mathcal E(G)\\ge 2(n-\\alpha(G))$ and supplies inequality (5), the local estimates behind $F_{uv}\\ge 0$, and the three lemmas the equality proof starts from.","marker":"[8]"},{"why":"Observes the known equality examples, balanced complete multipartite graphs and $K_2\\square K_r$, that the classification must cover and extends.","marker":"[1]"},{"why":"Provides the weighted bipartite matching theorem and LP duality facts used to turn cell adjacencies into the matching problem.","marker":"[10]"},{"why":"Supplies the standard spectrum of balanced complete multipartite graphs used in Lemma 7.1 to verify equality.","marker":"[4]"},{"why":"Gives the earlier lower bound with odd-cycle correction and its equality case, a direct forerunner the paper's result supersedes.","marker":"[12]"},{"why":"Characterizes equality for the related matching-number energy bound, showing why the vertex-cover equality needs a separate proof.","marker":"[5]"}],"fun_headline_variants":["Three graph families achieve the energy lower bound","Energy bound equality: only three component types","All extremal graphs for the energy lower bound","Equality in energy bound fully characterized","Tight energy bound: isolated, multipartite, and H_r(a,b)"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the earlier inequality about anti-neighborhood energies and its local per-edge estimate are both correct, because every step of the equality proof uses them; if either were false, the classification would not follow.","fun_headline_variants_meta":{"raw":{"variants":["Three graph families achieve the energy lower bound","Energy bound equality: only three component types","All extremal graphs for the energy lower bound","Equality in energy bound fully characterized","Tight energy bound: isolated, multipartite, and H_r(a,b)"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000742,"raw_usage":{"total_tokens":3268,"prompt_tokens":863,"completion_tokens":2405,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":479,"completion_tokens_details":{"reasoning_tokens":2331}},"tokens_in":479,"tokens_out":2405,"duration_ms":16769,"temperature":1.0,"reasoning_tokens":2331,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-08T18:54:34.542773+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Exact spectral enumeration of all connected graphs of order at most 10: compute $\\mathcal E(G)$ and $\\alpha(G)$ for each, and any graph with $\\mathcal E(G)=2(n-\\alpha(G))$ that is not a disjoint union of the three listed families would refute Theorem 1.3.","supporting_citations":[{"cited_title":"Energy and independence number","cited_arxiv_id":"2607.19817","evidence_quote":"Proves the lower bound $\\mathcal E(G)\\ge 2(n-\\alpha(G))$ and supplies inequality (5), the local estimates behind $F_{uv}\\ge 0$, and the three lemmas the equality proof starts from."},{"cited_title":"A graph energy conjecture through the lenses of semidefinite programming","cited_arxiv_id":"2509.05814","evidence_quote":"Observes the known equality examples, balanced complete multipartite graphs and $K_2\\square K_r$, that the classification must cover and extends."},{"cited_title":"Lov´ asz and M","cited_arxiv_id":null,"evidence_quote":"Provides the weighted bipartite matching theorem and LP duality facts used to turn cell adjacencies into the matching problem."},{"cited_title":"Wang and X","cited_arxiv_id":null,"evidence_quote":"Gives the earlier lower bound with odd-cycle correction and its equality case, a direct forerunner the paper's result supersedes."},{"cited_title":"Chen and X","cited_arxiv_id":null,"evidence_quote":"Characterizes equality for the related matching-number energy bound, showing why the vertex-cover equality needs a separate proof."}],"review_version":1}