REVIEW 5 minor 38 references
The role of expanders in the spectral geometry of metric graphs
T0 review · 0 major / 5 minor · reviewed 2026-08-02 · deepseek-v4-flash
Pith's one-line read Expander graphs prove that metric spectral gaps cannot be bounded by size, diameter, girth, or mean distance, and that a classical inequality is sharp.
desk verdict Clean, publishable expander-based disproofs of geometric spectral-gap bounds for metric graphs; two minor proof-hygiene issues in Section 4, but the central results hold. 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 engine is the discrete-to-continuous transfer principle for unilateral metric graphs: λ is a metric Laplacian eigenvalue exactly when 1−cos√λ is an eigenvalue of the discrete normalized Laplacian, so λ2(G) = arccos(1−ν2(G))^2. On a d-regular Ramanujan graph the nontrivial discrete eigenvalues stay within 2√(d−1)/d of 1, giving a degree-only lower bound on the metric spectral gap that persists as the number of vertices grows. The second engine is a set of comparison estimates showing that metric volume equals d#V/2 and that metric diameter, girth, and mean distance stay within O(1) of their combinatorial counterparts on d-regular graphs. For the Dirichlet results, the paper uses the close
What would settle it
Take a sequence of d-regular Ramanujan expander graphs with growing vertex count and compute, for their unilateral metric versions, the spectral gap λ2(G) via the scalar equation 1−cos√λ = ν2(G) and the metric mean distance ρ(G) by integrating the shortest-path metric. The paper predicts λ2 stays bounded below by a positive constant while ρ(G) grows logarithmically; if λ2 ρ^2 remains bounded, Theorem 3.8 is false. For the sharpness result, evaluate λ1(G;{v})T(G;{v})/|G| on the same graphs with one Dirichlet vertex: the claim is that these ratios approach 1, so finding a uniform constant C<1 th
Extended reading notes
Core claim
The central discovery is a negative one with a positive mechanism: expanding families of regular graphs, viewed as unilateral metric graphs, have a spectral gap λ2(G) that converges to arccos(2√(d−1)/d)^2 — a constant depending only on the graph degree d — while volume, diameter, girth, and mean distance all diverge (volume linearly, the other three logarithmically). Consequently λ2 cannot be bounded above by C times the inverse square of any of those quantities. A transfer principle converts the discrete spectral gap ν2 into the metric gap via λ2 = arccos(1−ν2)^2, and comparison lemmas show the combinatorial and metric versions of the geometric quantities differ by at most a constant. The s
Load-bearing premise
The load-bearing premise is that infinite families of d-regular Ramanujan graphs exist for each fixed degree with arbitrarily many vertices, and that the cited logarithmic lower bound on the average vertex distance of d-regular graphs is valid; if that mean-distance bound were weaker than logarithmic, the mean-distance counterexample would collapse.
Editorial extensions
If this is right
- No universal upper bound on λ2 in terms of volume, diameter, girth, mean distance, or any (−2)-homogeneous product of them can exist for unilateral metric graphs; an open problem on mean distance is settled negatively.
- For each fixed degree d, every sufficiently large Ramanujan unilateral metric graph has spectral gap essentially equal to arccos(2√(d−1)/d)^2, so the gap is determined by local degree alone, decoupled from global geometry.
- The Pólya–Szegő inequality λ1 T < |G| is asymptotically sharp: for any C<1 some metric graph with Dirichlet conditions violates λ1 T ≤ C|G|, so the strict inequality's constant 1 is optimal.
- The same expander argument also rules out universal lower bounds: when the product involves volume with positive exponent and a logarithmic quantity, no constant c>0 can bound λ2 from below for all unilateral metric graphs.
- The triameter, whose order of growth matches the diameter, inherits all the failure-of-upper-bound results.
Reading between the lines
- A likely upshot is that any successful upper bound on the metric spectral gap must encode information beyond global metric invariants—for example, the profile of local volumes or the geometry of the graph's 'filling'—since the expander examples show that size alone is irrelevant.
- The comparison lemmas suggest that for regular graphs the continuous and discrete mean distances are interchangeable up to an additive constant; this could allow future spectral-geometric estimates to be computed purely combinatorially.
- The double-asymptotic technique used for torsional rigidity (degree and vertex count both tending to infinity) might be adapted to test optimality of other strict inequalities, such as the conjectured failure of the inradius bound, where the paper's single-asymptotic method stalls.
- A direct numerical check on explicit small-degree Ramanujan graphs (e.g., d=3, moderate vertex count) computing λ2 and ρ from the edgewise eigenvalue equations would confirm the predicted logarithmic divergence of λ2 ρ^2, and would make the mechanism concrete.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper studies unilateral metric graphs (all edges of length one) and shows that expander constructions, especially LPS Ramanujan graphs, disprove natural upper bounds on the spectral gap in terms of volume, diameter, girth, mean distance, and products of these quantities. Its main results are: (i) Theorem 3.8, resolving an open problem of Baptista–Kennedy–Mugnolo by showing that no bound λ2 ≤ C/ρ(G)^2 can hold for the mean distance; (ii) Theorem 3.11, a unified statement about products of geometric quantities with the correct (−2)-homogeneity; and (iii) Theorem 4.6, which proves that the Pólya–Szegő inequality λ1 T < |G| is asymptotically sharp: no constant C < 1 can replace 1. The paper also discusses limitations of the expander method for Dirichlet inradius/mean-distance bounds, leaving two conjectures open.
Significance. The results, if correct, are significant for spectral geometry of metric graphs. They give a clean, systematic mechanism—Ramanujan graphs plus the von Below/Nicaise transference—for producing counterexamples to plausible bounds, and they settle a concrete open problem and a conjecture from the literature. A particular strength is that the main arguments do not fit parameters or use circular reasoning: the lower bound on λ2 comes from the Ramanujan property and Alon–Boppana, and the diverging geometric quantities come from standard logarithmic growth of diameter, girth, and mean distance. The probabilistic proof of Lemma 3.9 is elegant and correct, and the double-limit computation in Theorem 4.6 is nontrivial. The paper is largely self-contained in its review portions, and the external dependencies (LPS existence, Alon–Boppana, torsional rigidity formula) are well established.
minor comments (5)
- [§4, Lemma 4.3] The proof of the bipartite case is not correct as written. The vector u⊥ = (Id_V − #V^{-1}J_V)Eu does not vanish on V_D; at v ∈ V_D, Eu(v)=0 gives u⊥(v) = −#V^{-1}∑_{w∈eV} u(w), which is generally nonzero. The sentence about an 'eigenvalue −d' is also inconsistent with the normalized Laplacian spectrum lying in [0,2]. Since Theorem 4.6 uses non-bipartite LPS expanders, the central argument survives, but Lemma 4.3 and Corollary 4.5 need either a corrected bipartite proof or an explicit restriction to the non-bipartite case.
- [§2.2, Definition 2.3] Definition 2.3 bounds only ν2 (and ν#V in the non-bipartite case), while the proof of Lemma 4.3 invokes the estimate for all nontrivial eigenvalues. For non-bipartite graphs the two-point bound suffices by monotonicity of eigenvalues, and for bipartite regular graphs spectral symmetry would give the rest, but this should be stated explicitly so that (4.6) is justified for every i.
- [§4.2, Lemma 4.10(2)] The step '#N_{k+1} ≥ #N_k + #(N(N_k)\N_k) ≥ (1+ε)#N_k' uses the edge Cheeger constant, but the passage from edge expansion to vertex expansion requires a factor of 1/d (each new vertex can account for up to d boundary edges). The claim is plausible and can be repaired by replacing ε with ε/d, but the proof as written skips this point.
- [Remark 3.12(1)] The sentence citing '[19, Theorem 7.1](existence of an upper bound by 4|G|−3 diam(G) diam(G)^3 [6, Proposition 1.8 and Theorem 1.9]' is garbled and missing punctuation; please reformat the references and formulas.
- [§3.3] The notation 'girth(G) (resp., G)' is confusing; the metric girth should be defined clearly and distinguished from the combinatorial girth throughout.
Circularity Check
No significant circularity: negative spectral-gap results are driven by external Ramanujan/Alon–Boppana inputs and in-text lemmas, not by fitted or self-referential premises.
full rationale
The derivation chain is not circular. The central negative results (Proposition 3.3, Corollaries 3.4 and 3.7, Theorem 3.8, and Theorem 3.11) follow a uniform structure: fix the degree of an LPS/MSS Ramanujan family, invoke the von Below transference (Theorem 3.1, external) to turn the Ramanujan spectral gap into a fixed lower bound on λ2, and then use independently cited or in-text estimates showing that the relevant metric quantity grows without bound. Theorem 3.8's only external input is the logarithmic mean-distance bound [38, Formula (5)], a parameter-free theorem whose assumptions do not include the target result; Lemma 3.9, which connects the graph and metric mean distances, is proved in the text. Theorem 4.6 uses formula (4.13) from [31, Theorem 3.9]; although [31] shares an author, it is a published, externally checkable exact formula with fixed constants and stated assumptions that do not include the Pólya–Szegő sharpness claim. The subsequent inequalities (Lemma 4.3, Corollary 4.5, and the Loewner inversion in (4.14)) are derived in the text. No parameter is fitted and later renamed as a prediction, and no conclusion is assumed in the hypotheses. The ChatGPT-suggested Lemma 3.9 is fully proven rather than assumed. The conjectures in Section 4.2 are explicitly open and do not support any claimed result. Thus no step reduces by construction to its own input.
Assumptions & free parameters
assumptions (8)
- standard math von Below/Nicaise transference: for unilateral metric graphs, λ2(G)=arccos(1−ν2(G))^2, and λ1(G;V_D)=arccos(1−ν1(G;V_D))^2
- standard math LPS Ramanujan graphs exist with fixed degree d=p+1 and arbitrarily many vertices, with diameter, girth, and mean distance ≍ log_{d-1} #V
- standard math Alon–Boppana bound and Ramanujan spectral bound |1−ν2| ≤ 2√(d−1)/d
- standard math Lower bound on average distance of d-regular graphs, ρ(G) ≥ log_{d-1}(#V) − O(1)
- standard math Exact formula for torsional rigidity T(G;V_D)=d#V/24+(d/4)⟨L^{-1}_{G;V_D}1,1⟩
- standard math Elementary inequality arccos(1−x)^2 ≥ 2x + x^3/72 for x∈[0,1]
- standard math Girth of non-bipartite LPS expanders grows as ≍ 4/3 log_{d-1}(#V)
- standard math Expanders have a positive Cheeger constant ε > 0
Cite this review
Pith. "Pith review of The role of expanders in the spectral geometry of metric graphs." pith.science (2026). https://pith.science/paper/SKBDODPI
@misc{pith2026260714312,
author = {Pith},
title = {Pith review of: The role of expanders in the spectral geometry of metric graphs},
year = {2026},
howpublished = {\url{https://pith.science/paper/SKBDODPI}},
note = {Machine review of arXiv:2607.14312}
}
read the original abstract
Expanders are families of graphs that are sparse in edges but dense in connectivity. After reviewing combinatorial and spectral definitions of expanders, we use precise lower bounds on Ramanujan graphs to investigate upper bounds -- and, specifically, the lack thereof -- on the eigenvalues of the Laplacian on \emph{metric} graphs in terms of volume, diameter, girth, mean distance, and torsional rigidity, among others.
Reference graph
Works this paper leans on
-
[1]
N. Alon. Eigenvalues and expanders.Combinatorica, 6:83–96, 1986
1986
-
[2]
Alon and V.D
N. Alon and V.D. Milman. Eigenvalues, expanders and superconcentrators. In25th Annual Symposium on Founda- tions of Computer Science, 1984, pages 320–322. IEEE, 1984
1984
-
[3]
Alon and V.D
N. Alon and V.D. Milman.λ 1,isoperimetric inequalities for graphs, and superconcentrators.J. Combin. Theory Ser. B, 38:73–88, 1985
1985
-
[4]
Baptista, J.B
L. Baptista, J.B. Kennedy, and D. Mugnolo. Mean distance on metric graphs.J. Geom. Anal., 34:137, 2024
2024
-
[5]
von Below
J. von Below. A characteristic equation associated with an eigenvalue problem onc 2-networks.Lin. Algebra Appl., 71:309–325, 1985
1985
-
[6]
Berkolaiko, J.B
G. Berkolaiko, J.B. Kennedy, P. Kurasov, and D. Mugnolo. Impediments to diffusion in quantum graphs: geometry- based upper bounds on the spectral gap.Proc. Amer. Math. Soc., 151:3439–3455, 2023
2023
-
[7]
Bifulco and D
P. Bifulco and D. Mugnolo. On thep-torsional rigidity of combinatorial graphs.Nonlinear Anal. TMA, 251:113694, 2025
2025
-
[8]
Biggs and A.G
N.L. Biggs and A.G. Boshier. Note on the girth of Ramanujan graphs.J. Comb. Theory. Ser. B, 49:190–194, 1990
1990
Show all 38 references
-
[9]
Casazza and J.C
P.G. Casazza and J.C. Tremain. The kadison–singer problem in mathematics and engineering.Proc. Natl. Acad. Sci. USA, 103:2032–2039, 2006
-
[10]
F.R.K. Chen, V. Faber, and T.A. Manteuffel. On the diameter of a graph from eigenvalues associated with its laplacian.SIAM J. Disc. Math., 7:443–457, 1994
1994
-
[11]
Chung.Spectral Graph Theory, volume 92 ofReg
F.R.K. Chung.Spectral Graph Theory, volume 92 ofReg. Conf. Series Math.Amer. Math. Soc., Providence, RI, 1997
1997
-
[12]
A. Das. Triameter of graphs.Discuss. Math., Graph Theory, 41:601–616, 2021
2021
-
[13]
Doyle and J.E
J.K. Doyle and J.E. Graver. Mean distance in a graph.Discrete Math., 17:147–154, 1977
1977
-
[14]
M. Fiedler. Algebraic connectivity of graphs.Czech. Math. J., 23:298–305, 1973
1973
-
[15]
Friedlander
L. Friedlander. Extremal properties of eigenvalues for a metric graph.Ann. Inst. Fourier, 55:199–212, 2005
2005
-
[16]
Garijo, A
D. Garijo, A. M´ arquez, and R.I. Silveira. Continuous mean distance of a weighted graph.Results Math., 78:139, 2023
2023
-
[17]
R.J. George. Private communication, 2026
2026
-
[18]
J.B. Kennedy. A family of diameter-based eigenvalue bounds for quantum graphs. In F.M. Atay, P.B. Kurasov, and D. Mugnolo, editors,Discrete and Continuous Models in the Theory of Networks (Proc. Bielefeld 2017), volume 281 ofOper. Theory Adv. Appl., pages 213–240, Basel, 2020....
2017
-
[19]
Kennedy, P
J.B. Kennedy, P. Kurasov, G. Malenov´ a, and D. Mugnolo. On the spectral gap of a quantum graph.Ann. Henri Poincar´ e, 17:2439–2473, 2016
2016
-
[20]
Kirchhoff
G. Kirchhoff. Ueber die Aufl¨ osung der Gleichungen, auf welche man bei der Untersuchung der linearen Vertheilung galvanischer Str¨ ome gef¨ uhrt wird.Ann. Physik, 148:497–508, 1847
-
[21]
Kowalski.An introduction to expander graphs, volume 26 ofCours Sp´ ec
E. Kowalski.An introduction to expander graphs, volume 26 ofCours Sp´ ec. (Paris). SMF, Paris, 2019
2019
-
[22]
Lubetzky and Y
E. Lubetzky and Y. Peres. Cutoff on all Ramanujan graphs.GAF A, Geom. Funct. Anal., 26:1190–1216, 2016
2016
-
[23]
Lubotzky, R
A. Lubotzky, R. Phillips, and P. Sarnak. Ramanujan graphs.Combinatorica, 8:261–277, 1988
1988
-
[24]
Marcus, D.A
A.D. Marcus, D.A. Spielman, and N. Srivastava. Interlacing families I: Bipartite Ramanujan graphs of all degrees. Ann. Math., 182:307–325, 2015
2015
-
[25]
Marcus, D.A
A.D. Marcus, D.A. Spielman, and N. Srivastava. Interlacing families II: Mixed characteristic polynomials and the Kadison–Singer problem.Ann. Math., 182:327–350, 2015
2015
-
[26]
Marcus, D.A
A.D. Marcus, D.A. Spielman, and N. Srivastava. Interlacing families IV: Bipartite Ramanujan graphs of all sizes. SIAM J. Comput., 47:2488–2509, 2018
2018
-
[27]
Margulis
G.A. Margulis. Explicit construction of concentrators.Problemy Peredachi Informatsii, 4:71–80, 1973
1973
-
[28]
Margulis
G.A. Margulis. Explicit group theoretic constructions of combinatorial schemes and their applications for the con- struction of expanders and concentrators.Problemy Peredachi Informatsii, 24:51–60, 1988
1988
-
[29]
B. Mohar. Eigenvalues, diameter, and mean distance in graphs.Graphs and combinatorics, 7:53–64, 1991
1991
-
[30]
D. Mugnolo. What is actually a metric graph? arXiv:1912.07549, 2019
1912 arXiv
-
[31]
Mugnolo and M
D. Mugnolo and M. Pl¨ umer. On torsional rigidity and ground-state energy of compact quantum graphs.Calc. Var., 62:27, 2023
2023
-
[32]
S. Nicaise. Some results on spectral theory over networks, applied to nerve impulse transmission. In C. Brezinski, A. Draux, A. P. Magnus, P. Maroni, and A. Ronveaux, editors,Polynˆ omes Orthogonaux et Applications (Proc. Bar-le-Duc 1984), volume 1171 ofLect. Notes. Math., pag...
1984
-
[33]
S. Nicaise. Approche spectrale des problemes de diffusion sur les r´ eseaux. InS´ eminaire de Th´ eorie du Potentiel Paris, volume 8, pages 120–140. Springer-Verlag, Berlin, 1987
1987
-
[34]
S. Nicaise. Spectre des r´ eseaux topologiques finis.Bull. Sci. Math., II. S´ er., 111:401–413, 1987
1987
-
[35]
A. Nilli. On the second eigenvalue of a graph.Discrete Math., 91:207–210, 1991
1991
-
[36]
Pl¨ umer
M. Pl¨ umer. Upper eigenvalues bounds for the Kirchhoff Laplacian on embedded metric graphs.J. Spectral Theory, 11:1857–1894, 2021. 18 D. MUGNOLO
2021
-
[37]
N.T. Sardari. Diameter of Ramanujan graphs and random Cayley graphs.Combinatorica, 39:427–446, 2019
2019
-
[38]
N. Shimizu. The average distance and the diameter of dense random regular graphs.Electron. J. Comb., 27:#P3.62, 2020. Lehrgebiet Analysis, F akult¨at Mathematik und Informatik, FernUniversit¨at in Hagen, D-58084 Hagen, Germany Email address:delio.mugnolo@fernuni-hagen.de
2020
Reviewed August 2, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.