Critical Fluctuations for the Spherical Sherrington-Kirkpatrick Model in an External Field

Pax Kivimae

REVIEW 2 major objections 5 minor 22 references

Critical Fluctuations for the Spherical Sherrington-Kirkpatrick Model in an External Field

T0 review · 2 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The paper proves a critical external-field window in which spherical SK ground-state fluctuations interpolate between Tracy–Widom and Gaussian laws, via the stochastic Airy operator.

desk verdict The paper proves the conjectured critical fluctuations for the spherical SK model in an external field, with a repairable regularity gap in Lemma 3.7 and a scaling typo in Theorem 1.3. read the letter →

arxiv 1908.07512 v2 pith:SR2H2TGL submitted 2019-08-20 math.PR math-phmath.MP

Pax Kivimae This is my paper · ORCID

classification math.PRmath-phmath.MP MSC 60B2082B4460K3582B27

keywords sphericalSherrington-KirkpatrickmodelTracy-WidomdistributionexternalfieldcriticalfluctuationsstochasticAiryoperatorbetaensembleWeyl-Titchmarshm-functionrandommatrixtheory

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

External fields change the way the spherical Sherrington–Kirkpatrick model's ground-state energy fluctuates, and this paper pinpoints the transition. When the field is scaled as $N^{1/6}h_N\to h$, the paper proves that $N^{2/3}(1+\tfrac12 h_N^2-E^\beta_{N,h_N})$ converges in law to a one-parameter family $-TW^h_\beta$ that reduces to the Tracy–Widom law at $h=0$ and connects to the Gaussian regime for fixed fields. The same approach proves a stronger statement: the $N^{2/3}$-rescaled point process of the top critical values converges to an explicit limit $\Lambda^{h,k}_\beta$. A parallel theorem treats a joint critical regime with a Curie–Weiss ferromagnetic term, producing a three-parameter family $TW^h_{\beta,w}$. Both theorems are proved at the level of the $\beta$-ensemble analogue, so they cover the real, complex, and quaternionic spherical SK models.

What carries the argument

The engine is the stochastic Airy operator $$A^\$\beta$ = -\frac{$d^{2}$}{$dx^{2}$}+x+\frac{2}{\sqrt{\$\beta$}}B'_x$$ on $L^2(\mathbb R_+)$, with $w$-Robinson boundary condition $w\phi(0)=\phi'(0)$ (and $\phi(0)=0$ when $w=\infty$); its eigenvalues are the limiting edge eigenvalues of the rescaled $\beta$-Hermite ensemble. The paper's new input is to control the resolvent entry appearing in the Lagrange-dual formula for the ground-state energy. The central object is the Weyl solution $\phi^w_\lambda$, the unique meromorphic family solving $A^\beta\phi=\lambda\phi$ with normalization $w\phi^w_\lambda(0)+1=(\phi^w_\lambda)'(0)$, whose boundary value is the Weyl–Titchmarsh $m$-function. The discrete quasi-derivatives $D_N^{[1]},D_N^{[2]}$ and the compactness lemma that convert weak $L^2$ convergence of discrete resolvent columns into compact-uniform convergence of their first entries carry the proof; Proposition 3.3 is the analytic heart.

What would settle it

Construct a sequence $f_N$ in the discrete quasi-derivative setting such that $f_N$ and $D_N f_N$ converge weakly in $L^2$ to an $H^1_{\rm loc}$ limit $f$ with $f'\notin C^0$, yet the interpolating remainder $f_N-g_N$ does not vanish compact-uniformly; this would disprove the compact-uniform convergence used in Proposition 3.3. Alternatively, check whether Lemma 3.8 remains true for $H^1_{\rm loc}$ limits: if it does, the gap is repairable, and if it does not, the limiting point-process identification lacks a proof.

Watch

Extended reading notes

Core claim

The central claim is that the critical fluctuations of the spherical SK ground state are governed by the stochastic Airy operator, not just through its eigenvalues but through a boundary object: the Weyl solution. For each $\lambda$ off the spectrum, the relevant quantity is the boundary value or boundary derivative of the unique solution $\phi^w_\lambda$ of $A^\beta\phi=\lambda\phi$ with $w$-Robinson data, and the limiting fluctuation law is obtained by optimizing $\tfrac12(\lambda-h^2\phi^w_\lambda(0))$ over $\lambda$ below the first eigenvalue. The paper proves the compact-uniform convergence of the corresponding discrete resolvent entries to these Weyl data, and from that convergence derives the point-process limits in Theorems 1.1 and 1.3. In the Curie–Weiss case the same machinery yields the three-parameter family $TW^h_{\beta,w}$, subsuming the two-parameter family of the spiked random-matrix transition.

Load-bearing premise

The load-bearing step is the compact-uniform convergence of the discrete resolvent entry, and that step applies a compactness lemma (Lemma 3.8) to a limit function whose first derivative is only known to be locally square-integrable, while the lemma is stated for continuously differentiable functions; if that regularity gap cannot be closed, the identification of the limiting point process would need a new argument.

Editorial extensions

If this is right

For every $\beta>0$ and every $h$, the ground-state fluctuation law $-TW^h_\beta$ interpolates between the Tracy–Widom law at $h=0$ and the Gaussian fluctuations of the fixed-field regime.
The $N^{2/3}$ scaling applies to the entire top end of the critical-value landscape: the point process of the $k$ largest critical values has an explicit limiting law $\Lambda^{h,k}_\beta$, not just its minimum.
With a critical Curie–Weiss term, the limiting fluctuation is a three-parameter distribution $TW^h_{\beta,w}$, extending the two-parameter family previously known without an external field.
Because the theorems are proved for the $\beta$-ensemble analogue, they apply simultaneously to the real ($\beta=1$), complex ($\beta=2$), and quaternionic ($\beta=4$) spherical SK models.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

Extension beyond the paper: the resolvent-entry convergence method suggests that the same Weyl–Titchmarsh data control rank-one perturbations of general spiked tridiagonal ensembles at the edge, so the limiting laws should be universal across ensembles with the same stochastic Airy limit.
Testable extension: for $h_N=hN^{-1/6}$, empirical quantiles of $N^{2/3}(1+h_N^2/2-E_{N,h_N})$ should track the quantiles of $-TW^h_\beta$; this prediction is sharp enough for a direct simulation check.
Conjectural connection: the three-parameter family $TW^h_{\beta,w}$ plausibly interpolates between the spiked-random-matrix phase transition and the Tracy–Widom law, making the Curie–Weiss critical window a spherical analogue of that transition.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Bottom line: this is a real proof of a conjectured critical fluctuation theory, not a numerical or heuristic paper. It confirms the N^{1/6} h_N critical window predicted in [3,12], produces the interpolating Tracy-Widom family, and adds the joint Curie-Weiss/external-field transition. The main theorems look right in broad strokes, and the machinery is genuinely new: the resolvent-entry convergence in Proposition 3.3 is the key technical step, and it goes beyond the L^2 resolvent convergence in [18,4]. That alone is a solid contribution.

The paper earns credit for building the limit objects from eigenfunctions of the stochastic Airy operator rather than fitting finite-N data. The special cases h=0 and fixed h fall out, so the family is not reverse-engineered. The citation pattern is honest; the paper clearly states what was known and what was conjectured.

Now the soft spots, and they are exactly where the reader put them. Theorem 1.3 as printed has an inconsistent scaling: it states N^{2/3}(2 - Crit_{N-k,N}(H^beta_{N,mu,h})) with N^{1/2}(1 - mu_N) -> w, but the proof and the reference [4] use N^{1/3}(1 - mu_N) -> w. This is a typo-level issue but it is load-bearing for anyone trying to state the theorem; it must be fixed before publication.

More serious: the proof of Lemma 3.7 applies Lemma 3.8 to a limit f^{[1]} = f' - yf that is only known to lie in H^1_loc, while Lemma 3.8 is stated for C^1 limits and its proof uses sup|f'|. That is a genuine gap in the printed argument. The stress-test note is right that it is repairable: a standard compactness argument (weak L^2 convergence plus the discrete inequality |g_N - f_N| <= m_N^{-1}|g_N'| and Morrey embedding) gives uniform convergence for H^1 limits. But the manuscript needs to state and prove that weaker lemma. Since Proposition 3.3 depends on this step, the proof as written is not complete.

Neither issue undermines my belief in the result. The strategy is coherent, the technical difficulties are engaged head-on, and the gaps look like repairs, not holes. For a reader working on random matrix edge statistics or spherical spin glasses, this is worth serious engagement. For a referee, I would send it out, with instructions to demand the scaling fix and the H^1 version of Lemma 3.8.

Referee Report

2 major / 5 minor

Summary. This paper studies the spherical Sherrington-Kirkpatrick model in an external field, and a joint external-field/Curie-Weiss critical regime, for general inverse temperature β. The main results identify the fluctuations of the low-lying critical values of the Hamiltonian. Theorem 1.1 states that when N^{1/6}h_N→h, the top N-k critical values, shifted by N^{2/3}(1+h_N^2/2), converge to a point process Λ^{h,k}_β built from the β-stochastic Airy operator; in particular the ground-state fluctuation N^{2/3}(1+h_N^2/2-E^β_{N,h_N}) converges to -TW^h_β. Theorem 1.3 extends this to a joint critical regime with a ferromagnetic Curie-Weiss term, producing a three-parameter family Λ^k_{β,w,h}. The proof passes through a Lagrange-dual representation, a general convergence theorem for spiked tridiagonal ensembles (Proposition 3.1), a discrete-to-continuum resolvent-entry result (Proposition 3.3), and new stochastic-operator facts in Section 4.

Significance. If the technical gaps are repaired, the paper confirms predictions of Baik-Lee and Fyodorov-Le Doussal and extends the Bloemendal-Virág spiked-random-matrix transition to the spherical SK setting. The limiting distributions are constructed from eigenfunctions of the stochastic Airy operator, not fitted to finite-N data, and the known Tracy-Widom and Bloemendal-Virág limits arise as special cases. The point-process-level statement is stronger than ground-state convergence alone and covers all β>0, with complex and quaternionic versions via the Householder reduction. The proof strategy is coherent and builds seriously on [4,14,18]. However, two load-bearing issues in the printed version must be addressed: the stated μ-scaling in Theorem 1.3 is inconsistent with the proof, and the proof of Lemma 3.7 invokes Lemma 3.8 outside the hypotheses of that lemma.

major comments (2)

[Theorem 1.3, Eq. (22); Section 3.3] The stated scaling of the ferromagnetic coupling in Theorem 1.3 is inconsistent with the proof. The theorem requires N^{1/2}(1-μ_N)→w, but in the reduction in Section 3.3 the spike parameter is w_N=N^{1/3}(1−μ_N). Under the theorem's stated scaling, w_N→0 for every finite w, so the limiting process would be the w=0 member of the family Λ^k_{β,w,h}, not the claimed Λ^k_{β,w,h} with arbitrary w. The statement should read N^{1/3}(1−μ_N)→w, the scaling used in [4] and in the rest of the paper; with the current text, Eq. (22) and the subsequent application of Proposition 3.1 do not match.
[Section 3.5, Lemma 3.7; Lemma 3.8; Proposition 3.3] In the proof of Lemma 3.7, Lemma 3.8 is applied to f_N^{[1]} with limit f^{[1]}=f'−yf. At that point only f∈D_max is known, which gives f^{[1]}∈H^1_loc; the limit need not be C^1. Lemma 3.8 as stated assumes f∈C^1, and its proof uses sup_I|f'|, so the invocation is not justified as printed. This is load-bearing: the compact-uniform convergence of f_N^{[1]} is what upgrades the weak L^2 convergence of the discrete quasi-derivatives to the resolvent-entry convergence in Proposition 3.3, and hence to the limiting point process. The gap appears repairable by proving an H^1 version of Lemma 3.8 (for instance, combining |g_N(x)-f_N(x)|≤m_N^{-1}|g_N'(x)| with weak L^2 convergence and Morrey compactness, and replacing sup|f'| by y_2(x)m_N^{-1/2}||f'||_{L^2}); the manuscript should state and prove that version.

minor comments (5)

[Section 3.3, Eq. (65)] The displayed relation in Eq. (65) appears to be N^{2/3}(2−H/N)=L, but the preceding calculation gives N^{2/3}(1−H/N)=L. As printed the constant would affect the h_N^2/2 shift in Theorem 1.1; this should be corrected to match the derivation.
[Section 1, paragraph after Eq. (7)] The text states that for fixed h, √N(E_N,h−√(1−h²)) converges to a Gaussian limit. Given the immediately preceding statement that E_N,h→√(1+h²), the formula should be √(1+h²) rather than √(1−h²).
[Abstract and Introduction] The word 'quarternionic' should be 'quaternionic'.
[Section 1.1, notation] The notation R^* = R ∪ {∞} is potentially confusing because R^* is commonly used for the nonzero reals; a different symbol such as R_* would avoid ambiguity.
[Section 3.3, paragraph after Eq. (58)] The phrase 'a w-spiked (∞-spike) tridiagonal ensemble' is awkward; it should say that the pure SK case corresponds to w=∞, while the Curie-Weiss case corresponds to finite w when the scaling is corrected.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: limiting laws are derived from convergence to the stochastic Airy operator, not fitted to finite-N data.

full rationale

The paper's central limit theorems are not circular. The limiting point processes Lambda^{h,k}_beta and distributions TW^h_beta are defined in Sections 3.2 and 4 from Weyl solutions of the stochastic Airy operator (see (10)-(13) and Proposition 4.11), not from finite-N data or from the conjectured laws in [3,12]. The paper proves a resolvent-entry convergence result (Proposition 3.3) and derives the point-process convergence through Lagrange duality; the h=0 and w-only cases reduce to the known Tracy-Widom and Bloemendal-Virag limits as special cases, rather than being used as inputs to force the new h-dependence. The citations [4,18] provide the external stochastic-operator framework and are independent prior work, not self-citations, so they constitute real evidence rather than circularity. The proof gap noted in the reader's take concerning Lemma 3.7's invocation of Lemma 3.8 is a mathematical regularity issue, not a circularity: the compact-uniform convergence in question is derived from convergence of resolvent entries, not assumed as the conclusion. Overall, no load-bearing step reduces by definition, by fitted parameter, or by self-citation to its own inputs.

Assumptions & free parameters 0 free parameters · 4 assumptions · 1 invented entities

The central claim rests on prior results in stochastic operator theory, namely norm-resolvent convergence of beta-Hermite and spiked tridiagonal ensembles, and on Weyl-Titchmarsh theory. These are imported rather than re-derived. No numerical free parameters are fitted to data. The only new postulated objects are the limiting point-process families, which are derived and have independent falsifiable content.

assumptions (4)

domain assumption The w-spiked tridiagonal ensemble H_{N,w_N} converges to the stochastic operator H_w in the norm-resolvent sense (Proposition 3.2, cited to Krishnapur-Rider-Virag [14]).
This imported convergence is the starting point for the resolvent-entry analysis in Proposition 3.3 and hence for both main theorems.
domain assumption For beta = 1,2,4 the Householder reduction identifies the beta-spherical SK model with the original GOE/GUE/GSE model; for general beta the theorem concerns the beta-ensemble analogue.
The paper defines H^beta_{N,h} via the beta-Hermite tridiagonal matrix; the equivalence for beta=1 is from Dumitriu-Edelman [8] and is used to transfer results to the original model.
standard math Weyl-Titchmarsh theory for Sturm-Liouville operators with distributional potentials (Eckhardt et al. [9]) supplies the Weyl solutions phi^w_lambda and the identities d/dlambda phi^w_lambda(0) = ||phi^w_lambda||^2.
Used in Section 4 to define the limiting point processes and the m-function identities that drive the variational formulas.
domain assumption The beta-Hermite ensemble satisfies Assumptions 1-3 (tightness, growth and oscillation bounds) with limit y = x^2/2 + (2/sqrt(beta)) B_x.
Established in [18], Section 6; needed to apply the spiked-ensemble theory to the matrices B_N.

invented entities (1)

The limit families TW^h_beta and TW^h_{beta,w}, and the point processes Lambda^{h,k}_beta and Lambda^k_{beta,w,h} independent evidence
purpose: Describe the N^{2/3}-scale fluctuations of ground-state and low-index critical values in the critical external-field and Curie-Weiss windows.
They are defined independently of the finite-N model via eigenfunctions of the stochastic Airy operator, reduce to known Tracy-Widom and Bloemendal-Virag laws at h=0, and are falsifiable by simulation of finite-N critical values.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Critical Fluctuations for the Spherical Sherrington-Kirkpatrick Model in an External Field." pith.science (2026). https://pith.science/paper/SR2H2TGL

@misc{pith2026190807512,
  author       = {Pith},
  title        = {Pith review of: Critical Fluctuations for the Spherical Sherrington-Kirkpatrick Model in an External Field},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SR2H2TGL}},
  note         = {Machine review of arXiv:1908.07512}
}

abstract

We prove the existence of a critical regime for the fluctuations of the ground-state energy of the spherical Sherrington-Kirkpatrick model in an external field, confirming predictions given in [3,12]. We also establish a critical regime for the fluctuations in a model with a critical Ferromagnetic interaction term, producing a three-parameter family of distributions generalizing the two-parameter family given in [4]. These results are both established in the generality of a $\beta$-ensemble analogue of the spherical Sherrington-Kirkpatrick model, which subsumes the complex and quarternionic generalizations.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

22 extracted references · 22 canonical work pages

[4]

and Vir´ ag, B

Bloemendal, A. and Vir´ ag, B. (2013). Limits of spiked random mat rices I. Probability Theory and Related Fields , 156(3):795–825

work page 2013
[1]

Baik, J., Ben Arous, G., and P´ ech´ e, S. (2005). Phase transition of the largest eigenvalue for nonnull complex sample covariance matrices. Ann. Probab., 33(5):1643–1697

work page 2005
[2]

Spherical spin glass model with external field

Baik, J., Collins-Wildman, E., Le Doussal, P. and Wu, H. (2020). Sphe rical spin glass model with external ﬁeld. arxiv, cond-mat.dis-nn/2010.06123

work page Pith review arXiv 2020
[3]

and Lee, J

Baik, J. and Lee, J. O. (2017). Fluctuations of the free energy of the spherical Sherrington–Kirkpatrick model with ferromagnetic interaction. Annales Henri Poincar´ e, 18(6):1867–1917

work page 2017
[5]

and Sen, A

Chen, W.-K. and Sen, A. (2017). Parisi formula, disorder chaos and ﬂuctuation for the ground state energy in the spherical mixed p-spin models. Communications in Mathemat- ical Physics , 350(1):129–173

work page 2017
[6]

and Sommers, H.-J

Crisanti, A. and Sommers, H.-J. (1992). The spherical p -spin int eraction spin glass model: the statics. European Physical Journal B , 87:341–354

work page 1992
[7]

and Forrester, P

Desrosiers, P. and Forrester, P. J. (2006). Asymptotic corr elations for Gaussian and Wishart matrices with external source. International Mathematics Research Notices

work page 2006
[8]

and Edelman, A

Dumitriu, I. and Edelman, A. (2002). Matrix models for beta ense mbles. Journal of Mathematical Physics , 43(11):5830–5847

work page 2002

Show all 22 references

[9]

Eckhardt, J., Gesztesy, F., Nichols, R., and Teschl, G. (2012). W eyl-titchmarsh theory for sturm-liouville operators with distributional potentials. Opuscula Mathematica , 33:467— -563

2012
[10]

and Sutton, B

Edelman, A. and Sutton, B. D. (2007). From random matrices t o stochastic operators. Journal of Statistical Physics , 127(6):1121–1165

2007
[11]

Forsythe, G. E. and Golub, G. H. (1965). On the stationary va lues of a second-degree polynomial on the unit sphere. Journal of the Society for Industrial and Applied Mathe- matics, 13(4):1050–1068

1965
[12]

Fyodorov, Y. V. and Le Doussal, P. (2014). Topology trivializat ion and large deviations for the minimum in the simplest random optimization. Journal of Statistical Physics , 154(1):466–490

2014
[13]

M., Thouless, D

Kosterlitz, J. M., Thouless, D. J., and Jones, R. C. (1976). Sph erical model of a spin- glass. Phys. Rev. Lett. , 36:1217–1220. 27

1976
[14]

Krishnapur, M., Rider, B., and Vir´ ag, B. (2013). Universality of the stochastic airy operator. Communications on Pure and Applied Mathematics , 69:145–199

2013
[15]

Minami, N. (2015). Deﬁnition and self-adjointness of the stoch astic airy operator. Markov Processes and Related Fields , 21(3P):695–711

2015
[16]

and Talagrand, M

Panchenko, D. and Talagrand, M. (2007). On the overlap in the multiple spherical SK models. Ann. Probab., 35(6):2321–2355

2007
[17]

P´ ech´ e, S. (2006). The largest eigenvalue of small rank perturbations of hermitian random matrices. Probability Theory and Related Fields , 134(1):127–173

2006
[18]

Ramir´ ez, J., Rider, B., and Vir´ ag, B. (2011). Beta ensembles, stochastic airy spectrum, and a diﬀusion. Journal of the American Mathematical Society , 24(4):919–944

2011
[19]

and Simon, B

Reed, M. and Simon, B. (1981). Methods of Modern Mathematical Physics: Vol. I: Functional Analysis. Methods of Modern Mathematical Physics. Elsevier Science

1981
[20]

and Reed, M

Simon, B. and Reed, M. (1978). Methods of Modern Mathematical Physics: Vol.: 4. : Analysis of Operators . Methods of Modern Mathematical Physics. Academic Press

1978
[21]

Talagrand, M. (2006). Free energy of the spherical mean ﬁeld model. Probability Theory and Related Fields , 134:339–382

2006
[22]

Tracy, C. A. and Widom, H. (1996). On orthogonal and symplec tic matrix ensembles. Comm. Math. Phys. , 177(3):727–754 28

1996

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [4]

and Vir´ ag, B

Bloemendal, A. and Vir´ ag, B. (2013). Limits of spiked random mat rices I. Probability Theory and Related Fields , 156(3):795–825

work page 2013

[2] [1]

Baik, J., Ben Arous, G., and P´ ech´ e, S. (2005). Phase transition of the largest eigenvalue for nonnull complex sample covariance matrices. Ann. Probab., 33(5):1643–1697

work page 2005

[3] [2]

Spherical spin glass model with external field

Baik, J., Collins-Wildman, E., Le Doussal, P. and Wu, H. (2020). Sphe rical spin glass model with external ﬁeld. arxiv, cond-mat.dis-nn/2010.06123

work page Pith review arXiv 2020

[4] [3]

and Lee, J

Baik, J. and Lee, J. O. (2017). Fluctuations of the free energy of the spherical Sherrington–Kirkpatrick model with ferromagnetic interaction. Annales Henri Poincar´ e, 18(6):1867–1917

work page 2017

[5] [5]

and Sen, A

Chen, W.-K. and Sen, A. (2017). Parisi formula, disorder chaos and ﬂuctuation for the ground state energy in the spherical mixed p-spin models. Communications in Mathemat- ical Physics , 350(1):129–173

work page 2017

[6] [6]

and Sommers, H.-J

Crisanti, A. and Sommers, H.-J. (1992). The spherical p -spin int eraction spin glass model: the statics. European Physical Journal B , 87:341–354

work page 1992

[7] [7]

and Forrester, P

Desrosiers, P. and Forrester, P. J. (2006). Asymptotic corr elations for Gaussian and Wishart matrices with external source. International Mathematics Research Notices

work page 2006

[8] [8]

and Edelman, A

Dumitriu, I. and Edelman, A. (2002). Matrix models for beta ense mbles. Journal of Mathematical Physics , 43(11):5830–5847

work page 2002