REVIEW 4 major objections 5 minor 16 references
Dynamic Fingerprint of Controlled Structural Disorder in Artificial Spin Lattices
T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash
Pith's one-line read Static structural entropy predicts the GHz spin-wave complexity of disordered nanomagnet arrays.
desk verdict A clever sample series and a plausible qualitative story, but the headline regression rests on six simulated spectra and a subset-size mismatch; should go to review with the quantitative claim decoupled. 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 carrying objects are three Shannon entropies built from probability distributions: $S_{\mathrm{config}}$ bins nanobar orientations over 100 angular bins; $S_{\mathrm{connect}}$ bins the weighted degree of each nanobar, a sum of $1/r^3$ alignment-dependent interaction weights over neighbors; and $S_{\mathrm{spectral}}$ bins the normalized power among key spectral features (peaks and inflection points) of the simulated GHz response. Pair and orientational correlation functions $g_{\mathrm{pc}}(r)$ and $g_{\mathrm{oc}}(r)$ characterize the structural ordering that these entropies summarize. The quantitative link is made by Pearson correlations and a multiple linear regression of $S_{\mathrm{spectral}}$ on $S_{\mathrm{connect}}$ and $\langle|B_{\mathrm{local}}|\rangle$. Micromagnetic simulations, broadband ferromagnetic resonance, and Brillouin light scattering supply the dynamic spectra.
What would settle it
Simulate the same six arrays on multiple independent subsets and on progressively larger patches, then recompute the regression of $S_{\mathrm{spectral}}$ on $S_{\mathrm{connect}}$ and $\langle|B_{\mathrm{local}}|\rangle$; if the adjusted $R^2$ drops substantially or the regression coefficients change sign, the reported dynamic fingerprint is an artifact of the chosen subset rather than a property of the full array.
Extended reading notes
Core claim
The central claim is that dynamic spectral complexity, quantified as spectral entropy $S_{\mathrm{spectral}}$ computed from micromagnetic power spectra, is largely determined by two static structural metrics: connectivity entropy $S_{\mathrm{connect}}$, which captures heterogeneity of local dipolar environments, and the average local dipolar field strength $\langle|B_{\mathrm{local}}|\rangle$. Across six samples spanning periodic to random structure, these two metrics together account for roughly 95% of the variance in $S_{\mathrm{spectral}}$ (adjusted $R^2 \approx 0.948$, $p = 0.0054$). The authors further separate the two disorder channels: positional disorder alone lifts spin-wave degeneracies through modified dipolar coupling, while rotational disorder is the dominant driver that proliferates modes into a dense spectral manifold. They interpret this evolution as a dynamic fingerprint of an increasingly complex energy landscape, and show that thermal probing reveals richer mode diversity than driven inductive probing.
Load-bearing premise
The dynamic entropy is computed from micromagnetic simulations of a small 3.5 micrometer square subset containing roughly a hundred nanobars, while the structural metrics are computed from the full arrays of more than 60,000 nanobars; the central correlation assumes the small simulated patch statistically represents the full array.
Editorial extensions
If this is right
- A purely static characterization of a fabricated array suffices to predict a large share of its GHz spectral complexity, so fabrication tolerances can be screened by structural imaging.
- Rotational disorder can be used as a design knob to enrich spin-wave spectra, pointing toward reservoirs with more diverse mode sets for neuromorphic computing.
- The same entropy metrics can be applied to other artificial spin ice geometries to test whether the disorder-to-complexity mapping is universal.
- Thermal excitation provides a more complete view of the available microstates than driven excitation, an insight relevant to reading out the state of disordered magnonic systems.
Reading between the lines
- Because the regression uses only six samples, the reported $R^2$ and $p$-value are suggestive rather than decisive; a natural extension is to fabricate intermediate disorder levels and test whether the linear relation persists.
- The connectivity entropy relies on an ad hoc effective interaction radius, so recomputing the metrics with different cutoffs would test how sensitive the correlation is to that modeling choice.
- If confirmed on larger simulated subsets, $S_{\mathrm{spectral}}$ could serve as a low-cost dynamic fingerprint for classifying disordered magnetic arrays without full spectral measurement.
- The framework could be extended by varying the nanobar material or dimensions, which would test whether the entropy link survives changes in the underlying interaction strength.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper introduces artificial spin lattices of Ni81Fe19 nanobars with controlled positional and rotational disorder, realized as six samples spanning periodic (S-F) to fully random (S-A). It characterizes static structural disorder through configurational entropy Sconfig, connectivity entropy Sconnect, and average local dipolar field <|Blocal|>, and characterizes dynamics through VNA-FMR, thermal Brillouin light scattering, and micromagnetic simulations. The central quantitative claim is a multiple linear regression Sspectral = β0 + β1 Sconnect + β2 <|Blocal|> with adjusted R^2 ≈ 0.948 (F = 46.92, p = 0.0054), interpreted as a dynamic fingerprint of the energy-landscape complexity. The authors further argue that rotational disorder is the dominant driver of spectral mode proliferation and that thermal BLS accesses more microstate diversity than driven FMR.
Significance. If the quantitative correlation is robust, the platform would be a valuable controlled testbed connecting structural disorder to collective high-frequency dynamics, with plausible relevance to glassy physics and neuromorphic magnonics. The paper is transparent about metric definitions, uses independently defined static and dynamic measures (so the reported correlation is not tautological), and combines experiment with simulation in a way that makes several falsifiable trends explicit. However, the central regression currently rests on only six samples and on simulated spectra obtained from small subsets whose representativeness is not demonstrated; the significance is therefore conditional on additional convergence and robustness evidence.
major comments (4)
- [Supplemental Material: Micromagnetic Simulation Details; main text Fig. 2(c)-(d)] Sspectral is computed from OOMMF runs on a 3.5 μm × 3.5 μm square subset containing roughly 100 nanobars, while Sconnect and <|Blocal|> are evaluated on full arrays with over 60,000 nanobars. No convergence test with box size, no comparison of the subset's structural statistics to the full layout, and no averaging over disorder realizations are reported. Because Sspectral is the dependent variable in the central regression, a size-dependent shift in the simulated spectra could produce or inflate the reported adjusted R^2. The authors should demonstrate representativeness (e.g., by computing Sconnect and <|Blocal|> on the same subset, simulating at least two box sizes, or simulating multiple disorder realizations) and quantify the resulting sensitivity of Sspectral.
- [Supplemental Material: Spectral Entropy (Sspectral)] Feature-detection thresholds for Sspectral are sample-specific: the Savitzky-Golay polynomial order is 3 for most samples but 4 for S-C, and the inflection-point power threshold is 0.05 for S-A, S-B, S-D, S-E, and S-F but 0.02 for S-C. No sensitivity analysis is provided. Since the number of detected key features directly determines Sspectral, the correlations in Fig. 2(d) may partly reflect these hand-adjusted choices rather than intrinsic spectral complexity. A robustness check over reasonable threshold values, or an alternative estimator of spectral entropy that does not rely on feature detection, is needed.
- [Supplemental Material: Connectivity Shannon Entropy (Sconnect)] The connectivity entropy uses hand-tuned parameters: effective interaction radii of 550 nm for closely aligned nanobars (δθ < 9°) and 500 nm otherwise, a 1/r^3 interaction weight, and 100 bins for the weighted-degree distribution. No sensitivity analysis is provided for these choices. Because Sconnect is one of the two predictors in the central regression, the authors should show that the reported correlation is stable under reasonable variations of Reff, the alignment threshold, and the bin count.
- [Main text Fig. 2(d) and Supplemental Material: Correlation and Regression Analysis] The regression is based on n = 6 samples with two predictors and no reported error bars or metric uncertainties. With such a small sample, adjusted R^2 ≈ 0.948 is not by itself evidence against overfitting; in-sample fits can be inflated even after adjustment. The authors should report coefficient estimates with standard errors and confidence intervals, and ideally include leave-one-out or bootstrap validation, while acknowledging that n = 6 limits the strength of the quantitative claim.
minor comments (5)
- [Main text, Eq. (1)] The phrase 'Ashell(r) is is the exact area' contains a duplicated word and should be corrected.
- [Main text, section discussing Smit-Beljers formalism] The spelling 'Smit-Beljer formulation' should be 'Smit-Beljers formulation' for consistency with the Supplemental Material and references.
- [Abstract] The sentence 'thermal probe via thermal Brillouin light scattering reveal significantly richer microstates diversity' is grammatically awkward; consider phrasing such as 'thermally excited Brillouin light scattering reveals significantly richer microstate diversity'.
- [Supplemental Material: Spectral Entropy] The spectral entropy is defined at H = -200 mT in the SM while the main text reports H = -2 kOe for the same quantity; please use consistent units or state the conversion explicitly.
- [Figure captions for Figs. 3 and 4 and SM Figs. S2-S3] The 'LN' naming format is described separately in each caption; defining it once at first occurrence and referring back thereafter would improve readability.
Circularity Check
No circularity: the static and dynamic metrics are independently defined, and the regression is a descriptive fit, not a prediction built from its own outputs.
full rationale
The derivation chain is not circular. The static metrics (Sconfig from the orientation histogram, Sconnect from the weighted-degree distribution with 1/r^3 weights, and ⟨|Blocal|⟩ from point-dipole sums) are computed directly from nanobar coordinates, while Sspectral is computed from OOMMF micromagnetic power spectra (SM 'Micromagnetic Simulation Details' and 'Spectral Entropy'). No quantity is defined in terms of another: Sspectral is not derived from Sconnect or ⟨|Blocal|⟩, and the correlation/regression is fit to the same six samples as a descriptive statistic, not presented as a held-out prediction. The only self-citation (ref. 30 / SM ref. 3) supports the Smit–Beljers resonance formula, which is independently sourced to Smit, Suhl, and Vittoria; it is not load-bearing for the central correlation. The subset-size mismatch (3.5 µm OOMMF boxes vs. full >60,000-bar arrays) and sample-dependent spectral feature thresholds are robustness/validity concerns, not evidence that the result reduces to its inputs by construction. Therefore no circular step is present.
Assumptions & free parameters
free parameters (4)
- Sconnect effective radius and alignment threshold =
R_eff = 550 nm if delta_theta < 9 degrees, else 500 nm
- Spectral feature detection thresholds =
Peak prominence 1e-4, height 1e-5; inflection thresholds 0.05 (0.02 for S-C); S-G order 3 (4 for S-C)
- |Blocal| integration cutoff =
100 nm to 470 nm
- Entropy bin count =
M = M' = 100
assumptions (4)
- domain assumption Material parameters for permalloy (A = 1.3e-11 J/m, Ms = 8e5 A/m, alpha = 0.01) are accurate for these nanobars.
- domain assumption The 3.5 micrometer by 3.5 micrometer simulation patch is representative of the full 60,000+ bar array statistics.
- domain assumption Point-dipole approximation with a distance cutoff captures the relevant dipolar interactions.
- domain assumption Spin-wave modes in the simulated patch correspond to the collective modes of the full array.
Cite this review
Pith. "Pith review of Dynamic Fingerprint of Controlled Structural Disorder in Artificial Spin Lattices." pith.science (2026). https://pith.science/paper/I25EPJDZ
@misc{pith2026250607007,
author = {Pith},
title = {Pith review of: Dynamic Fingerprint of Controlled Structural Disorder in Artificial Spin Lattices},
year = {2026},
howpublished = {\url{https://pith.science/paper/I25EPJDZ}},
note = {Machine review of arXiv:2506.07007}
}
read the original abstract
Investigating the emergence of complexity in disordered interacting systems, central to fields like spin glass physics, remains challenging due to difficulties in systematic experimental tuning. We introduce a tunable artificial spin lattice platform to directly probe the connection between controlled structural disorder and collective spin-wave dynamics. By precisely varying positional and rotational randomness in Ni81Fe19 nanobar arrays from periodic to random, we map the evolution from discrete spectral modes to a complex, dense manifold. Crucially, we establish a quantitative correlation between information-theoretic measures of static disorder and the dynamic spectral complexity derived from the GHz spin-wave response. This correlation provides a dynamic fingerprint of an increasingly complex energy landscape resulting from tuned disorder. Furthermore, thermal probe via thermal Brillouin light scattering reveal significantly richer microstates diversity in disordered states than driven probe using broadband ferromagnetic resonance. Our work presents a unique experimental testbed for studying how the ingredients of glassy physics manifest in high-frequency dynamics, offering quantitative insights into the onset of complexity in interacting nanomagnet systems.
Figures
Figures from the paper (1 more)
Reference graph
Works this paper leans on
-
[1]
J. Smit, H. Beljers, and S. Vonsovskii, Rep 1, 113 (1955)
work page 1955
- [2]
-
[3]
S. Lendinez, M. T. Kaffash, and M. B. Jungfleisch, Nano Lett ers 21, 1921 (2021)
work page 2021
-
[4]
Vittoria, Magnetics, Dielectrics, and Wave Propagation with MATLAB ® Codes (CRC Press, 2023)
C. Vittoria, Magnetics, Dielectrics, and Wave Propagation with MATLAB ® Codes (CRC Press, 2023)
work page 2023
-
[5]
Aharoni, Journal of applied physics 83, 3432 (1998)
A. Aharoni, Journal of applied physics 83, 3432 (1998)
work page 1998
- [6]
-
[7]
C. E. Shannon, Bell System Technical Journal 27, 379 (1948)
work page 1948
-
[8]
J. M. Coey, Magnetism and magnetic materials (Cambridge university press, 2010)
work page 2010
Show all 16 references
-
[9]
Savitzky and M
A. Savitzky and M. J. Golay, Analytical chemistry 36, 1627 (1964)
1964
-
[10]
Hansen and I
J.-P. Hansen and I. R. McDonald, Theory of Simple Liquids , 4th ed. (Academic Press, Oxford, 2013)
2013
-
[11]
D. A. McQuarrie, Statistical Mechanics (University Science Books, Sausalito, CA, 2000)
2000
-
[12]
De Gennes and J
P.-G. De Gennes and J. Prost, The physics of liquid crystals , 83 (Oxford university press, 1993)
1993
-
[13]
P. M. Chaikin, T. C. Lubensky, and T. A. Witten, Principles of condensed matter physics , Vol. 10 (Cambridge university press Cambridge, 1995)
1995
-
[14]
Bray and K
C. Bray and K. Gubbins, Theory of molecular fluids. volume 1 : Fundamentals (1984)
1984
-
[15]
G. W. Snedecor and W. G. Cochran, Statistical Methods, 8th ed. (Iowa State University Press, Ames, 1989)
1989
-
[16]
M. H. Kutner, C. J. Nachtsheim, J. Neter, and W. Li, Applied Linear Statistical Models , 5th ed. (McGraw-Hill/Irwin, Boston, 2005)
2005
Reviewed August 7, 2026 · model on record in the stance chip above.
Discussion (0). Sign in to comment.