A pattern beneath different shapes
An analysis of 21 marine diatom genera found a common statistical pattern in the way their structures are arranged, even though the sampled forms differ. Every analyzed genus showed a positive long-wavelength spectral slope, which the study interprets as suppressed density fluctuations at large scales. The result is a structural-statistical comparison, not a phylogenetic explanation of how the pattern arose.
That pattern is described as hyperuniformity. In ordinary terms, it means density fluctuations are suppressed at long wavelengths. The study looked for it in two complementary ways: one tracked fluctuations across spatial wavelengths, while the other measured how local volume-fraction variance decayed as the observation window grew. The first result was summarized by a spectral-density slope, called alpha, and the second by a direct-space exponent, called beta.
Built from images, with two kinds of measurement
The dataset covered 21 marine diatom genera from three taxonomic classes. Fifteen genera were represented by original in-house scanning electron microscope images, while the remaining genera came from curated public SEM material. Images were manually selected for clarity and taxonomic traceability, cropped to individual valves and binarized with adaptive thresholding using Otsu or local-mean methods.
The binary view retained pore size, connectivity and porosity. A second, point-pattern view isolated positional correlations. That allowed the analysis to compare organization at more than one scale, rather than reducing each image to a single visual impression.
One broad pattern, several kinds of order
On the spectral-density measure, a positive slope counted as hyperuniform behavior. A slope above 1 placed a genus in Class I; exactly 1 placed it in Class II; and a value between 0 and 1 placed it in Class III. The analyzed set was described as approximately divided between predominantly Class I and predominantly Class III behavior.
The direct-space analysis varied by genus. Its exponent was greater than the spatial dimension for every genus except Planktoniella. Roperia had a reported value of 2.97 and Coscinodiscus 2.84, while Porosira and Thalassionema had intermediate values of 2.55 and 2.53. A synthetic Poissonian pattern was included as a non-hyperuniform reference.
The two measures should not be treated as interchangeable scores. Finite, heterogeneous and anisotropic images, along with segmentation and projection effects, meant that the direct- and reciprocal-space exponents did not always follow their ideal asymptotic relationship. The authors therefore treated them as complementary descriptions of structures recovered from finite images.
Other measures showed that order was not a single thing. Local hexagonal order was stronger in Roperia and Planktoniella, while Porosira and Azpeitia had lower mean local-order values of about 0.68. More ordered Class I architectures generally showed slower correlation decay and lower gamma, whereas predominantly Class III architectures tended to lose correlations faster and had higher gamma. Correlation persistence varied continuously across genera.
A map rather than a ladder
To compare the measurements together, the analysis used three standardized descriptors: the spectral exponent, the inverse sign of the correlation-persistence measure and the local hexagonal-order value. It left beta out because of redundancy and finite-image robustness. The first two principal components, a way of compressing several measurements into fewer dimensions, accounted for 87.5% of the variation between genera. The first component alone accounted for 55.7% and was primarily associated with spectral-density scaling.
The second component accounted for 31.8% and contrasted local order with correlation persistence. The reported fit between those two descriptors had an R-squared value of 0.19, indicating that they were only weakly related in this dataset. The measurements therefore captured partly independent aspects of organization rather than one master order axis.
The results describe a continuous structural spectrum across morphologically and taxonomically diverse genera, rather than a pattern confined to one lineage or visual design. That supports the study's aim of using hyperuniformity as a common statistical framework for comparing different diatom architectures. It remains a comparative map, however, not a phylogenetic account of how those architectures arose.
The evidence has a boundary
The analysis was limited to two-dimensional SEM representations of a limited sample. It provides a comparative rather than phylogenetic map, and the authors identify three-dimensional measurements and explicit phylogenetic analyses as needed next. The document is an arXiv version-1 preprint dated 27 August 2026.
For now, the clearest conclusion is narrow: across 21 genera, the analysis found positive long-wavelength scaling, while local order and correlation persistence varied across the set. The preprint presents hyperuniformity as a shared statistical language for comparing these images, with its descriptors capturing partly independent aspects of organization.
Paper data and sources
Original title: Hyperuniformity as a unifying organizational principle across diatom architectures
Authors: Eric Ballestero, Chiara Gazzola, Raj Kumar Pal et al.
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-27
DOI: Not available
Original paper · Full text