A computer-based analysis points to local rings and pores as important features shaping the reflectance of a disordered photonic network. But the result does not identify a single biological structure: diamond-like and ctn-like models both matched the biological spectrum closely.
The study asks whether an inverse method can recover structural properties from reflectance spectra while testing the possible roles of local resonators and photonic-band-gap mechanisms.
Turning simulated color back into structure
The researchers generated disordered three-dimensional networks from periodic crystalline starting points using an extended Wooten-Weaire-Winer algorithm. Disorder was introduced through bond-switch moves, with the parameter β used to vary short-range disorder.
For each network type, the computational set contained 25 realizations at each of 13 β values, producing 325 configurations. Of those, 195 were used to calculate average spectra and 130 were used for testing. Reflectance was calculated with three finite-difference time-domain simulations for each generated network, along the (100), (010) and (001) directions.
The optical model represented network bonds as cylinders with refractive index 1.5 and radius R = 0.39. Diamond-like and ctn-like networks had a volume fraction of about 0.4.
What changed the simulated spectrum
In the diamond-like simulations, reflectance began increasing in the resonant regime at a frequency of roughly ν ≳ 0.2 and peaked near ν ≈ 0.3. As β increased, the peak broadened, a pattern consistent with greater bond-length and ring disorder.
Lower mean coordination was associated with larger rings and pores and with a red-shifted reflectance peak. The researchers interpret rings and pores as local resonators whose dimensions help shape the interaction with light.
Across the simulated library, the average misfit between a target spectrum and a model spectrum detected both broadening associated with β and shifts associated with ring and pore-size differences. Within the tested set, the measure supported inference of network type and short-range order.
A close match that still leaves two answers
The biological comparison showed why a good spectral match is not the same as a unique structural diagnosis. At R = 0.39 and β = 0.001, the listed average misfit was 0.00029 ± 0.00005 for the diamond-like model and 0.00036 ± 0.00005 for the ctn-like model. Both reproduced the biological spectrum closely, while the other network types had minimum misfits more than twice as large.
The biological network had a critical pore radius δc = 0.31 and a mean ring radius rs = 0.71, measured in the model’s length unit d. Models with pore and ring scales closer to those biological values provided better spectral agreement, while primitive-cubic-like and gyroid-like models had much farther-apart scales.
The biological misfit did not show a distinct minimum. The analysis therefore could not settle on one network type or one value of β for the biological sample, even though it identified diamond-like and ctn-like models as the closest matches.
Local geometry takes the lead, for now
Some measures of broader structural order appeared to matter less in this dataset. At β = 0.001, the hyperuniformity metric α was 0.55 ± 0.21 for diamond-like networks and 0.12 ± 0.15 for ctn-like networks, compared with α = 0.95 for the biological network. The small misfit difference between the β = 0.001 hyperuniform and β = 0.2 nonhyperuniform comparisons suggested only a weak spectral effect from hyperuniformity, and primitive similarity was judged minor.
The analysis found no reported correlation between misfit and bond-length or bond-angle disorder. It also found that diamond-like models reproduced the biological spectrum more accurately than ctn-like models despite comparable reported coordination statistics.
Taken together, the authors propose an inverse-photonic-glass picture in which pores and rings behave as correlated local resonators in a disordered network. That analogy is interpretive and was not independently validated in fabricated materials.
The comparison depended on a specific optical model: cylindrical bonds, a refractive index of 1.5, a bond radius of 0.39 and selected volume fractions. The pore geometry was also treated through a simplified model rather than a full biological reconstruction.
The study’s central uncertainty remains the biological match itself. Because the biological misfit lacked a distinct minimum, the results do not uniquely identify the biological network type or its disorder parameter.
The manuscript identifies itself as an arXiv preprint. It reports support from the European Research Council, the Adolphe Merkle Foundation, the Swiss National Science Foundation, CNPq, CAPES, FAPERJ, Research Ireland and the European Regional Development Fund under the AMBER award.
Paper data and sources
Original title: From rings to resonance: an inverse method links biophotonic structural color to inverse photonic glasses
Authors: Florin Hemmann, Matthias Saba, Ullrich Steiner et al.
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-26
DOI: Not available
Original paper · Full text