Damselflies overcome color saturation barriers of photonic glasses via pigment loading and refractive index modulation

Significance

Photonic glasses are a promising class of noniridescent material but suffer from weak color saturation due to particle polydispersity. Here, we reveal two strategies employed by blue-tailed damselflies to generate strikingly saturated colors from a photonic glass structure, circumventing these limitations. First, pigment doping simultaneously suppresses off-resonant scattering and enhances the local refractive index near the reflectance peak. Second, crystallinity is delicately controlled to maintain an inverse correlation between nanosphere size and its refractive index. This means that the Mie resonance scattering peak becomes nearly size independent—avoiding polydispersity-induced broadening. These strategies highlight how evolution arrives at elegant solutions to optical constraints that yield unexpected principles for producing saturated photonic glasses.

Abstract

Biological strategies for manipulating light have revealed new concepts in light scattering, inspiring the design of sustainable photonic materials. While iridescent optical systems have been extensively studied, many applications require noniridescent structural colors which are much more difficult to achieve. Photonic glasses, comprising randomly arranged dielectric spheres, offer a promising solution toward such structural colors. However, their intrinsic disorder and particle size polydispersity typically lead to poor color saturation. Here, we identify two strategies employed by certain damselflies to generate unexpectedly vivid, tunable angle-independent colors from a photonic glass. First, doping of transparent pteridine nanospheres with yellow pigments strengthens blue–green reflectance resonances by simultaneously absorbing off-resonant wavelengths and enhancing the refractive-index near the reflectance band. Second, the refractive index of the nanospheres is modulated, via changes in crystallinity, to be almost exactly inversely correlated with nanosphere size. Thus, variations in nanosphere size, that ordinarily broaden reflectance resonances, resulting in poor color saturation, are compensated for by a correlated change in their refractive index. This ensures that even in a polydisperse ensemble, a consistent Mie scattering size parameter is maintained, strengthening short-range correlations and Mie scattering resonances. Finally, we show how damselflies tune these structural colors during maturation by precisely modulating the average size of the nanospheres, which arises naturally during the development of the pigment cells due to the densification and crystallization of the nanospheres. These findings reveal design strategies for overcoming limitations in the saturation of disordered photonic systems.

From iridescent butterflies (1–3) to ultrawhite beetles (4) and shrimp (5), biology’s optical systems have revealed new fundamental principles in scattering physics which inspire the development of sustainable optical materials. While biological iridescent materials, composed of ordered photonic crystals, have been widely explored (6–9), disordered photonic systems, which generate angle-independent structural colors, have received less attention. Angle-independent structural colors are widespread in biology (10, 11) and offer distinct advantages for applications in paints, displays, and coatings (12, 13). Photonic glasses, comprising random arrangements of dielectric spheres, offer a promising solution for generating such structural colors. Here, noniridescent color emerges from short-range order, resonant scattering, and possibly absorption (11, 14–16).

However, a major limitation of photonic glasses is that their disorder typically leads to low color saturation (17–20). A key obstacle in obtaining high color saturation in both photonic glasses and photonic crystals is the polydispersity of the scattering spheres (21, 22). In photonic glasses, variations in particle size degrade narrowband reflectance resonances by disrupting short-range scattering correlations between neighboring particles (23). The strength of such correlations is determined by the Mie scattering size parameter, which determines the peak wavelength of the Mie scattering resonance; x = πnd/λ (24)—the ratio of the optical path along the particle circumference (πnd) and the wavelength of incident light (λ) (d; diameter, n; refractive index). Maintaining a constant Mie size parameter across disparate particle sizes would be highly desirable for obtaining high color saturation. Yet, this is practically impossible to achieve through refractive index (n) dispersion alone since dispersion of transparent materials is typically very weak.

Although relatively few photonic glasses in animals have been described (25–30), a striking case is found in blue-tailed damselflies (Ischnura elegans) (31–35). During the maturation of juvenile males, the color of the thorax changes from vivid green to blue (33, 36). Prum et al. (32) showed that blue colors of many dragonflies and damselflies are produced from coherent scattering by arrays of nanospheres within epidermal chromatophore cells. The composition of these nanospheres has not been determined, but they are thought to be composed of pteridines (33, 35). Henze et al. (33) hypothesized that changes in the relative concentrations of xanthopterin and erythropterin pteridines may be responsible for tuning damselfly (Ischnura elegans) colors during development. Thus, while this coloration phenomenon is thought to be primarily structural in nature, the optical mechanisms and underlying materials enabling damselflies to produce and tune highly saturated, angle-independent structural colors remain a mystery.

Here, we show that blue-tailed damselflies precisely modulate the average size of pteridine nanospheres to tune the color of a photonic glass during maturation. Two surprising strategies are used to dramatically enhance color saturation. First, doping of transparent leucopterin nanospheres with xanthopterin, a yellow pigment, enhances blue–green reflectance by simultaneously absorbing off-resonance light and boosting the refractive index in the vicinity of the reflectance band (leading to high material dispersion). Second, we found that the refractive index of the nanospheres is inversely proportional to their size across a broad diameter range (280 to 400 nm)—a phenomenon we term “structural dispersion”. Thus, variations in particle diameter, normally deleterious to short-range scattering correlations, are compensated by inverse changes in refractive index—large particles with a lower refractive index have a similar Mie resonance peak to small particles with a higher refractive index. This ingenious compensation mechanism imparts the system with significant tolerance against reflectance broadening caused by polydispersity—maintaining high color saturation. We demonstrate that the correlation between refractive index and nanosphere diameter is controlled via modulation of nanosphere crystallinity. Less dense, less crystalline particles are larger and have a lower refractive index and vice versa. Remarkably, the modulation of crystallinity (and thus particle size and refractive index) is driven by the ontogenetic development of the chromatophore pigment cells—where the pteridine nanospheres undergo a process of densification, crystallization, and metabolic maturation.

Results

The Blue Shift in Damselfly Colors Is Correlated With a Decrease in Nanosphere Size.

Reflectance spectra from the thoraxes of over 100 male I. elegans damselflies (Fig. 1 A and B), exhibit a reflectance peak shift from 530 to 510 nm as they transform from green to blue during maturation (Fig. 1 C i-iv). The blue tail exhibits a reflectance peak at 470 nm (Fig. 1C, Tail) which remains unchanged throughout ontogeny. Conversion of these spectra into CIE color space (37) (Fig. 1D) shows that the damselfly colors are highly saturated (close to the boundary of the most saturated colors in the sRGB color space) and undergo a gradual blue shift during maturation (Fig. 1E). We note that the CIE color space quantitatively describes colors based on human perception and excludes the ultraviolet range which is relevant for insect and bird vision.

Fig. 1.

Highly saturated thorax colors in blue-tailed damselflies. (A) The male blue-tailed damselfly (Ischnura elegans). Photo credit: Ian Kirk. (B) Juvenile thoraxes during the development (i–iv). Right; tail region. (C) Reflectance spectra from the thoraxes and tail also exhibiting an intensifying shoulder peak at 430 nm. The black arrow indicates the color progression over ontogeny. (D) The spectra in (C) plotted on a CIE chromaticity diagram. w; white point of standard illuminant D65 corresponding to average daylight (saturation increases with distance from the white point). Dotted triangle; the sRGB color space, dashed rectangle; enlarged region of interest in (E). (E) Reflectance spectra collected from ~100 damselflies, converted to chromaticity coordinates and plotted on a CIE chromaticity diagram, exhibiting a gradual color change from green to blue.

Cryogenic scanning electron microscopy (cryo-SEM) shows that the distal epidermis in the damselfly thorax contains a 5-10 µm thick array of membrane-bound nanospheres. The nanospheres are randomly arranged in a photonic glass (14, 15, 38), lacking long-range positional ordering (see SI Appendix, Fig. S1 for FFT analysis). Underlying this, the proximal epidermis contains melanin/ommochrome pigment granules, which likely enhance contrast by absorbing stray light (33, 35) (Fig. 2A). The thorax ultrastructure remains the same throughout damselfly maturation, indicating that changes in nanosphere ordering or pigment cell arrangement are not responsible for the color change (SI Appendix, Fig. S2). Rather, as the damselflies transition from green to blue during maturation, the average size of nanospheres in the thorax decreases from 339 to 306 nm (Fig. 2 B and C). In contrast, in tails, where the color remains blue throughout maturation, the nanosphere size remains constant at ~282 nm. The polydispersity (Fig. 2B) is up to 10% for all samples. This correlation strongly indicates that the color change in the thorax is driven by a progressive decrease in nanosphere size during maturation.

Fig. 2.

Changes in nanosphere size during ontogeny. (A) Cryo-SEM micrograph of a damselfly tail showing a cross-section through the epidermis. Yellow arrowheads; direction of incident light. (B) Box chart showing the sizes of nanospheres extracted from individuals with different thorax colors as well as tails. The average diameter and the SD measured in each specimen (nm) is displayed on the chart. (C) Cryo-SEM images of pteridine nanospheres in the distal epidermis in thoraxes (i–iv) and a tail (v). The colors of the box chart in (B) and panel bars in (C) are the CIE representation of the reflectance measured from the specific tail/thorax analyzed.

Crystallization Underlies the Size and Refractive Index Modulation of Nanospheres.

To rationalize why and how the nanospheres decrease in size during ontogeny, we investigated their structural properties. At early developmental stages, when the thorax is green, the nanospheres display a smooth texture (Fig. 3 A, i and SI Appendix, Fig. S3; extended sample size), becoming granulated as development progresses (Fig. 3 A, ii). Later, radial spokes appear (white arrowheads, Fig. 3A), initially occupying the center of the nanosphere, with the edge containing an amorphous material (white brackets, Fig. 3A). Densification of the spokes occurs concomitantly with a contraction of the nanosphere membrane (white arrows, Fig. 3A) and a decrease in its diameter (Fig. 3 A, ii–iv). Finally, a further decrease in size occurs as the spoke-like texture transforms to a concentric lamellae (onion-like) texture (Fig. 3 A, v)—resembling crystalline isoxanthopterin nanospheres in crustacean visual systems (25, 39, 40).

Fig. 3.

Changes in the structural and optical properties of nanospheres during ontogeny. (A) Cryo-SEM micrographs nanosphere cross-sections exhibiting different stages of formation during ontogeny. Arrowheads; radial spokes, brackets; layer of amorphous material, arrows; nanosphere limiting membrane. (B) Electron diffraction from nanospheres extracted from (Top) a damselfly early in development and (Bottom) later in development. (Scale bar, 200 nm.) (C) TEM micrograph and corresponding selected area electron diffraction performed on different locations along the nanosphere edge exhibiting diffraction arcs (arrowheads). (D) Scatter plot showing the aggregate measured tangential refractive index (n_t) of nanospheres with varying diameters from all developmental stages. Color panels (A and B) and colored data points (D); CIE representation of the reflectance measured from the specific tail/thorax analyzed.

These textural changes indicate that the nanospheres undergo a process of formation during damselfly maturation—likely related to an ongoing crystallization process. Electron diffraction confirms this interpretation. In early juveniles, the nanospheres exhibit an extremely weak, single diffraction ring with a d spacing of ~3.2 Å—associated with stacking of pteridine molecules (Fig. 3B) (5, 41). As the damselflies mature, the diffraction intensity of the nanospheres increases (Fig. 3B and SI Appendix, Fig. S4). In agreement with textural observations (Fig. 3A), this indicates that the crystallinity and density of the nanospheres increase during maturation, causing a reduction in their volume.

Selected area electron diffraction along the nanosphere edge exhibits diffraction arcs (Fig. 3C), demonstrating the nanospheres are composed of 1-dimensionally ordered, radially oriented, stacked molecular assemblies. This spherulitic arrangement (commensurate with the “spoke” like texture observed in the cryo-SEM images, Fig. 3A) closely resembles isoxanthopterin and uric acid spherulites in cleaner shrimp (5) and fish (41). Previous studies (5, 39, 40, 42, 43) showed that isoxanthopterin spherulites exhibit extreme birefringence, with a high tangential refractive index (n_t = 1.96) and low radial refractive index (n_r = 1.4). Birefringence results from the preferred orientation of the planar isoxanthopterin molecules, which have a highly anisotropic molecular polarizability (44). The magnitude of the birefringence is dictated by the degree of molecular alignment and thus, crystallinity. In damselflies, poorly crystalline nanospheres (in early juveniles) are expected to be weakly birefringent and thus have a lower tangential refractive index. As development progresses and crystallinity increases, both the birefringence and the tangential refractive index of the nanospheres (which determines the Mie resonance) are also expected to increase.

To test this premise, we measured the refractive index of individual nanospheres extracted from differently colored damselflies using a previously described scattering method (42). Intriguingly, we observed an almost perfect inverse correlation between the tangential refractive index (n_t) of nanospheres and their diameters (Fig. 3D and SI Appendix, Fig. S5). As the nanosphere size decreases from ~400 nm to just over 280 nm, their tangential refractive index at the resonant scattering wavelength increases from 1.63 (corresponding to an isotropic sphere) to well above 2 (corresponding to a nearly fully crystalline, birefringent sphere). This behavior is dominated by the increasing alignment of pteridine molecules in the nanospheres as they crystallize during chromatophore development and has some contribution from material dispersion (as the resonance blue shifts for smaller spheres). Thus, the modulation of nanosphere refractive index appears to be driven by the change in their crystallinity. We term this type of refractive index dispersion, “structural dispersion.”

The Relative Abundance of Different Pteridine Pigments Does Not Determine Thorax Color.

To investigate whether there is a correlation between thorax color and the pteridine content of the nanospheres, we quantified pteridine concentrations from the distal epidermis of 45 damselflies using liquid chromatography–mass spectrometry (LC–MS). K-means clustering was used to group the samples based on the concentrations of the 7 most abundant pteridines, yielding two clusters, representing groups of individuals with a similar pattern of pteridine abundance (Fig. 4A) (Materials and Methods and SI Appendix, Supplemental Methods). The reflectance spectra from the same thoraxes are plotted on a CIE diagram, with the red or blue data points denoting their cluster classification (Fig. 4B). Cluster 1 predominantly incorporates green samples, and cluster 2, the rest of the samples. The most abundant pteridines in both clusters are leucopterin (colorless), xanthopterin (yellow pigment), and isoxanthopterin (colorless) (Fig. 4C). The relative amount of yellow pigment (xanthopterin), compared to the colorless pteridines, decreases only slightly from 29 ± 3% in green thoraxes to 23 ± 5% in blue thoraxes, concomitant with a relative increase in leucopterin from 48 ± 7% to 61 ± 7% (Fig. 4D). However, the tails from the same individuals (which remain blue throughout their lifetime) exhibit the same trend in the relative abundances as the thoraxes (Fig. 4 D and E). This supports our finding that particle size, rather than pigment variation controls color change, ruling out the earlier hypothesis of Henze et al. (33). This also raises the question of what function, if any, xanthopterin plays in damselfly coloration, particularly given that other pteridine-based biological photonic glasses are composed of transparent pteridines and do not contain pigments (5, 25, 39, 40).

Fig. 4.

Pteridine composition of the thorax and tails. (A) K-means clustering of pteridine concentration data from damselflies of different colors. Each cluster represents a group of individuals with a similar pattern of pteridine concentrations. (B) CIE chromaticity diagram denoting reflectance spectra from the thoraxes of each damselfly in the analysis. Red and blue data points; classification to clusters 1 and 2 from (A). w; white point of standard illuminant D65 corresponding to average daylight. (C) Average abundance (pg mL⁻¹ mg⁻¹ of tissue) of pteridines in damselfly thoraxes of each of the clusters in (A). (L; leucopterin, X; xanthopterin, I; isoxanthopterin, 7,8; 7,8-dihydroxanthopterin, P-6; pterin-6-carboxylic acid, P; pterin, B; biopterin). IQR; interquartile range. (D) Relative abundance of pteridines in thoraxes and tails of each cluster (mean ± SD %). (E) Average abundance (pg mL⁻¹ mg⁻¹ of tissue) of pteridines in damselfly tails of each of the clusters in (A).

The relative increase in leucopterin and decrease in xanthopterin and 7,8-dihydroxanthopterin (Fig. 4D) in both the thoraxes and tails suggests that the pigment cells undergo a metabolic maturation process during ontogeny, where precursors, xanthopterin, and 7,8-dihydroxanthopterin are converted enzymatically (via xanthine dehydrogenase) into the less soluble leucopterin—an end product of the pteridine biosynthetic pathway in insects (45–47).

Increase in the Absolute Abundance of Pteridines Provides a Driving Force for Nanosphere Densification.

Important observations are made when considering the absolute abundance (per mass of tissue) of pteridines in the samples (Fig. 4C). When the thoraxes transition from green (cluster 1) to blue (cluster 2), the absolute abundance of leucopterin, xanthopterin, and isoxanthopterin increases by 94%, 33%, and 21%, respectively (the mass of tissue analyzed in the specimens of the two groups does not differ significantly). The significant increase (P = 0.00014, see SI Appendix, Supplemental Methods for statistical analysis) in overall pteridine abundance in the thorax during maturation indicates the constituent pteridine organelles are in a state of high metabolic activity associated with their formation and provides an underlying metabolic cause for the densification and crystallization of the nanospheres during maturation. In contrast, in the tails, the overall pteridine abundance (per mass of tissue) does not change significantly during maturation (P = 0.61, see SI Appendix, Supplemental Methods for statistical analysis). This suggests the nanospheres in the tail may have already attained a more metabolically mature and thus dense, crystalline state.

Taking the morphological, structural (Figs. 2 and 3), and compositional (Fig. 4) observations together, the ontogenetic color change in damselflies is associated with a reduction in nanosphere size which appears to be caused by densification of pteridines during crystallization. This crystallization process is correlated to an increase in the absolute pteridine content of the nanospheres and in their tangential refractive index (n_t). This interpretation is consistent with the observation of Hinnekint et al. (36) who reported that damselfly color changes during maturation are correlated with the dehydration of the tissue—another possible driver of densification/crystallization.

Pigment Loading Enhances Color Saturation.

Photonic glasses typically produce low color saturation due to multiple scattering effects (17–20). To rationalize how damselflies produce and tune highly saturated colors, we used molecular dynamics (MD) simulations to generate random particle assemblies and the finite difference time domain (FDTD) method to simulate their reflectance spectra (Materials and Methods). Simulations were performed using experimentally measured nanosphere sizes, with a 55% filling fraction in a water medium (Fig. 5). The refractive indices of the particles were made inversely dependent on size based on our experimental measurements (Fig. 3D and SI Appendix, Table S1). The simulations represent a simplified model of the biological photonic glass that enables us to vary and thus test the key structural and chemical parameters that determine the experimental spectral profile—i.e., the positions and widths of the reflectance peaks which contribute to color hue and saturation. We note that this is by no means an effort to accurately reconstruct measured spectra as the model is still quite simplistic.

Fig. 5.

Simulated reflectance of nanosphere assemblies. (A and B) FDTD simulated reflectance spectra of MD generated 5 × 5 × 5-μm³ photonic glass slabs comprising nanospheres with 5% SD, 55% filling fraction, varying diameters, size-dependent refractive index and wavelength-dependent complex refractive index (A; leucopterin nanospheres, B; 30% xanthopterin, 70% leucopterin nanospheres). (C) The reflectance spectra in (A and B) plotted on a CIE chromaticity diagram exhibiting the increasing saturation of the reflected colors. (D and E) FDTD simulated reflectance spectra of MD generated 5 × 5 × 5-μm³ photonic glass slabs comprising nanospheres of 300 nm average diameter, 55% filling fraction, varying polydispersity and (D) a uniform refractive index or (E) size-dependent refractive index. (F) The reflectance spectra in (D and E) plotted on a CIE chromaticity diagram. Black arrows in (D and F) indicate the direction of increasing polydispersity and decreasing saturation. Saturation values of all spectra are specified in the legends. w; white point of standard illuminant D65 corresponding to average daylight. (G) Snapshots of 1.5 × 1.5 × 1.5-μm³ subregions of the MD-simulated particle configurations in (E) with color indicating the n_∞ value used to calculate refractive index.

Our initial simulations explored the reflectance properties of a photonic glass composed of leucopterin nanospheres (the most abundant molecule in the nanospheres). To account for leucopterin’s absorption in the UV-A (315 to 400 nm) and refractive index dispersion, we incorporated the wavelength-dependent real (n) and imaginary (κ, extinction coefficient) components of the complex refractive index (n+iκ) into the simulation (Fig. 5A). The extinction coefficient (κ) was calculated from a scaled absorbance measurement (Eq. 1 and SI Appendix, Fig. S6 A and B). The real part of the refractive index (n) was computed via the Kramers–Kronig relation (48, 49) (Materials and Methods, Eq. 2 and SI Appendix, Fig. S7A). The resulting simulations reveal a structural peak in the visible which blue shifts from ~560 to ~520 nm as the nanosphere size decreases (Fig. 5A)—closely matching the experimentally observed trend. The slight red shift (~20 nm) compared to the experimental spectra likely results from a small overestimation of the nanospheres’ refractive index (elaborated on in the Materials and Methods and SI Appendix, Fig. S5). These results confirm that the green-to-blue ontogenetic color shift is driven by a decrease in nanosphere size. However, these simulated spectra exhibit lower saturation (s = 0.27 to 0.37) compared to the experimental spectra, likely due to the shoulder peak at 460 nm.

We then tested how xanthopterin (which accounts for 23 to 30% of the total pteridine content of the nanospheres) influences color saturation compared to leucopterin-only nanospheres. Xanthopterin is a yellow pigment with strong visible absorbance (λ_max = 388 nm) and significant refractive index dispersion (SI Appendix, Fig. S6 A and B). We simulated assemblies composed of 30% xanthopterin and 70% leucopterin (vol. fraction) (Fig. 5B). The complex refractive index of pure xanthopterin was calculated as described for leucopterin (SI Appendix, Fig. S7B), and that of the mixture was calculated using the Lorentz-Lorenz mixing rule (Eqs. 3–8 and SI Appendix, Fig. S8 A and B) (50). The reflectance spectra exhibit the same peak positions as the leucopterin-only assemblies, but with a significantly sharper reflectance peak and thus, a marked increase in color saturation (Fig. 5C) (s = 0.42 to 0.50)—similar to the experimental saturation values.

These results demonstrate that xanthopterin doping functions to enhance color saturation. This phenomenon arises from the interdependent effects of pigment absorption (dampening of the shoulder peak at ~445 nm) and greater refractive index dispersion which enhances structural peaks in the green (51). We note that the decrease in color saturation during damselfly maturation in the experimental spectra was not accounted for in our simulations. This may be caused by a small decrease in the relative xanthopterin content of the nanospheres or because the contribution of multiple scattering is more pronounced at shorter wavelengths (23).

Nanosphere Size-Refractive Index Dependence Mitigates Against Reflectance Broadening.

To investigate how the crystallization induced refractive index-size dependency (Fig. 3D) influences coloration, we compared simulations of xanthopterin-loaded assemblies (300 nm average diameter) with varying polydispersity (0 to 20%) where the particles i) all have the same complex refractive index (Fig. 5D) and ii) where refractive index depends on size (Fig. 5E). For i), the reflectance peak broadens as the polydispersity increases—an expected outcome of increased disorder, reducing saturation from 0.48 to 0.37 (Fig. 5D). In contrast, little reflectance broadening is seen when refractive index is (inversely) correlated with size, and the saturation remains high (s = 0.45) even at 20% polydispersity (exceeding the observed polydispersity Fig. 5 F and G). We note that in photonic glasses and photonic crystals, polydispersity values >10% are typically associated with broadband, whitish reflectance (5, 21, 22). The same trend was observed for all nanosphere sizes between 280 to 340 nm (SI Appendix, Fig. S9). Size-dependent refractive index thus mitigates the dampening effect of polydispersity and allows the system to tolerate broader distributions of particle sizes while maintaining highly saturated colors.

Although the reflectance peak arises from collective interference from the assembly (structure factor), its spectral sharpness crucially depends on single-particle scattering (form factor) (18, 23, 52). The robustness of the reflectance peak to polydispersity can thus be interpreted using Mie scattering: The spectral position of a given Mie resonance is governed by the figure of merit x = πnd/λ (SI Appendix, Fig. S10), such that polydispersity broadens the reflectance spectrum of the assembly because particles with different diameters resonate at different wavelengths (24). Here, the inverse relationship n(d)∝1/d observed in the experiment (Fig. 3D) approximately preserves the optical path length nd (and therefore x) across differently sized particles, keeping the dominant scattering resonance aligned in wavelength and suppressing disorder-induced broadening (SI Appendix, Fig S11).

Discussion

We show that the vivid colors of damselflies are tuned via precise modulation of the size of pteridine nanospheres in a photonic glass. The application of photonic glasses as alternatives to conventional absorbing dyes and iridescent materials is limited by their low color saturation (17–20). Several strategies for enhancing saturation have been proposed, including the use of core-shell particles (20, 53), control of layer thickness and packing density (54, 55), dual-size particle systems (54, 56), and the incorporation of narrowband (57) or broadband (17, 51, 58–61) absorbers. Here, we report two “ingenious” strategies used by damselflies to enhance color saturation. First, doping of transparent nanospheres with xanthopterin, a yellow pigment, dramatically enhances saturation by the absorption of blue-wavelengths coupled with refractive index (material) dispersion near the absorption band due to Kramers–Kronig relations (51, 62). Sai et al. (57) predicted theoretically that loading polystyrene particles with beta carotene can expand the color gamut and saturation of disordered photonic assemblies. However, to the best of our knowledge, this is the only known “real” system employing narrowband pigment loading within colloidal particles. The scarcity of synthetic examples utilizing this powerful strategy is likely due to the challenges of synthesizing heterogeneous composite particles. Second, the correlation between refractive index and particle size (structural dispersion), which occurs through tuning of the crystallinity of the particles, introduces a tolerance against reflectance broadening—maintaining high color saturation even in very disordered, polydisperse systems. To the best of our knowledge, this structurally programmed compensation mechanism has not been previously observed or created in natural and or synthetic photonic glasses, nor suggested theoretically as a design principle.

Remarkably, the tuning of nanosphere crystallinity, and the ensuing particle size and particle size-refractive index control, arise spontaneously from nanosphere formation during ontogeny. As the thorax transforms from green to blue, densification of the nanospheres causes them to shrink, and their increasing crystallinity results in an increase in refractive index. The increase in absolute pteridine concentration, caused by ongoing pteridine biosynthesis (45–47) in the nanospheres, appears to drive densification and crystallization. Pigment loading is also an inevitable consequence of pteridine biosynthesis and the ongoing metabolic activity during pigment cell formation. Xanthopterin is a biosynthetic precursor to leucopterin (45–47) and will thus be naturally present in the nanospheres during leucopterin biosynthesis. Interestingly, Zhang et al. (63) reported an analogous phenomena in lizards, where the optical properties of the system are tuned during ontogeny by a gradual maturation of an existing pigment cell rather than its replacement with a new pigment cell type.

Body coloration in blue-tailed damselflies is known to function as an external indicator of sex, maturation stage, metabolic condition, and hormonal state (36). Behavioral studies demonstrate that males identify and select mates based on body coloration (64), indicating that optimization of color saturation may perform a functional role in intraspecific signaling. While this study is on male blue-tailed damselflies, the same optical mechanism is likely responsible for violet, blue, and green coloration in female morphs of blue-tailed damselflies and other species of the Odonata order (31–33). More broadly, biological photonic glass structures, whether pigment-loaded or purely dielectric, are also probably more widespread than currently known. They may occur in other insect orders, where pteridines are abundant and function as pigments and excretory substances (45), but likely extend across a wide range of taxa, morphologies, and material compositions.

While the formation of guanine, the most abundant biogenic organic crystal, has been described in detail (65–71), this study presents insights into the formation of optically functional biogenic pteridine crystals (25, 39, 40). The gradual onset of molecular ordering in the pteridine nanospheres is consistent with a nonclassical crystallization mechanism (72). This increase in ordering correlates with an increase in pteridine biosynthesis and metabolic activity in the organelles during ontogeny. The absence of fully crystalline nanospheres in damselflies [as opposed to crystalline counterparts in crustacean’s visual systems (25, 39, 40)] is likely because the large number of different pteridines in the organelles may hinder crystallization. Understanding how biology forms birefringent spherulites may inspire new strategies for improving the development of synthetic mimics to these materials—as alternatives to inorganic particles in photonic assemblies (73).

Concluding Remarks

We report two biological strategies for enhancing color saturation in photonic glasses that have not been conceived or practically manifested in synthetic optical devices. The degree of tolerance of color saturation toward variations in particle size is a strong demonstration of how biology’s complex and sophisticated solutions in optics offer understanding in scattering physics that can inspire the development of efficient and sustainable organic replacements to conventional synthetic pigments. This quest is becoming increasingly important as concerns grow over the toxicity and environmental impact of conventional pigments in foods, cosmetics, pharmaceuticals, and textiles (74–77). Several questions remain about how photonic structures are biologically controlled in organisms, including the degree of genetic regulation and the mechanisms of precursor synthesis and transport to the transforming tissues.

Materials and Methods

Specimen Collection.

Damselflies were collected during June-August 2024 in two locations in Israel (River Park, Beersheva, and Ekron River, Mazkeret Batya). Use of insects for research purposes does not require animal ethics approval in Israel, but animals were handled with established best practices.

Optical Microscopy.

A Zeiss Discovery.V20 stereomicroscope equipped with an Axiocam 305 color camera was used to take images of the damselflies.

Microspectrophotometry.

The reflectance was measured from the thorax area of each male damselfly on the day of collection or the following day while the damselflies were alive. Reflectance was collected from 3 to 5 different locations on the thorax of each damselfly, and the measurements were averaged. To correlate between structural/chemical properties and damselfly color, each damselfly was labeled and stored for further analyses (electron microscopy or LCMS). Reflectance spectra were obtained using a Zeiss AX10 microscope equipped with a Zeiss X10/0.25 NA HD DIC objective. The specimens were epi-illuminated with an LEJ HXP 120-V compact light source. The reflected light was collected from the same objective into an optical fiber (QR450-7-XSR, Ocean Insight) and the spectra measured using an Ocean Insight FLAME miniature spectrometer. The spectra were normalized using a white diffuse reflectance standard (Labsphere USRS-99-010, AS-01158-060). Background spectra were measured when no light was applied. Reflectance spectra were obtained over the range of 180 to 880 nm. The integration time ranged from 10 to 40 s, with 1 to 10 averaged scans for each measurement.

CIE Chromaticity Diagrams.

International Commission on Illumination (CIE) diagrams were created using Origin(Pro) (version 2023b, OriginLab Corporation, Northampton, MA). The CIE 1931 color space and D65 illuminant (represents average daylight illumination) were used.

Cryogenic-SEM Sample Preparation and Imaging.

On the day of collection or the following day, the area of interest (first section of the abdomen or tail) was dissected and chemically fixed with 4% paraformaldehyde (PFA, CAS 50-00-0, Thermo Scientific Chemicals) and 2% glutaraldehyde (GA, CAS 111-30-8, Sigma-Aldrich) in 1xPBS for 3 to 4 h. The chemically fixed tissues were embedded in 7% agarose gel and sectioned to 180-microns thick slices using (LEICA VT1000 S Vibratome). The sections were sandwiched between two aluminum discs in a 15% dextran (CAS 9004-54-0, Sigma-Aldrich) in 1X PBS solution and cryo-immobilized in a high-pressure freezing device (EM ICE, Leica). The frozen samples were then mounted on a holder under liquid nitrogen in a specialized loading station (EM VCM, Leica) and transferred under cryogenic conditions (EM VCT500, Leica) to a sample preparation freeze-fracture device (EM ACE900, Leica). Samples are fractured at −120 °C by hitting the top disc carrier with a tungsten knife at a speed of 135 to 140 mm/s, this exposed a clean fracture plane which can be imaged. The samples were etched at −110 °C for 3:00 to 3:40 min to evaporate excess water and coated with 3 nm of Pt/C. The samples were imaged in an HRSEM Gemini 300 scanning electron microscope (Zeiss) by a secondary electron in-lens detector while maintaining an operating temperature of −120 °C.

Filling fraction (ff) was estimated from cryo-SEM images by measuring the 2D area of nanospheres in cells and dividing by the total area of interest. Assuming the sample is isotropic, and particles are roughly spherical, then 2D ff ≈ 3D ff.

Nanosphere Extraction for TEM Imaging.

The nanospheres were extracted from the thorax area by washing the specimen with ultrapure water and dissecting the colored areas into a vial with hexane or ethanol. The vials were sonicated briefly to release the nanospheres. The resulting suspension was drop-casted on a carbon-coated Cu-meshed TEM grid and allowed to dry.

TEM Imaging (Measuring Nanosphere Diameters) and Electron Diffraction.

Extracted nanospheres were imaged with a ThermoFisher Scientific (FEI) Tecnai T12 G2 TWIN TEM operating at 120 kV. Images and electron-diffraction patterns were recorded using a Gatan 794 MultiScan CCD camera. Electron-diffraction analysis was done using Gatan DigitalMicrograph software with the Diffpack module. Nanosphere diameters of samples with different colored thoraxes (reflectance from the thorax was measured as described above) were measured from TEM images of extracted nanospheres using Gatan DigitalMicrograph software. Each sample was extracted from an individual damselfly, 70 to 100 nanospheres were measured.

Refractive Index Measurements By Dark Field Scattering Experiments.

Scattering measurements of single nanospheres were performed using a commercial inverted optical microscope (Zeiss, Axio Observer 5) conjugated with a custom-built optical setup. The sample was illuminated with a “white” light-emitting diode in a dark-field configuration and collected using a low NA objective to reduce unscattered light collection (N-Achroplan, ×20, NA^.0.45). The light was then directed outside the microscope into the additional setup that further magnified the image (×6) and coupled into a fiber (105 to 150 μm in diameter, depending on nanospheres size) connected to a spectrometer (Shamrock 303i, Andor). The nanospheres were illuminated using an annulus (Θi = 25°). Remaining unscattered light was filtered out using an iris placed at the Fourier plane (imaged by a Blackfly S USB3, FLIR). The experimentally obtained spectrum was normalized according to the background signal and source spectrum and was then compared with refractive index–dependent calculations of the scattering cross-section to extract the refractive index. The calculations were done using MATLAB according to the dimension or each nanosphere as extracted from TEM images of the specific measured nanosphere using ImageJ. Blurred edges of certain nanospheres due to remains of organic residues around them may have resulted in an underestimation of their diameters, leading to an overestimation of their refractive index as discussed in the main text. Detailed information about the optical setup and calculations of the scattering cross-section is available in ref. 42.

Pteridine Pigment Extractions.

Damselfly specimens were killed humanely by plunge freezing in liquid nitrogen on the day of collection or the following day after measuring their reflectance. The specimens were stored at −80 °C until the extraction. All solvents used for extractions were of LC–MS grade. The sample set included 45 damselflies (45 thoraxes and 35 tails, some tails were used for other analysis or lost) which were randomized and extracted in batches on six separate days. The extraction method was developed for this experiment, partially based on refs. 78, 79. The thorax and tail were dissected and weighed separately using a Radwag XA 6.4Y.M PLUS Microbalance. Each specimen was transferred to a labeled 2 mL Eppendorf vial and 0.5 mL 1%NH₄OH was added. Small scissors were used to lyse the tissue. Each sample was spiked with 5 μL of the 100 μg/mL internal standard biopterin-d3 (CAS 1217838-71-5, Santa Cruz Biotechnology). The samples were sonicated for 5 min. 0.5 mL methanol was added. The samples were incubated at −20 °C for 60 min. Next, the samples were sonicated for 2 min and filtered through a 0.45-micron syringe nylon filter (424-FNY422013, Jet Biofil) into clean 2 mL Eppendorf vials. 100 μL 1% formic acid and 0.9 mL of methanol were added. The samples were fully dried in a SpeedVac and the pellets were stored at −80 °C until the LC–MS analysis. The samples were resuspended on the day of LC–MS analysis in 50 μL 1% NH₄OH and 60 μL acetonitrile (ACN), sonicated for 5 min and vortexed briefly. Then 10 μL of 1% formic acid was added to neutralize the pH. The samples were centrifuged and decanted twice, then transferred into LC vials for analysis.

Liquid Chromatography–Mass Spectrometry (LC–MS).

The commercially available analytical standards of the following pteridines were used to build a calibration curve for the analysis: isoxanthopterin (CAS 529-69-1, Santa Cruz Biotechnology), xanthopterin (CAS 5979-01-1, Santa Cruz Biotechnology), pterin (CAS 2236-60-4, Sigma-Aldrich), leucopterin (CAS 492-11-5, Biosynth), 7,8-dihydroxanthopterin (CAS 1131-35-7, BOC Sciences), biopterin (CAS 22150-76-1, AA Blocks), 7,8-dihydrobiopterin (CAS 6779-87-9, Aaron Chemical), pterin-6-carboxylic acid (CAS 948-60-7, Sigma-Aldrich), and sepiapterin (CAS 17094-01-8, A2B Chem). Biopterin-d3 (CAS 1217838-71-5, Santa Cruz Biotechnology) was used as internal standard. See SI Appendix, Supplemental Methods for further information on the creation of the calibration curve.

Pteridines were analyzed using Waters ACQUITY UPLC I-Class Plus System coupled with Thermo Exploris 240 high-resolution mass-spectrometer. A separation of the selected pteridines was achieved on Waters BEH Amide column (1.7 µm, 2.1 × 100 mm) using the following gradient of the mobile phase A (10 mM ammonium formate in 10% acetonitrile, pH 4.0) and mobile phase B (10 mM ammonium formate, pH 4.0 in 90% acetonitrile): 0 to 2 min 5% A, 2 to 15 min 50% A, 15 to 15.5 min 80% A, 15.5 to 17 min 80% A, 17 to 17.5 min 5% A, 17.5 to 20 min 5% A. The injection volume was 5 µL, the column temperature was kept at 30 °C, and the flow rate was 0.25 mL/min. The total run time was 20 min.

High-resolution mass spectra were acquired using electron spray ionization (ESI) in positive full scan mode (70 to 800 m/z) with a resolution 24000 full width at half-maximum (FWHM). The MS parameters (ion spray voltage, sheath gas, aux gas, sweep gas, ion transfer tube temperature, and vaporizer temperature) were set up according to the manufacturer recommendations. Data were acquired under the control of Xcalibur software, v3.5 (Thermo Scientific). Xcalibur software was also used for the data analysis to extract targeted ions (EIC) (80).

The concentration of each compound was calculated using the calibration curve built using Quan browser for Xcalibur. The obtained results were normalized (divided by) to the sample weight, log₁₀ transformed and subjected to K-means cluster analysis (Fig. 4A) (see SI Appendix, Supplemental Methods for further information on K-means clustering). The optimal number of clusters (two) was found using the Partitioning Around Medoids algorithm using R software (81) (SI Appendix, Fig. S12). K-means cluster analysis (82) was performed using Origin(Pro) (version 2023b, OriginLab Corporation, Northampton, MA, USA.). The initial cluster centers were not specified, and the maximum number of iterations was 10.

Optical Constant Calculation using Kramers–Kronig Relation.

The real (n) and imaginary (κ, extinction coefficient) parts of the complex refractive index (n+iκ) of leucopterin and xanthopterin were calculated from their absorbance spectra using the Kramers–Kronig relations. Absorbance of commercial leucopterin and xanthopterin was measured using a UV-visible spectrophotometer (Thermo Fisher Scientific Evolution 220). Leucopterin (CAS 492-11-5, Biosynth) and xanthopterin (CAS 5979-01-1, Santa Cruz Biotechnology) powders were dissolved in 1% NH₄OH and filtered with a 0.22 μm mesh filter. The absorbance was measured between 190 to 1000 nm. All calculations were performed using a custom Python script (Python 3.11, NumPy, SciPy, Matplotlib, and Pandas libraries). The script was developed with the aid of Gemini (Google) and ChatGPT (OpenAI).

To estimate the imaginary part of the refractive index (κ), the absorbance spectrum was normalized such that the maximum absorption coefficient (α) in the range of 320 to 700 nm matched a target value of α_max = 2.5 × 10⁶ m⁻¹ (2.5 μm⁻¹). α_max was estimated based on measurements from ref. 62. The absorption coefficient α(λ) was then converted to κ(λ) using the relation:

$$\kappa \left(\lambda \right)=\frac{\alpha \left(\lambda \right)\ast \lambda }{4\pi },$$ [1]

where λ is the wavelength in meters.

The real part of the refractive index (n) was computed via the Kramers–Kronig relation (48):

$$n\left(\lambda \right)=n_{\infty }+\frac{2}{\pi }\mathcal{P}{\int }_0^{\infty }\frac{{\lambda }^{\prime }\kappa \left({\lambda }^{\prime }\right)}{{\lambda }^{\prime 2}-{\lambda }^2}d{\lambda }^{\prime },$$ [2]

where n_∞ is the refractive index at infinite wavelength (a constant representing the electronic contributions at very high frequencies, or essentially the refractive index far from any absorption bands), $\mathcal{P}$ denotes the Cauchy principal value of the integral. The integral was approximated numerically using the trapezoidal rule. n_∞ values were estimated based on the experimental measurements of the tangential refractive index of the nanospheres, n_t. To account for the refractive index varying with nanosphere diameter, a linear fit of the experimental measurements of n_t or a reciprocal function (Fig. 3D) were used to estimate a range of n_∞ values corresponding to increasing nanosphere diameters (SI Appendix, Table S1). The real part of the refractive index, n, was computed for each n_∞ value.

Calculation of the Complex Refractive Index for a Two-Component Mixture.

To compute the complex refractive index spectra of a two-component mixture, we applied the Lorentz–Lorenz mixing rule (50) to the complex dielectric permittivity of each component. The real (n) and imaginary (κ) components of the refractive index were obtained for each pure compound over 190-1000 nm wavelength range as previously described. The same value of n_∞ was used for the two components in the mixture. The real (n) and imaginary (κ) components of the refractive index were computed for each n_∞ value corresponding to varying nanosphere diameters.

The complex refractive index N = n+iκ was squared to obtain the complex dielectric function ϵ = N². For each pure component (denoted as 1 and 2), the Lorentz–Lorenz transformation of the dielectric function was computed as:

$$f\left(ϵ\right)=\frac{ϵ-1}{ϵ+2}.$$ [3]

The effective Lorentz–Lorenz factor of the mixture was calculated via a linear volume-fraction-weighted average:

$$f\left({ϵ}_{\mathit{mix}}\right)={\varphi }_{1}f\left({ϵ}_1\right)+{\varphi }_{2}f\left({ϵ}_2\right),$$ [4]

where ϕ₁ and ϕ₂ = 1−ϕ₁ are the volume fractions of the two components.

This expression was inverted to solve for the complex dielectric function of the mixture:

$${ϵ}_{\mathit{mix}}=\frac{\left(1+2·f_{\mathit{mix}}\right)}{\left(1-f_{\mathit{mix}}\right)}.$$ [5]

Finally, the complex refractive index of the mixture was obtained from the square root of the dielectric function:

$$|{ϵ}_{\mathit{mix}}|=\sqrt{\left(Re{\left({ϵ}_{\mathit{mix}}\right)}^2+Im{\left({ϵ}_{\mathit{mix}}\right)}^2\right)},$$ [6]

$$n_{\mathit{mix}}=\sqrt{\frac{\left(\left|{ϵ}_{\mathit{mix}}\right|+Re\left({ϵ}_{\mathit{mix}}\right)\right)}{2}},$$ [7]

$${\kappa }_{\mathit{mix}}=\sqrt{\frac{\left(\left|{ϵ}_{\mathit{mix}}\right|-Re\left({ϵ}_{\mathit{mix}}\right)\right)}{2}}.$$ [8]

To ensure physical validity, the terms inside the square roots are clamped to be non-negative.

Saturation of Color.

Saturation was calculated as in ref. 83:

$$S_{\mathit{ab}}=\frac{\sqrt{a^{\ast 2}+b^{\ast 2}}}{L^{\ast }},$$ [9]

where L* represents the lightness, and a*, b* are the chromaticity coordinated in CIELAB space. All reflectance spectra were normalized (divided by the maximum value) before the coordinates were calculated so the simulations and experimental data are comparable.

Simulated Reflectance Spectra.

MD simulations were carried out in similar manner to ref. 5 using HOOMD-blue (version 2.9.4) (84), with the Langevin integrator (kT = 0.05) and repulsive polydisperse 12 to 0 (LJ 12-0) pair-potential:

$$\varphi \left(r\right)=\left\{\begin{array}{c}{v_0}_{\left[{\left(\frac{{\varphi }_{\mathit{ij}}}{r}\right)}^{12}-{\left(\frac{{\varphi }_{\mathit{ij}}}{r}\right)}^0\right]+{\sum }_0^2{C}_k{\left(\frac{r}{{\varphi }_{\mathit{ij}}}\right)}^{2k},r/{\sigma }_{\mathit{ij}}\le r_c}\\ 0, otherwise\end{array}\right.,$$ [10]

$${\varphi }_{\mathit{ij}}=\frac{1}{2}\left({\varphi }_i+{\varphi }_j\right)\left(1-ϵ\left|{\varphi }_i-{\varphi }_j\right|\right),$$

where v₀ = 1, ϵ_ij = 0.01, σ_a is the diameter of particle a, and r_c = 1.5μ₁ is the cutoff distance, using the PolydisperseMD (85) plugin. Particle sizes were randomized from a log-normal distribution, with mean ${\mu }_2=\text{log}\left(\frac{{\mu }_1^2}{\sqrt{{\mu }_1^2-{\sigma }_1^2}}\right)$ and ${\sigma }_2^2=\text{log}\left(1+\frac{{\sigma }_1^2}{{\mu }_1^2}\right)$, and the values outside range [100 nm, 600 nm] were rounded to the nearest boundary to avoid the chance of extreme particle sizes. In the simulations, mean particle sizes μ₁ = (280 nm, 300 nm, 320 nm, 340 nm) and SD of σ₁ = (0, 5%, 10%, 20%) were used. Refractive index was assigned based on particle size (SI Appendix, Table S1). If the particle size is beyond (280 to 400 nm), RI was assigned to the closest limit to avoid unrealistic RI values. Simulations were initiated with random placement of particles in a 20 × 20 × 20-μm³ box with periodic boundary conditions (PBC). Particle overlaps were relaxed with repulsive Gaussian potential for 2000 time steps, and the systems was then squeezed to a final 5 × 5 × 5-μm³ volume during the rapid quench, in 5 × 10⁵ time steps while using the LJ12-0 pair potential. Particle coordinates and diameters were then imported to Lumerical FDTD (Ansys) release 2025 R1.3. In the FDTD simulation set-up, PBC boundaries in the x and y directions were used and a plane-wave source and reflection monitor were placed 6.5 and 7.5 μm above the sample (in the z direction), respectively. A perfectly matched layer (PML) boundary was used in the bottom z direction, which mimics an absorbing layer under the structure. The recorded reflectance was integrated in Lumerical over the 5 × 5-μm² monitor area. In all FDTD simulations, the nanospheres were embedded in a surrounding medium of water (n = 1.33) representing the cytoplasm. Absorption was neglected (k = 0) over the simulated spectral range (290 to 800 nm), where water is essentially nonabsorbing.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Dataset S03 (XLSX)

Dataset S04 (XLSX)

Dataset S05 (XLSX)

Code S01 (TXT)

Code S02 (TXT)

Data, Materials, and Software Availability

All study data and code are included in the article and/or supporting information.