Inspired by natural structural color such as the wings of butterflies and beetles, artificial micro- or nano-scale structure with transparent materials can generate structure color with intricate design [1,2]. It is a new environment friendly method and will not fade as traditional dying color will. Principles of obtaining structural colors include thin film [3,4], photonic crystal [5,6], metal nanostructure [7–10], and dielectric nanostructure configurations [11–16]. Thin film and photonic crystal structures are typically produced by atomic layer deposition (ALD) [17], chemical vapor deposition (CVD) [18], and physical vapor deposition [19]. Metal and dielectric nanostructures are primarily created through electron beam lithography (EBL) [20,21] and focused ion beam (FIB) [4] lithography. Besides the pattern of micro/nanostructures, information also can be carried or encoded by controlling polarization and phase, or by altering materials via stretching, electrical stimulation, or vapor deposition, thereby enabling encryption and steganography [5,6,22–31]. These techniques usually depend on metasurface structures and involve complex, expensive processes. The plasmonic resonance effect on metal, which is dominantly adopted for structural color, has advantages of angle independence but is easily damaged by oxidation and applying protective passivation layers may alter the color. Thus, an economical, durable, and simple structure fabrication method is in demand. Two-photon polymerization (TPP) technology provides a possible resolution with high precision, customization, and flexibility in nanostructure creation [32–35]. The Mie resonance is strong in high-refractive-index media, resulting in high color saturation. Polymer with a lower refractive index (compared to titanium dioxide, silicon, etc.) is light, stable after solidification, affordable, eco-friendly, and easy to mass-produce, which makes it an ideal material for color production.

In this study, structural color is generated by nanostructures with larger periods to simplify the fabrication process. At this scale (d≥λ/n, d is the diameter, λ is the wavelength, and n is the refractive index), the choice of wavelength is mainly derived from Mie scattering [36,37]. And due to the low refractive index (1.52) in this paper, the Mie resonance is very weak [38,39]. We theoretically and experimentally investigate the transmitted light of various scattering structures (cylindrical arrays [32,33], long arrays [26], cross-state arrays [40], gratings [31], grids [41]) under cross-polarized transmission illumination. Subsequently, the color of these structures is observed under bright-field transmission illumination. The optical performance of these structures is analyzed with multipole decomposition, and the transmission spectra of each structure are simulated, showing good agreement with the experimental results. Based on the polarization characteristics of these structures, a cylindrical array structure with minimal transmitted light and a grating structure with maximum transmitted light are chosen to fabricate a full-color panel. Subsequently, the full-color panel is “translated” and a database correlating the structural parameters with colors is established based on the colors observed with cylindrical arrays and gratings under bright-field illumination. To realize steganography, the pattern to be hidden is written using gratings, while the remaining part is written using cylindrical arrays, achieving a colored hidden pattern. In addition, a large-scale color pattern with hidden messages is printed. The proposed method enables color printing with embedded encrypted information. By leveraging the ability of 2D patterns to store more information than text, it significantly increases information density. Patterning also provides faster information retrieval compared to dynamic tuning.

Substrates (glass and silicon wafers) are exposed to distinct 10-min each ultrasonic baths in acetone and isopropanol solutions, respectively, succeeded by a 10-min ultrasonic bath in distilled water. Subsequently, nitrogen is used for purging the substrates for 15 s, and a flat heating stage at 200°C heats the substrates for 20 min. Nano-array printing is performed using a two-photon laser direct-writing device (Moji-NanoTech, China, PROME-Uni) with a femtosecond laser at a wavelength of 532 nm (40 MHz) focused via an immersion objective lens (Olympus 60×, NA 1.43), and polymerization and cross-linking occurred in the liquid resin ATP-DIP (Moji-NanoTech, China, index 1.52). The nano-arrays are printed by the vector mode or ultra-high-speed mode of the device; afterwards the nano-arrays are developed by acetone immersion for 5 min, isopropanol immersion for 3 min, and n-hexane immersion for 2 min.

Pixel palette photographs and art paintings are observed under an optical microscope (Suzhou Aseet Tech, China, AST-BH200M) and photographed using a charge-coupled device (Suzhou Aseet Tech, China, AST-T1200). A 5× objective lens (Soptop, 5×, NA 0.13) is employed to collect the transmitted light. Figures 5(a) and 6 are observed indoors under white light and taken using a cell phone (IQOO neo7, China). Scanning electron microscope (SEM) images are captured using a bench-top scanning electron microscope (NeoScope, JCM-7000).

Finite element analysis software is used to simulate the scattering field and transmission spectra of nanostructures. Figure 3(b) depicts the geometry structure, with the central portion comprising a photoresist having a refractive index of 1.52. The lower part demonstrates the substrate (indices of glasses are 1.5226 and 1.43, index of sillicon is 3.49), and the middle part is air. The multipole scattering simulation wraps the perfectly matched layer (PML) around the perimeter of a single structure. Far-field radiation, reflection, and transmission calculations place only the PML at the top and bottom, with periodic conditions added around the perimeter. In the transmission calculation, the light source propagates along the positive z-axis, and the reflected light propagates along the negative z-axis, with wavelengths ranging from 400 nm to 750 nm. The transmission spectrum is obtained by capturing the zeroth-order diffraction light.

Figure 1(a) shows all structures with different parameters fabricated by TPP printing. Each array of structures is defined by three parameters: height, linewidth, and period. Laser power, writing speed, and focal depth can vary the writing height and linewidth; the exposure time affects the height and linewidth depending on the structure [Fig. 1(b)]; the period of the structure arrays is encoded in the printing program. After printing, the structures are observed using a 5× objective lens as shown in Fig. 1(c). At this observation height, the microscope primarily captures the zeroth-order diffraction light from the structures, while higher-order diffraction is filtered out. Under bright-field transmission illumination, structures with specific periods, linewidths, and heights exhibit various colors accordingly. Under cross-polarized illumination, the transmittance of different structures will vary when the orientation of the polarizer is at an angle of 45° to the structure, as shown in the right side of Fig. 1(c). This feature enables the design of encrypted patterns in cross-polarized transmission mode, which remain unaffected in bright-field illumination mode. By tuning the parameters such as period, linewidth, and height, colorful patterns can be composed, and hidden information can be retrieved in cross-polarized transmission mode.

We first investigate the color performance of nanostructure arrays under cross-polarized illumination. In this section, all samples are fabricated with the photoresist on a glass substrate (n=1.5229, Schoot, B270i). The laser power is tuned as the output ratio in our manufacturing platform so the real power is tested previously as shown in Fig. 2(a). In the left panels in Figs. 2(c)–2(g), the parameters for characterizing the five structural arrays are defined as p (period), h (height), d (linewidth, and for cylindrical arrays, d refers to the diameter), and l (length of the long and cross arrays). One should pay particular attention that d for the grid structure is divided into d (lateral) and d (longitudinal) since the machine produces different linewidths in the lateral and longitudinal directions during writing, which should be ascribed to the discrepancy of the sweeping speed between x- and y-directions of the galvanometer. Since the periods between structural arrays have an effect on saturation [34], the sizes of a single test square (cite), squares of the color palette, and individual pixels of patterns are all designed to be 5μm×5μm.

Figure 2(i) shows the simulated transmittance of the five structures from Table 1 under varying angles of cross-polarized transmission. The cylindrical structure is polarization-insensitive and does not have polarization conversion properties. It can be observed that, with the exception of the cylindrical structure, the other structures exhibit the highest transmittance at a cross-polarized angle of 45°. Given that the majority of structures appear brighter at this angle, our study employs a cross-polarized angle of 45° for observation. Figure 2(a) compares the simulated transmittance of the five structures in Table 1 for 45° cross-polarized transmission, with the grating transmittance being the highest and the cylindrical transmittance being the lowest almost coinciding with the horizontal axis. Observing the diagram in Fig. 2(h), we can see that the direction of the electric field in the cylindrical structure is almost constant. The grating structure changes significantly in the direction of the electric field at a wavelength of 520 nm, resulting in high transmittance at this wavelength. The right panels in Figs. 2(c)–2(g) show the photochromes of these five arrays under cross-polarized illumination, where the period of the structures in the matrix decreases from left to right, and the power increases from bottom to top, with each small square representing a structure array with specific parameters. The transmittance ranking of each structure is similar to that of the simulation results. The difference in transmittance between cylindrical and grating arrays is the largest under cross-polarized illumination, an advantage that will help in hiding patterns.Table 1.

Parameters of the Partial Array Structure

In this section, bright-field illumination is perpendicularly incident from below the samples of all five types of arrays as previously mentioned. Simulations for the cylindrical arrays yield the scattering spectra [Fig. 3(a)] and transmission spectra [Fig. 3(c)] of multilevel decomposition. Figure 3(a) shows the scattering, electric dipole (ED), magnetic dipole (MD), electric quadrupole (EQ), and magnetic quadrupole (MQ) of the four structures under linearly polarized light along the x-axis and y-axis. The parameters of the cylindrical, cross-shaped, and long structures are consistent with those in Table 1. Considering computational limitations, the parameters of the asymmetric cross are set as p=1μm, d (horizontal) = 600 nm, d (vertical) = 400 nm, and h=2μm, to explore the polarization sensitivity of the grid structure. The polarization-sensitive characteristics of the grating can be inferred from the long structure. It can be observed that, due to the symmetric nature of the cylindrical and cross-shaped structures, they exhibit good polarization insensitivity, with nearly identical results under different polarization states of the input light. When the symmetry is broken, the results of the asymmetric cross and long structures change under different light sources. The changes in MQ are shown to influence the scattering spectra (as indicated by the red and blue arrows in the figure), and the variation in MQ contributes to the polarization insensitivity of the structures.

From the multipole analyses, we can observe that the structures exhibit weak electric and magnetic dipole resonances [37,42,43], the Mie resonance is weak, and the magnetic quadrupole contributes more among the four poles. At this time, the interference of the magnetic quadrupole with other higher-order modes leads to Mie scattering, which is dominated by forward scattering. Figure 3(b) shows the electric field diagram of the cylindrical structure. Since one part of the light is transmitted in the column and some is in the air, the two parts of the light will interfere due to the presence of phase difference [32]. Thus diffraction is generated. Combined with the far-field radiation pattern in Fig. 3(c), it can be seen that only zeroth-order diffracted light will enter the objective lens due to the objective lens collection angle. At 440 nm the scattering cross section is high but less light can enter the objective, and the opposite is true at 680 nm, resulting in differences in scattering cross section and peak transmission spectra. The left sides of Figs. 3(c)–3(g) show the bright-field transmission images of cylindrical arrays, long arrays, cross-shaped arrays, gratings, and grids under various fabricating power and period settings, among which the color of cylindrical arrays appears to be the brightest. The right sides present simulated transmission spectra for these five structures. Considering the performance of each structure under cross-polarized transmission illumination, cylindrical arrays and grating structures are ultimately selected for printing large-scale color palettes and creating dual-color patterns for steganography.

Based on the previous results, large-scale color panels with hidden information are fabricated using cylindrical arrays and gratings, and corresponding simulations are conducted. Under our experimental conditions, a bigger discrepancy of the refractive index between glass substrate and photoresist leads to more accurate results of the substrate-photoresist interface detection. Therefore, a glass substrate with a refractive index of 1.43 (SIPT, quartz) is selected for printing the full-color palette and images in the following section.

As a new glass substrate is adopted, the optimal relative interface depth, referred to as the focus depth as shown in Fig. 4(d), is explored correspondingly. At the interface depth of 1 μm, detachment during the development process is as in the area of the blue dashed line frame. Conversely, when the interface relative depth is −1μm, only a small fraction of the nanopillars undergo polymerization with the photoresist, showing fading structural color within the orange dashed frame. Therefore, the full-color panel is printed at a relative interfacial depth of 0 μm.

In Figs. 4(f) and 4(h), the red dashed areas in the top left corners indicate the regions being skipped during the palette printing process. This is because, during the fabrication process, excessively high power and very small periods will result in nanopillar adhesion, accumulation, and overexposure. It includes basic colors such as red, yellow, and blue [Fig. 4(c)]. Figures 4(a) and 4(b) show that with increasing height, the color shifts towards red, and an increase in diameter also results in a red shift, consistent with the actual experimental outcomes. The simulated spectra for nanocolumn arrays with different periods [Fig. 4(e)] demonstrate the same transmitting peaks, which means their colors are the same but those with steeper peaks show higher brightness/saturation of color. This result can be explained: as the period increases, less scatterer units stand in the same area, so less light with the specific color is generated when each scatterer works individually, resulting in lighter color.

However, for the experimental results, the color palette obtained shows not only changes in saturation but also a trend of color shifting towards the blue end with increasing period size, as observed by the solid gray box in Fig. 4(f). This is mainly due to processing defects: when printing the current nanopillar, the local solidification effect of the absorbed energy combines with the residual energy from the last one if they are close enough, resulting in larger diameters of nanopillars. Slowing down the fabrication speed may alleviate this issue, but it would compromise processing efficiency. Additionally, this issue will not affect the subsequent pattern fabrication. Figure 4(e) selects nanopillar arrays with diameters of approximately 600 nm and heights of about 2 μm, as outlined by the gray dashed box in Fig. 4(f), where the color observed within the box [exp. in Fig. 4(e)] shows that the color obtained experimentally becomes lighter with increasing period, under conditions of consistent height and diameter.

Based on the full-color palette in Fig. 5, a color database is established correlating colors to manufacturing parameters (height, period, linewidth) correspondingly. For the target image to be printed, a pixel-by-pixel analysis is conducted to extract the color of each pixel and find a matching color and corresponding parameters in the color database. If a pixel requires information to be concealed, it is replaced by a grating array with determined parameters; otherwise, it is replaced by a cylindrical array. After traversing the entire image, a printed file containing structural information is obtained. Printing is performed using a photoresist with a refractive index of 1.52 on a glass substrate with a refractive index of 1.43.

Due to the limited working range of the scanning galvanometer during printing, a translation stage is used under the substrate for stitching when printing millimeter-sized images. Figures 5(b), 5(e), and 5(f) illustrate the results of writing with a smaller scanning range for the galvanometer (5μm×5μm). Due to the frequent movement of the stage, printing a 1mm×1mm pattern [Fig. 5(b)] takes 1 h. Printing with a 50μm×50μm (10×10 pixels array) galvanometer range for the same size pattern [Fig. 5(c)] takes only 4 min. However, the former range has the advantage of more uniform color and smaller stitching traces. In consideration of time efficiency, print with a large scan range. Ultimately, we produce a 2mm×1.5mm dual-information pattern [Figs. 5(g) and 5(h)].

Structural coloration on silicon substrates with nanostructures has been previously established [35], and we can produce structural colors with a 1.52-index photoresist on silicon similar to that on the glass. In Fig. 6(a) it can be seen that under different substrates, the glass substrate is close to the refractive index of the structure, and the energy continues to propagate in the substrate [38], resulting in low reflectivity. Silicon improves the light confinement ability of the structures by enhancing boundary reflection, so as to bring strong reflected light. Similar to the above research on a glass substrate, we demonstrate structural color using cylindrical arrays of varying sizes on a silicon substrate (Xi'an Jinxing Dianzi Tech, index 3.4975) under ambient light and hidden dual patterns observable at specific angles and under intense light. Surprisingly, we find that when diameters and heights of nanorods are reduced on the silicon substrate, they become less perceptible by naked eyes, and become visible only under strong illumination or from specific angles, which enables the concealment of information. We print a large color pattern of 3.2cm×2.4cm [Fig. 6(c)] using ultra-high-speed mode (the vector mode used dot-by-dot printing earlier; ultra-high speed is not as granular as vector mode) and using 8 h. Different colors can be seen from different angles under ambient light. The bottom left of Fig. 6(c) displays cylindrical structures with a diameter of 150 nm, their small volume effectively hiding the pattern. Figure 6(b) reveals the pattern viewed from different angles, indicating that text hidden in the top right corner of the map becomes visible at certain angles [Fig. 6(d)]. Under intense light, the colored text “TECH” becomes distinctly visible [Fig. 6(e)]. The supplementary video demonstrates that as the light source moves, the complete hidden text becomes clearly visible (Visualization 1).

In this paper, we introduce a method for creating colorful dual patterns by nanostructures written with two-photon polymerization technology and explore a steganography technique based on structural color. We extensively explored the color performance of five types of structures—cylindrical arrays, long arrays, cross-state arrays, grating, grid—under different linewidths, heights, and periods, both under cross-polarized transmission and bright-field transmission illumination. Based on characteristics of these structures, the transmittance difference of cylindrical arrays and grating structures is adopted for steganography with dual-pattern encoding. To stretch the application of structural color, we also developed full-color panels for these two structures and determined the optimal structural parameters through a comparison of simulation and experimental results. Experimentally, we successfully printed 1mm×1mm color patterns and 2mm×1.5mm dual patterns using these structures, achieving multi-level control of color and information. In addition, a large-scale pattern measuring 3.2cm×2.4cm has been developed with concealed information on a silicon substrate and the steganography is realized based on the visibility depending on the size of the nanostructure. This color preparation method, based on structural design and two-photon polymerization technology, embodies sustainability and environmental friendliness, providing innovative ideas and methods for the fields of color science. The capability of large-scale pattern building stretches the application of structural color to a normal life dimension. The developed steganography techniques provide novel means for information security and storage.