1. Introduction

As a promising alternative to conventional chemical dyes-based colorant pigment, structural color has been extensively investigated due to its potential advantages, such as ultra-high resolution, long-standing stability, and environmental friendlies [1–4]. Currently, various structural designs are demonstrated to produce wide gamut and vivid colors by reflecting or transmitting a certain spectral region of visible light. After a summary and analysis, it is found that there are mainly three physical mechanisms for the generation of different colors: Fabry-Perot (FP) resonance based on multiple layer metal/dielectric stacking structure [5], plasmonic resonance mode generated by metallic nanostructures [6], and the combination of magnetic and electric dipole resonances in all-dielectric materials [7]. With the growing requirement of practical application of structural color, angle-insensitive characteristic has become a key feature of structural color filter, that is, optical spectral features including center wavelength and transmission/reflection efficiency retain unchanged over a wide angle range. However, these have been great difficulties to satisfy this demand. This is because optical resonant modes producing color in most reported nanostructures is commonly associate with angle for fulfilling the momentum matching condition. To achieve angular-independent property a great deal of structural designs capitalizing on a highly absorbing medium-based thin-film nano-resonator or phase compensating covering have been researched [8–19]. Besides angle-invariant feature, the shielding of short-wavelength infrared light is of crucial significance for the potential application of structural color in some special application scenarios, such as glass smelting, steel forging and the like. These special scenarios generally can produce intensive near-infrared radiation, which is not friendly to human eyes, and potentially increases the risk of eye diseases of radioactive cataracts for long-term exposure of human eyes [20–24]. Architectural decorative coatings, both demonstrating colors and blocking near-infrared light, has a protective effect for workers in some harsh environments, such as glass smelting, steel forging, and the exposure to the outdoors for a long time. Consequently, it is desirable to develop a new type of reflective color filter with both functions of angle-invariance and near-infrared shielding to protect human eye in some harsh workplace.

In this paper, we demonstrate an angle-insensitive reflective color filter with near-infrared light shielding function, constructed by four-layer alternately (silicon-titanium-silicon-titanium) planar nanostructure with the requirement of a cost-effective and simple deposited process. The strong optical interference generated by asymmetric FP nano-cavities is responsible for dual-function. Experimental measurements show that by controlling the thickness of the silicon (Si) layers, the color filter can achieve reflective three primary colors of cyan, yellow, and magenta (CYM). Minimal propagation phase shift changes are offset by nontrivial reflected phase shift, and equips this color filter with angle-insensitive characteristics. The intensity of the resonant reflection peak is greater than 0.7, which endows the structural color with the characteristics of high brightness. The color filter features a non-iridescent performance over a wide angle of incidence up to 60°, over which the center wavelength is shifted by below 1.8%. The color filter under three primary colors (CYM) has an 70% average absorption for near-infrared light, which is much higher than the traditional silver (Ag)-based and aluminum (Al)-based color filters. Finally, the thermal stability of the proposed reflective color filter is investigated in the harsh environment with high temperature up to 500°C, demonstrating its robust thermal stability for potential practical applications.

2. Results and discussion

Figure 1(a) shows schematic diagram of the proposed color filter. The color filter consists of a three-layer metal-dielectric-metal (Ti-Si-Ti) nanocavity resonator with an integrated dielectric Si overlay. In order to obtain high-performance reflective color filter, the thickness of two Si layers is designed to be the same and denoted by the symbol t. A thin Ti layer with the thickness m of 20 nm is used as a broadband absorber layer. The thick Ti layer with the thickness M of 200 nm at the bottom is choose as an opaque reflection layer. Figure 1(b) shows the cross-sectional scanning electron microscope (SEM) image of the yellow filter prepared on the silicon wafer, which clearly demonstrates the four-layer structural feature of the fabricated sample. As depicted in Figs. 1(c)–1(e), reflective three primary colors of cyan, yellow, and magenta are generated with the thickness of the Si layer t = 30 nm, 60 nm and 70 nm used, whilst their corresponding simulated and measured reflectance spectra are also demonstrated. Simulated reflection peak position/intensity of three primary colors in visible light are 429 nm/79%, 607 nm/78%, and 672 nm/77%, respectively, which almost coincide with experimental results of 418 nm/43%, 672 nm/69%, and 748 nm/76% for cyan, yellow, and magenta color generation. The discrepancy between simulated and measured results may be attributed to the refractive index (RI) variation between the simulated and actual materials, especially for thin Ti layer. The difference in the reflection peak intensity of the cyan color filter is larger as compared to the other colors. This is because the actual Si material has a high extinction coefficient near the wavelength of 400 nm. Furthermore, the calculated average absorptivity of the whole near-infrared light from 780 nm to 1500 nm for cyan, magenta, and yellow color primaries are 81%, 79%, and 77%, respectively, and their respective experiment results are 79%, 72%, and 58%. It is clearly observed that the average absorptivity of three primary color filters for the whole near-infrared light is 70%, which is of crucial significance for color filter with near-infrared light shielding. The color patches on the right of insets of Figs. 1(c)–1(e) are calculated from the simulated reflectance spectra, which are comparable to the experimental ones on the left. Vivid and bright images taken using the fabricated three primary color filters are displayed on the bottom of Fig. 1(b). We further convert reflective spectra into color coordinates, plotted in the CIE1931 chromaticity diagram [25], as shown in Fig. 1(f). The chromaticity coordinate distribution shows that the simulated results are very close to the measured ones.

 

Fig. 1. Structure design and characteristic spectra of angle-invariant eye-friendly color filter. (a) Schematic diagram of angle-invariant eye-friendly color filter. (b) The cross-sectional SEM image of a representative fabricated four-layer structure on a silicon substrate. The Ti layers are marked by yellow false-color to clearly depict the interfaces between alternating layers. Scale bar: 100 nm. Insets show photographs of the fabricated three square-shape color samples taken by mobile phone camera. The measured and calculated reflection spectra of the designed four-layer stack structure with Si spacer thickness of (c) 30 nm, (d) 60 nm and (e) 70 nm at the normal incidence with unpolarized light. Insets are their displayed colors calculated by using their respective reflection spectra (D50/2°). (f) Measured (black circle symbol) and calculated (red triangle symbol) color coordinates in the CIE1931 chromaticity diagram with data from c, d and e.

Download Full Size | PDF

For further elucidate physical origin of the spectral characteristics of the proposed angle-invariant eye-friendly color filter, the yellow color filter is taken as a typical example for the sake of illustration. Figure 2(a) shows the simulated reflectance spectra of the proposed four-layer stack structure with Si spacer thickness of 60 nm. Clearly, two reflection peaks appear at the wavelengths of 607 nm (peak 1) and 401 nm (peak 2), accompanied by two reflection valleys located at the wavelengths of 855 nm (dip 1) and 474 nm (dip 2). We further investigate the effect of the Si layer thickness on its reflection spectra in Fig. 2(b). Obviously, the reflection minimum strips increase gradually with the studied wavelength range as the Si layers thickness increases, the variation tendency of which is consistent with that of a Fabry-Perot (FP) cavity resonances. Different reflection minimum strip corresponds to FP resonance with different order [26]. It is generally acknowledged that the cavity resonances occur when the net phase shift is equal to a multiple of 2π. The net phase shift accumulated includes three parts: two reflection phases shifts and a round-trip propagation phase shift. Two reflection phases shifts acquired upon the internal reflection at the top and bottom interfaces of each Si layer and the round-trip propagation phase shift is accumulated by propagation within Si layer [27]. The round-trip propagation phase shift and two reflection phases shifts are given by the following Eq. (1) and Eqs. (2)-3, respectively.

 

Fig. 2. Physical mechanism of angle-invariant eye-friendly color filter. (a) Calculated reflectance spectra of the yellow filter (t = 60 nm) at the normal incidence with unpolarized light. (b) Reflection spectra as a function of the thickness of Si layers, with the thickness of Ti layer is m = 20 nm. Different types of phase shifts associated with the (c) top and (d) middle Si layers as a function of incident wavelength for the yellow filter. Here, reflection phase + represents the reflection phase shift of the upper interface, reflection phase - is the counterpart of the lower interface. Distributions of the electric field and absorbed optical power for (e) λ = 855 nm, (f) λ = 474 nm, (g) λ = 607 nm, and (h) λ = 401 nm.

Download Full Size | PDF

To investigate the generated mechanism underlying resonant effect in reflection spectra, different types of phase shifts associated with the top and middle Si layers are presented in Figs. 2(c) and 2(d), respectively. It is noteworthy that, for the top Si layer, the net phase shift at the wavelength of 907 nm gets zero, indicating the fundamental FP resonance mode is excited in this Si layer. Similarly, the zero net phase shift in Fig. 2(d) is observed at the wavelength of 760 nm, corresponding to the fundamental FP cavity resonance inside the middle Si layer. It is important to note that the dip 1 at the wavelength of 855 nm in the reflectance spectra is justly located at the middle between above two resonances. This is because dip 1 in the proposed four-layer stack structure stems from their overlapped resonant effect. Furthermore, it is dramatic that the net phase shift approaches to zero in the wide near-infrared wavelength range from 1200 nm to 1500 nm. Therefore, the fundamental FP resonance in the top Si layer is the main reason for the high absorption of near-infrared length for the proposed structural color, which are usually explained as the anti-reflection (AR) effect. Figure S2 in the Supporting Information compares the reflection spectra and optical admittance of the yellow color filter with and without the top Si overlay, validating the AR function of the top Si overlay in near-infrared range. Furthermore, the first-order FP resonance is also excited in the top and middle Si layers with the net phase shift of 2π, corresponding to the wavelength of 468 nm and 429 nm, respectively. Likewise, the first-order FP cavity resonances in the top and middle Si layers contribute altogether to the generation of dip 2 in the reflectance spectra. In addition, we note that the net phase shift of π/3π at the top and middle Si layers occur near the wavelength of 600 nm/400 nm, which is almost consistent with the reflection peak of 607 nm/401 nm. Therefore, the reflection peak 1 and peak 2 originates from the combined contributions of the anti-resonant effect inside the top and middle Si layer. As a comparison, the peak before the 1200 nm with low reflection intensity only originates from the anti-resonant effect in the middle Si layer.

In order to gain further insight into the optical resonant phenomenon in the proposed structure color, Figs. 2(e)–2 h depict the normalized intensity distributions of electric field and absorbed optical power inside the whole structures of the designed yellow color filter for two dips and two peaks in Fig. 2(a). It is obvious that, for high absorption dip 1, the strong electric field is mainly concentrated in both the top and middle Si layers where there is no node, which corresponds to the fundament FP cavity resonance. Although electric field intensity at the position of the thin Ti layer is relatively lower than the Si layer, the high absorption performance at the dip 1 is attributed to the middle thin Ti and the bottom opaque Ti layers due to its large extinction coefficient of the Ti materials, verified by the absorption power distribution on the right side of Fig. 2(e). For the electric field distribution of dip 2 in Fig. 2(f), there is one node observed in both Si layers, which corroborates the generation of the first-order FP cavity resonance in both Si layers. It is notable that this high absorption at dip 2 originates from the cooperative effect of the Ti and Si materials (the right panel of Fig. 2(f)). This is because the large absorption coefficient of the Si material at the short wavelength as compared to the one at the long wavelength. For the cases of peak 1 and peak 2 in Figs. 2 g and 2 h, electric field intensity from the fundamental and first-order FP cavity resonance in both the top and middle Si layer is greatly weaken as compared to those of dip 1 and dip 2. And the intense electric filed is located at the air region and is prevented into the proposed stack structure. In other words, both the top and middle Si layers play a significant in the generation of anti-resonance effect. Additionally, the absorption of the Ti materials in peak 1 is the main reason for the reduction of reflectivity.

Next, we investigate the wide-angle invariant characteristics of the proposed color filter. To illustrate this point, the cyan color filter with the 30 nm-thick Si layer is used as a typical example and the low RI magnesium dioxide (MgF₂)-based four-layer stack structure with the 200 nm-thick MgF₂ layer are employed to highlight the angle-invariant performance of the proposed color filter. Figures 3(a) and 3(b) respectively present photographs under the incident viewing angle of 0° to 60° for the high RI Si-based structure and the low RI MgF₂-based structure. Intuitively, the proposed Si-based color filter is indistinguishable to the naked eye at different observation angles. However, the MgF₂-based color filter has a pronounced color change from green to cyan with the observation angle from 0° to 60°. The above experimental results corroborate that the high RI materials in the proposed color filter can considerably improve the angle insensitivity of color. Furthermore, Figs. 3(c)–3(d) thoroughly investigate the measured angle-resolved reflectance spectra, reflection peak intensity, and the shift of peak wavelength of the proposed Si-based color filter. As a direct comparison, the corresponding results of the MgF₂-based color filter is shown in Figs. 3(e) and 3(f). It can be seen that the measured reflection spectra profile of the fabricated Si-based color filter almost remains unchanged as the incident angle increases to 60° (Fig. 3(c)), exhibiting good agreement with the simulated one excepting for reflection intensity due to the fabricated imperfect and different material parameters (refer to Fig. S3 in Supporting Information). Specifically, When the incident angle changes from 10° to 60°, the peak wavelength in the reflectance spectra has a small blue shifts from 433 to 425 nm, whilst it’s intensity reduced from 0.46 to 0.39. In order to further quantitatively describe the angle invariance, the peak shift is defined as the ratio of Δλ to λ_i, where Δλ represents the difference between the initial and central wavelengths at different angles, and λ_i stands for the initial wavelength (at 10° incidence here). Consequently, the maximum peak shift at 60° is only 1.8%. In stark contrast, the corresponding MgF₂-based color filter has an obvious blue shift of the resonance peak, as shown in Fig. 3(e). The maximum peak shift reaches up to 16% at the incidence of 60° with reflection peak intensity (wavelength) changes from 0.43 (514 nm) to 0.41 (431 nm). The above results indicate that the proposed Si-based color filter has excellent angle robustness characteristics.

 

Fig. 3. Wide-angle non-discoloration test of color filters of high and low RI materials. The photographs of (a) high RI Si-based and (b) low RI MgF₂-based color filter at the incident angles of 0°, 30° and 60°, respectively. Measured angle-resolved reflection spectra, reflection peak intensities, and peak shifts of (c)-(d) Si-based color filter and (e)-(f) MgF₂-based color filter.

Download Full Size | PDF

By exploiting the nontrivial reflection phase shifts that occur at the interfaces between semiconductor and metal/air, the colors are able to be created over a wide angle of incidence [28]. To illustrate this, we calculated thoroughly net phase shifts as a function of wavelength, and different types of phase shifts as a function of the incident angles at the resonance wavelength for both the Si-based and MgF₂-based filters in Fig. 4. The reflectance spectra of the two color filters at normal incidence are also calculated in Fig. S4 of the Supporting Information. The fundamental mode resonances of the Si-based cyan filter are located at 791 nm and 528 nm at the top and middle cavities in Fig. 4(a), while the counterparts of MgF₂-based filter are 1129 nm and 681 nm in Fig. 4(d). Compared to the low RI MgF₂ layer, the variation of propagation phase shift in the Si layer is very small since the change of refraction angle is insignificant. Meanwhile, a small change in the propagation phase shift can be further offset by the nontrivial reflection phase shift. Contributed from these two effects, the change in the sum of the two reflection phase shifts and the propagation phase shift is almost zero over the wide range of incident angles, which is the reason of a greatly improved angle dependent performance.

 

Fig. 4. The physical origin of wide-angle non-discoloration. (a) Calculated net phase shifts in the top and middle Si layers as a function of wavelength for the Si-based filter. Different types of phase shifts as a function of the incident angles at the resonance wavelength in (b) top Si and (c) middle Si layers. (d)-(f) The corresponding results of the MgF₂-based color filter. The dielectric layer thickness of the Si-based and MgF₂-based color filters is 30 nm and 200 nm, respectively, with other parameters remaining the same.

Download Full Size | PDF

As above we presented, due to their high reflectivity in the visible region, metals such as Ag and Al become a traditional choice for manufacturing FP nanocavity-based color filters. As shown in Fig. 5, we further highlight the performance of the proposed titanium (Ti) -based color filter as compared to traditional silver (Ag)-based and aluminum (Al)-based color filters. Ag-based structure means that all metal layers are Ag, including the bottom thick metal layer and the middle thin metal layer. The corresponding Al-based and Ti-based structures are also the same as above. Figures 5(a)–5(c) respectively show color palettes under different metal and medium thicknesses for three structures. The thickness of dielectric layer t is changed from 30 nm to 70 nm in steps of 5 nm, in parallel with the thickness of metal layer from 5 nm to 30 nm. All three color filters can obtain full color tinting, but the Ti-based color filter has significantly higher saturation. All the palette colors are plotted into CIE1931 chromaticity diagram, as shown in Figs. 5(d)–5(f). Obviously, within the same thickness variation range, the Ti-based color filter obtains a larger color gamut. The near-infrared average absorptivity of all color filters are calculated as Figs. 5(g)–5(i). It is found that the average absorptivity of all Ti-based color filters for near-infrared light is 76%, while the corresponding values of Ag-based and Al-based color filters are 18% and 38%. Therefore, this Ti-based color filter is distinguished in both color and near-infrared shielding.

 

Fig. 5. Property comparison of Ti-based and other metals-based color filters. Color palette presentation of color filters based on (a) Ti, (b) Ag, and (c) Al materials. Color gamut distribution of color filters based on (d) Ti, (e) Ag, and (f) Al materials. The Near-infrared average absorption of color filters based on (g) Ti, (h) Ag, and (i) Al materials.

Download Full Size | PDF

Finally, for glass processing, steel forging, and outdoor scenes with strong near-infrared rays, the designed angle-insensitive eye-friendly color filter is required to be thermally stable. We heat the designed magenta color filter (t = 70 nm) sequentially from room temperature to 300 °C and 500 °C for half an hour, and then the photographs of the samples are obtained with a mobile phone camera. As shown in Fig. 6(a), the samples still maintain a vivid magenta color after being sequentially heated at 300 °C and 500 °C, reflecting the excellent thermal stability. This stems from the high melting points of the designed materials Ti and Si. We further measure the normalized reflection spectra of the color filters treated at different temperatures, as shown in Fig. 6(b). The reflection spectral profiles at the three temperatures are almost identical. Finally, the designed structural color is utilized to prepare the school emblem of “Dalian University of Technology” on a transparent silica substrate. As shown in Fig. 6(c), the diameter of the circular emblem is 20 mm. The high-quality cyan color on the school badge reflects the great potential of the color filter for mass production.

 

Fig. 6. Thermal stability of color filters. (a) The photographs and (b) measured reflectance spectra of magenta filters at different temperatures. (c) The photograph of the “Dalian University of Technology” logo fabricated by using mask cover technique with the angle-invariant eye-friendly colors.

Download Full Size | PDF

3. Methods

Numerical Simulation: The transfer matrix method is used to calculate the spectral characteristics and phase shifts [29]. The spatial distribution of electromagnetic field is calculated by finite-difference time-domain (FDTD) Solutions commercial software. In the simulation, period boundary conditions are used along the x and y directions, and perfectly matched layers are used in the z direction. The grid sizes along the x, y, and z directions are 1 nm × 1 nm × 1 nm, respectively. A frequency-domain field profile monitor covers the structural unit to obtain the spatial distributions of electric field and absorption power. The RI of Si and Ti are from E. D. Palik’s database [30]. The RI of MgF₂ is obtained from Refs. [31]. The corresponding RI curves are plotted in Fig. S5 of the Supporting Information.

Fabrication of structure: The Si and Ti layers are alternately deposited via direct-current magnetron sputtering (Quorum-Q300TD plus) onto the substrate. The Ti layer is deposited with a purity of 99.9% of Ti target, a beam current of 100 mA, and the sputtering rate of 0.3 nm/s. While the Si layer is to be a purity of 99.9% Si target, a beam current of 50 mA, and the sputtering rate of 0.5 nm/s. The MgF₂ layers are deposited via radio frequency magnetron sputtering (JPG450) onto the substrate. The MgF₂ target with a purity of 99.9% works at a power of 100 W, and the sputtering rate of the MgF₂ is 0.8 nm/min. The chamber pressure is 5 × 10⁻³ Pa, the sputtering pressure is 0.5 Pa, and argon flowing is 20 sccm. The used substrates include the square-shaped silicon of 10 mm × 10 mm and the square-shaped fused quartz of 25 mm × 25 mm.

Optical Measurement: The reflective spectra of the proposed color filters are recorded using a homemade optical setup, where the combined end of bifurcated optical fiber jumper connects the optical fiber collimator, and the halogen lamp based broad spectrum light source (HL-2000, Ocean Optics) and fiber-optic spectrometer are connected to two branch ends of a bifurcated optical fiber jumper. The two fiber-optic spectrometers (AvaSpec-Mini4096CL and AOTF-R2) work together to measure wavelengths from 400 nm to 1500 nm. the normalized reflection spectra of the sample are obtained by using the following formula,

4. Conclusions

In summary, we propose an angle-invariant eye-friendly color filter, processing the ability to protect human eyes in some harsh environments with intensive near-infrared radiation, such as glass manufacturing, steel forging, and outdoors. The experimental results demonstrate that the color filter can achieve full-color display including but not limited to cyan, yellow, and magenta, and concurrently possess an average absorption of 70% for near-infrared light. Compared with the low RI MgF₂-based color filter, the designed Si-based color filter has the advantage of excellent angle insensitivity, even the maximum peak shift at the incident angle of 60° is only 1.8%. Compared with the color filters composed of traditional metal Ag or Al, the proposed Ti-based color filter has obvious advantages in saturation, color gamut, especially near-infrared absorption. In addition, high-temperature tests further verifies that the color filter is suitable for extreme environments.