Primary and Secondary Reflective Color Realized by Full‐Solution‐Processed Multi‐Layer Structures

A full solution‐based method is reported to fabricate multilayer thin film stacks for structural color applications. This is in contrast to the conventional fabrication methods that require high vacuum processes such as physical vapor deposition and magnetron sputtering, which significantly limits their practical use due to the high cost and long processing time. Copper/silicon dioxide/copper (Cu/SiO2/Cu) and copper/titanium dioxide/copper (Cu/TiO2/Cu) are chosen as the model system due to their simple structure and wide color tunability. A systematic investigation is carried out for each layer to ensure good film quality as well as its compatibility with all previous layers. Especially, a particulate Cu layer having optical properties distinct from that of bulk is obtained through such solution deposition, where the refractive index of Cu can be continuously tuned with deposition time. Both primary and secondary colors are achieved with a continuous manipulation of both the dielectric thickness and top Cu morphology on different substrates (e.g., silicon, glass, plastics, etc.).


Introduction
Color pigments of all types have been used extensively in our daily life. In contrast to the natural dye pigments that usually suffer from long-term degradation, artificial structural colors made of inorganic elements are often much more durable, robust, and environmental-friendly. However, conventional fabrication of structural colors either relies heavily on vacuum-based deposition techniques [1][2][3][4] or involves complicated patterning procedures, [5][6][7][8][9] which greatly limits their applications for largescale and cost-effective production. Therefore, much effort has DOI: 10.1002/adom.202300456 been devoted to alternative, solution-based processes. Though photonic crystals synthesized through nanoparticles [10][11][12][13][14] or block-copolymer self-assembly [15][16][17][18] with various colors have been extensively explored, their color properties are largely compromised when being ground into micro/nano-flakes. Besides, these photonic crystals usually take several to hundreds of microns thickness in order to achieve a high color brilliancy. In comparison, simple layered structures such as metal-dielectricmetal (MDM) structure and its variant [19][20][21][22] could easily give strong coloration and only require thickness within several hundreds of nanometers. Reflection color is produced when light at a certain wavelength resonates inside the thin film cavity and gets absorbed, giving a substrative color (i.e., secondary color). Though the optical properties of such simple tri-layer structure have been widely studied in the past decades, only quite recently [23,24] have researchers realized the usage of nanoparticles as a top broadband absorbing layer, where primary reflection peak can be achieved. Since, almost all current fabrication strategies involve some level of vacuum-based deposition, a more cost-effective way of multilayer film fabrication is preferred. Previously, our group has reported the first effort of utilizing electrochemical deposition to fabricate a layered gold (Au)/cuprous oxide (Cu 2 O)/Au structural color. [25] Both film thickness and roughness are well controlled with the film nucleation and growth rate. However, in the electrochemical deposition process, a conductive substrate is required for each add-on layer, and the cost of gold is also a big concern. In this work, we focused on developing a general solution-based strategy for MDM structure fabrication for both reflective primary and secondary color on various substrates, and do not require conductive substrate that was needed for electrochemical deposition. The optical properties of the metal layer due to different morphology can also be tuned by the deposition process.
In our experiment, inexpensive copper (Cu) was chosen as the top and bottom metals and was electroless-deposited on substrates. Full solution-processed dielectrics with different refractive indices, such as silicon dioxide (SiO 2 ) or titanium dioxide (TiO 2 ) from sol-gel process were chosen as the middle dielectric layer to tune different colors. Appropriate deposition condition for each layer has been successfully investigated with an aim of not only good morphology or thickness control, but also chemical compatibility with all previous layers. All depositions were www.advancedsciencenews.com www.advopticalmat.de Scheme 1. An overview of full solution process of MDM structural color fabrication.
driven with the inherent chemical potentials of the solution, without any external power (i.e., electric, photonic, mechanic, etc.) input. Different substrates including silicon wafer, glass, polyethylene terephthalate (PET), acrylonitrile butadiene styrene (ABS) were tested to demonstrate the generality of our full-solutionbased method. Color coatings from orange to cyan were produced by varying the thickness of the dielectric layer. Angle-dependent spectra responses were also observed for SiO 2 -based samples and gave the sample an iridescent appearance due to the low refractive index of SiO 2 . This fabrication strategy can be extended to the deposition of other metals (e.g., nickel (Ni), silver (Ag), etc.) and dielectrics (zirconium dioxide (ZrO 2 ), etc.) as well, which greatly enriches the material library. We also expect that such facile and cost-effective fabrication can be easily scaled up for mass production and leads to an extensive usage of structural color in diverse fields such as color displays, cosmetics, solar cells, aesthetic decorations, etc

Results and Discussion
Scheme 1 gives the MDM structure fabrication overview, with each layer being deposited additively. The bottom Cu layer was deposited through a Cu electroless deposition process. Silicon or glass substrate underwent a silanization [26,27] step with 3aminopropyltrimethoxysilane (APTMS) to ensure an amine terminated surface, which serves to anchor the catalyst particles. For plastic substrate such as ABS, PET, etc, where silanization is not possible, stearyltrimethylammonium chloride (SC) [28] was adopted for surface treatment to help with palladium (Pd) nanocolloids adsorption due to the positively-charged amine group. The treated substrate was then immersed in a presynthesized Pd nanocolloidal solution for Pd absorption. The autocatalytic Cu electroless deposition process was then carried out on the silicon surface, where Cu complex was reduced by formaldehyde: [29] [Cu (II) − Tar] + 2HCHO + 2 OH − → 2 HCOOH+H 2 where Tar [2]− is the tartrate anion. A strong basic solution (pH > 11.5) is typically required since the formation of methylene glyco-late anion is favored at high pH. The Cu color gradually appeared on the top of the silicon substrate, indicating an increasing Cu layer thickness with immersing time. As shown in Figure 1a, the Cu layer grows at the rate of 0.27 nm s −1 , following a nucleation and growth mechanism ( Figure 1c). The Pd nanocolloids serve as an autocatalytic center as well as a nucleation center, where Cu nanoparticles start to form and merge as the deposition goes on. Such mechanism could be clearly seen from the scanning electron microscopy (SEM) images ( Figure 1c) as well as the refractive index changes with increasing Cu thickness ( Figure S8, Supporting Information). Since the bottom copper serves as a good reflector, we adopted a longer deposition time of 120 s for ≈40 nm thick Cu where the reflection spectra hardly change thereafter. The resulted shiny Cu film has a root-mean-square (RMS) surface roughness of 5.8 nm ( Figure S1a, Supporting Information), which ensures the smoothness and uniformity for the bottom reflector in the MDM structure. The second dielectric layer controls the resonance wavelength, which is used to tune the color appearance. Hence, a smooth dielectric layer with good thickness control is highly desired. A typical sol-gel process [30][31][32] was carried out, where tetraethyl orthosilicate (TEOS) and titanium isopropoxide (TTIP) solution were first prepared to ensure hydrolysis (Equations S1 and S3, Supporting Information). The pH of solution was carefully tuned to 3 for SiO 2 coating to maximize TEOS hydrolysis kinetics while minimizing bulk condensation. The substrate deposited with Cu was then immersed into the solution for several minutes followed by widthdrawing the substrate from the TEOS solution with a constant speed. As the solvent quickly evaporates upon withdrawing, condensation reactions (Equations S2 and S4, Supporting Information) take place among the pre-hydrolyzed precursors, forming a thin coated oxide layer. The thickness can be controlled by either varying the precursor concentration or changing the withdrawing rate. A linear concentrationdependent thickness increase was observed (Figure 2a,b) with TEOS concentration increasing from 14.8 to 35.6 wt.% and TTIP concentration from 13.1 to 34.1wt.% at a constant withdrawing rate of 200 μm s −1 . Accordingly, the coated samples appeared to show a color from pale orange to pale yellow with a negligible dip across the spectra ( Figures S2 and S3, Supporting Information). Another facile way to tune the dielectric film is investigated through the control of withdrawing rate. As shown by Landau and  Levich [33] and Faustini, [34] a change in the withdraw rate leads to various deposition layer thickness: where U is the withdraw rate, E, L, and D stands for the solvent evaporation rate, width of the film, and physicochemical constant of the solution, respectively. In Figure 2c and 2d, we have observed a similar relation between the deposited dielectric thickness and the withdrawing rate from 100 μm s −1 to 500 μm s −1 , as shown by the fitted curves. The photos in Figure S12 (Supporting Information) show samples of several cm 2 area, limited by the size of the beaker used in our experiment. Scale bar in the pictures is 1 cm. Withdrawing is in the vertical direction for both of the Cu/SiO 2 /Cu (left) and Cu/TiO 2 /Cu (right) samples. The color difference at the edge of the sample is due to the faster evaporation of solvent. The bottom edge color variation is caused by the remaining meniscus liquid on the sample surface when the substrate leaves the solution. Note that there is no fundamental limitation on color uniformity with the dip coating process as long as rigorous engineering controls are employed.
The resulted SiO 2 and TiO 2 films have refractive indices of 1.45 and 1.81 @550 nm by ellipsometry measurement, respectively ( Figure S4, Supporting Information), which are lower than those vacuum deposited ones. We attributed the reduction of the refractive index to the low density of the solution-deposited materials, which is likely due to nanopores generated within the film during the condensation reaction. A very small RMS surface roughness of 1.8 nm (Figure S1b, Supporting Information) was obtained across the sample surface except for the drying front where a meniscus formed once the substrate left the solution. The smoothly and uniformly coated dielectric layer provides a solid foundation for the top Cu layer coating as well as ensures a good overall color performance.
A similar repetition of the top Cu layer plating was carried out after silanization of the sol-gel deposited dielectric layer and Pd nanocolloids activation. However, Cu plating recipe developed for the bottom Cu layer could not be applied due to the residue hydroxyl groups left in the dielectric layer, which gives a local high basic environment (pH > 11.5). Because such high pH would further trigger the condensation reaction inside the dielectric layer, leading to an internal stress build-up across the film. The stressed film can crack and delaminates from the substrate. Hence, a milder Cu plating recipe was developed to work under neutral pH [35][36][37] with minimum heating. Instead of the formation of methylene glycolate at high pH with formaldehyde, copper-boron (Cu-B) phase with dimethylamine-borane (DMAB) is more favorable for Cu plating at lower pH value.
The growth of the top copper layer also follows a nucleation and growth mechanism (Figure 1c and Figure S7, Supporting Information), where the refractive index changes with increasing deposition time. Such tunable property can lead to new colors that is not possible with the conventional MDM structure. As shown in Figure 3a, a decrease in the real part of the refractive index was observed with time, while the imaginary part shows an opposite trend, especially in the red end of the spectrum. We noticed a moderate imaginary part of the refractive index leads to a lossy metal, i.e., a too large imaginary part gives too much reflection (like a shiny metal) but very limited absorption, while a too small imaginary part leads negligible absorption across a thin film. Therefore, 1 min electroless-deposited copper was chosen as the top absorbing layer due to its lossy nature in the spectra simulation with various dielectric layer (i.e., SiO 2 ) thicknesses ( Figure 3b). It can be seen that a broadband absorber showed up with thin dielectric layer thickness followed by a reflection peak (Figure 3d inset) from blue to red end when the resonance builds up. In comparison to a continuous metal (i.e., 3 min Cu electroless deposition, Figure 3c) where a typical secondary color is observed, the color gamut is greatly expanded (Figure 3d).
As extracted from ellipsometry data, the discontinued copper layer showed a very distinctive refractive index than the bulk copper, where the imaginary part k is greatly reduced in the long wavelength range (> 600 nm). Especially, the reduction in the k values makes the Cu extremely lossy toward longer wavelength and results in a broadband absorption above 600 nm (Figure 3a, blue curve). This is in contrast with bulk Cu, which is highly reflective toward NIR wavelength range. The sample gave a reflective blue color with only 30 s for top Cu layer plating due to this stronger absorption in red, followed by a dark purple color at 1 min (Figure 4a). These colors are not obtainable for a typical MDM structure and offer great potential in expanding the possible color gamut of the simple tri-layered structures. Prominent peaks show up when the dielectric layer becomes thicker. This can be understood as the particulate Cu morphology broadens both the first-order and second-order resonance at longer and shorter wavelengths respectively. Figure 4d shows the non-trivial color generated with 1 min Cu electroless deposition on top of the SiO 2 dielectric surface, giving reflection peaks across the visible spectra (Figure 4e, black arrows indicate the peak position).
Here, we treated the discontineoues Cu as effect medium [38,39] made of Cu islands filled with air. As it is hard to distinguish the host and inclusion for the discontinuous Cu layer, we adopted a Bruggeman model with appropriate Cu filling fraction to simulate the spectra of these non-conventional colors, which matches the measured spectra accurately (Figure 4b,c). The simulated reflection spectra (Figure 4f) match fairly well with the experimental results with the top Cu layer refractive index extracted from ellipsometry.
Secondary (or subtractive) colors can be obtained with longer deposition time (> 3 min) for a continuous Cu layer with ≈25 nm thickness where the structure behaves like that of a typical MDM. The cross-section SEM image (Figure 5a) clearly showed the metal-dielectric-metal stack in a layer-by-layer fashion. Colors (Figure 5b and Figure S9, Supporting Information) from orange to cyan (depends on dielectric layer thickness) were observed immediately after taking the substrate out of the plating bath. A red shift in the resonant wavelength in the reflection spectra (Figure 5c-d and Figure S10a  size or the uniformity of the fabricated sample as long as rigorous engineering controls are employed. A potential continuous fabrication of large films could also be achieved with a roll-to-roll setup. We also showed that this MDM structure can be coated on various substrates ( Figure S6) including glass, ABS plastics, and PET films, to name a few. The ability to coat either a high index dielectric or low index dielectric gave us more flexibility in tuning the color performance in practical applications It is interesting to note that the Cu/SiO 2 /Cu samples demonstrated an iridescent color (Figure 6a) by changing the viewing angles, while the color change in the Cu/TiO 2 /Cu samples was less noticeable. As clearly shown in Figure 6b-d, the resonance wavelength blue shifted for all samples with SiO 2 as the middle dielectric. The angle-dependent color can be understood with Bragg's law [40] (assuming air as the incident medium), where the resonance wavelength of the cavity depends on the incident angle : where n, d, and N are the refractive index of the cavity material, cavity length and an integer, respectively. It is also easy to note that an increase in the refractive index could reduce the angle-dependent feature: where A = − 2Ndsin( )cos( ) and thus, the color of the TiO 2 samples ( Figure S5, Supporting Information) is much less anglesensitive than that of the SiO 2 .

Conclusion
In summary, we have developed a full-solution-based additive approach toward both primary and secondary reflective structural color fabrications. The novel fabrication process at ambient condition with minimum heating (50˚C for faster top copper layer deposition rate) distinguish our work from all previous works with vacuum-based deposition technologies. An elaborate tuning of the metal and dielectric layer together gives colors from orange to cyan with a good control of metal layer morphology and dielectric layer thickness. The color angle-sensitivity could also be easily tuned by switching the dielectric between SiO 2 and TiO 2 under an almost identical fabrication process. Other metals (e.g., Ni or Ag) and dielectrics (e.g., ZrO 2 ) can be deposited in similar manner, which enriches the material library. Colors beyond the conven-  tional MDM color gamut were achieved by carefully tuning the top Cu layer morphology, where an effective layer plays an important role in modifying the imaginary part of the refractive index. Further pigmentation and packaging strategies (including antiscratching coating) will be implemented in the future for scaleup productions but is out of the scope of this work. Taking advantage of the electroless deposition and dip-coating techniques, we expect the approach presented here can be scaled up to a continuous fabrication in the near future. All these aspects help to reduce the cost and accelerate the implementation of MDM structural colors for practical applications.
Substrate Preparation: All substrates (i.e., Si wafer, glass, PET, ABS) were sequentially sonicated for 5 min in acetone, methanol, isopropanol, and water before any surface treatment. Si wafer and glass substrate were immersed into a 10 wt.% APTMS solution (ethanol as solvent) for 1 h, followed by a rinse with ethanol. PET and ABS substrates were immersed into a 0.1 wt.% SC solution for 1 min followed by a rinse with water. The pretreated substrates were then immersed into the Pd nanocolloidal solution for 10 min for surface activation followed by rinsing with water. The Pd nanocolloidal solution was prepared as follows: a) dissolve 1 g of sucrose and 10 mg of dextran (MW = 200000) in 92.5 mL water as solution A; b) solution B contains 20 mM PdCl 2 and 100 mM NaCl; c) solution C contains 40 mM NaBH 4 ; d) then add 2.5 mL solution B and 5 mL solution C into solution A followed with heating at 80°C for 24 h. The solution was then cooled down to room temperature and stored as the Pd nanocolloidal solution. Note the nanocolloidal solution should be good to use for at least 6 months with proper handling.
Deposition of the Bottom Cu Layer: The electroless plating bath consist of three parts: solution A contains 2 wt.% CuSO 4 ·5H 2 O; solution B contains 26 wt.% potassium sodium tartrate dihydrate and 8.5 wt.% NaOH; solution C contains 37 wt.% formaldehyde solution. The plating bath was then mixed with equal weight of solution A, B, and C in sequential order. The Pd nanocolloid activated substrate was then immersed into the plating bath for 120 s followed by a rinse with water. A shiny Cu appearance could be observed.
Deposition of the Dielectric Layer-SiO 2 Deposition Solution: TEOS solution was prepared by mixing 35.5 mL ethanol, 5 mL water, TEOS, and MTES in order. 0.1 M HCl was used to the solution pH to 3 for hydrolysis. The amount of TEOS added increased gradually from 6.4 mL to 22.4 mL with 3.2 mL interval, marked as #1 to #6 (Table S1, Supporting Information). The MTES was added in proportion to TEOS with a volume ratio of 4:1 (TEOS:MTES). The Cu coated substrate was then immersed into the TEOS solution for several minutes and withdraw at a constant rate of 200 μm s −1 using a linear actuator (Z6254B 25 mm motorized actuator, Thorlabs Inc.). Various of withdrawa rates from 100 μm s −1 to 500 μm s −1 were also carried out with recipe #4.
Deposition of the Dielectric Layer-TiO 2 Deposition Solution: The TTIP solution (Table S2, Supporting Information) was first obtained by mixing 30 mL ethanol, AcAc, and TTIP for 30 min. The amount of TTIP was increased from 4.44 mL to 17.76 mL with 4.44 ml interval, marked as #7 to #10. The AcAc was added in proportion to TEOS with a volume ratio of 2.9:1 (TTIP:AcAc). Then 1.2, 2.4, 3.6, and 4.8 mL water was added dropwise to the TTIP solution #7 to #10 respectively. The final solution was stirred for another hour to ensure hydrolysis. The Cu-coated substrate was then immersed into the TTIP solution for several minutes and withdrawn at a constant rate of 200 μm s −1 using a linear actuator. Various of withdraw rates from 100 μm s −1 to 500 μm s −1 were also carried out with recipe #9.
Deposition of the Top Cu Layer: The top Cu layer plating bath contained 50 mM CuCl 2 , 50 mM EDTA-2Na, 0.1 m boric acid and 0.9 m NaOH and tuned to neutral pH with 1 m NaOH. The plating solution was then heated to 50°C with the addition of DMAB (0.4 g/50 mL plating bath). The dielectric/Cu coated substrate was then immersed in solution for 3 min and an MDM structural color sample was obtained (Z6254B 25 mm motorized actuator, Thorlabs Inc.).
Optical Simulation: Simulated reflectance spectra were calculated based on the transfer matrix method. Refractive index of each layer was extracted from spectroscopic ellipsometer (M-2000, J. A. Woollam Co.) as inputs.
MDM Film Characterization: Reflectance spectra at normal incidence were measured using a thin-film measurement instrument integrated with a spectrometer (HR4000CG-UV-NIR, Ocean Insight) and a white halogen light source (HL-2000-FHSA, Ocean Insight). Grazing incidence XRD was taken with Rigaku Smartlab XRD from 0-5˚for the dielectrics. Angle-resolved reflectance spectra of as-prepared MDM stacks were measured with an unpolarized light source with the UV-Vis-NIR spectrometer (Lamda 1050, PerkinElmer Inc.) and determination of the thickness and refractive index of each layer were performed with a spectroscopic ellipsometer (M-2000, J. A. Woollam Co.). Cross-sectional SEM was performed with an FE-SEM (TFS Helios 650 Nanolab SEM/FIB) with a Schottky field emitter operated at 5 keV beam voltage. The surface morphology of each layer was also investigated by tapping mode AFM (TESPA-V2 tip and Dimension icon AFM, Bruker Corporation).

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.