Ultrathin sputter-deposited plasmonic silver nanostructures

In this study, ultrathin silver plasmonic nanostructures are fabricated by sputter deposition on substrates patterned by nanoimprint lithography, without additional lift-off processes. Detailed investigation of silver growth on different substrates results in a structured, defect-free silver film with thickness down to 6 nm, deposited on a thin layer of doped zinc oxide. Variation of the aspect ratio of the nanostructure reduces grain formation at the flanks, allowing for well-separated disk and hole arrays, even though conventional magnetron sputtering is less directional than evaporation. The resulting disk–hole array features high average transmittance in the visible range of 71% and a strong plasmonic dipole resonance in the near-infrared region. It is shown that the ultrathin Ag film exhibits even lower optical losses in the NIR range compared to known bulk optical properties. The presented FDTD simulations agree well with experimental spectra and show that for defect-free, ultrathin Ag nanostructures, bulk optical properties of Ag are sufficient for a reliable simulation-based design.


Introduction
The scientic research in ultrathin metals is mainly driven by applications as exible transparent electrodes, combining high transparency, low sheet resistance and mechanical stability. Typical metals employed are gold (Au), copper (Cu) and silver (Ag), which are deposited by sputtering and evaporation techniques. 1 In terms of optical properties, Ag is considered by far the best choice owing to the low optical losses in the visible and near infrared region. 2 However, due to the high free surface energy of silver, thin lm growth starts with the nucleation of discrete silver islands, 3 which hampers the fabrication of extremely thin continuous lms. Various strategies have been developed to lower the threshold thickness for a continuous lm, such as the deposition on top of dielectric buffer layers, 4 use of metal seeding layers, 5 surfactants 6 and metal doping. 7 For instance, continuous Ag layers with thickness as low as 6 nm have been realized by a partially oxidized Ag wetting layer 8 or by oxygen doping of Ag on ZnO-coated substrates. 9 Ag lms doped with Cu showed continuous lm formation at 6 nm on a glass substrate. 10 Ag lms doped with Al have been shown to form continuous 7 nm-thick lms on a fused silica substrate, however optical characterization revealed signicantly higher optical losses compared to pure Ag layers. 11 The simulation-based design of nanostructures incorporating an ultrathin metal requires the precise knowledge of the optical properties of the metal. The optical properties of ultrathin Ag lms can vary strongly from bulk optical properties. 12,13 The reason for deviations is commonly found to be the noncontinuity of the metal and therefore the excitation of localized surface plasmon resonances that occur at grains or in voids. In addition, surface roughness of otherwise continuous lms is suggested to alter the optical properties of thin lms, 14,15 as the surface scattering of conduction electrons leads to increased losses compared to the bulk. 16 More fundamentally, band calculations with density functional theory for thin Au lms predicted a signicant change of optical properties with signicant anisotropy arising. 17,18 However, the nanostructuring of ultrathin metals holds potential for transparent electrodes with high transmittance or for introduction of additional functionalities, such as surfaceplasmon-induced lter properties. Examples of nanostructured ultrathin metals in the literature include a 14 nm thick Au nanohole array 19 and a one dimensional plasmonic colour lter with 9 nm Ag on a 1 nm Al seed layer. 20 Let us note that the two investigated geometries exhibited a signicant difference between simulation and experiment. In another example, the deposition of a continuous, 8 nm-thick Ag on a periodic, nanodome polymer structure, pretreated with oxygen plasma, resulted in a transparent electrode with high transmittance, leading to increased power conversion efficiency when applied in organic solar cells. 21 Precise patterning of periodic plasmonic nanostructures can be realized by various lithography techniques, such as electron beam lithography (EBL), focused ion beam (FIB) milling and nanoimprint lithography (NIL). EBL and FIB are very elaborate and costly fabrication techniques and are only employed for small prototype test structures, while NIL allows for large-scale realization of nanostructures. Nevertheless, NIL requires additional etching and li-off steps when isolated periodic metallic nanoparticles are required. An alternative route for NILfabricated metallic nanostructures is the directional metal deposition on an NIL-structured resist without further li-off or etching. In case the imprinted resist is a periodic disk or hole array, metal deposition results in the formation of nanodisks, separated from a perforated metal hole array by a metal-free gap. These kinds of structures have been investigated as plasmonic colour lters, 22,23 in plasmon-enhanced uorescence, 24 as well as for plasmonic biosensors. 25,26 The present work aims to combine the aforementioned aspects of simple nanostructure fabrication by NIL (without further li-off or etching steps) with ultrathin Ag deposition to form plasmonic disk-hole arrays with high visible transmission and strong plasmonic resonance in the near-infrared range.
Therefore, thin lm growth by conventional magnetron sputter deposition on different planar and structured substrate congurations is studied to obtain defect-free Ag lms with thickness below 10 nm. An ultrathin disk-hole array is realized without any additional li-off steps, enabling an easy-toproduce, highly transmissive silver plasmonic nanostructure.
Numerical simulations of planar and patterned designs are used to analyse the optical properties of ultrathin Ag and to discuss localized plasmonic modes present in the disk-hole array.

Experimental
Thin lm deposition All AZO (Al-doped ZnO) and Ag lms are prepared using a direct current (DC) magnetron sputtering system (Leybold Univex 450C, 3SC Leybold, Germany) with a base pressure of 1.9-7.0 Â 10 À8 mbar. AZO is deposited from a 4 inch target of zinc oxide (ZnO) with 2 wt% aluminum oxide (Al 2 O 3 ) in a pure argon (Ar) atmosphere at 1.0 Â 10 À3 mbar and with a sputter power of 60 W, resulting in a sputter rate of 0.294 nm s À1 . Ag is deposited in pure Ar at a pressure of 2.0 Â 10 À3 mbar from a 4 inch target of Ag at 120 W sputter power, yielding a rate of 1.809 nm s À1 .

Fabrication of the nanostructures
The nanostructures were realized by nanoimprint lithography. A silicon (Si) master structure was obtained from EULITHA AG (Switzerland) having periodic pillars with 90 nm in height, 120 nm in diameter and a period of 200 nm. The master was silanized with trichloro(1H,1H,2H,2H-peruorooctyl)silane (448931, Sigma-Aldrich, Austria) to reduce adhesion of the NILresist and ensure complete delamination. Subsequently, a drop of 1 ml of UV-resist (OrmoStamp®, Micro Resist Technology GmbH, Germany) was placed on the master and a cleaned sodalime glass substrate (25 mm Â 25 mm) was placed on top without applying additional pressure. UV-curing of the resist was done with a UV lamp (Polylux500, Dreve Optoplastik GmbH, Germany) with a wavelength range of 315-400 nm for 5 min, followed by a 30 min hard-bake on a hotplate at 150 C. The resulting structure is an array of cylindrical holes. Aerwards the structure was gradually altered in height and diameter using Ar ion beam etching (IBE) (IonSys 500, Roth & Rau AG, Germany) at a beam-to-substrate-angle of 40 for different etching times of 0 min to 3 min (at 500 V accelerator voltage and microwave power of 375 W). Subsequent thin lm sputter deposition of AZO and Ag (as described before) resulted in the formation of a disk-hole array without post-deposition li-off or etching steps.

Morphological and optical characterization
The nanostructures were imaged with an atomic force microscope (AFM) (Molecular Imaging PicoPlus) in tapping mode, using SSS-NCHR-10 (Nano world AG) tips. The surface morphology and metal growth were studied using scanning electron microscopy (SEM) (Zeiss, SUPRA 40) at 5 kV acceleration voltage with an in-lens detector. Optical transmittance and reectance spectra were obtained using a Fourier transform infrared spectrometer (FTIR) (Vertex 70, Bruker Corporation), equipped with an additional visible light source. All measurements were performed with unpolarized light. Transmittance was measured at normal incidence and referenced to air, while reectance was determined at an incidence angle of 13 from the coated side of the sample and referenced to a calibrated mirror (STAN-SSH-NIST, Ocean Optics).

Simulation
Simulations of planar thin lms were performed by the transfer matrix method (TMM). 27,28 The complex refractive index of AZO was taken from the literature. 29 For the Ag thin lm, data from three frequently used sources, namely Palik, 30 Johnson and Christy 31 and Rakic et al., 32 were compared with experimental values to nd the best tting data for the complex refractive index.
TMM simulations of transmittance were performed at normal incident light, while reectance was calculated at 13 to match experimental conditions. The complex permittivity of 6 nm thick Ag lm on 10 nm thick AZO lm was retrieved from transmittance and reectance measurements in combination with a TMM algorithm. 33,34 Simulations of periodic nanostructures were performed by the nite-difference time-domain (FDTD) method (FDTD Solutions from Lumerical Solutions, Inc., Canada). A single unit cell of the nanostructure with periodic boundary conditions in the x-and y-direction and perfectly matched layers (PMLs) in the zdirection was used for simulation of transmittance and reectance for normal incident light. The unit cell height was xed to 500 nm in the z-direction and 16 layers of PMLs were used. A broadband plane wave light source was used and is incident from the air-side. The optical properties of AZO were taken from the literature 29 and the complex permittivity of Ag was taken from Rakic et al. 32 The glass/OrmoStamp substrate was assumed with constant refractive index n ¼ 1.516 over the complete spectral range and considered innite in the minus z-direction.
The surface charge density at the Ag/air interface was calculated from the divergence of the electric eld at the respective spectral resonance position. A uniform 1 nm-mesh in the x-, y-and zdirection was employed for all simulations. Mesh convergence was checked by an additional simulation with a 0.5 nm uniform mesh, where no additional improvement of the spectrum was achieved. For simulation of angle-dependent transmittance for different polarizations, periodic boundary conditions in the xdirection were replaced by Bloch-periodic boundary conditions.
The uncoated air/glass interface, which is otherwise not considered in the TMM and FDTD simulations, due to practical reasons, was taken into account using simple analytical equations. 15 This allows for a direct comparison between experiment and simulation.

Results and discussion
The realization of ultrathin lms is oen restricted by dewetting of the deposited lms. Due to the high free surface energy of silver, thin lm growth starts with the nucleation of discrete silver islands followed by cluster coalescence until the percolation threshold is reached and nally a continuous lm is formed. 35 The progress of nucleation, island growth and coalescence strongly depends on the surface and interface energy of the substrate affecting the wetting properties, adhesion strength and cluster mobility of the adsorbed silver. 3 With decreasing energy difference between the Ag and the substrate, the Ag atoms become more strongly bound to the substrate than to each other, yielding enhanced wetting properties and a more stable lm. 35 Since the morphology of the metal lm strongly inuences the optical properties, 10,36 an investigation of Ag growth on different planar substrate congurations, namely soda-lime glass, NIL-resist (OrmoStamp) and 10 nm AZO layer, is performed prior to implementation in nanostructures.
In Fig. 1, the evolution of ultrathin Ag layers on these substrates is presented. While the deposition on bare glass offers a comparison with other studies, 4 the Ag growth on the planar NIL-resist (OrmoStamp) is essential to investigate for the following nanostructuring process. ZnO has been used as a supporting material for improved wetting, 37-39 hence the introduction of the AZO layer is expected to change the adhesion properties of Ag while maintaining the transparency of the substrate. All Ag layers are deposited under the same sputter conditions, where a high sputter power of 120 W was chosen to support continuous lm growth even at low thicknesses. 40 It should be noted that there is no difference in Ag growth irrespective of whether the AZO layer was deposited on the planar resist or directly on glass (see Fig. S1 †).
4 nm Ag samples on glass and resist show small isolated islands, while the Ag lm on AZO is already coalesced. The increased wettability of AZO with respect to glass cannot be explained by higher substrate surface energy (for contact angle Nanoscale Advances measurement data see Table S1 in the ESI †). Instead, the chemical bonding of Ag atoms to the underlying AZO surface as well as a benecial crystallographic arrangement is considered as a contributing factor for the improvement in Ag wetting. 35 The polycrystalline AZO lm with a prominent (002) wurtzite peak and a weaker (013) peak in the XRD pattern (see Fig. S2 †) may promote the stable arrangement of Ag with (111) and (202) orientation, compared to the amorphous glass substrate. When increasing the Ag thickness to 6 nm, the islands on glass and the resist start to coalesce, whereas on AZO the lm is already continuous, exhibiting only a few occasional holes. Further increase of the Ag thickness to 8 nm yields a defect-free lm on AZO and a lm with sporadic holes on glass that is similar to the 6 nm Ag layer on AZO. The 8 nm thick Ag layer on the resist has enlarged meandering gaps compared to 6 nm Ag, showing the tendency of Ag migration and high surface mobility. Therefore, Ag adatoms on the resist coalesce with each other rather than sticking on the substrate and consequently the lling of voids is delayed. This behaviour is conrmed by contact angle measurements, where the planar resist exhibits a signicantly higher contact angle (CA) than AZO or glass (see Table S1 †). The signicantly improved adhesion of Ag on AZO, compared to its deposition on the resist, makes AZO a necessary base layer for the realization of an ultrathin Ag lm on nanostructures. Experimental transmittance and reectance spectra of 6 nm Ag on top of 10 nm AZO layer are shown in Fig. 2(a). TMM simulations of the layer system were performed with 3 frequently used Ag dispersion relations from the literature, while the optical properties of AZO were taken from a previous study. 29 Even though the ultrathin Ag layer has occasional holes, which are not considered in the simulation, a good agreement with simulation and experiment is observed. The good agreement motivated us to calculate the complex electrical permittivity, despite the defects in the lm. The comparison of the derived complex permittivity of the 6 nm Ag and the literature values is given in Fig. 2(b). The real part of the experimental permittivity agrees best with the Ag permittivity from Rakic et al.  the near-infrared with high plasmonic quality factors. 2 To our knowledge no other study has shown so good agreement between ultrathin Ag lm below 10 nm and the bulk optical properties. Fig. 3(a) schematically presents the conguration of the nanostructures investigated in the following, where the NILresist is patterned with cylindrical holes and AZO and Ag are subsequently deposited. The sketch represents an ideal situation, where no material is adsorbed on the sidewalls, resulting in a perforated Ag lm raised above an array of Ag disks. Directional metal deposition techniques, typically applied for such nanostructures without a li-off step, are thermal-and egun evaporation in ultrahigh vacuum. Nevertheless, even directional deposition techniques cannot avoid the formation of metallic grains on geometrically shadowed regions due to metal diffusion from adjacent layers. 24,42,43 These grains can interact with local electric elds of excited surface plasmons and alter the optical properties of plasmonic structures. Consequently, detailed investigation of grain formation is important. In contrast to highly directional deposition techniques, magnetron sputtering exhibits a cosine emission prole and higher collision rates in the gas phase. 44 Therefore the incident Ag particles have a noticeable velocity component parallel to the surface and are thus able to directly adsorb on the sidewalls. However, the narrow holes lead to shadowing effects reducing the deposition rate on the sidewalls as well as on the bottom disks. Fig. 3(b) shows SEM-images presenting the Ag growth on structured substrates with a 10 nm AZO base layer and varying Ag thicknesses. At 6 nm thickness the deposited material on the sidewalls grows as small, separated grains, similar to the lm growth on planar substrates at very small thicknesses. There is a clear separation of the disk array and the hole array that originates from substantial shadowing, which amplies perturbations in the surface 45 as well as surface energy minimization by dewetting the corners. For increasing Ag thickness, the grain size on the sidewalls increases continuously. While the disks are still completely separated at 10 nm Ag thickness, the grains start to connect with the disks on the bottom, as well as with the hole array at 15 nm Ag thickness. Further increase to 20 nm Ag results in almost complete stepcoverage and hence electrical connection of the disks and holes. Consequently, the excitation of localized surface plasmons in the disks and holes will be less pronounced. Therefore, a small Ag thickness is not only benecial for high optical transparency but also essential for the exploitation of localized plasmonic resonances, ensuring the separation of disk and hole arrays.
Another interesting feature arises from the comparison of the structured AZO 10nm /Ag 6nm sample with the same layer conguration on a planar substrate (see Fig. 1(b)). While the planar 6 nm Ag lm is still disrupted by occasional holes, the perforated hole array lm, as well as the disks themselves is free of these defects (see Fig. S3 †). It is noted here that the holes appear on the planar substrate regardless of whether the deposition is done on plain glass or on glass covered with the resist (Fig. S1 †). Although contact angle measurement data (see  S1 †) show different wetting properties for the planar and structured samples, it cannot explain the improved Ag growth on the structured sample. The higher contact angles on the structured congurations probably come from macroscopic dewetting effects due to the nanostructured surface, where the drop does not cover the same effective surface area as in the planar conguration. Another possibility is that the distance that Ag adatoms (or clusters) may diffuse is modied by the spatial restrictions posed by the nanostructured substrate, with a smaller diffusion length leading to faster island percolation. The reported large diffusivity of Ag on various weakly interacting substrates supports this argument. 46,47 In any case, it seems that the silver growth on AZO is improved by the presence of the structure, making stable nanostructured Ag lms as thin as 6 nm possible. The formation of Ag grains on the side walls of the nanostructures was further investigated by change of the geometry of the nanostructured resist, prior to deposition of 10 nm AZO and 6 nm Ag. The geometry parameters of the nanostructured resist were changed by IBE with different exposure times (see Fig. 4 with 0 min, 1 min, 2 min and 3 min etching time from the top to bottom row, respectively). The beam-to-substrate angle is chosen in a way that the diameter and the height of the cylindrical holes are increased and decreased, respectively. Fig. 4(a) shows the SEM images of the coated structure with different aspect ratios (AR ¼ height/hole diameter) for different etching times. With decreasing aspect ratio, the number of grains is signicantly reduced while the separation of disks and holes is clearly maintained or even improved. At the lowest aspect ratio nearly all grains are merged with the upper rim of the hole array. The corresponding experimental optical spectra are shown in Fig. 4(b) and are compared to FDTD simulations (with Ag bulk properties taken from Rakic et al.) of a disk-hole array without consideration of the grains. The geometrical parameters for the simulation are deduced from SEM normal view images (disk diameter and hole diameter) and from AFM measurements (height of the nanostructure, data not shown) and are indicated in the insets. The spectra show a multitude of optical modes which manifest as dips in transmittance and peaks in reectance and absorption. The comparison of simulated and experimental spectra reveals that with decreasing AR a better agreement between experiment and simulation is achieved. A possible explanation for this behaviour is the drastically reduced number of grains for lower ARs, which are not accounted for in the Fig. 4 (a) SEM images at normal and 30 tilted view for different aspect ratios (ARs) of the disk-hole array with 10 nm AZO and 6 nm Ag. (b) Experimental (blue) and FDTD-simulated (red) transmittance, reflectance and absorption spectra of the corresponding structures. Geometry parameters used for the simulations are indicated by the schematic insets. The grey solid line in the last row represents the experimental spectra of the planar structure with same layer thicknesses.
simulations. Especially at the highest AR of 0.82 the deviation between simulation and experiment is signicant in the complete wavelength range, while for the disk-hole array with low AR of 0.34 the optical modes in the near-infrared region agree well in spectral position and magnitude. A possibly reduced disk thickness in the hole due to the shadowing effect in the high AR case (AR 0.82) has been ruled out as explanation for the strong deviation by additional simulations (see Fig. S4 †). All samples exhibit lower experimental transmittance and higher absorption in the visible range compared to simulation. This can be explained by the higher imaginary part of the complex permittivity in the visible range as discussed earlier. The low optical losses of the ultrathin Ag in the NIR, already discussed for the planar case, can be especially well observed in the absorption spectra of the AR 0.34 structure above 1050 nm, where the absorption spectrum of the experiment is lower than the simulation employing bulk optical properties.
The most prominent optical mode in the NIR shows a spectral dependence on the change of geometry parameters, while two optical modes well distinguished in the absorption spectra and marked with two green arrows are surprisingly stable towards change of parameters.
In the last row, the spectra of the planar structure (grey line) with same layer thicknesses of AZO and Ag are additionally shown. Comparing average transmittance in the visible range (here considered from 380 nm to 700 nm) yields 71% for the disk-hole array and 77% for the planar layers. Above 700 nm the transmittance of the disk-hole array drops signicantly faster compared to the planar layer. At the main resonance at around 1100 nm, experimental transmittance drops as low as 4% for the disk-hole array, while the planar counterpart shows a transmittance of 59%. Above 1360 nm the transmittance of the diskhole array rises above the transmittance of the planar structure. Considering that the solar spectrum above this wavelength contributes only app. 11% to the total solar intensity (see the ESI of Bauch et al. 29 ), these properties qualify the presented structure with an aspect ratio of 0.34 as an easy-to-produce lter to block solar radiation in the NIR. Therefore, the aging of the ultrathin silver lm (for AR 0.38) when exposed to air was observed over several weeks, showing the known weakening and red-shi of the resonant transmittance dip (see Fig. S5 †). 48 However, the tarnishing of the silver and with it the reduction of optical performance happened much slower than expected from other studies. 11,49,50 In the rst week aer deposition no Nanoscale Advances signicant change was observed. Aer 4 weeks of air exposure, the transmittance minimum of the strong NIR-mode shied by 40 nm to higher wavelengths and weakened about a total of 2% compared to the as-deposited measurements. A detailed analysis of the optical modes present in the diskhole array is shown in Fig. 5. The disk-hole array is decomposed into its main components, a disk array and a hole array, to facilitate mode analysis and investigate possible coupling between disks and holes. The transmittance spectra and charge density maps of the disk array are shown in Fig. 5(a). A strong plasmonic dipole resonance is observed in the near-infrared region around l ¼ 1000 nm and a much weaker multipole excitation in the visible range. Multipole excitations in nanoparticles have been frequently observed for a variety of geometries including metallic nanodisks. [51][52][53] An increase in disk diameter (see Fig. 5(b)) results in a typical red shi of the localized surface plasmon resonance of nanodisks. 22,51,54 The change of disk array period (see Fig. 5(c)) has only a minor inuence on the dipole resonance position. With decreasing period, a slight red shi is observed due to near-eld interaction between the disks. 55 The transmittance spectrum and charge density of the hole array can be seen in Fig. 5(d). Two multipole hole array modes 56 can be observed. These modes are of localized nature as the change of incident angle for TE and TM polarization does not inuence resonance position (data not shown). Furthermore, the localized nature of the hole array modes has been investigated in one of our previous studies. 29 Interestingly, a dipole resonance in the hole array, as observed by other authors, 23,56 is not readily seen. Only a decrease in hole diameter (see charge distribution in Fig. 5(e)) reveals a clear dipole hole mode. It seems that the formation of a clear dipole mode in the hole array is inhibited when the hole diameter becomes large with respect to the hole array period.
In contrast to the disk array, the spectral position of the hole array resonances is only weakly inuenced by a change in diameter. This explains the stable resonance positions of the hole array modes observed in Fig. 4 (green arrows) with a xed period. The hole array resonances are strongly inuenced by the period of the array, even for a non-diffractive array, as reported by Parsons et al. 57 The simulation of the disk-hole array in Fig. 5(g) (same parameters as the experimental structure in Fig. 4 with AR 0.38) reveals no signicant coupling between the disk array and hole array as resonance positions compared to the individual components do not exhibit a spectral shi. The charge density of the disk-hole array shows same charge distribution as the individual components with the respective counterpart showing the induced image charges. 58 The angle wavelength maps for transverse electric (TE) and transverse magnetic (TM) polarization of the disk-hole array (see Fig. 5(h) and (i)) show angle independent modes and therefore conrm the localized nature of all observed plasmon modes. 57

Conclusion
This study reports ultrathin, Ag-based plasmonic nanostructures, fabricated by nanoimprint lithography and DC magnetron sputtering without li-off steps. With its ease of fabrication, this process can be easily adapted for highthroughput and large-scale roll-to-plate or roll-to-roll production for rigid or exible substrates. Indeed, each processing step (application of the resist, nanostructuring by the mold roller, UV-curing of the resist, sputter deposition, etc.) can be scaled up for large surface areas. While 6 nm Ag on a planar, ZnO-based substrate conguration yielded a continuous lm with occasional holes, the deposition of the same lm on a nanostructure resulted in a defect-free Ag lm. The comparison of optical properties of the at 6 nm Ag with Ag bulk optical properties from the literature yielded a good agreement with slightly increased optical losses in the visible range, while losses in the NIR region were reduced. These results hold great potential for plasmonic applications of ultrathin Ag lms with high quality factors in the NIR region. The comparison of experimental optical spectra of the nanostructure and FDTD simulations demonstrated likewise a good agreement, when grain formation at the nanostructure anks can be avoided. It is shown that the bulk optical properties of Ag are in rst approximation sufficient to predict optical properties of nanostructures as thin as 6 nm, when a continuous lm can be reached. Therefore, the results indicate that ultrathin Ag nanostructures with high quality factors in the NIR can be achieved and reliably designed with simulation tools. The presented nanostructures show a high transmittance of 71% in the visible range due to the low Ag thickness, while a strong plasmon resonance with increased reectance in the NIR is observed. This behaviour allows for possible application of the ultrathin Ag nanostructure as a solar control coating or an NIR-lter.

Conflicts of interest
There are no conicts to declare.