Ultrasmooth ultrathin Ag films by AlN seeding and Ar/N₂ sputtering for transparent conductive and heating applications Open Access

Silver (Ag) is widely used for transparent heat regulation applications considering its low bulk electrical resistivity and visible transparency in the ultrathin film regime. Given the modern architectural trend of favouring high window-to-wall ratios in building facades, heat regulation of windows plays an increasingly dominant role in improving energy efficiency in the buildings as well as in the automobile sectors.^1,2 To achieve high IR reflectance and maximal visible transmittance, ultrathin silver films (UTSFs) are of great interest. However, the fundamental challenge in depositing UTSFs (and other metals) is their natural tendency to grow in disjointed island forms on many dielectrics,³ most notably on amorphous SiO₂⁴ but also on TiO₂,⁵ Al₂O₃,⁶ ZnO,⁷ SnO₂,⁷ CaF₂,⁸ and Si.⁹ This phenomenon, known as the Volmer-Weber mode,¹⁰ creates a high percolation threshold thickness for UTSF continuity and a rough surface that contributes to optical absorption loss via localized surface plasmon resonance (LSPR).^11,12 In the application of spectrally selective coatings, also referred to as solar control coatings (SCCs), absorption losses reduce the desired reflectance in the near infrared (NIR) region as well as the desired transmittance in the visible region.

The most well-known method for promoting thin film continuity and smoothness is to deposit a seed layer with high surface energy to promote wetting of the thin film.¹⁰ Effective seed layers for Ag include Ge,¹³ Cu,¹⁴ PEDOT:PSS,¹⁵ Cr,¹⁶ Ni,¹⁷ Al,¹⁸ and AgO.¹⁹ With a seed layer, the root-mean-square (RMS) roughness can be pushed to the sub-nanometer regime, while without a seed layer, the RMS roughness tends to be in the nanometer regime. However, with the exception of PEDOT:PSS, all of the above seed layers have non-negligible absorption in the visible and NIR spectrum that deteriorates the desired transparency or reflectivity in SCCs by virtue of being a less reflective metal compared to Ag or, in the case of AgO, having an absorption peak that overlaps with the visible spectrum.²⁰ PEDOT:PSS, while transparent,²¹ requires spin coating of aqueous-phase media, a process that is not easily integrated with industrial-scale sputtering processes. Because of these limitations, a visibly transparent seed layer that can be produced by vacuum deposition would be advantageous over the currently reported seed layers for applications of UTSF in SCCs and transparent heat mirrors (THMs).

A less well-known method for promoting smoothness of UTSF is the introduction of nitrogen (N₂) gas alongside argon (Ar) gas during sputtering without the use of any seed layer. Unlike most transition metal nitrides like copper nitride, silver nitride is an unstable compound with a +314 kJ/mol free energy of formation.²² To date, there has been no evidence of silver nitride formation during the sputtering of Ag regardless of the N₂-to-Ar gas ratio.²³ Bulíř et al.²⁴ first reported the use of Ar/N₂ sputtering at 7.1% N₂ during Ag sputtering onto silica to decrease the RMS roughness of UTSF from 3.3 nm to 1.4 nm without any seed layer. They also noticed a change in preferential Ag crystal orientation with the introduction of N₂, which matched the result from a previous study on thick Ag films sputtered by Ar/N₂.²³ To the best of the author’s knowledge, there have been no other studies performed on Ar/N₂ sputtering of UTSFs.

To produce the UTSFs with optimal optical properties needed for SCCs, we present the first report on the use of aluminum nitride (AlN) as a seed layer for Ag as well as the combination of AlN-seeding with Ar/N₂ sputtering of Ag, and thus demonstrate improvements to the UTSF wetting and optical properties. AlN, which can be reactively sputtered at room temperature,²⁵ is a visibly transparent dielectric with an optical bandgap of 5.8 eV²⁶ and shows strong potential as a seed layer candidate due to its high surface energy.²⁷ Finally, we integrated our UTSF into a Ag/a-C:H multilayer SCC.

The substrate used for all film depositions is a 1.1-mm-thick Corning Eagle XG glass. The substrates were cleaned in an ultrasonic bath of acetone, isopropanol alcohol, and deionized water sequentially for 15 min each and then dried off with flowing N₂ gas. The radio frequency (RF) (13.56 MHz) magnetron sputter deposition chamber and the RF (13.56 MHz) PECVD deposition chamber were pumped down to below 5 × 10⁻⁷ Torr before film depositions. All film depositions were performed at room temperature.

For sputter deposition of Ag films, 300 W of RF power was supplied to a 99.99% pure Ag target (Kurt J. Lesker, 76 mm in diameter) with sputter gas flow of 20 standard cubic centimeters (SCCM) of pure Ar or Ar/N₂ at a process pressure of 5 mTorr. The thickness of the Ag films was determined using a quartz crystal monitor. For the seeding layer AlN, a 99.99% pure Al target (Kurt J. Lesker, 76 mm in diameter) was reactively sputtered at 150 W with 8 SCCM of Ar and 12 SCCM of N₂ at a process pressure of 5 mTorr for 4 min. The thickness of AlN following 4 min of deposition was determined to be ∼1 nm from cross-sectional SEM analysis of a thicker AlN film. The target-to-substrate distance for both Ag and Al is 18 cm. Hydrogenated amorphous carbon films (a-C:H) were grown on substrates placed on the cathode of the RF PECVD system.²⁸ A RF power of 3 W was supplied with 20 SCCM of 99.99% pure methane gas at a chamber pressure of 60 mTorr. The deposition rate was 3 nm/min.

Four conditions of UTSF fabrication were used to make the comparisons. In condition 1, Ag was sputter deposited with 20 SCCM of Ar onto glass at a rate of 0.37 nm/s. In condition 2, the seeding layer AlN was first deposited. The sputtering chamber was allowed to pump down to below 10⁻⁶ Torr (<0.5% process pressure) to remove the N₂ gas prior to the sputter deposition of Ag with 20 SCCM of Ar onto AlN at a rate of 0.37 nm/s. In condition 3, Ag was sputter deposited with 8 SCCM of Ar and 12 SCCM of N₂ (60% N₂) onto glass at a rate of 0.30 nm/s. In condition 4, the seeding layer AlN was first deposited, and thereafter Ag was sputter deposited with 8 SCCM of Ar and 12 SCCM of N₂ onto AlN at a rate of 0.30 nm/s.

The resistivity of UTSFs was measured using a coplanar four-point probe configuration (Nanometric Hall Effect Measurement System HL5500PC) in quadruplicate by rotating the sample by 90° after each measurement. SEM images of UTSFs were taken using an FEI Quanta FEG 250 ESEM. AFM scans of UTSF were taken using 10-nm-diameter tips in a tapping mode at 8 ms per point and a resolution of 256 points per line in an area of 1 μm × 1 μm (Nanonics MultiView 2000). Specular transmittance and specular reflectance of UTSFs and SCCs were measured using a UV-vis-NIR spectrophotometer from 2600 nm to 200 nm in steps of 5 nm (PerkinElmer Lambda 1050). Mid-infrared (MIR) reflectance and emissivity of SCCs were measured using an FTIR spectrometer from 2 μm to 26 μm.

The transfer matrix method was used to simulate the optical responses of single-layered and multi-layered coatings.^29,30 The Drude-Lorentz model was used to extract the complex refractive indices of our UTSFs. The complex permittivity is given by Eq. (1),

The first two terms contain the Drude parameters where ϵ_∞ is the permittivity at infinite frequency, ω_p is the plasma frequency (or equivalently expressed as the plasma wavelength λ_p), and γ is the damping coefficient. The summation term contains the parameters for an arbitrary number of Lorentz peaks where A_L is the peak amplitude, ω_L is the peak resonance frequency (or equivalently expressed as the peak resonance wavelength λ_L), and γ_L is the peak shape parameter. The calculations for solar heat gain coefficients (SHGC) of SCCs were carried out in compliance with the National Fenestration Ratings Council standards using the WINDOW program from the Lawrence Berkeley National Laboratory.

Figure 1 shows the resistivity of UTSFs corresponding to each of the four deposition conditions, and the sharp drop in the resistivity indicates the percolation threshold. The percolation threshold is estimated to be 8 nm, 5 nm, 4 nm, and 3 nm for conditions 1 to 4, respectively. The lowering of the percolation threshold indicates better wetting of the UTSF. The results indicate that either AlN seeding or Ar/N₂ sputtering alone improve the wetting of the UTSF, and that wetting is most improved when both AlN seeding and Ar/N₂ sputtering are used. The differences in wetting of UTSFs in each condition are most obvious at the percolation threshold as observed in the surface morphology micrographs of Figs. 2(a)–2(d); specifically, SEM images of UTSFs were obtained corresponding to each condition for the films deposited near their respective percolation threshold, that is, at a film thicknesses of 8.1 nm, 4.5 nm, 4.0 nm, and 3.4 nm corresponding to conditions 1 to 4, respectively. The scale bar is 200 nm long. All of the UTSFs are semi-continuous but differ in the dimensions of both the islands and the channels connecting individual islands. The UTSF morphology in condition 1 can be described as large islands (white circular spots) connected by long channels (gray oblongs between white circular spots), while the UTSF morphologies in condition 2 to 4 contain smaller islands and shorter channels. Importantly, the amount of Ag in Fig. 2(a) is much greater than the amount of Ag in Figs. 2(b)–2(d), so there is a large reduction in the vertical dimensions of the islands and channels in condition 2-4 compared to condition 1. In condition 4, the small variation in the vertical dimension of the UTSF reduces the contrast between peaks and troughs of the UTSF to the point that noise or image blurring is far more noticeable in this micrograph.

Wetting of UTSFs can be broadly understood from the thermodynamic perspective. AlN has a higher surface energy than Ag, while glass has a lower surface energy than Ag. As such, the global energy minimum of the Ag-AlN system occurs where the surface area of AlN is minimized while the global energy minimum of the Ag-glass system occurs where the surface area of Ag is minimized. The former encourages high 2D Ag coverage on AlN, while the latter encourages 3D clustering of Ag on glass. Therefore, the lowered percolation threshold of AlN-seeded UTSF in condition 2 compared to unseeded UTSF in condition 1 can be understood as a modification of the substrate to a state that thermodynamically favours wetting of the UTSF. In the case of Ar sputtered UTSF versus Ar/N₂ sputtered UTSF, it is not as obvious whether surface energy plays a role in changing the growth morphology. It is possible that during Ar/N₂ sputtering, nitrogen is incorporated onto the top few nanometers of the glass substrate and modifies its surface energy. Indeed, nitridation of silicon dioxide by N₂ plasma has been reported in the literature, but typically under conditions of high N₂ plasma density, high temperature, or long exposure time.^31–33 Our Ar/N₂ sputtering process does not match such nitridation conditions. Moreover, XPS depth-profiling of 10-nm-thick Ar/N₂ sputtered UTSF onto glass showed no detectable nitrogen signals. We thus speculate that nitridation of the glass surface, if present, is minimal and does not play a large role in determining the growth mode of the UTSF.

While thermodynamic considerations to UTSF growth are important, kinetic considerations during the growth process must also be considered. One such consideration is the deposition rate of Ag. It is well-known that high deposition rates discourage the island growth mode of Ag and flatten the film surface.³⁴ Under a high influx of Ag atoms, migration of Ag adatoms onto existing Ag islands is disrupted by the formation of Ag aggregates with the impinging Ag atoms, reducing both the size and height of Ag islands. Surprisingly, we observe the opposite trend in the change in growth morphology as would be predicted by the difference in the deposition rate of Ar sputtered UTSF versus Ar/N₂ sputtered UTSF. Ar/N₂ sputtered UTSF have smaller Ag islands and a smoother surface despite being grown at a lower rate of 0.30 nm/s compared to the rate of 0.37 nm/s for Ar sputtered UTSF. We propose that a different kinetic consideration in the form of kinetic constraints is the explanation for the observed change in growth morphology. Kinetic constraints suggest that, while the thermodynamically favoured state is 3D clustering of Ag on glass, the pathway to such a state may be kinetically unfavourable. Specifically, for 3D islands to form, individual Ag atoms must repeatedly overcome the large activation potential barrier of upstepping to a higher plane within each individual island. In the case of UTSF growth on glass, an Ag adatom on glass can lower its energy by upstepping to a higher Ag plane but only if it has enough initial energy or within a sufficiently high temperature bath. This kinetic constraint is especially prominent for Ag step atoms or kink atoms which have more nearest-neighbours binding them tightly. Since the temperature is not changed in our sputtering process, we posit that Ar/N₂-sputtered Ag atoms have an energy profile lower than that of pure Ar sputtered Ag atoms due to the poor energy transfer between N₂ sputtering species and Ag compared to that between Ar sputtering species and Ag. As such, Ag atoms sputtered by N₂ are kinetically unfavoured to cluster into 3D islands.

Surface roughness of even a continuous UTSF is associated with LSPR absorption losses. AFM scans of the UTSFs having the same thickness of 15 nm for each condition are shown in Fig. 3. RMS surface roughness of the UTSFs corresponding to conditions 1 to 4 are 2.8 nm, 1.2 nm, 0.7 nm, and 0.6 nm, respectively. The peak-to-peak values show similar trend as the RMS surface roughness. Both AlN-seeding and Ar/N₂ sputtering alone resulted in improvements in UTSF smoothness. Ar/N₂ sputtering produces a smoother UTSF compared to AlN-seeding, while combining both techniques produces an even smoother UTSF.

The impact of UTSF surface roughness morphology on its optical response is presented in Fig. 4, showing the transmittance, reflectance, and absorption of 15-nm-thick UTSFs grown on glass corresponding to each condition. Immediately, we see that Ar sputtered UTSFs, whether seeded with AlN or not, have greater absorption, lower transmittance, and lower reflectance in the visible and NIR region compared to Ar/N₂ sputtered UTSFs. We note here that the optical response measurements of 15-nm-thick UTSFs using Ar/N₂ sputtering at 20%, 40%, 80%, and 100% N₂ were also carried out, and it was observed that their transmittance and reflectance are virtually identical to those of 60% N₂ (data not shown). The presence of this absorption is consistent with the LSPR explanation since UTSFs grown using Ar sputtering in general are rougher than UTSF grown using Ar/N₂ sputtering (Fig. 3). Furthermore, the reduction in lateral sizes of Ag islands by Ar/N₂ sputtering compared to Ar sputtering would blueshift the LSPR and away from the visible and NIR spectrum (Fig. 2). From the above arguments, the best UTSF with optimal optical properties to integrate into a multilayer coating is one that is seeded with AlN and sputtered by Ar/N₂.

In order to model and optimize a multilayer coating containing UTSFs, we apply the Drude-Lorentz model to fit the transmittance and reflectance data of 15-nm-thick UTSFs in each condition so as to extract their complex refractive indices and permittivities. Within this model, the optical response of UTSFs is described using Drude parameters for the visible and IR response and Lorentz oscillators for the interband transitions (IBT) of Ag in the UV and LSPR, if any. The modelled fits are shown in Figs. 5–8 corresponding to conditions 1 to 4, respectively, and the matching fit parameters are given in Table I. The residuals standard errors for conditions 1 to 4 are 0.0046, 0.0044, 0.0034, and 0.0044, respectively. As can be most easily seen in the imaginary component of permittivity, Ar/N₂ sputtered UTSFs follow the Drude model closely in the visible and NIR region [Figs. 7(e) and 8(e)], while Ar sputtered UTSFs deviate from the Drude model in the visible and NIR region due to the presence of a Lorentz peak [Figs. 5(e) and 6(e)]. Somewhat unintuitively, the consequence of the addition of the LSPR absorption manifests in the real part of the refractive index n rather than the imaginary part of the index k. There is an increase of up to 0.4 in n in the visible and NIR region from Ar/N₂ sputtered UTSFs to Ar sputtered UTSFs [Figs. 7(b) and 8(b) in contrast to Figs. 5(b) and 6(b)], while there is a difference of less than 0.1 in k in the visible and NIR region among all four UTSFs [Figs. 5(c), 6(c), 7(c), and 8(c)].

Interestingly, the film electrical resistivity in all four conditions converge to approximately the same value once the percolation threshold is passed (Fig. 1) in contrast to the work reported by Bulíř et al. where they describe observing an order of magnitude increase in resistivity for their Ar/N₂ sputtered UTSF compared to Ar sputtered UTSF.²⁴ Our UTSFs from 10 nm to 15 nm in all four conditions have similar resistivities and, in particular, at 15 nm thickness, the resistivities lie between 5 μ$Ω$ cm and 6 μ$Ω$ cm which are some 3 to 4 times the resistivity of bulk Ag (1.6 μ$Ω$ cm). These resistivity measurements are independently validated through our Drude-Lorentz model fits of the optical responses where the resistivity is extracted from the damping coefficient. The resistivities given by the Drude-Lorentz model are between 6 μ$Ω$ cm and 9 μ$Ω$ cm, in reasonable agreement with our measurements.

Commercial high-performance SCCs incorporate two UTSFs in the multilayer coating, commonly in the configuration substrate/D/M/D/M/D instead of a single UTSF in the substrate/D/M/D configuration commonly reported in the literature.³⁵ The dual UTSF configuration allows for a higher light-to-solar-gain (LSG) ratio, a metric for spectral selectivity defined as the quotient of T_vis (visible light transmittance weighted by the photopic sensitivity) to SHGC (the solar heat gain coefficient). Following the preceding discussion on the conditions that produced a UTSF with optimal optical properties, we integrated Ar/N₂ sputtered AlN-seeded UTSFs into a multilayer SCC using a-C:H thin dielectric films. As we have previously reported, the optical bandgap and refractive index of a-C:H depend on the supplied RF power during PECVD.²⁸ The a-C:H thin films used in this study have an optical bandgap of 4.0 eV and an n of 1.55 at 550 nm. Our SCC structure is glass/UTSF/a-C:H/UTSF/a-C:H where the thicknesses are 15 nm/109 nm/10 nm/41 nm. The design of the multilayer can be intuitively understood as a Fabry-Pérot resonator of UTSF/a-C:H/UTSF and a further anti-reflection a-C:H layer at the top. The total Ag thickness of 25 nm in the SCC was selected to produce <10% emissivity, while the a-C:H film thicknesses were selected through an iterative optimization routine using the sum of T_vis and IR reflection as a metric alongside the constraint of >70% T_vis. The visible and NIR transmittance and reflectance of our SCC are shown in Fig. 9(a), and the MIR reflectance is shown in Fig. 9(c). We obtained a T_vis of 0.72, a SHGC of 0.38, a LSG ratio of 1.9, and an emissivity of 0.084. The transparency of the SCC is illustrated in the photograph Fig. 9(b).

Herein we have reported on the investigation of two methods of producing ultrathin silver films (UTSFs) with minimal optical absorption loss. These are the use of AlN as a seed layer with Ar sputtered deposition of Ag, or using Ar/N₂ sputter gas combination for the deposition of Ag, or combining both methods (seeding with AlN and using Ar/N₂ sputter gas combination) to deposit the Ag film. We find that the optimal procedure of attaining the lowest percolation threshold and lowest RMS roughness is the combination of both methods. The use of Ar/N₂ sputtering by itself reduces the percolation threshold of Ag, indicating kinetic constraints dominating over thermodynamic influences in determining the initial film growth mechanism. At 15 nm thickness, an UTSF with an RMS roughness of 0.6 nm and a resistivity of 5 μ$Ω$ cm (approximately three times the bulk resistivity of Ag) was obtained. Integration of the optimized UTSF in a solar control coating device was demonstrated as an example of its application as a transparent conductive film.

The authors acknowledge the support of the Natural Sciences and Engineering Research Council of Canada (NSERC) through the Discovery and Idea-to-Innovation grant programs and The Edward S. Rogers Sr. Department of Electrical and Computer Engineering at the University of Toronto. The authors also acknowledge Nishant Bhatt for helping model the refractive index of a-C:H, Jessica Miller for carrying out preliminary resistivity measurements, Mubarek Abdela for assisting with optimization algorithms, and Pratish Mahtani and Houfar Daneshvar for assisting with the deposition of a-C:H based spectrally selective coatings.