1. Introduction

Nano multilayers (NMs) offer a wide range of tunability, as they demonstrate the ability to achieve desired properties through modulation of layer thicknesses, interfaces, and composition [1,2]. Optical coatings for both single layer and multilayer systems have been widely studied, mostly focusing on the achievable optical behavior such as transmittance or reflectance, where highly transparent NMs are typically composed of layers with alternating high and low indices of refraction [3,4]. These types of transparent NMs have attracted interest due to the possibility of introducing additional functionalities, such as low sheet resistance [5,6] or favorable mechanical properties [7] while retaining transparency in the UV-Vis and NIR wavelengths. However, before designing additional functionalities, it is imperative to better understand the juxtaposition between microstructure and interfaces to the optical behavior. Currently, methodologies do exist for optimizing transmittance in optical NMs [7], but the impact of tuning layer thickness for maximized transparency on the microstructural and interfacial character against the backdrop of multifunctionality remains relatively unexplored.

In transparent NMs, the relationship between interfaces and optical performance has been investigated by focusing on minimizing interfacial roughness to avoid scattering effects [8]; however, the impact of nanoscale microstructural features on optics has been studied to a lesser extent, particularly focusing on the effects of varying layer thicknesses. For example, crystalline-crystalline-type interfaces with varying degrees of coherency in non-transparent NMs have been investigated and shown to follow classical interface interactions, where the interfaces act as defect sources and sinks, as well as barriers to defect mobility, thus dominating plastic deformation behavior [9,10]. In contrast, amorphous-crystalline interfaces, which are common in transparent NMs, have been subject to more limited studies, and are expected to play essential roles in optical performance and functionality [11,12]. Therefore, in order to expand upon current studies beyond solely optics and into functionality, this study focuses on the synthesis and subsequent characterization of a model AlN/SiO₂ multilayer amorphous-crystalline system. A comprehensive exploration into the effect of varying layer thicknesses and layer ratios in optically optimized (aperiodic) and baseline repeated bilayer (periodic) multilayers uncovers the role of microstructure and interface character in highly transparent NMs.

2. Experimental methods

AlN/SiO₂ multilayers were deposited on Corning Eagle 2000 glass substrates by DC and RF magnetron sputtering using two targets. AlN was deposited from a 99.999% Al target under an Ar flow of 37.5 sccm and a reactive N₂ flow of 12.5 sccm (1.2 Pa working pressure) at 150 W of DC power. SiO₂ was deposited from a 99.99% SiO₂ target under an Ar flow of 0.6 Pa working pressure at 40 W of RF power. The on-times for deposition were controlled with a pneumatic shutter, allowing for precise control of layer thickness and deposition from one source material at time. An XP-2 profilometer (AMBiOS) was used to measure surface profiles for calculation of sputtering rates and an estimation of residual stresses of the films using Stoney’s equation.

Two different types of multilayer samples were synthesized in this study: samples designed for maximum optical properties and those with repeated bilayer thicknesses as a baseline. The repeated bilayer samples were AlN(50nm)/SiO₂(50nm) and AlN(100nm)/SiO₂(100nm), where the parenthesis indicate the layer thicknesses held constant throughout the total 1 µm film thickness. From this point forward, the ‘nm’ designation will be dropped for concision. These constant bilayer thickness samples have the same total number of layers, and therefore the same number of interfaces, as the optically optimized samples, and provide a comparative standard for optical, microstructural, and mechanical properties.

The optically optimized samples synthesized for maximized optical transmittance were designated “MBI 10 Layers” and “MBI 20 Layers,” indicating the total number of layers throughout the same 1 µm total film thickness. Optimized AlN/SiO₂ layer configurations were calculated for maximized optical transmittance in the UV/Vis wavelengths (300-800 nm) using the Multiple Beam Interference (MBI) Recursive method [7,13]. MBI formulas were implemented in an in-house MATLAB optimization scheme [14] with experimental optical constants measured from the as-sputtered AlN and SiO₂ films. The spectroscopic index of refraction data as measured from ellipsometry for the monolithic AlN and SiO₂ films synthesized at the same deposition conditions. The refractive indices were approximately n_550nm= 2.18 and n_550nm= 1.43 for AlN and SiO₂, respectively. The calculated layer thicknesses required non-constant bilayer thickness, and the 10 Layer and 20 Layer sample configurations are detailed in Table 1. The optically optimized samples exhibited good agreement between MBI-calculated predictions, within 2% for the 10 Layer sample (94.0% experimental vs. 96.1% predicted) and within 1% for the 20 Layer sample (95.3% vs. 96.2% predicted). The high transparency confirms good agreement to the optical predictions and precise deposition of layer thicknesses across all multilayered samples in this study.

Table 1. Summary of layer thicknesses of MBI 10 Layer and MBI 20 Layer Samples

View Table | View all tables in this article

Sample microstructural characterization, including energy dispersive spectroscopy (EDS), was performed using a FEI Talos 200s (Thermo Scientific) transmission electron microscope (TEM). The TEM foils were prepared via focused ion beam (FIB) lift-out preparation using a Zeiss Auriga Dual Beam FIB. Hardness and reduced elastic moduli measurements were performed by nanoindentation using a Hysitron Triboindeter (Buehler) and a force-controlled, constant loading rate function with a 100 nm Berkovich tip.

3. Results and discussion

In general, studies in the field focus in non-optical NMs with repeated bilayers [15,16] or optical behavior with varying layer thicknesses [5], both of which do not provide a direct link between interfaces and microstructural properties in high transparency NMs. Thus, this study provides a connection beyond optics by integrating a comprehensive study of the microstructural variations in both repeated bilayer and optically optimized NMs. Figure 1 presents a summary of the overall characteristics of the synthesized multilayer films. Figure 1(a) illustrates a schematic of a nano multilayered sample configuration, where Fig. 1(b) highlights a representative TEM cross-section of the as-sputtered AlN/SiO₂ multilayers noting the clear interface between the layers. The accompanying selected area electron diffraction (SAED) pattern (Fig. 1(b)) indicates nanocrystalline spots in the AlN layer and a diffused amorphous pattern in the SiO₂ layer for an overall amorphous/crystalline structure. The experimental % transmittance plot, as presented in Fig. 1(c), confirms much higher optical transmittance in the MBI samples than the repeated bilayer samples. The corresponding top views of the as-sputtered samples (Figs. 1(d)–(1)(g)) indicate the high transparency in the UV/Vis region of the optically optimized samples. Moreover, the experimental transmittance of the MBI 10 Layer and 20 Layer samples are 94.0% and 95.3% from 300 to 800 nm, respectively. This is a marked increase as compared to the repeated bilayer samples, which both have experimental transmittance values below 80%.

 

Fig. 1. Overview of AlN/SiO₂ nanomultilayers, where (a) is a schematic of alternating and (b) representative cross-sectional TEM nanocrystalline AlN and amorphous SiO₂, as indicated with the inset SAED patterns. (c) Experimental % transmittance curves, where layer thicknesses of repeated bilayer samples are indicated in parentheses in nanometers. Top views of as-deposited samples (d-g): (d) AlN(100 nm)/SiO₂(100 nm), (e) AlN(50 nm)/SiO₂(50 nm), (f) MBI 10 Layers, and (g) MBI 20 Layers.

Download Full Size | PDF

To investigate the repeated bilayer microstructures as a baseline and to examine the implications of varying layer thickness, an AlN(50)/SiO₂(50) sample is displayed in Fig. 2(a). Uniform layers were observed through the entire film thickness with columnar grains extending through the 50 nm AlN layers, a common morphology observed in sputter-deposited nanocrystalline films. This columnar grain structure can affect optical behavior of dielectric materials, such as increasing spectral selectivity and refractive index anisotropy [17,18]. The inset SAED pattern shows nanocrystalline rings with some preferential texturing as indicated by the brighter spots on the rings. At higher magnifications, Fig. 2(b) features a crystalline-amorphous AlN-SiO₂ interface with the highly columnar AlN grains. Within the layers, the amorphous nature of the SiO₂ is confirmed by the lack of SiO₂ peaks in the radial intensity profile (Fig. 2(c)) which indicates hcp and fcc AlN peaks. Previous studies have found that a formation of mixed phase hexagonal (wurtzite) and cubic (rock salt and zinc blende) crystal structure AlN is possible by reactive sputtering, especially when deposited at low discharge powers, which were employed in this study [19]. Additionally, some studies have demonstrated that with increasing AlN layer thickness, a rock-salt (cubic) to hexagonal transformation or zinc blende (cubic) to hexagonal transformation is observed [20,21].

 

Fig. 2. Overview of AlN(50 nm)/SiO₂(50 nm) repeated bilayer sample, (a) BF cross-sectional TEM and inset SAED pattern of the multilayer film with yellow box indicating the region of (b) BF HRTEM of the AlN-SiO₂ interface region. Red dotted line indicates the path of the 1D compositional line profile shown in (d). (c) Integrated radial intensity profiles interpolated from the SAED pattern and (d) the 1D compositional line profile as measured by STEM EDS, showing atomic % fraction of elements across the multilayers.

Download Full Size | PDF

Beyond layer microstructure, the role of the interface character must be examined through both interface roughness and composition. The amorphous-crystalline interface of AlN(50)/SiO₂(50) is highlighted in Fig. 2(b), where there is a sharp interface with clear delineation between the AlN and SiO₂ layers. Upon closer examination, there appears to be varying termination of AlN grains, suggesting non-isotropic columnar growth in the AlN layer a type of interfacial deviations which could affect multilayer response, such as increasing resistance to deformation [11]. Furthermore, the AlN-SiO₂ interfaces are compositionally discrete, as evidenced by the EDS line profiles in Fig. 2 where the average slope of the compositional gradient [22] is calculated from the Al line profile and is ≈3.48 atomic fraction/nm. Thus, the repeated AlN(50)/SiO₂(50) configuration exhibited distinct layer interfaces in addition to a strong columnar AlN grain morphology.

Increasing the bilayer thickness in a nanolayered structure has implications on the microstructure and interface character, and it is expected that the columnar grain size will scale with layer thickness [23], thus leading to an increase in the anisotropic columnar growth in the AlN layers. Therefore, a periodic AlN(100)/SiO₂(100) multilayer was investigated as an additional baseline configuration, and it presents uniform layers (Fig. 3(a)) with more distinct bright spots in the inset SAED patterns. The AlN columnar grains observed in the HRTEM (Fig. 3(b)) have higher grain widths than in the AlN(50)/SiO₂(50) multilayers, increasing from around 10 nm to 20 nm. The same mixed crystalline phase AlN is observed (Fig. 3(c)) with a higher peak-to-valley ratio, which indicates more preferential texturing in the increased AlN layer thickness. Additionally, the average slope of the EDS line profile decreases to ≈3.05 atomic fraction/nm, which suggests more mixing in the thicker bilayer sample.

 

Fig. 3. Overview of AlN(100 nm)/SiO₂(100 nm) repeated bilayer sample, (a) BF cross-sectional TEM with inset SAED pattern and (b) BF HRTEM of the AlN-SiO₂ interface region. (c) Integrated radial intensity profiles interpolated from the SAED pattern and (d) the 1D compositional line profile.

Download Full Size | PDF

The repeated bilayer configurations limit the achievable maximum transmittance values, but in view of these results, the optically optimized samples allow further investigation of the nanoscale features that potentially lead to improved transmittance in non-constant bilayer thickness samples. Figure 4 highlights the MBI 20 Layer sample with transmittance 95%, which has the same total number of layers and interfaces as the baseline AlN(50)/SiO₂(50) sample. The cross-section TEM shown in Fig. 4(a) suggests uniform layers with concentric diffused rings in the SAED pattern, indicating an overall amorphous morphology. HRTEM in Fig. 4(b) features an AlN layer within the amorphous SiO₂. At 5 nm, the AlN layers remained nanocrystalline, with a smaller grain size than in the repeated bilayer samples. However, the radial intensity plot in Fig. 4(c) illustrates the loss of an observed mixed crystalline phase AlN structure due to the thin layers. In nitride/SiO₂ NMs it has been reported that amorphous SiO₂ can only crystallize and grow epitaxially with sufficiently small (SiO₂ < 0.6 nm) layer thicknesses due to templating effects from the hcp nitride layers, but beyond this the layers break coherent growth and transform back into the amorphous state [24]. This agrees with what was observed in the present study, where the amorphous structure in SiO₂ was maintained across all samples due to the sufficiently thick SiO₂ layers. Additionally, due to the thin AlN layers, the grain structure confined to the nanocrystalline region does not develop into a highly aligned columnar structure as other studies have reported [25].

 

Fig. 4. Overview of AlN/SiO2 MBI 20 Layer sample, (a) BF cross-sectional TEM with inset SAED pattern and (b) BF HRTEM of the AlN-SiO₂ interface region. (c) Integrated radial intensity profiles interpolated from the SAED pattern and (d) the 1D compositional line profile.

Download Full Size | PDF

As for the effect on the interface due to varying layer thickness, there is less evidence of variation in grain termination as a result of reduced columnar growth in these thinner AlN layers, leading to decreased grain morphology-induced structural roughness [11]. However, there is increased compositional mixing of the AlN and SiO₂ layers at the interface, as evidenced by the slope of the EDS line plot in Fig. 4(d), where the average slope is ≈1.20 atomic fraction/nm, a notable decrease from the repeated bilayer samples. Furthermore, the O₂ atomic fraction remains above ∼30% within the AlN layers, where this trace oxygen from the SiO₂ leading to formation of compositional transition regions has been observed in AlN deposited on amorphous SiO₂ substrates [25]. This increased interdiffusion and transition region at the AlN-SiO₂ interface could lead to interfacial scattering, but the agreement between the MBI calculations and experimental transmittance suggests a minimal effect on achieving high transparency.

Decreasing the total number of layers in a constant overall thickness could have implications on optical properties such as decreasing the incidents of scattering at interfaces. Thus, an overview of the optically optimized MBI 10 Layer sample with half the total layers as the MBI 20 is presented in Fig. 5. Comparing the MBI 10 Layer sample to the MBI 20 Layer, the cross-sectional TEM in Fig. 5(a) reveals more uniform layers with an absence of the previously observed layer waviness and fewer nanocrystalline spots in the inset SAED pattern. Figure 5(b) presents sharp interfaces similar to the previous sample and ∼5 nm nanocrystalline grains. The microstructure within the AlN layers and at the AlN:SiO₂ interface are consistent with what was observed in the MBI 20 Layer sample. However, as the total number of layers decrease, there is less interdiffusion at the AlN-SiO₂ interface, which is confirmed by the EDS slope increasing to ≈1.67 atomic fraction/nm.

 

Fig. 5. Overview of AlN/SiO₂ MBI 10 Layer sample, (a) BF cross-sectional TEM with inset SAED and (b) BF HRTEM of the AlN-SiO₂ interface region. (c) Integrated radial intensity profiles interpolated from the SAED pattern and (d) the 1D compositional line profile.

Download Full Size | PDF

The previous results confirm that as the layer thicknesses are tuned for maximized optical properties, there are changes in the layer structure and interface properties, as confirmed by deviations in crystallinity and interface discreteness. These fluctuations also lead to changes in multifunctional behavior and two methodologies were employed to verify these effects: residual stress analysis and nanoindentation measurements which are summarized in Table 2. From these results, it is evident that varying the layer thicknesses for optimized transmittance impacts the residual stresses in the films, with repeated bilayer films having moderate tensile residual stresses (∼300-500 MPa). Dominating compressive residual stresses of the SiO₂ layer and reduced AlN crystallinity lead to the shift from tensile to compressive stresses in the optically optimized samples [26]. Similarly, microstructural variations due to tuning bilayer thickness affects hardness of the multilayer films. The measured nanoindentation hardness values span from ∼7 GPa to 9.8 GPa. For reference, AlN hardness is approximately 12 GPa [27] and SiO₂ hardness is approximately 4.5-9.5 GPa [28,29], depending on crystallographic structure. In the MBI 10 and 20 Layer samples, the fraction of SiO₂ is much higher, leading to lower hardness values. Overall, the changes in multifunctional properties were measurable, and these differences are in part due to composition, but are also driven by changes in interfacial character.

Table 2. : Results of experimental mechanical and optical properties, including total film thickness, residual stresses as measured by profilometry, nanoindentation results, and experimental % transmittance.

View Table | View all tables in this article

In summary, microstructural evaluation of AlN/SiO₂ NMs has shown that optimizing layer configurations for high optical transparency (94-96%) in the UV/Vis region induces nanoscale variations within the layers and at the interfaces. Based on the current study, designing multilayer configurations for maximized transparency leads to changes in the microstructure such as the observed columnar to nanocrystalline grain morphology in the AlN layer as well as reduction of sharp compositional transitions at the interfaces. Furthermore, nanoindentation and residual stress results for the varying NM configurations, highlight a correlation between optimization for high optical transparency and physical attributes. An understanding of these interfacial and microstructural changes in relation to optical behavior facilitates eventual introduction of additional functionalities to optical NMs.