Introduction

The global energy demand is ever increasing and is projected to rise nearly 50% by 2050 [1]. In the view of the current climate crisis, energy harvesting inevitably steers toward environmentally conscious sources [2]. However, this type of energy supply is inherently fluctuating and often depends on multiple environmental factors such as season and weather conditions. Effective and reliable storage of the energy captured during the peak production time for later use is therefore crucial. Time separation of the energy production and energy consumption stages is enabled through battery technologies, where solid-state batteries are of particular interest due to improved safety and higher energy density compared to their liquid state counterparts. To date, solid-state batteries based on Li have been investigated most extensively and dominate the global market since their commercialization in 1991, the current development mainly driven by electric vehicle adoption. However, recent studies on resource assessment modeling predict that the global expansion of the electric vehicles fleet powered by Li ion batteries appears to be unrealistic due to geographically concentrated and finite Li supplies being the limiting factor [3, 4]. An emerging alternative pathway for the next-generation solid-state batteries could be Li replacement with another alkaline metal: Na [5, 6]. In contrast to Li, Na is an abundant, geographically evenly distributed, and inexpensive element. Development of a suitable well-performing electrolyte is the current focus in advancing Na-based solid-state batteries. The solid-state electrolyte NASICON (Na super-ionic conductor) type materials attracted specific attention due to high ion-conductivity resulting from fast diffusion of Na ions within a three-dimensionally interconnected interstitial sites network in the crystal lattice [7]. Na1+xZr2SixP3−xO12 in particular has been explored, owing to the combination of superior ion-conductivity and excellent electrochemical stability [8]. To date, most research on NASICON materials has been done employing bulk synthesis methods. Thin film solid-state batteries, on the other hand, allow manufacturing of miniaturized devices for use, e.g., in medical monitoring, biosensors, and wearable electronics [9]. Magnetron sputtering of NASICON thin films within the composition space Na2O-SiO2-P2O5-ZrO2 was investigated in a series of publications by Horwat et al. Reactive sputtering from a single Na3Zr2Si2PO12 and composite Na3PO4/ZrSiO4 targets in an Ar/O2 mixture suffered from target instabilities such as cracking and Na ion migration or resulted in chemically heterogeneous films, respectively [10]. The film quality could be improved by reactive co-sputtering of Na3PO4 and Zr0.52Si0.48 alloy targets [11,12,13]. Room temperature (RT) deposition followed by annealing in air at 700 °C was used to transform amorphous films into crystalline NASICON phase [11,12,13,14].

In this study, we increase process flexibility through using elemental Zr and Si targets in combination with a Na3PO4 target and employ high-throughput characterization methods to explore the Na2O-ZrO2-SiO2-P2O5 composition space, realized in form of thin film material libraries with focus on composition gradients of the Na1+xZr2SixP3−xO12 NASICON phase.

Materials and methods

NASICON thin film material libraries (MLs) were synthesized by magnetron co-sputtering in confocal geometry from two 50 mm diameter commercial compound Na3PO4 (99.5% purity) and elemental Zr (99.99% purity) targets and one 100 mm diameter Si (99.999% purity) target in a high vacuum sputter system (Creavac GmbH, Germany). Na3PO4 was sputtered using radio frequency (RF), and Zr and Si using direct current (DC) power sources in power-controlled mode at 40 W, 26 W, and 7 W, respectively. In order to account for the substantially different sputter rates of the target materials, the lowest possible powers were set for Zr and Si in combination with a relatively high power for Na3PO4, taking the ceramic target susceptibility to thermal shock into account. In addition, the standard sputter chamber configuration was modified to further enhance the Na and P fluxes arriving to the substrate by lowering the substrate to the target-substrate distance of 8 cm and horizontal translation toward the Na3PO4 target. The depositions were carried out on 100 mm diameter single crystal Al2O3(0001) wafers for 3 h at RT. Before depositions, the sputter chamber was evacuated to a base pressure of 1.1 × 10–5 Pa. The depositions were performed at 2 Pa pressure maintained with a 6N grade Ar (99.999%) flow of 50 sccm. Prior to the depositions, the substrates were cleaned in acetone and isopropanol ultrasonic baths for 5 min each. Al2O3 was selected as a non-reactive substrate stable at high processing temperatures.

As-deposited MLs were post-treated in two steps. First, they were annealed in the deposition chamber at 460 °C in O2 at 100 Pa and 10 sccm flow for 1 h, followed by a second annealing in an air furnace at 900 °C for 1 h. Images of as-deposited and post-treated MLs were taken using an in-house built photo setup [15].

A high-throughput approach was implemented for investigations of the thin film libraries [16]. X-ray diffraction (XRD) and energy-dispersive X-ray spectroscopy (EDX) characterization was conducted in automated mode on a regular grid defining 342 individual measurement areas (MAs) with edge lengths of 4.5 mm × 4.5 mm. The numbering of the MAs starts with MA1 in the lower left corner of the ML. Symmetric θ-2θ XRD measurements were performed with a Bruker D8 Discover diffractometer equipped with a Cu Kα (λ = 0.154 nm) IμS microfocus source (collimated to a 1 mm point focus) and a VÅNTEC-500 area detector. A 2θ range from 5° to 60° was covered by collecting two frames for 30 s each at 2θ of 20° and 40°. The resultant two-dimensional (2D) frames were integrated into one-dimensional (1D) data sets using a DIFFRAC.EVA (Bruker) software. A Python script was implemented to automate the conversion process. Elemental composition maps were obtained using a JEOL 5800 scanning electron microscopy (SEM) system. The EDX spectra were collected for 40 s at 20 kV acceleration voltage and 10 mm working distance. The data analysis was performed implementing a standard ZAF correction process provided by the INCA Energy software (Oxford Instruments). The surface morphologies were examined using a JEOL JSM-7200F SEM equipped with an EDX detector for correlative composition analysis.

Based on the XRD and EDX results, specific MAs were identified for further investigation with transmission electron microscopy (TEM). Cross-sectional samples for TEM analysis were prepared by standard focused ion beam (FIB) milling in FEI Helios G4 CX dual-beam system operated at 30 kV using a Ga ion beam. Prior to the FIB processing, the samples were coated with a ~ 70 nm Au layer to avoid excessive charging and minimize the ion beam damage. Due to general electron and ion beam sensitivity of NASION-type materials, thicker FIB lamellas than the standard thickness (typically ≤ 100 nm) were prepared. TEM images were recorded using a JEOL JEM ARM200F Cs-probe corrected TEM operated at 200 kV.

Results

Figure 1 shows photographs of the as-deposited and post-treated ML. The as-deposited film appears in brown–yellow color and is transparent. The region closest to the Zr target displays the most contrast. After the first annealing at 460 °C for 1 h in O2, this contrast has reduced, but the ML still retained the brown–yellow color. The second annealing at 900 °C for 1 h in air resulted in complete color loss and film transformation into a transparent state. Photos of the ML were taken against a patterned background representing a grid of 342 MAs. Additionally, a black background was used for the photo after the second annealing to enhance the optical contrast. Three regions along the Zr-Na/P gradient, distinct in terms of color and surface roughness, can be distinguished on the ML in the final state, denoted as A, B, and C in Fig. 1.

Figure 1
figure 1

Photographs of as-deposited and subsequently annealed ML along with the target configuration relative to the substrate. Black background was used to enhance the contrast and reveal three optically distinct regions which developed after the second annealing, marked as A, B, and C.

Full size image

EDX elemental maps of Na, P, Si, and Zr in the final state of the ML, i.e., after the second annealing, are presented in Fig. 2. Oxygen signal was recorded and included into the data analysis, but not quantified due to superposition with a much stronger O peak originating from the Al2O3 substrate. With the four elements Na, P, Si, and Zr used in quantification, the covered composition ranges are 11–61 at. %, 0–48 at. %, 1–52 at. %, and 3–80 at. %, respectively. MAs with the targeted composition Na1+xZr2SixP3−xO12, with 0 ≤ x ≤ 3, are highlighted in Fig. 2b. In Fig. 2c, ± 5 at. % deviation of x is allowed, in order to also include over- and under-stoichiometric NASICON compositions. The sputter powers on the individual targets were adjusted to ensure that the region of the aimed NASICON compositions would be located in the central part of the ML, thus taking full advantage of the composition gradients in the NASICON phase as well as elucidating on the competing phases forming alongside the NASICON phase at off-stoichiometric conditions in the surrounding regions. MA166, located in region B, represents a stoichiometric NASICON composition of Na2.4Zr2Si1.4P1.5O12 and, together with MA273 from region C, has been selected for further TEM analysis. The compositions of the MAs indicated in Fig. 2c are normalized by Zr = 2 and assume stoichiometric occupancy of O.

Figure 2
figure 2

EDX of the ML in its final state after the second annealing. a Elemental maps of Na, P, Si, and Zr along with MAs corresponding to Na1+xZr2SixP3−xO12 composition with b 0 ≤ x ≤ 3 and c allowing ± 5 at. % deviation in order to account for over- and under-stoichiometry. End compositions along the Si gradient as well as compositions of MA166 and MA273 selected for further TEM analysis are also indicated.

Full size image

The previously discussed formation of the three distinct optical regions developed on the ML after the second annealing can be correlated to the elemental EDX maps in Fig. 2a. The region A coincides with the highest Zr content, while the region C can be related to the highest contents of Na and P. The region B located in-between corresponds exactly to the NASICON compositions, as shown in Fig. 2b, except for the lower part closest to the Si target and the resultant oversupply of Si.

XRD measurements carried out on a few selected MAs (close to each target and in the center of the ML) after the first annealing indicated amorphous material. After the second annealing, full XRD mapping has been performed. Each of the three optically distinct regions A, B, and C displays a characteristic XRD pattern, shown in Fig. 3. In region A, the main phase is cubic ZrO2 (ICSD-36700), although some NASICON phase (ICSD-473) has also formed. In region B, exclusively NASICON phase was detected. The NASICON peaks have grown in intensity and exhibit narrower peak widths, indicating larger crystallite sizes. Full diffraction rings observed in 2D XRD patterns in both regions A and B reveal the polycrystalline nature of the film. In contrast, in region C the NASICON phase develops preferential texture. Furthermore, formation of SiO2 (ICSD-41412) and Na3PO4 (ICSD-91565) in the region C can also be concluded. Figure 3e and d shows the correlation between the phase distribution on the ML and the EDX composition gradient. ZrO2 is the dominant phase in the Zr-rich region, while Na3PO4 emerges in Na and P excess conditions. Presence of SiO2 is a consequence of no Si being incorporated in the textured NASICON phase in region C, as revealed by TEM–EDX analysis. In the central region B, the relative elemental contents are in correct proportions to support formation of phase-pure NASICON film, except for the Si-rich part closest to the Si target.

Figure 3
figure 3

Representative 2D XRD along with the integrated patterns obtain from the regions a A (MA221), b B (MA230), and c C (MA293) with the selected MAs marked in d the XRD overview map and e EDX pie-chart diagram indicating the relative elemental compositions of Na, Zr, Si, and P at each of the 342 MAs in the final state of the ML.

Full size image

Surface morphologies of the two selected MAs from representative NASICON regions B and C are compared in Fig. 4. MA166 exhibits fine surface roughness with occasional surface cracks up to 4 μm in length. SEM imaging induced significant electron beam damage, particularly in MA166. An example of a post-imaged area is shown as an inset in Fig. 4a. The film has swelled forming surface bubbles, some associated with protruding metallic Na features (as determined by correlative EDX) preferentially forming at the perimeter of the investigated area. Solid electrolytes are typically sensitive to electron and ion beams due to high mobility of ions under the influence of electric fields, resulting in structural and composition alterations [17]. Relating to a textured 2D XRD, MA273 displays a distinctively different surface with grain outlines clearly visible. The film is composed of well-distinguished individual irregular hexagonal shape platelet-like grains up to 10 μm in width, preferentially oriented with the flat facet lying parallel to the surface.

Figure 4
figure 4

SEM surface overview images of a MA166 and b MA273. The inset in a illustrates swelling of the film and formation of metallic Na features (bright contrast) caused by electron beam damage for the imaged area.

Full size image

Overview TEM images of MA166 and MA273 are shown in Fig. 5. Both coatings are dense and well adherent to the substrate. As concluded from SEM, microstructures composed of fine randomly oriented grains versus preferentially textured large platelet-like grains are revealed in cross section for MA166 and MA273, respectively. The film thickness is 500 nm for MA166 and 975 nm for MA273, a difference which is to be expected on a ML with composition gradients and related thickness gradients. The thickness of the large grains in MA273 measures up to 810 nm, i.e., a single grain may extend almost throughout the entire film thickness. The black layer on top of the film in MA166 is a protective Au layer, which has been partially removed in MA273 during FIB sample preparation. The layered contrast within the film in MA166 arises from the deposition process which was interrupted mid-way.

Figure 5
figure 5

TEM cross-section overview images of a MA166 and b MA273.

Full size image

Qualitative scanning (S)TEM–EDX maps in Fig. 6 show the constituting elements and their distribution in the films. In accordance with literature [17], even though the film structural integrity has been retained, FIB sample preparation with Ga ion beam led to severe Na loss as compared with SEM–EDX results, including Na ion migration into the protective carbon layer. The major qualitative difference between the two MAs is Si incorporation into the NASICON films. In MA166 Si is homogeneously distributed throughout the NASICON film with no signs of other Si-containing phases in neither (S)TEM–EDX elemental maps nor XRD. In contrast, the NASICON grains in MA273 comprise Na-Zr-P-O. Si is not accommodated in the NASICON structure and accumulates as SiO2 into a continuous ~ 66 nm layer at the film/substrate interface as well as in the spaces between the grains. The grain displaying the brightest contrast in Fig. 5b MA273 is Na- and P-rich, which, taking into account ~ 60% Na loss during standard RT FIB sample preparation [17], is in agreement with the composition of Na3PO4 phase identified in XRD.

Figure 6
figure 6

Qualitative (S)TEM–EDX maps showing constituent elements of the NASICON phase in a MA166 and b MA273.

Full size image

Discussion

With the aim to improve the process flexibility and explore large compositional ranges, we performed combinatorial depositions using Zr and Si elemental targets instead of both elements contained in a Zr-Si alloy target, as was the case in all previous Na2O-ZrO2-SiO2-P2O5 thin film studies [11,12,13,14]. A Na3PO4 compound target was used as a source for both Na and P, which resulted in a fixed Na : P ratio throughout the ML and, consequently, attainable stoichiometric NASICON compositions falling into a certain predefined range. However, separating Na and P into elemental sputtering targets is not feasible for a magnetron sputtering process.

In addition to standard structure and composition characterization with XRD and EDX, we present TEM analysis complimented with qualitative EDX elemental maps of Na1+xZr2SixP3−xO12 thin films. Precautions have been taken against ion and electron beam exposure damage during FIB sample preparation, TEM imaging, and (S)TEM–EDX spectra acquisition; however, some progressive damage of the films has still occurred. Even though quantitative (S)TEM–EDX evaluation could not be performed due to severe Na loss, qualitative observations still illustrate the importance of local composition analysis of individual grains, as opposed to relying solely on the global composition and phase constitution characterization methods. Here, (S)TEM–EDX reveals difference in the constituting elements, Si specifically, of the NASICON phase formed in two separate regions on the ML. This is of importance due profound Si effect on Na1+xZr2SixP3−xO12 performance as solid-state electrolyte and could otherwise not be concluded based on XRD and SEM–EDX results alone.

The difference between the two investigated MAs in terms of Si accommodation within the NASICON structure may relate to their respective elemental compositions. Si : P ratio at MA273 is 1 : 8.7, while almost equal Si and P fluxes arrive at MA166. In the Na1+xZr2SixP3−xO12 structure, Si and P share the same site on the crystal lattice. P being in oversupply in MA273 may act as the preferred element for the shared site, while the little amount of available Si is expelled to the grain boundaries.

Conclusions

The composition space Na2O-ZrO2-SiO2-P2O5 was explored using co-sputtered thin film MLs with focus on formation of compounds with stoichiometry Na1+xZr2SixP3−xO12. The NASICON phase was observed for all compositions synthesized using optimized conditions, displaying a phase-pure region in the central part of the ML. Closest to Zr and Na3PO4 targets, the phases ZrO2 and Na3PO4 were also present, respectively. The film microstructure and Si incorporation into the NASICON phase were strongly influenced by the incoming elemental fluxes. The phase-pure NASICON region in the central part of the ML consists of fine grains and covers a Si gradient of x = 0.4—2.9. In contrast, Na-P excess conditions lead to formation of large textured Si-free NASICON grains.

Conflicts of interest

The authors declare no conflict of interest.