Challenges and solutions in RF sputtering of superconducting Nb for nanostructuring processes Open Access

In basic research as well as in applied physics or engineering, there exists an increasing demand to interface low-dimensional systems, such as graphene with superconducting materials for nanostructures or superconducting electrical electrodes.^1–5 A typical superconducting material candidate is aluminum. Aluminum has a melting point of only 660 $°$C and can be deposited with relative ease by physical vapor deposition (PVD) from a heated crucible in a final processing step onto a two-dimensional material. However, the choice of a superconductor depends on the operating conditions of the fabricated, final device, such as experimental temperature $T$ and magnetic field $B$⁠. While a critical temperature $T c$ close to $T$ can offer high photodetection sensitivity in transition-edge sensors that operate at the boundary between the superconducting and normal states,^6,7 for the experimental study of a Josephson junction, $T c$ should exceed $T$⁠. Superconducting materials with experimentally more convenient critical fields and critical temperatures, such as niobium (Nb), niobium nitride (NbN), niobium titanium nitride (NbTiN), or molybdenum nitride (MoN), can be deposited via molecular beam epitaxy (MBE),^8,9 chemical vapor deposition (CVD),^10,11 atomic layer deposition (ALD),^12,13 or sputtering.^14–18 Here, the growth parameters determine the superconducting properties, such as the critical temperature $T c$ and critical magnetic field $B c$⁠.

Superconducting thin films are widely used in the development of novel nanostructured quantum devices. In many cases, the superconducting material must be deposited as one of several fabrication steps. Sputter deposition, a subtype of PVD, has more compatible growth parameters and is less disruptive to the thin van der Waals material or the resist than most deposition methods due to much lower process temperatures and shorter deposition times. Instead of using precursors or evaporating the base material in a crucible, sputter deposition is based on the ejection of the material from a source target by ion bombardment.^19–21 The substrate is often heated to high temperatures to obtain higher adatom mobility and subsequently larger grains and higher thin-film quality.^22–24 Analogous to CVD and ALD processes, these temperatures are incompatible with typical lithographic fabrication processes.

Here, we give practical and simple solutions for the sputter deposition of high-quality nanostructured superconducting Nb thin films. Our systematic study highlights that the quality of thin films and their superconducting properties are dictated by the sputter parameters and lithographic preparation, rather than by a high-end deposition system. The introduction of a lithographically defined mask structure by a PMMA resist generally tends to deteriorate the superconducting film quality due to contamination from outgassing. However, we show that relatively simple modifications to the lithographic process can restore thin-film quality. Our approach is aimed at fundamental research with limited access to advanced and high-end growth and characterization facilities. We provide guidelines to produce and optimize high-quality thin films suitable for nanostructuring, even with limited resources, that are transferable to other sputtered superconducting materials and applicable to most 2D materials. Further optimization is required for designing specific nanostructures, such as junctions, interfaces, or electrical contacts.

Nb films are prepared by RF sputtering in a modified Torr International Compact Research Coater System (CRC 600) as schematically shown in the inset of Fig. 1(b). For all sputtering processes, small (0.5 cm $2$⁠) pieces of an $n$-doped silicon wafer with native oxide are ultrasonically cleaned in acetone and isopropanol, followed by mild oxygen plasma treatment to remove any organic contaminants. The plasma treatment was conducted a few days before the deposition to prevent the introduction of reactive oxygen species into the deposition process. The substrates are placed on a rotating metallic sample holder inside the sputtering chamber, which is pumped below $5× 10 − 5$ Torr prior to the sputter deposition. At ambient temperature, films are deposited from a 2 in. pure Nb target (99.95  $%$⁠) in an Argon (Ar) atmosphere provided by the flow control unit. After plasma ignition, the chamber pressure and the plasma power are brought to deposition conditions and stabilized for several minutes before the actual deposition. This period serves to clean the Nb target and condition the chamber, which is important to reduce run-to-run variations.^28,29

The deposition rate is regulated by setting power and argon flow and ranges from 30 to 70 nm/min. Lower deposition rates facilitate manual operation of the shutter to achieve the desired thickness. Deposition rates have been determined by measuring film thickness on test runs with a Bruker Dektak surface profilometer at a step between the bare substrate and the film. No lithographic preprocessing is needed; an area of the substrate is simply covered with a marker pen, lifted off post-deposition to create a film edge, and the thickness is measured through multiple height profiles along this step.

For the fabrication of superconducting Nb nanostructures, we use standard electron beam lithography (EBL) with a PMMA resist (AR-P 672.045, Allresist). In an alternative lithographic process that prevents the contamination by organic polymers through the resist, we use aluminum (Al) instead of a PMMA mask. First, we cover the entire sample with approximately 800 nm of Al using thermal evaporation. The sample is then coated with PMMA, and the desired structure is written by EBL. After development, the alkaline TMAH-based photoresist developer (MF-319, microresist technology) is used to remove Al and impose the pattern of the PMMA into the Al layer. After removing the PMMA, Nb sputter deposition can proceed without any organic polymers on the sample. The finished superconducting Nb structures are obtained after lift-off in the same alkaline developer, which selectively etches the Al.

Nb thin films are electrically characterized in a physical property measurement system (PPMS) DynaCool by Quantum Design from 300 K down to 2 K and up to 9 T using a standard four-terminal measurement scheme [the inset in Fig. 1(a)] to determine the critical temperature, $T c$⁠, and an upper critical magnetic field, $B c$⁠, that serve as the figures of merit for the film quality. An exemplary measurement to evaluate the critical temperature of a deposited thin film is presented in Fig. 1(a).

When the thickness of a superconducting film is comparable or less than the coherence length and the superconducting penetration depth, quantum size effects affect the superconducting properties because Cooper pairs cannot form effectively in confined dimensions. The thickness dependence shown in Fig. 1(b) is a combination of increasing grain sizes that exhibit properties closer to the bulk value^25,26,30,31 and of stress and strain effects when niobium thin films are grown on substrates such as silicon.^32–34 While most reports on sputtered Nb (and NbN films) use 200 nm or larger^35,36 to reduce these effects, for our studies, we consider niobium films with thicknesses larger than 60 nm, which is a common thickness for electrical contacts.⁴ For the Nb film with the largest thickness, the $T c$ of 7.6 K is closest to that of single crystal bulk Nb, which has a $T c$ of 9.25 K.^25–27 The deviation from the bulk value could be the result of residual oxygen content in the sputter chamber, which is known to impact the superconducting properties of sputtered Nb films.³⁷ 

Each data point in Fig. 2 represents the best result obtained for a fixed $P$ but within a parameter range for the layer thickness (60–100 nm) and argon pressure (3.4–6.6 mTorr). These slight re-adjustments for each $P$ are required due to a complex interdependence of the growth parameters to optimize the films. We found that lower argon pressures (3.4–4.3 mTorr) are required for a lower power to achieve good Nb thin films. The quality and reproducibility of most deposition techniques, including sputter deposition of superconducting Nb, are also significantly influenced by the chamber’s condition and growth history. We will discuss the relevance of these factors on the obtainable $T c$ in more detail later.

The increase of the RF power in Fig. 2 results in a steady rise of the critical temperature, indicating a continuous improvement in film quality. The cleanest Nb thin films and, thus, the highest $T c$ are achieved at 200 W, followed by a slight decrease at higher powers. The power dependence can be understood via the growth kinetics: at ambient temperature, adatom mobility is rather low and their binding affinity is comparably high, which generally promotes island growth.³⁸ Increasing the kinetic energy of the sputtered particles leads to more uniform films as it allows for a redistribution of adsorbed atoms, thus effectively reducing the likelihood of island growth. At very high powers, however, detrimental effects, such as re-sputtering and amorphization,^39,40 begin to deteriorate the quality of the films.

The room temperature resistance $R 300 K$ of the sputtered films is dominated by electron–phonon scattering, while the resistance at low temperatures above $T c$ is dominated by impurity scattering that affects the superconducting properties. The residual resistance ratio (RRR), defined as $R 300 K R 15 K$ [Fig. 2(b)], is commonly used and provides a standardized method to compare different films.^14,27 While bulk Nb can have a RRR above 2000, Nb films often display a lower RRR $<10$⁠.^25–27, Figure 2(a) demonstrates that RRR shares a similar dependence on power with the $T c$⁠, indicating that $R 300 K R 15 K$ can indeed serve as a quality figure of merit for the thin superconducting films. However, the experimental determination of this ratio requires a cryogenic system, offering no real advantage over directly determining $T c$⁠. While previous reports on the effect of residual oxygen on Nb films observed a clear dependence between $T c$ and $R 300 K$⁠,^26,37 we found that for our study of RF power dependence, $R 300 K$ alone is not a reliable quality indicator, which could result from other effects beyond residual oxygen, such as grain size. However, its ratio with $R 77 K$ (measured simply in liquid nitrogen) is already sufficient to evaluate the superconducting properties, as shown by Fig. 2(c).¹⁴ Hence, in scientific environments where access to liquid helium or cryogenic systems is limited, characterizing the resistance at 77 K can already screen suitable thin films. In the following, we will not focus on RRR but on the more commonly used $T c$⁠.

Our study is driven by the need to achieve high-quality superconducting thin films in hybrid nanostructures, such as electrical contacts for graphene devices encapsulated in hexagonal boron nitride (hBN).² Self-aligned one-dimensional edge contacts⁴¹ to these devices can be nanofabricated by EBL to define contact areas within a layer of a PMMA resist, followed by reactive ion etching (RIE) to locally expose the graphene layer,⁴² and subsequent sputter deposition of Nb. Typically, the thickness of metallic contact materials to graphene devices is below 100  nm,^41,43,44 thereby also setting our superconducting film thickness range between 60 and 100 nm, as indicated in Fig. 1(b).

With the standard parameters (200 W RF power and 6.5 mTorr Argon), we consistently obtain a critical temperature of approximately 6 K for large lateral thin films, which we now define as pristine. Repeatedly, we also encountered a significantly lower $T c$ for samples that were sputtered with the same parameters as shown in Fig. 3(a). We can pinpoint the deterioration of the superconducting properties to the lithographic process and the history of the chamber. In Fig. 3, we compare the superconducting properties and the resistive behavior of pristine samples to reference samples.

In materials science, impurities and contamination generally lead to issues, such as reduced mobility in the MBE growth of III/V semiconductors or high resistive, non-linear metallic electrical contacts in PVD. Notably, the observed significant sensitivity of Nb to impurities³¹ from a PMMA resist during sputtering dramatically impacts the critical temperature, limiting the temperature range for superconducting graphene devices to below the liquid helium temperature of 4.2 K. The reproducibility also depends on the history of the chamber and growth frequency. Despite a cleaning and conditioning procedure described in Section II prior to deposition to remove oxides and most other contaminants from the target surface, Fig. 3(a) illustrates a clear dependence on the frequency of deposition. Best and most consistent results are generally obtained at frequent depositions. Less frequent and intermittent depositions yield significantly lower critical temperatures. As discussed earlier, the niobium films are also very sensitive to residual oxygen levels in the sputter chamber.³⁷ We, thus, suspect that frequent deposition keeps the target conditioned, with minimal oxidation, leading to more consistent results and higher film quality. For less frequent and intermittent depositions, the conducted initial cleaning procedure is not sufficient to yield ideal starting conditions. Our results highlight the importance of a well-conditioned chamber. Conditioning protocols prior to sputter deposition can significantly reduce run-to-run variations.^28,29 Heating of the sample stage, which is often included in such conditioning protocols,²⁹ as well as the application of a sample bias voltage to reduce oxygen content in the Nb films³⁷ are not available in our experiments.

Increasing the growth frequency can indeed enhance the superconducting properties. However, addressing issues associated with lithographic processes, which are required for nanostructured graphene devices, for example, are much more challenging. When we compare the results of the pristine sputter depositions to that of PMMA reference samples, we find that those samples always show inferior properties, independent of the growth frequency. Note that the reference samples PMMA were neither coated with PMMA nor nanostructured in any way. The PMMA-labeled samples are, thus, niobium films on clean silicon substrates that were sputtered together with other PMMA-coated substrates.

In Fig. 3(b), the critical temperatures for a pristine (thin dashed lines) and a PMMA sample (thick solid lines) are determined at different magnetic fields. The two blue traces representing 0 T show a significant decrease of approximately 30 % in $T c$ from 5.5 K (pristine) to roughly 3.8 K for the PMMA. While the PMMA resist retains its integrity during the sputter deposition and can be easily removed in a lift-off process, outgassing of PMMA or an interaction of Nb species having high kinetic energies in the plasma with the PMMA leads to enhanced incorporation of impurities in the superconducting thin film. The impurities not only decrease $T c$ but also affect the sensitivity to magnetic fields. The $T c$ of the pristine sample in Fig. 3(b) decreases more in response to the applied magnetic field compared to the $T c$ of the PMMA sample. Nevertheless, the magnetic field dependence in Fig. 3(c) shows a significantly higher upper critical field for the pristine sample. The underlying physical details are complex and not the scope of this paper. However, the different magnetic field dependencies in Fig. 3 are consistent with previous reports and relate to impurity scattering that affects the Cooper pair formation and the interplay of flux pinning at defects under external magnetic fields.^45,46

To mitigate the degradation of Nb thin-film quality, in Fig. 4(a), we assess the feasibility of different nanostructuring approaches by comparing the averaged critical temperatures $T c , avg$ and film stress, determined from an ensemble of samples. On average, a pronounced drop of almost 40% is observed in the $T c , avg$ between a pristine sample and sputter processes involving a PMMA resist, displayed as blue and red data points, respectively. However, $T c , avg$ significantly increases when the PMMA resist is hard-baked (PMMA+HB) previous to the Nb deposition (orange data point). This suggests that resist outgassing is responsible for the degradation as additional baking reduces the amount of solvent residues and hardens the resist. We could not observe any further improvement of the superconducting properties when using alternative EBL resists that are advertised as resilient against higher temperatures and plasma processes.

The differences in the crystallinity and morphology of the Nb films can directly be visualized by atomic force microscopy (AFM). Figures 4(c) and 4(d) compare $5×5$  $μ$m $2$ AFM images from a pristine film and a film that was sputtered together with a PMMA-coated substrate. While the topographic images reveal dense and homogeneous films for both samples, visibly larger grains and a higher root mean square (RMS) are found in the pristine sample. We attribute the larger surface roughness of the pristine film to larger crystallites and, thus, a higher level of crystallinity,³² which is in agreement with our resistance and XRD measurements.

Achieving high-quality superconducting Nb thin films through sputter deposition can be challenging, especially when incorporated into larger nanostructuring schemes for two-dimensional materials. Our study focused on preserving the quality and superconducting properties of RF-sputtered Nb thin films, intended for the application in nanostructuring processes that involve 2D materials such as graphene. The feature sizes and demands for specific superconducting nanostructures, such as electrical contacts or small junctions, can vary considerably and will require further adjustments of the process parameters. However, we demonstrated that advanced growth and characterization facilities are not required to obtain competitive films, and we provided guidelines that are transferable to other superconducting materials and 2D material systems and hybrid superconducting devices.

Here, we investigated the electrical and structural properties of RF-sputtered Nb thin films using electrical transport, AFM, and XRD measurements, which demonstrate their sensitivity to small changes in growth parameters, chamber history, and lithographic processes. However, very simple and practical solutions can be employed to restore or maintain high-quality films. A standard lift-off technique based on a PMMA resist leads to contamination and dramatic deterioration of film quality; therefore, we propose an aluminum mask-based approach to address this issue. By comparing the $T c$⁠, the $R 300 K R 15 K$ ratio and the resistance ratio $R 300 K R 77 K$⁠, we also demonstrate that preliminary screening of Nb films at liquid nitrogen temperatures is already sufficient to identify samples for a more detailed characterization in a cryogenic environment. Reproducibly fabricating high-quality superconducting films and nanostructures is not reserved for high-end sputtering facilities; rather, it depends on careful tuning of growth conditions.

The authors would like to especially thank Professor Koziej and D. Derelli for their help and support with the XRD measurements and R. Venugopal for conducting the AFM measurements. We would also like to thank S. Haugg and T. Finger for their expertise in operating the sputter chamber. We acknowledge support by the Deutsche Forschungsgemeinschaft (DFG) with Grant No. BL-487/14-1 and the BMBF project “NOVALIS” under Grant No. 05K22GUD.

The authors have no conflicts to disclose.

Vincent Strenzke: Conceptualization (equal); Data curation (lead); Formal analysis (lead); Investigation (lead); Writing – original draft (lead); Writing – review & editing (lead). Annika Weber: Data curation (equal); Formal analysis (equal); Investigation (equal); Writing – review & editing (supporting). Peer Heydolph: Data curation (equal); Formal analysis (equal); Investigation (equal); Writing – review & editing (supporting). Isa Moch: Data curation (equal); Formal analysis (equal); Investigation (equal); Writing – review & editing (supporting). Isabel González Díaz-Palacio: Investigation (supporting); Writing – review & editing (equal). Wolfgang Hillert: Project administration (supporting); Supervision (supporting); Writing – review & editing (supporting). Robert Zierold: Conceptualization (supporting); Funding acquisition (equal); Supervision (supporting); Writing – review & editing (equal). Lars Tiemann: Conceptualization (lead); Funding acquisition (equal); Supervision (lead); Writing – original draft (equal); Writing – review & editing (equal). Robert H. Blick: Conceptualization (supporting); Funding acquisition (equal); Project administration (lead); Supervision (equal); Writing – review & editing (supporting).

The data that support the findings of this study are available from the corresponding author upon reasonable request.