A Coaxial Nozzle Attachment Improving the Homogeneity of the Gas Flow Sputtering

Abstract

The Hollow Cathode Gas Flow Sputtering (GFS) provides special plasma conditions and is of extensive interest as a more affordable alternative to the high vacuum sputtering techniques. In the case of the tubular cathode a circular outlet symmetry stipulates homogeneity issues for both metallic and reactive deposition regimes. Using the results of Direct Simulation Monte Carlo (DSMC), we propose an external coaxial attachment which is manufactured and examined in a nozzle and a diffuser positioning. The impact on the homogeneity of Ti and ${\mathrm{TiO}}_2$ films is examined using profilometry and spectral ellipsometry. Our results demonstrate that the use of the nozzle attachment significantly enhances film homogeneity from about 3 ${\mathrm{cm}}^2$ to more than 12 ${\mathrm{cm}}^2$. It also secures better process control in terms of oxygen stoichiometry and film thickness. Some crucial general issues of the reactive GFS process are discussed.

1. Introduction

Sputtering is one of the most significant deposition methods utilized for depositing thin films across various applications (industrial coatings on glass, wear-resistant coatings, photovoltaics, electronics and others) [1,2,3]. Among the sputtering techniques, magnetron sputtering is widely employed in the industry but is prone to target poisoning due to hysteresis behavior during reactive sputtering, which also impacts the deposition rate [4,5,6]. Gas flow sputtering (GFS), based on hollow cathode discharge, demonstrates several key advantages over conventional low-pressure sputtering techniques. Among the practical ones are much higher deposition rates and much lower costs of systems construction, the reactive gas inlet is placed outside the cathode, which prevents poisoning of the cathode surface (Figure 1b) [7,8,9,10]. GFS takes advantage of the thermalization of sputtered atoms and subsequent transport to the substrate by forced convection, there is no high energy impact to the substrates used and layers deposited, which is usually assessed as plasma damage [11,12,13]. It provides a high plasma density compared to magnetron sputtering and an adjustable, comparably low kinetic energy of ions (about kT up to few eV) using pulsed plasma operation [13,14,15,16,17,18,19,20]. High-pressure metallic mode sputtering combined with gas flow assistance facilitates exceptionally high deposition rates compared to alternative sputtering methods [1,8,21].

In typical hollow cathode GFS, the laminar argon flow has the highest velocity in the middle of a cathode [22]. This causes inhomogeneous mass transport across the outlet and radially thinning metallic films [9,23,24]. This problem is well-known and addressed in the literature to the multiple cathode systems or to the optimized cathode dimensions [25,26,27], which makes the system more complex with added costs. A larger target-to-substrate distance can improve the homogeneity of the films but simultaneously causes their larger porosity [24,28].

Moreover, the overall process control deteriorates due to (i) a significant drop in sputtering gas pressure at the outlet and (ii) a spread or unsteady gas stagnation point.

Direct simulations Monte Carlo (DSMC) are widely used in industry to optimize gas dynamics in vacuum processes [29,30], to simulate charge transfer in plasmas [31,32,33], and to upscale deposition systems [29,34,35]. Pressure ranges from low pressures < 1 Pa to high pressures > 100 Pa can be simulated [36,37,38]. We exploited this tool to conduct speculative homogeneity improvements. Various coaxial accessories were of interest. Titanium oxide was taken as one of the most investigated and widely used oxides in technology. Its microstructure and oxygen stoichiometry are known to correlate with optical properties strongly.

This study is aimed to optimize gas dynamics in the GFS deposition process. The use of different coaxial accessories matching hollow cylindric targets is simulated by DSMC and benchmarked. To approve the simulated findings, the homogeneity of Ti and ${\mathrm{TiO}}_2$ films is evaluated by profilometry and spectral ellipsometry.

2. Methodology

This work used the hollow cathode gas flow sputtering system as shown in Figure 1, the argon gas supply is behind the sputter source (1) the substrate is fixed on a heatable substrate holder (2) which can be also heated from the treatment station (3). Figure 1b shows the Principle of Gas Flow Sputtering, by applying a voltage between the hollow cathode and the anode, a glow discharge with a high plasma density occurs inside the hollow cathode. The sputtered material is effectively transported by argon gas flow to the outlet of the hollow cathode towards the surface of the substrate, where ${\mathrm{O}}_2$ is realized by reactive deposition (which prevents interaction with the target surface), and the sputtered atoms can react with some of the offered ${\mathrm{O}}_2$ and build the film at the substrate surface. A diffuser and nozzle attachments were fixed at the outlet of the GFS source as shown in Figure 2. The parameters for titania deposition were based on our previous experience [11]. Since the diffuser accessory blocks the reactive gas inlet (Figure 2b), we did not use it further. A 3.1 mm thick low-iron soda lime glass type Diamant (Saint-Gobain Glass, Cologne, Germany) and 50 × 50 ${\mathrm{mm}}^2$ size was used as a substrate. The glass substrates were cleaned with acetone and then dried with nitrogen gas before deposition. Depositions were conducted without intentional heating. High-purity (99.99% Ti) tubular targets of 60 mm length, 50 mm outer, and 40 mm inner diameters were taken. A discharge power was 500 W in pulse mode with 100 kHz frequency, 2 µs pulse duration, and +50 V reverse voltage. These settings were found to deliver denser titania films at low deposition temperatures. Base and process pressures were 1.4 × ${10}^{-2}$ mbar and 0.35 mbar, respectively. Argon gas flow of 1000 sccm was used for sputtering. Oxygen gas flow was set after optimization based on optical transmittance to 80 sccm for the blank outlet and to 5 sccm for the nozzle outlet. The distance between oxygen inlets and a substrate was set to 25 mm in both cases. All films were deposited for 150 s.

The thicknesses of deposited Ti thin films were evaluated by profilometer (DektakXT, Bruker, Billerica, MA, USA), where it is necessary to have a step in the film profile on the substrate, which was achieved using marked with edding before the deposition and removed afterward with acetone. The optical properties of the ${\mathrm{TiO}}_2$ films were characterized by the Sentech Ellipsometer SE 850 with a µ-spot (which focuses the beam to 200 microns in diameter at the polarizer unit), the measurements were carried at 70° incident angle, and the data were measured at a spectral range between 280 and 850 nm, the measured refractive index is shown at $\lambda$ = 550 nm.

The Direct Simulation Monte Carlo has been applied for the GFS deposition process. For this purpose, the Linux 96xCPU cluster and the elaborated simulation software developed by Dr. A. Pflug (Fraunhofer IST, Braunschweig, Germany) [39] were used. The geometric model of our system was built by the finite-element mesh generator: Gmsh^® [40]. The Direct Simulation Monte Carlo (DSMC) to model the gas flows and distribution of neutral particles was used [41]. We used a 2D model instead of a 3D model to reduce the simulation time and adapt the simulated deposition parameters to this circumstance. This approach was validated in our previous works [38,42]. Following this approach, we set 2D-flows as 31.83 sccm for Ar gas and 2.547 sccm for ${\mathrm{O}}_2$ gas. The relevant length of a substrate holder was set to 100 mm and the target-to-substrate distance was the same as it was in depositions.

3. Results and Discussion

To improve the homogeneity of metal films, the so-called diffuser accessory (see Figure 2b) was proposed and approbated using DSMC. The optimum opening angle ($\alpha$) was found to vary in the range 60–80° [43], but overall homogeneity even degraded compared to the bare source (as will be seen later in the Deposition section). The opening angle $\alpha$ used in this work was 67°. Considering the former experience of successful use of the coaxial gas nozzle in metal-organic chemical vapor deposition (MOCVD) reactors [44], we applied the very same accessory, positioning it inversely (Figure 2c) which resulted in much better homogeneity. Furthermore, this geometry made the reactive sputtering possible as well.

3.1. Simulation

It can be seen from the gas simulation (Figure 3) that the argon pressure inside the hollow cathode becomes higher if the nozzle attachment is used. The oxygen pressure over the substrate surface is also higher in this case. Higher pressure implies, larger diffusion gradients toward the substrate and more intense bombardment of the surface by ions.

To characterize the spreading of gases more quantitatively we extracted the (p(${\mathrm{O}}_2$)/p(Ar)) ratio profile from the simulations and measured the oxygen partial pressure (p(${\mathrm{O}}_2$)) directly (Figure 4).

The latter was conducted with the help of oxygen Frontal ZIROX E2010.B0 which was placed over the substrate at various radial distances from the center. Since the measuring head was 10 mm large, we implied around 5 mm remoteness from the substrate surface. We note that the (p(Ar)) represents with some approximation a delivery of titanium species at a fixed discharge power. Both estimations demonstrate a remarkably effective and homogeneous delivery of oxygen to the substrate when the nozzle is used.

3.2. Deposition

Figure 5a presents the thickness values measured by profilometer across differently deposited metal titanium films (excluding the use of reactive gas). The x-axis represents a radial distance from the center of the sample. For quantitative comparison, the impacts of the outlet accessories on homogeneity are given using criterium h:

$$h=\frac{d_c-d_{20}}{d_c}\times 100\%$$

(1)

where $d_c$ is the thickness in the film’s center and $d_{20}$ is the thickness at a 20 mm distance from the center. According to Figure 5a the thickness decreases at 20 mm distance from the center by 43% using the diffuser and by 28%, without any outlet. The best homogeneity of a 6% difference in thickness was achieved by using the nozzle accessory.

A Ti target was used to deposit metallic titanium and titania thin films. Since we used a reactive GFS process, the oxide forms right on the substrate surface during film growth. This fact stipulates two features: (i) no plasma damage by accelerated oxygen ions and (ii) the importance of the mixing degree of argon and oxygen over the substrate. The insufficient homogeneity of deposited layers still hinders the wider application of the GFS method [17,23]. The problem of the metal-to-oxygen ratio in the middle of a deposition zone was signified in our previous work on titania films [11]. Figure 5b shows the ellipsometry measurements of the ${\mathrm{TiO}}_2$-films. A two-layer model was used for fitting the corresponding ellipsometry spectra in SpectraRay software. The model layer stack included one dense Brendel layer for the ${\mathrm{TiO}}_2$ film and one effective medium approximation (EMA) layer, consisting of 50% ${\mathrm{TiO}}_2$ and 50% air, which usually describes the roughness of the film. It can be seen that the ${\mathrm{TiO}}_2$-films deposited with the nozzle attachment revealed within a 40 mm circular area, remarkably better thickness homogeneity than the film deposited ordinarily. It is also valid for the refractive index. The refractive index is higher for the ordinary film in the center probably due to its oxygen under stoichiometry. Indeed, the refractive index is known to correlate with ${\mathrm{TiO}}_2$ density which increases with oxygen deficiency [45,46]. The absorption index k was found to be lower (0.00639) for titania deposited using the nozzle compared to the ordinary one (0.00917) in the frame of the optical model (see Figure 5b), which supports the suggested difference in the oxygen stoichiometry sample [47,48,49,50]. In sum, the nozzle attachment increased the homogeneous area of titania films from 3 ${\mathrm{cm}}^2$ to more than 12 ${\mathrm{cm}}^2$ and allowed to strongly decrease the reactive gas consumption.

4. Conclusions

In conclusion, we demonstrated a successful implementation of the gas pressure simulations to improve the GFS process. The coaxial nozzle attachment was found to significantly improve the film’s homogeneity in both metallic and reactive deposition modes. This concept was transposed from a 1 cm nozzle width in MOCVD to a 4 cm diameter of the tubular hollow cathode in GFS. It was surprisingly auspicious and allowed improving the film’s homogeneity from about 3 ${\mathrm{cm}}^2$ to more than 12 ${\mathrm{cm}}^2$. Much better process control in terms of oxygen stoichiometry and film thickness was reached for ${\mathrm{TiO}}_2$. This improved gas flow distribution leads to more uniform film thickness and refractive index. The question if any parasitic sputtering from the inner surface of such attachment occurs will be addressed in the upcoming investigations.