Study of process parameters and characteristics properties of W coatings deposited by rf plasma sputtering Open Access

In recent years, the interest in metallic tungsten (W) films is gradually increased, and this is motivated by their useful properties such as high electrical conductivity, high mechanical strength, good metal barrier behavior, high melting temperature, and simple patternability.¹ For these reasons, W coating is considered a versatile material, and it is characterized by a wide applicability range including magnetic sensors,² diffusion barrier coatings in CMOS technology,³ coatings in MEMS application,⁴ and coatings for the corrosion resistance for materials in the field of electrical engineering;⁵ very recently, W has been proposed as a possible alternative to Cu for semiconductor metallization.⁶ Moreover, the application field in which this material is mostly used is nuclear fusion, where W and its alloys have emerged as an important plasma-facing material (PFM) for the first wall and divertor coating due to their high sputtering threshold energy, low sputtering rate, low deuterium/tritium retention rate, low tritium permeability, in addition to the already mentioned mechanical and high temperature properties.^7–10 

Radio frequency (rf) plasmas used for thin-film metallic deposition are an emerging research field aimed at different industrial applications. As a consequence, characterization studies on rf discharges are fundamental to select the optimal plasma conditions for specific applications. Due to the importance and versatility merely described regarding depositions of W, this work focuses on a study to characterize the plasma in terms of electron density and electron temperature, during the formation of W films in the rf magnetron sputtering system. There are different characterization methods for the evaluation of plasma parameters. One of the most important diagnostic tools to do it is the Langmuir probe, LP.^11,12 This in situ intrusive characterization technique consists of a small bare wire inserted in the working plasma mixture. When the applied voltage is switched from a negative to a positive potential, the probe, which is characterized by well-defined length and diameter dimensions, collects charged carriers formed in the adjacent region to the wire called “sheath.” By appropriate theories,¹³ the collected I–V characteristic curve supplies quantitative information about electron temperature, electron density, plasma potential, the Debye length, and electron energy distribution function. In order to know the basic phenomena that occur in the apparatus used during the W deposition, some macroscopic parameters such as average cathode voltage and deposition rate as a function of the rf power input and working pressure were studied. Then, using LP technique, the electron temperature and density as a function of sputtering power and operative pressure were evaluated. Regarding the rf input power, an optimal working condition was found, and the film properties were evaluated as a function of pressure. The deposition of thick W films (in the range of hundreds of nanometers) was carried out to analyze how the plasma conditions modify the film properties (i.e., morphology, microstructural, and electrical characteristics).

The sputtering plant consists of a cylindrical stainless-steel vacuum chamber with a volume of 0.088 m³ (cylinder: diameter of 0.4 m and height of 0.7 m) equipped with two magnetron sputtering cathodes (balanced type,¹⁴ produced by Angstrom Sciences, Inc.) tilted at an angle of about 10° with respect to a vertical axis (Fig. 1). The cathodes are water cooled and connected to two separate rf (ω_rf/2π = 13.56 MHz) power supplies (300 W, TRUMPF Hüttinger), which operate in a steady state, coupled with an automatic impedance matching unit to minimize the reflected power. For all investigated experimental conditions, it generally does not exceed 2 W. The substrate holder, facing the targets, is grounded and rotating. A tungsten (W) target with a diameter of 3″ (purity: 99.9%, thickness: 0.25") was installed on the working cathode. Argon plasma (with a constant flux rate of 20 SCCM) in different pressure and power operative conditions was used. The working Ar flux and pressure were monitored by a mass flow controller (MKS) and a vacuum gauge (Pfeiffer single Gauge), respectively. The sputtering chamber was evacuated by a high vacuum pumping system, composed of a series of rotary (TRIVAC 24 m³/h, Leybold) and turbo-molecular (TURBOVAC 150 l/s, Leybold) pumps, which provided a base pressure of 1 × 10⁻⁴ Pa.

A 90° tilted Langmuir probe (Hiden ESPION) was used to characterize plasma parameters at different operative conditions. Measurements were performed at about 80 mm from the target using a negligible B field. The probe consists of a tungsten tip with a 0.15 mm diameter and 10 mm length. The probe is rf compensated and is designed for high frequency plasma. Hiden’s ESPION software was used for collecting the experimental results and analyzing different plasma parameters. A potential ramp from −20 to 35 V with a resolution of 0.2 V and a precleaning step before starting the data acquisition process were used. Each I−V trace was acquired in 5 ms (precleaning for 95 ms). The maximum probe current was limited to 100 mA. For this kind of plasma characterization, operative pressure was tuned from 1 to 5 Pa with different input power values (30, 50, 70, 90, 110 W). Plasma parameters such as electron density and electron temperature were carried out from five scans and averaged operation of each current–voltage (I−V) curve through the semi-automatics data analysis mode.

W coatings were deposited on silicon substrates [Si, p-type (100), 1 × 1 cm², thickness 400 μm], and no heating of the sample holder was used. The thickness of the deposited coatings was evaluated by covering with a silicon mask a portion of the samples during the process. Upon removing it, it was possible to measure the step formed between deposited and nondeposited portions of samples with a P15 surface profiler (KLA Tencor, San Jose, CA, USA) using a KLA profilometer (Tencor model). The surface morphology of the samples was characterized by Atomic Force Microscopy (AFM) (Core AFM, Nanosurf GmbH, Langen, Germany) in the dynamic mode. Scanning electron microscopy (SEM) (Tescan mod. MIRA III, Brno, Czech Republic)¹⁵ was used to analyze the surface and the cross section of the coatings. A x-ray diffractometer (Bruker mod. D8 Advance) was used with Cu-kα radiation source (λ = 1.54 Å) to study the grazing incidence x-ray diffraction (GI-XRD) patterns. The DC electrical resistivity (ρ_300K) of the W coatings was evaluated at room temperature by the four points probe measurement technique (the van-der-Pauw method) on a 1 × 1 cm² specimen.¹⁶ The coatings studied had the same thickness of 700 nm. Repetitive tests (four times for the sample) were executed to measure the electrical resistivity of the coatings using an HP Hewlett Packard 3457A.

In the first part of this study, the rf power (capacitively coupled) delivered to the target was changed in the range of 10–110 W. In this type of gas discharge plasmas, the term “capacitively coupled” refers to the way of coupling the input power into the discharge, i.e., by means of two electrodes and their sheaths forming a kind of capacitor. The primary sustaining mechanism is ionization due to electrons in the bulk plasma (so-called “volume ionization”). The electrons gain energy from the oscillating rf-electric fields (i.e., the rf sheath expansion and contraction), i.e., by the so-called Ohmic and stochastic heating. At the rf frequency used in the experiment, the electrons and ions due to their difference in mass follow the electric field in a different way. The electrons can follow the instantaneous electric field produced by the applied rf voltage, while the ions (due to the higher mass and therefore lower mobility) follow the time-averaged electric field. The average balance of electron and ion fluxes to the powered electrode results in a time-averaged negative bias at the rf-powered electrode.^17,18 Obviously, this effect phenomenon also happens at the grounded electrode, but it is much smaller. This average voltage is also referred to as the DC-bias or self-bias voltage V_dc, and it changes significantly with the power. As reported in the literature,¹⁹ a higher rf power gives rise to a higher self-bias voltage V_dc, which produces larger kinetic energy of the incident ions. Figure 2 shows V_dc as a function of the rf input power P in the argon discharge. In our experimental conditions, V_dc is independent of the operating pressure in the range of 0.4–2 Pa,²⁰ and it exhibits a dependence on P^1/2, a similar trend was also reported by other authors.^21–23 At higher pressure (for example, at 5 Pa), V_dc exhibits a dependence on ln P. As can be seen, V_dc at low pressure is, in general, higher (it is clear from 50 W and up) than that at high pressure. This is attributable to a low density of neutrals at lower pressures and, consequently, to a larger electron–neutral collision mean free path, which implies the demand for larger kinetic energy (which results in a higher V_dc) for the ionization of sputter gas atoms.

As mentioned before, since the effect of V_dc during sputtering is known to enhance the kinetic energy of positive ions (sputter gas), the deposition rate will tend to increase as a function of the power. Indeed, V_dc increase can lead to the enhancement of the momentum transfer to the atoms on the target, which means a larger sputtering rate effect, and, therefore, a greater deposition rate at a fixed pressure. The experimental data of Fig. 3 corroborate this view (below is an example using a working pressure of 2 Pa). It can be further noted a nonlinear trend of the data at higher powers, which may be attributable both to a nonlinear V_dc dependence at higher powers (Fig. 2) and the nonlinearity due to the energy-dependent sputtering yield.²⁴ 

In a similar way, the value of the V_dc also changes as a function of working pressure at a fixed power and it could influence the deposition rate, but in this case (90 W), the variation of V_dc in the range 1–5 Pa is quite low (a few volts); therefore, we correlate the variation in the deposition rate (Fig. 4) to the collision frequency and density of particles as a function of pressure. At low sputtering pressures ≤1 Pa, the deposition rate was found to be relatively low due to the fact that there exists a low density of sputtering gas, less bombardment, and, hence, a lower sputtering rate. At pressures around 2–3 Pa, there is a lot of energetic bombardment due to the high sputter gas density and, hence, an enhanced sputtering rate resulting in the greater deposition rate as described in theory.²⁵ At higher pressures >4 Pa, although there is a high density of sputter gas, the ion scattering phenomenon becomes predominant, sputtering rate decreases, and this justifies the deposition rate decreases. At the same time, due to increased collisionality with increasing pressure, the transport efficiency of sputtered atoms to the substrate decreases, consequently, a reduced deposited flux is expected at higher pressures.²⁶ 

In fact, for the sake of completeness, it should be taken into account that the deposition rate dependence on pressure is the result of more concurrent processes (a) the variation in density of particles, (b) the variation of the average electron energy, and (c) the variation of the ion energy.

As is well known, plasma parameters directly affect the feature of produced coatings. Plasma diagnostics is fundamental for defining plasma parameters and generating reproducible plasma in sputtering processes. Below measurements of electron density (n_e) and electron temperature (T_e) during the W sputtering processes are reported and correlated with the coatings deposition rate (nm/min). In order to study the influence of pressure on n_e and T_e, the pressure was changed in the range 1–5 Pa with a fixed power of 90 W. From the I–V characteristic, by means of the probe theory, n_e as a function of working pressure has been calculated (Fig. 5) by the electron saturation current at the plasma potential.²⁷ As a first observation, the trend of the density as a function of pressure (at fixed rf power of 90 W) reminds the trend of the deposition rate as a function of pressure (Fig. 4). For n_e, the trend as a function of pressure shows at first an increase up to a maximum at around 2 Pa and, subsequently, a decrease with increasing pressure. In the range 1–2 Pa, we could attribute this trend to the increase of Ar ions bombarding the target with a consequent increase of secondary electrons emitted from the target.^28,29 At the same time, the ionized particles are sufficiently energetic to sputter the target material with a high sputtering rate. As the plasma density showed a higher value at around 2 Pa, a better erosion rate (and consequent deposition rate increases) of the target resulting from increased ion flux (evaluated by the Bohm criterion³⁰) was expected.

As expected, the increase in pressure over 2 Pa leads to an increase in the collision frequency between electrons and neutral atoms with a consequent reduction of n_e and as not all the particles are sufficiently energetic to sputter the target material, the sputtering rate falls with increasing pressure.

Figure 6 shows n_e as a function of rf power at a fixed pressure of 2 Pa. The n_e behavior is characterized by a nonlinearly increment during power increasing (this trend is the same in the pressure range 1–5 Pa), showing good accordance with literature scientific results in these operative pressure conditions.^31,32 Even in this case, the deposition rate (Fig. 3) correlates with increasing density (ion flux), which leads to a better erosion rate of the target resulting in an increased deposition rate. Obviously, as indicated in Sec. III A, the deposition rate dependence is also related to the average kinetic energy of ions striking the target, which increases with V_dc (Fig. 2). It can be noted that a further increase in power (above 90 W) leaves the n_e substantially unchanged (n_e tends to saturate). This can be partially attributed to the decrease of the power transfer efficiency due to power losses in the matching network,³³ that in our experiment can be quantified in the order of a few percentage points.

As regards, the T_e, which is obtained from the slope of the logarithm of the electron current of I–V curve,^34,35 it can be seen that no substantial change is found in the T_e values as a function of power (Fig. 7), it decreased slightly because the increase in power led to a higher n_e in the plasma and, consequently, to more electron collisions, which induced higher ionization rates (resulting in electron energy loss).^36,37 Concerning the decreasing trend of T_e observed as a function of pressure (Fig. 7), it is mainly due to reduced electron mean free path (Fig. 8) by an increase of inelastic collision with neutral gas (the cooling effect).³⁰ 

Based on the results of plasma diagnostics (Sec. III B), in order to have an efficient process, a sputtering input power of 90 W (maximum n_e) was fixed, and the coatings were deposited as a function of pressure.

Figure 9 exhibits AFM surface pictures of the W coatings deposited (thickness of coatings at around 300 nm) at different sputtering pressures. The surface morphology is predominantly characterized by small granular structures. The RMS roughness (R_avg) shows a maximum value of around 5 nm at 2 Pa [Fig. 9(b)], while the coatings deposited to all the other pressures show a similar grain formation, with a similar roughness in the range of 1.3–1.9 nm. This is because the size of granular structures is strictly correlated to the coating deposition rate (Fig. 4). In fact, at 2 Pa, larger flux of sputtered atoms with the same average energy (V_dc is about constant as a function of pressure) leads to lower ad-atoms mobility, which promotes the formation of larger grains.³⁸ Instead, at lower or higher pressure than 2 Pa, as demonstrated by Langmuir probe analysis (Fig. 5), a lower plasma density is measured, which implies a lower density of Ar⁺ ions in the plasma, with a consequent reduction of the sputtering rate of W atoms. Obviously, at higher pressures, we also have to consider the component of the scattering effect (due to collisions in the plasma) of the sputtered atoms, leading to low mobility of ad-atoms.

The cross-sectional SEM images of W films on Si substrates for specimens deposited at various sputtering gas pressure are shown in Figs. 10(a)–10(e). All coatings exhibited a columnar grain morphology. As already observed on other thin-film materials,^39,40 in general, the grain densification of the coatings seems to decrease with pressure increase, at the same time, the density of defects appears to increase. In our experimental work, SEM results support AFM results with a strong correlation to the deposition rate (Fig. 4); in fact, the columnar grain size is higher at 2 Pa, where the density of plasma is higher and a reduced surface mobility of ad-atoms is expected.⁴¹ 

The variation of microstructure for W films was also supported by the x-ray diffraction measurements. The spectra of 500 nm thick films are displayed in Fig. 11. In all W films, only the α-W phase is observed, suggesting that the sputtered atoms arriving at the substrate have enough energy to produce the more stable α-W phase. GI-XRD patterns of the film deposited show three diffraction peaks at 2θ = 40.1°, 58.5°, and 73° that are assigned to (110), (200), and (211) crystallographic planes in cubic α-W coatings, respectively [Crystallography Open Database, COD 9012433]. With reference, in particular, to the main orientation (110), the experimental data showed a peak intensity reduction as the pressure increased up to 5 Pa. This agrees with the Thornton model,⁴² which predicts a T zone microstructure, i.e., columnar grains mainly grown along a preferred orientation, for metallic coatings prepared at room temperature and/or low pressure. The full width at half maximum (FWHM) was also determined, and the size of crystal (D) was calculated for the (110) plane by the Scherrer formula⁴³ (Fig. 12). The average crystal size decreased with the pressure increase. This means that the crystallinity of the coatings improves as the gas pressure is decreased. These results could be expected because the pressure increasing reduces the mean free path of the sputtered atoms due to the collisions (scattering) between particles; therefore, sputtered atoms will arrive at the substrate surface with lower kinetic energy, and, consequently, the mobility of ad-atoms on the surface decreases and slows down the dynamics of crystallization.

Figure 13 shows the results of the electrical resistivity measurements executed at 300 K. As expected, the resistivity values depend on the process pressure and vary in the range of 1.4 × 10⁻⁶ to 3 × 10⁻⁵ Ω m. Higher ρ is reached at high sputtering gas pressures. This trend mostly could be explained by the improvement of the crystallinity (increase in the crystal size) decreasing the pressure. Larger crystals lead to enhanced carrier mobility due to less scattering at grain boundaries and, thus, reduced ρ of the coatings.⁴⁴ On the other hand, by comparing the images of the SEM, a partial contribution to the ρ change could be expected from the packing density and defects, in particular, from SEM cross-sectional views, the less packing density of columnar and the greatest number of defects as a function of increasing pressure could explain the increase in ρ. Indeed, ρ is heavily dependent on the mobility of carriers, and the lower packing density (associated with increased defects concentration) will produce more scattering events for carrier motion, resulting in an increase in ρ. This result is consistent with other studies on thin-film sputtering deposition using targets of refractory metals.⁴⁵ Regarding the W coatings, the obtained resistivity values are consistent with those reported by other authors.⁴⁶ 

A planar rf magnetron sputtering discharge in the low-pressure range was characterized by electrical and Langmuir probe measurements. Tungsten (W) crystalline thin films were produced as a function of process pressure. The plasma discharge characteristics were determined and related to the W coating characteristics. Our results indicate that T_e decreased both with increasing power and pressure, while n_e increased with increasing power and decreased, in general, with increasing pressure. The effect of the pressure variation (1–5 Pa) on the structure, morphology, and electrical resistivity was studied. A relationship between n_e and the deposition rate (Fig. 5) correlated with the characteristics of the deposited coatings was found. For both these quantities, the profile was bell-shaped peaked around 2 Pa. The deposition rate profile determined was due to the variation of electron density (ion flux to the target), resulting in a different sputtered flux and subsequent atomic deposition. In the considered pressure range, the structure and morphology of W coatings look pretty similar, a columnar geometry is generated with a main orientation thermodynamically stable of the α-W phase, which could be attributable to the same directionality of the sputtered atoms impinging on the growing films. At the lowest sputtering gas pressure, a more crystalline and compact microstructure is achieved (Figs. 10–12), which results in a lower film resistivity. On the other hand, at higher pressures, due to both the decrease in the degree of crystallinity and the density of coatings (or increase in defects), an increase of the electron scattering could be induced, thus explaining the higher resistivity obtained.

The authors have no conflicts to disclose.

Espedito Vassallo: Conceptualization (equal); Investigation (equal); Methodology (equal); Supervision (equal); Writing – original draft (equal); Writing – review & editing (equal). Matteo Pedroni: Conceptualization (equal); Investigation (equal); Writing – review & editing (equal). Marco Aloisio: Investigation (equal); Resources (equal); Writing – review & editing (equal). Daniele Minelli: Investigation (equal); Writing – review & editing (equal). Antonio Nardone: Investigation (equal). Hao Chen: Investigation (equal). Silvia Maria Pietralunga: Investigation (equal); Writing – review & editing (equal). Andrea Stinchelli: Investigation (equal); Writing – review & editing (equal). Fabio Di Fonzo: Investigation (equal).

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