2 Experimental set-up

As shown in figure 1, the experimental set-up consists of an cross-shaped ultra high vacuum (UHV) chamber in which the 6 flanges are connected to the pumping device (primary pump $+$ Adixen ATH500M for a base pressure of $10^{-7}$  mbar), the gas mass flowmeters (Ar, Kr or Xe), a 2 inch magnetron with unbalanced magnetic configuration and linear shift mechanisms on which the QCM-GEA apparatus is fixed. The 2 inch magnetron can be powered by either a HiPIMS generator (Hüttinger TPHS 4002) or by a direct current (DC) power generator (advanced energy pinnacle plus). Each 2 inch target (Pt, Au, Pd, $\text{Pt}_{50}\text{Au}_{50}$ and $\text{Pt}_{5}\text{Pd}_{95}$ used in this study) have a normalized thickness of $1524~\unicode[STIX]{x03BC}\text{m}$ and are bonded to a $3175~\unicode[STIX]{x03BC}\text{m}$ thick copper backing, thus enabling us to make a comparative study with an isomagnetic topology for each target mounted on the magnetron device.

Figure 1.

(a) Picture of the experimental set-up for the magnetron sputtering process. (b) Three-dimensional view of the QCM-GEA system placed in front of the magnetron device.

Working processes are carried out with the injection of pure gas and without gas mixing. A UHV gate valve ensures pressure regulation in the vacuum chamber, fixed here to $10^{-2}$  mbar. The different gas flows are standardized on the argon flow for reaching the set pressure inside the vacuum chamber, here by 4 standard cubic centimeters per minute (sccm) in Ar.

The discharge conditions are regulated for all experiments by using the HiPIMS voltage. The negative $\text{U}_{\mathit{HiPIMS}}$ peak voltage ranges from an absolute value of 700 to 2000 V, the pulse duration and frequency are respectively fixed to $100~\unicode[STIX]{x03BC}\text{s}$ and 50 Hz. The current and the voltage characteristics of the discharge are measured by means of a high voltage probe (Tektronix P6015A) and a current probe (Stangenes CT0.5-0.1W) placed at the back of the magnetron system.

Figure 2.

(a) Schematic of the quartz balance and the polarization system. (b,c) Assembling polarization system with quartz microbalance to form the QCM-GEA.

Figure 2 presents the QCM-GEA device based on a commercial Maxtek TM-400 QCM on which a homemade grid system is fixed. Figure 2(b,c) exhibit the compactness of the QCM-GEA system. The QCM resolution can reach $0.01~\text{nm}~\text{s}^{-1}$ for the deposition rate and 0.01 nm for the thickness. The GEA consists of a stack of two stainless steel wire mesh grids spaced by a $900~\unicode[STIX]{x03BC}\text{m}$ thick piece of milar (blue spacer) and electrically isolated (red spacer) from the body of the quartz microbalance. With a diameter of 12 mm, the GEA system opening is lightly larger than the 8 mm diameter of the quartz surface collection. The first grid, with $81~\text{mesh}~\text{cm}^{-2}$ and wire diameter $200~\unicode[STIX]{x03BC}\text{m}$ , is grounded to the experiment vessel and acts as a screen grid electrically separate from the plasma post-discharge and the second grid. This second grid, with $900~\text{mesh}~\text{cm}^{-2}$ and wire diameter $50~\unicode[STIX]{x03BC}\text{m}$ , allows us to discriminate the ionized part of the metal vapour by the application of a positive potential, ( $V_{d}$ ), as a discriminating grid. Depending on the potential applied between the screen grid and the discriminating grid, the quartz microbalance behaves like an ion/neutral filter. When mounting the GEA system, a laser alignment of the grids is not necessary since the screen grid and the discriminating grid do not have the same mesh dimension. The total transmission of the QCM-GEA system is then 18 %.

The distance between the surface of the sputtering targets and the QCM-GEA ranges from 40 to 75 mm during the experiments, which corresponds to a classical range of distance for a such 2 inch sputtering magnetron device.

A homemade Labview software allows control of the QCM-GEA system in order to collect the measurements performed by the QCM and to manage the potential applied to the discriminating grid. Depending on the experiments, a voltage ramp or a fixed voltage can be applied to the discriminating grid. The acquisition time is equal to 75 s when a voltage ramp is applied to $V_{d}$ (scan rate of $1~\text{V}~\text{s}^{-1}$ between $-15$ and $+60$  V). For a fixed voltage $V_{d}$ , the acquisition time of one measurement is set to 10 s. These durations are a compromise between the signal to noise ratio and the consumption of precious metal targets. For other inexpensive materials, acquisition times should be increased in order to improve significantly the signal to noise ratio and therefore to obtain more accurate results.

First experiment measurements are done with the sputtering of the platinum target in argon gas by applying a negative HiPIMS voltage ranged between 1000 and 1900 V and with a voltage step of 300 V. In this case the QCM-GEA position is successively placed at different distances from the target, between 40 and 75 mm with 5 mm increments.

The second experiment gives the deposition rates obtained with the QCM-GEA set at 50 mm from the magnetron device in the HiPIMS regime for the Pt, Au, Pd, $\text{Pt}_{50}$ - $\text{Au}_{50}$ and $\text{Pt}_{5}\text{Pd}_{95}$ targets and for different pure gases, namely Ar, Kr and Xe.

3 Results and discussion

This study begins with the sputtering of a platinum target in argon gas. Figure 3 reports the evolution of the Pt deposition rate as a function of the voltage applied to the GEA discriminative grid ( $V_{d}$ ) for the four negative HiPIMS voltages: (a) 1000 V, (b) 1300 V, (c) 1600 V and (d) 1900 V. The fifth and last figure (e) corresponds to the Pt deposition rate performed in the DCMS regime for an average power of 50 W. In order to do a comparative study, all the measurements of deposition rate are arbitrarily normalized to the deposition rate obtained with a grounded discriminating grid ( $V_{d}=0$  V) in the sputtering HiPIMS regime, for a voltage of 1500 V which is equivalent to an average power of 50 W for our HiPIMS discharge conditions.

Figure 3.

Deposition rate on the QCM-GEA as function of the discriminative voltage $V_{d}$ for various QCM-GEA/target distances and different $\text{U}_{\mathit{HiPIMS}}$ : 1000, 1300, 1600, 1900 V. Deposition rates arbitrary normalized to the deposition rate of the $\text{U}_{\mathit{HiPIMS}}$ equal to 1500 V and for a $V_{d}$ null. Image at the bottom right corner shows results for a 50 W DCMS process.

Multiple results can be derived from figure 3. Firstly, the increase in voltage applied on the discriminative grid up to a threshold value (approximately 20 V for each HiPIMS case) allows us to separate the platinum ions from the total metallic vapour. This separation is characterized by a decrease in the deposition rate from 20 V whatever the distance between the QCM-GEA and the sputtering target, and regardless the applied HiPIMS voltage. Between 0 and 20 V, the atoms and the ions of platinum are collected on the quartz, while beyond 40 V, only the atoms are collected at the surface of the quartz. Since argon is a non-reactive gas, it is not necessary to worry about argon ions (also repelled) which cannot intervene in the increased mass at the top surface of the QCM quartz. This phenomenon is observed for all HiPIMS cases. For example, for $\text{U}_{\mathit{HiPIMS}}=1000$  V and a distance of 40 mm, the normalized deposition rate goes from 0.67 to 0.62 when the discriminative grid voltage changes from 10 to 40 V. According to the formalism previously expressed, we obtain $(\unicode[STIX]{x1D6E4}i+\unicode[STIX]{x1D6E4}n)_{\mathit{normalized}}=0.67$ and $\unicode[STIX]{x1D6E4}n_{\mathit{normalized}}=0.62$ . The ionization fraction $\unicode[STIX]{x1D6E9}$ is approximately 0.08 which means that platinum ions compose 8 % of the metal vapour assuming that the sticking coefficient of the ions and neutrals on the quartz are identical. This decrease of the deposition rate for a threshold voltage is observed for all graphs concerning the HiPIMS regime. These results highlight the formal presence of platinum ions in metal steam in the HiPIMS regime. The increase of the HiPIMS discharge voltage does not seem to influence the proportion of ions in the metal vapour, which remains between 8 % and 10 %. Moreover, $\unicode[STIX]{x1D6E9}$ does not vary with the distance between the QCM-GEA and the sputtering target. This result indicates that platinum ions do not undergo any inelastic collisions as volumic recombinations in the considered distance range. Since volumic recombinations are linked to mean free paths, the selected working pressure for our HiPIMS process seems to be in good agreement with the metal ions’ used during the deposition.

Secondly, looking at the grid voltage, $V_{d}$ , necessary to discriminate $\unicode[STIX]{x1D6E4}i$ , it would seem that the global energy of the metal ions remains stable between the two extreme positions of the QCM-GEA (40 and 75 mm). This suggests that the metal ions also do not undergo elastic collisions on their trajectories. We also note a division by a factor of approximately 2 of the deposition rate between the 40 and 75 mm positions. With the DCMS discharge at 50 W, the decrease of the deposition rates do not take place when increasing the discriminating grid voltage. Therefore, the metal vapour seems to be only composed of Pt atoms in DCMS. This comparison of HiPIMS/DCMS highlights the different modes of production of species between the two regimes. However, it is important to note that the deposition rate is higher in the DCMS than in the HiPIMS regime at the same average power of 50 W for the sputtering of the Pt target. This result corroborates those obtained with other elements in the HiPIMS regime (Alami et al. Reference Alami, Sarakinos, Mark and Wuttig2006; Lundin et al. Reference Lundin, Larsson, Wallin, Lattemann, Brenning and Helmersson2008; Anders Reference Anders2011; Čapek et al. Reference Čapek, Hála, Zabeida, Klemberg-Sapieha and Martinu2013).

Figure 4.

Target currents at $-1500~\text{U}_{\mathit{HiPIMS}}$ , pulse duration $100~\unicode[STIX]{x03BC}\text{s}$ , frequency 50 Hz, for different targets and gases: respectively Au, Pt, Pd, $\text{Pt}_{50}\text{Au}_{50}$ , $\text{Pt}_{5}\text{Pd}_{95}$ and Ar, Kr, Xe.

This study continues with the measurement of the deposition rates in the HiPIMS regime for Pt, Au, Pd, $\text{Pt}_{50}\text{Au}_{50}$ and $\text{Pt}_{5}\text{Pd}_{95}$ allied targets. The working pressure, the duration of the HiPIMS pulses and their frequency remain unchanged. Figure 4 gives an example of the measured target currents during a HiPIMS pulse voltage for each gas, here at $\text{U}_{\mathit{HiPIMS}}=1500$  V. The position of the QCM-GEA is set at 50 mm from the magnetron. The deposition rates are also normalized to those obtained by HiPIMS with Ar gas and reporting targets (Pt, Au, Pd, $\text{Pt}_{50}\text{Au}_{50}$ or $\text{Pt}_{5}\text{Pd}_{95}$ ), for $\text{U}_{\mathit{HiPIMS}}=1500$  V. As above, $\unicode[STIX]{x1D6E9}$ is determined by the evolution of the normalized deposition rates depending on the positive bias  $V_{d}$ .

Figure 5.

QCM-GEA normalized deposition rate (full lines) and ions/neutrals ratio $\unicode[STIX]{x1D6E9}$ (dashed lines) as a function of $\text{U}_{\mathit{HiPIMS}}$ , for different targets and gases: respectively Au, Pt, Pd, $\text{Pt}_{50}\text{Au}_{50}$ , $\text{Pt}_{5}\text{Pd}_{95}$ and Ar, Kr, Xe.

Figure 5 illustrates all the measurements done for Ar, Kr and Xe gases. Each graph is attributed to a specific target material: Pt, Au, Pd, $\text{Pt}_{50}\text{Au}_{50}$ or $\text{Pt}_{5}\text{Pd}_{95}$ . Normalized deposition rates are represented by the full lines while the different $\unicode[STIX]{x1D6E9}$ values (noted as a percentage (%) on the secondary vertical value axis) are represented by dashed lines. For the first time and by correlation with the pulse discharge currents from figure 4, three major observations on normalized deposition rates can be drawn:

1. (i) the deposition rate quasi-linearly increases with the HiPIMS voltage for all cases except for the Pd target (and $\text{Pt}_{5}\text{Pd}_{95}$ to a lesser extent) where a saturation is observed from 1600 V with Ar and Kr gases (and from 1800 V for Xe gas). A relationship of proportionality between deposition rate and target current exists for each material as a function of a distinct plasmogen gas, as it was already reported by Sarakinos, Alami & Konstantinidis (Reference Sarakinos, Alami and Konstantinidis2010).

2. (ii) The nature of the gas plays an important role in the deposition rate in the HiPIMS regime for each metal target. This result is globally visible if we compare the different pulse discharge currents collected on a considered target of figure 4 with the respective deposition rate of figure 5. In some cases, the deposition rate values found for distinct gases are different, even if the pulse discharge currents collected on the considered target are similar, therefore, for a similar mean power.

3. (iii) Argon (pink) allows us to get the highest deposition rate for a fixed HiPIMS voltage. Kr (dark blue) comes in second position, followed by Xe (cyan). This result is reversed in Kr and Xe for a HiPIMS voltage higher than 1800 V in the case of a Pt, $\text{Pt}_{50}\text{Au}_{50}$ or Pd sputtering target.

For the second time, all the ion proportions could not be measured. In some cases, increasing the voltage Vd on the GEA causes the appearance of a secondary plasma near the screen grid, affecting the initial $\unicode[STIX]{x1D6E9}$ parameter in the metal vapour. This phenomenon mainly occurs for Xe, and to a lesser extent for Kr for high HiPIMS voltage values. Indeed, increasing the voltage $V_{d}$ to positive values allows us to attract and to give energy to a few secondary electrons (initially repelled by the first grid) which can generate this visible secondary plasma. This difficulty could have been overcome by the simple addition of a so-called repeller grid sandwiched between the screen grid and the discriminative grid. This would have been enough to repel the electrons coming from the magnetron plasma. The counterpart of this alternative is the decrease of the GEA transparency and thus the decrease of particle flux coming onto the quartz surface. This was not done due to the sputtering target prices. Therefore, this solution reduces the resolution of the detection of the deposition rate and therefore mechanically, that of $\unicode[STIX]{x1D6E9}$ . The alternative is to work with systems that are more developed such as those mentioned in the introduction (e.g. gridless quartz microbalance of Hubička et al. Reference Hubička, Kment, Olejníček, Čada, Kubart, Brunclíková, Kšírová, Adámek and Remeš2013). Figure 5 only shows the valid measurements of the ion proportion.

With the Pt target, a maximum ion proportion of approximately 10 % is found with the metal vapour in Ar gas. The proportion of ionized material $\unicode[STIX]{x1D6E9}$ appears to increase slightly with the HiPIMS voltage, then it starts to decrease from 1800 V. The use of the Kr gas makes $\unicode[STIX]{x1D6E9}$ lower than 5 % over all the HiPIMS voltage range. With Xe gas, the proportion of ionized material is much lower than with Kr and tends to 0 %. The Au target gives similar results with nevertheless the lowest $\unicode[STIX]{x1D6E9}$ in Ar gas (between 5 % and 10 %), compared to the Pt target. The use of the Pd target gives a high ratio of metal ions in Ar gas with a $\unicode[STIX]{x1D6E9}$ reaching more than 50 % at 1600 V. This proportion then declines to 32 % for a 2000 V HiPIMS voltage. It appears that increasing the HiPIMS beyond a 1600 V voltage does not promote the increase of the proportion of Pd ions in the post-discharge metal vapour. A return of these metal ions in the direction of the negatively polarized target is a track considered to explain this fall in the ratio. The $\text{Pt}_{50}\text{Au}_{50}$ target reflects similar $\unicode[STIX]{x1D6E9}$ to those found in the case of the pure Pt and Au targets. Nevertheless, the ratio is slightly higher in the case of the mixed target in Ar gas, with a maximum value of approximately 19 %. In contrast, the $\text{Pt}_{5}\text{Pd}_{95}$ target gives a proportion of ions with the use of Ar gas well above those found with the pure Pd target. In this case $\unicode[STIX]{x1D6E9}$ is close to 90 %. The 5 % atomic Pt doping seems to play a key role in the increase of the value of the ratio $\unicode[STIX]{x1D6E9}$ for a fixed HiPIMS voltage. This result seems to be confirmed for a few measurements done with Kr gas (without the ignition of the secondary plasma near the QCM-GEA). As for the pure Pd target, the ratio of ions decreases from 1600 V in Ar gas.

It should be noted finally that, experimentally, the use of Kr and Xe gases do not favour the production of metal ions nor, in most cases, an increase of the deposition rates in the HiPIMS post-discharge. However, they present higher masses (83.80 and 131.29 amu) than Ar (39.948 amu) and a lower first ionization threshold energy (13.9996 and 12.1298 eV) than Ar (15.7596 eV). Future tests should be done on Ne and He gases to see if the trends are reversed and to point out the fine physical mechanisms involved in the obtained $\unicode[STIX]{x1D6E9}$ ratios.