Plasma Diagnostics in Reactive High-Power Impulse Magnetron Sputtering System Working in Ar + H₂S Gas Mixture

Abstract

A reactive high-power impulse magnetron sputtering system (HiPIMS) working in Ar + H₂S gas mixture was investigated as a source for the deposition of iron sulfide thin films. As a sputtering material, a pure Fe target was used. Plasma parameters in this system were investigated by a time-resolved Langmuir probe, radio-frequency (RF) ion flux probe, quartz crystal monitor modified for measurement of the ionized fraction of depositing particles, and by optical emission spectroscopy. A wide range of mass flow rates of reactive gas H₂S was used for the investigation of the deposition process. It was found that the deposition rate of iron sulfide thin films is not influenced by the flow rate of H₂S reactive gas fed into the magnetron discharge although the target is covered by iron sulfide compound. The ionized fraction of depositing particles decreases from r ≈ 40% to r ≈ 20% as the flow rate of H₂S, Q_H2S, changes from 0 to 19 sccm at the gas pressure around p ≈ 1 Pa in the reactor chamber. The electron concentration n_e measured by the Langmuir probe at the position of the substrate decreases over this change of Q_H2S from 10¹⁸ down to 10¹⁷ m⁻³

1. Introduction

Reactive sputtering of sulfide thin films is still not a fully explored field. A lot of effort was devoted to the research of direct-current (DC) or radio frequency (RF) sputtering methods of various sulfides [1,2,3]. These can be semiconductor materials for example iron disulfide FeS₂ in the pyrite phase reported as a p-type semiconductor [1,2,4] or an n-type semiconductor suitable as a photoanode in solar water splitting cells when the pyrite phase was mixed with a small fraction of marcasite [5]. FeS₂ was investigated for various applications such as absorbing material in solar cells, material for photocatalysis or electrocatalysis [6]. The high absorption coefficient, α = 6 × 10⁵ cm⁻¹ for λ = 633 nm [7], which is several orders of magnitude larger than that of crystalline silicon, and the composition of abundant, cheap, and non-toxic elements makes pyrite an interesting material for solar cells and further photonic applications [1,8,9,10,11,12,13]. FeS₂ thin films with promising semiconductor properties were also prepared by reactive magnetron sputtering of an Fe target in Ar + H₂S gas mixture [1,2,4].

Iron mono sulfide FeS is an interesting material that exhibits transition from metal to semiconductor at a certain temperature close to 147 °C [14,15]. Below this temperature, FeS is a semiconductor and above this temperature it behaves as a metal. Based on this material, devices triggered by heat can be realized. Furthermore, FeS films were found to be excellent low-friction materials working like solid state lubricants in tribological applications [16].

High-power impulse magnetron sputtering system (HiPIMS) is a relatively novel approach for the deposition of thin films [17,18,19,20]. This method utilizes very high discharge current densities during the pulse j_D > 1 A cm⁻², but, simultaneously, average power applied in the HiPIMS discharge is similar to DC magnetron sputtering. These conditions result in a high degree of ionization of sputtered particles. Thus, the deposited films have higher density, better adhesion, and 3D objects such as trenches, etc. and can be more uniformly coated as well [21,22,23]. Recently, reactive HiPIMS in various gas mixtures has been used for the deposition of oxides, nitrides, and other materials [24,25,26,27,28,29,30,31].

In this paper, we will present the first study of plasma parameters in the reactive HiPIMS in Ar + H₂S gas mixture whose results could be further used for the optimization of deposition of iron sulfide thin films or other types of sulfides.

2. Experimental

The experimental setup of the reactive HiPIMS in Ar + H₂S gas mixture can be seen in Figure 1. The system is equipped with a pure iron Fe target with an outer diameter of 50 mm and a thickness of 0.8 mm. The stainless-steel reactor vacuum chamber is continuously pumped by a combination of turbomolecular (pumping speed S_t = 250 L·s⁻¹) and rotary vane pumps (pumping speed S_r = 25 m³ h⁻¹). The gas flow rates of Ar (99.9999%) and pure H₂S (99.5%) are independently controlled by two mass flow controllers. It is useful to note that H₂S is an inflammable and highly toxic gas. Consequently, the use of H₂S in a laboratory is subject to approval by a respective official body. The pressure in the reactor chamber can be controlled by a gate valve placed between the reactor chamber and the turbomolecular pump. The magnetron cathode is connected with the HiPIMS power supply, which has a maximum current pulse of 200 A and maximum voltage of 1000 V at the output. The time-resolved Langmuir probe, the quartz crystal monitor (QCM) modified with an electron filter and biased collecting electrode, and the RF ion flux probe are placed at position of the substrate. The distance, l_s, between the target and the substrate is 70 mm for all presented experiments.

Emission spectroscopy of HiPIMS plasma was carried out by a TRIAX JY spectrometer with focal length 0.5 m and grating 1200 gr/mm. The CCD detector provided time-averaged spectra over several seconds in the spectral range 200–1000 nm. The entrance slit and integration time of the CCD detector was kept constant for a particular spectral range in order to be able to perform approximate quantitative comparison of particular conditions.

The HiPIMS magnetron discharge was always operated with the constant length of the active pulse, T_ON = 100 μs, and pulsing frequency, f_p = 100 Hz. The average power, P_av, applied in the discharge was slightly changed in different cases and was unambiguously related with presented pulse power, P_p, according to the formula:

$$P_{av}=P_p·T_{on}·f_p.$$

(1)

The current and voltage waveforms of HIPIMS discharge can be seen in Figure 3 and will be discussed later.

2.1. Langmuir Probe Measurement

The plasma parameters such as plasma density, n_e, and electron effective temperature, T_eff, were measured by the Langmuir probe. The Langmuir probe consists of fine tungsten wire (diameter 50 µm and length 2.2 mm) mounted on an ultra high vacuum (UHV)-compatible probe hardware from Scientific Systems, Inc. The probe was placed 70 mm axially from the target and radially above the target racetrack. An in-house developed electronic acquisition unit was capable of setting the probe voltage from −50 to 50 V with a step of 0.1 V and measuring the probe current with time resolution better than 5 μs [32]. The grounded chamber’s walls served as a reference electrode for the probe’s electrical circuit. The effect of the magnetic field on the probe measurement was negligible because the balanced magnetic field strength was less than 1 mT resulting in much larger Larmor radius of electrons in comparison with the radius of the probe.

Measured probe characteristics were evaluated under collisionless space charge sheath conditions because high-density HiPIMS plasma implies Debye length in the order of tens of µm whilst the mean free path of neutral-charged particle collision is in order of tens of mm [33]. The measured I–V characteristics were evaluated by using an in-house developed Langmuir probe software [34,35]. The software calculates the second derivative of the Langmuir probe characteristics numerically using the ‘sliding polynomial approximation’ method. This represents approximation of the selected odd number M = 2m + 1 (m = 1,2, …) of adjacent data points (equidistantly spaced by h) around selected x_l by a second order polynomial p(x) = a₀ + a₁x + a_{2 ×} ² (for x_l - m.h ≤ x ≤ x_l + m.h). The second derivative at x_l is hence $y^{\left(2\right)}\left(x_l\right)=2a_2$. Since the coefficient a₂ represents a weighted sum of the ordinates y_i (for l - m ≤ I ≤ l + m) this method may be regarded as a digital filtering. As a result, by selecting a proper n, the noise on the computed second derivative could be reduced down to an acceptable level. The fact that we cannot calculate the second derivative at m points at the start and the end of the characteristic can be compensated by collecting an adequate number of measured data points.

The electron energy distribution functions (EEDFs) in HiPIMS systems cannot be, in most cases, characterized by Maxwellian distributions [36]. Consequently, we had to calculate the basic plasma parameters, namely the electron density, n_e, and the electron mean energy, <ε>, directly from the measured EEDF. The EEDF, F(E), is related to the measured second derivative by the known Druyvesteyn relation [37,38], the formulae applicable in practice are given in [39]. Often the electron energy probability function, f(E), is introduced (EEPF) as f(E) = F(E)E^−1/2 where E denotes the energy [40].

As follows from [39], the second derivative of the electron component of the probe current I_e with respect to the probe voltage U_p, $d^2{I}_e/d{U}_p^2$, is related to f(U_p) by the relation:

$$\frac{d^2{I}_e}{d{U}_p^2}=\frac{q_0^{3/2}}{2^{3/2}}\frac{1}{\sqrt{m_e}}{n}_e{A}_{p}f\left(U_p\right)$$

(2)

where q₀ and m_e are the electron charge and mass respectively, A_p is the probe surface area. The EEPF is normalized to one, i.e., ${{\displaystyle \int }}_0^{\infty }f\left(U_p\right)U_p^{1/2}d{U}_p=1$.

We start the probe data processing by computing the second derivative of the measured probe characteristic with respect to the probe voltage using the method described above. We assume that the second derivative of the positive ion component of the probe current can be neglected. Then we continue by estimating the electron density from the integral of the calculated second derivative:

$$n_e={\left(\frac{2}{q_0}\right)}^{3/2}\frac{m_e^{1/2}}{A_p}{{\displaystyle \int }}_0^{\infty }{U}_p^{1/2}\frac{d^2{I}_e}{d{U}_p^2}d{U}_p$$

(3)

The electron mean energy <ε> (in electron-volts) is estimated from the normalized f(U_p) by the relation:

$$\frac{〈\epsilon 〉}{q_0}={{\displaystyle \int }}_0^{\infty }{U}_p^{3/2}f\left(U_p\right)d{U}_p=\frac{{{\displaystyle \int }}_0^{\infty }{U}_p^{3/2}\frac{d^2{I}_e}{d{U}_p^2}d{U}_p}{{{\displaystyle \int }}_0^{\infty }{U}_p^{1/2}\frac{d^2{I}_e}{d{U}_p^2}d{U}_p}.$$

(4)

The effective temperature T_eff (in electron-volts) is then given by the relation: T_eff = 2/3 <ε>.

2.2. Ion Flux Density Measurement by a Planar RF Ion Probe

Voltage at RF frequency, f_M = 350 kHz, and amplitude, U_RF = 50 V, was applied on the planar ion flux probe placed at the position of the substrate (see Figure 1). The RF signal was connected to the probe through a blocking capacitor, C, in order to measure only RF currents through the probe. The RF voltage and RF current waveforms were measured by a RF current transformer wound on a Teflon core and a capacitive voltage probe respectively. The signals from these probes were fed to the digital oscilloscope DSO7104B Agilent technologies (1 GHz, 4 GSa) Santa Clara, CA USA. The current transformer and capacitive probe were homemade and calibrated at frequencies in the region from 100 kHz up to 4 MHz. The calibration was done by the use of professional dummy load 50 Ω (1.5 kW VHF-UHF Dry dummy load 1-650 MHz Model MFJ-264, MFJ Enterprises, Inc. Starkville, MS USA). The complex correction function was created by this calibration procedure and the measured signal was corrected by this function after discreet Fourier transformation (DFT) done in the software OriginPro 2015, produced by OriginLab Corporation Northampton, MA USA. The accuracy of these measurements was tested on selected passive complex impedances with known real and imaginary part of impedance in the range 100 kHz–4 MHz. The accuracy of this measurement configuration was estimated to be below 5%. The probe was a planar disc with the area of approximately S = 1 cm². The RF substrate voltage and current waveforms, U_RF and I_RF, were measured by RF current and voltage probes connected to the digital storage oscilloscope. The ion flux density of positive ions j_S to the substrate was calculated from these waveforms by the method developed by Sobolewski [41,42]. The example of these RF waveforms on the planar ion flux probe during HiPIMS pulse can be seen in Figure 2. The current to the RF probe consists of three components, I_p(t) = I_i(t) + I_e(t) + I_c(t). All those currents follow the probe voltage, U_RF. We assume that I_i(t) = I₀ is saturated, i.e., at sufficiently negative probe voltage, I_i does not depend on U_RF. The probe ion current, I₀, can be determined from I_RF at the time when the RF voltage U_RF has a peak of negative value because here dU_RF/dt = 0 [41,42]. At this time, the capacitive current through the probe I_c = 0 and all electrons are repelled from the surface of the probe. It is necessary to remind here that this approach is valid only in case f_M is well below ion plasma frequency ω_i >> f_M [41,42]. The ion flux density, j_SP, can be calculated from I₀ as j_SP = I₀/S where S is the area of the ion flux probe. We were able to get in this way the time evolution of the substrate ion flux density during the HiPIMS pulse. An advantage of the Sobolewski method for the ion flux calculation is the independence of this value on the resistivity of the deposited film on the probe since only AC values of I_RF and U_RF are used for the j_SP calculation.

2.3. Ionization Fraction of Depositing Particles Measurement by a Modified QCM with a Magnetic Electron Filter

A modified quartz crystal monitor with a magnetic electron filter (m-QCM) was already used in the HiPIMS system for the purpose of measuring the ionization fraction of depositing particles and was described in details elsewhere [43,44]. This system uses an auxiliary constant magnetic field placed at the working collecting electrode of the QCM in order to eliminate the major fraction of electron flux on this working electrode. The working electrode can be either biased by the positive voltage or connected to ground. In the case of applied positive bias on the working electrode, we can measure only the neutral flux, i_n, of depositing particles whilst the positive ions are repelled from the working electrode. When the bias on the working electrode of QCM is zero we can measure the total flux, i_tot (neutral + ionized), of depositing particles. Thus, we can calculate the ratio r:

$$r=\frac{i_{ion}}{i_{ion}+i_n}$$

(5)

where i_ion is the flux of positively ionized depositing particles and i_n is the flux of neutral particles. The flux of positive particles, i_ion, is not directly measured and is calculated according to the formula:

$$i_{ion}=i_{tot}-i_n.$$

(6)

2.4. Iron Sulfide Thin Film Deposition

In order to show that iron sulfide thin films can be deposited by the reactive HiPIMS, the first deposition experiments of these thin films were done at selected conditions. Iron sulfide films were deposited on glass substrates. Chemical composition was measured by an electron microprobe Bruker XFlash 6 |10 and the structure of deposited films was determined by an XRD in grazing incidence configuration (Figure 8).

SEM (TESCAN LYRA3 RISE) photographs of surfaces of the films S1–S4 are presented as well (Figure 9). The distance between the target and the substrate was 70 mm for all the deposition conditions. The substrate was heated during the deposition to 250 °C by external heater. The deposition conditions were similar to those used for plasma diagnostics and presented in Figure 3. The gas pressure in the reactor chamber was held at the constant value p = 1 Pa. The flow of argon was for all the samples Q_Ar = 20 sccm (= 3.3 × 10⁻² Pa m³·s⁻¹). The changed parameters, such as the flow of H₂S, Q_H2S, pulsed power during active HiPIMS discharge P_p, and the cathode voltage U_c measured in the middle of active HiPIMS pulse, are mentioned in

Table 1. Deposition conditions of deposited iron sulfide thin films (Q_HS – H2S gass flow rate, P_p pulse discharge power, U_c cathode voltage).

Sample No.	QH2S [sccm]	QH2S [m3 s−1 Pa]	Pp [kW]	Uc [V] (Middle of Pulse)
S1	3.8	6.3 × 10−3	32.1	556
S2	7.6	1.26 × 10−2	34.1	567
S3	11.4	1.9 × 10−2	37.3	575
S4	15.2	2.53 × 10−2	39.8	510
S5	19	3.17 × 10−2	36.0	600
S6	22.8	3.8 × 10−2	37.3	575

. Other HiPIMS discharge conditions for all the samples were the same as in the plasma diagnostic experiments. Chemical composition of deposited films measured by electron microprobe can be seen in

Table 2. Chemical composition measured by electron microprobe.

Sample No.	Fe at.%	S at.%	O at.%
S1	40.7	54.8	4.5
S2	40.1	54.7	5.2
S3	39.4	56.8	3.8
S4	41.0	56.0	3.0

.

3. Results and Discussion

The results of current and voltage waveforms measurement of reactive HiPIMS in Ar + H₂S gas mixture can be seen in Figure 3 for different mass flow rates of H₂S gas, Q_H2S, in the range from 0 up to 19 sccm and at the constant argon mass flow rate Q_Ar = 20 sccm. The total gas pressure in the reactor chamber was held at the constant value p = 1 Pa by the control gate valve. The pulse power applied to the HiPIMS discharge during the “ON” time was in the range P_p = 33–36 kW as marked at the particular graphs (see Figure 3).

The HiPIMS discharge current waveforms for different Q_H2S qualitatively differ as it can be seen in Figure 3a–d. The case of Q_H2S = 0 sccm (see Figure 3a) looks like a typical HiPIMS metallic iron sputtering with a gas rarefaction in the vicinity of the target causing the discharge current decreases during the pulse. The situation changes when H₂S reactive gas is added into the discharge and the target is very probably covered by iron sulfide compound. The graph in Figure 3b shows that the current waveform character changes even for small Q_H2S = 3 sccm. In this case, the discharge current after initial fast growth shortly drops and after that slowly grows until the end of the active plasma pulse. A similar situation appears for higher Q_H2S (Figure 3b–d). It means the discharge current slightly decreases in the first part of the pulse and after that again increases until the end of pulse. In order to better explain this phenomenon we have to discuss at first the deposition rate measured by modified QCM and results of emission spectroscopy.

Figure 4a shows the ionization fraction, r, of depositing particles for different Q_H2S as measured by modified QCM with the electron filter (left axis). We can also see the deposition rate on the right axis of the graph at Figure 4a. The maximum ionization fraction, r, was found to be r = 43% for pure Fe deposition without H₂S. As Q_H2S grows, the ionization fraction, r, gradually decreases to the value r = 20%. On the other hand, the deposition rate increases twice for even the smallest Q_H2S = 3 sccm and remains constant up to high values of Q_H2S = 19 sccm. On the contrary, we can see the substantial decrease of deposition rate for higher pressure (see Figure 4b) and decrease of ionization fraction for gas pressure, p = 4 Pa.

The results of optical emission spectroscopy can be seen in Figure 5 and Figure 6. We can see spectra for different Q_H2S at constant pressure p = 1 Pa. Strong lines of Fe⁺ and Fe at the spectrum for Q_H2S = 0 sccm can be identified. The presence of strong Fe⁺ and Fe lines was still observed with similar intensity for cases of high mass flow rate of H₂S up to Q_H2S = 19 sccm. Furthermore, we can observe intensive sulfur lines in the optical emission spectra as shown in Figure 6 for different Q_H2S. It can be seen that intensity of sulfur spectral lines and, thus, the sulfur atom concentration in the plasma grows for higher Q_H2S but the argon line intensity decreases with increasing Q_H2S. These sulfur atoms are generated probably as a product of dissociation of H₂S gas in the plasma, and a certain fraction probably comes also from the target covered by iron sulfide. Intensities of measured lines Fe, Fe+, Ar, and S for particular Q_H2S are in

Table 3. Emission spectra of observed elements Fe, Fe+, S, and Ar.

Element	Line Wavelength [nm]	Intensity, Arb. Units QH2S = 0 sccm	Intensity, Arb. Units QH2S = 7.8 sccm	Intensity, Arb. Units QH2S = 19 sccm
Fe+	273.69	5862	3587	3304
Fe+	273.95	6832	5559	4774
Fe+	275.32	7497	6009	5143
Fe	302.40	5059	6653	6665
Ar	922.45	10,646	5809	4987
S	921.28	0	7492	11,829
S	922.80	0	5230	8071
S	923.75	0	3164	4753

.

Time-resolved measurement with the Langmuir probe was carried out for different mass flow rates, Q_H2S, and standardly at the pressure p = 1 Pa except the case of Q_H2S = 19 sccm when we also measured characteristics at p = 3 Pa. The obtained plasma parameters such are plasma density, n_e, and electron effective temperature, T_eff, are shown in Figure 7. Since the process of time-resolved measurements of the I–V probe characteristic is rather complicated, we assessed the general error of the plasma parameters estimated from the probe data as 20%. The H₂S mass flow rate varied between 0 and 19 sccm. As we can see in Figure 7a, the highest plasma density slightly above n_e ≈ 10¹⁸ m⁻³ was reached for pure argon discharge during the second part of the plasma pulse. Adding a small amount of H₂S to the working gas resulted in a notable decrease of n_e. Further H₂S mass flow rate increase led to slight decrease of the n_e whilst temporal evolution of n_e during the plasma active pulse and even during the plasma off-time was practically unchanged. The higher pressure resulted in lower n_e. The most probable reason of electron concentration, n_e, decrease for higher partial pressures of H₂S in the plasma can be attributed to the fact that H₂S is electronegative [45]. The electron attachment to H₂S is dissociative, producing negative ions $H{S}^{-}$, $H^{-},$and $S^{-}$ [45]. That causes loss of electrons from the tail of EEDF due to dissociation of H₂S molecules. Further, H₂S molecules can be dissociated by an electron impact and by interaction with other species (Ar) present in the plasma. Dissociation of

H₂S in plasma was already described for example in [46,47]. The H₂S molecule can be dissociated in plasma by a direct electron impact as follows from [46,48]:

e + H₂S → HS + H + e.

Further mechanism of H₂S dissociation is a two-step process as follows, e.g., from [46,48]. First the H₂S⁺ positive ion is formed by an electron impact, by a charge transfer with ionized argon or by a reaction with a metastable excited argon. In principle, in place of argon we can also imagine another specie present in plasma.

e + H₂S → H₂S⁺ + 2e, or Ar⁺ + H₂S → H₂S⁺ + Ar, or Ar*+ H₂S → H₂S⁺ + Ar + e

(7)

The H₂S⁺ molecular ion, however, quickly recombines with electrons by dissociative recombination: H₂S⁺ + e → HS + H.

H₂S⁺ ions can be dissociated also by reaction with metastable excited species present in the plasma, e.g., Ar* + H₂S → HS + H + Ar (Ar* metastable states have an energy around 11.5 eV [49,50], well above the ionization energy of H₂S, which lies around 10.3 eV [51]). The ion chemistry of H₂S⁺ is rather complex and has been studied, e.g., in [51]. Therefore, we were able to give here just the qualitative discussion.

Apart from the energy lost by dissociation, the total energy contained in the discharge can also be distributed among the excitation of rotational and vibrational states of the molecular species added and, at the end, the efficiency of ionization, i.e., the electron density, is reduced. The sputtered Fe atoms can also contribute to the electron cooling since they are ionized by electron impact. However, this effect is also present without introducing H₂S, and we do not have clear evidence that the addition of H₂S would increase the sputtering of the iron target.

Figure 7b clearly demonstrates significant impact of Q_H2S on the mean energy of plasma electrons. The decrease of the T_eff during the first 40 µs is consistent with other observations in non-reactive and reactive HiPIMS systems. Plasma pulse onset is accompanied by beam-like electrons with energy of roughly 100 eV at the probe position, which are reflected from the negative space charge sheath around the cathode [4]. The minimum of T_eff within the plasma pulse having magnitude between 1.0 eV and 1.2 eV (depending on the H₂S mass flow rate) is probably connected with massive ionization of sputtered particles that have much lower ionization potential than argon atoms (Fe ionization potential is around 7.9 eV). Nevertheless, after 40 µs from the pulse onset, we can observe increase in the T_eff. This is probably connected with the fast recombination reactions of the (molecular) ions created after the pulse onset. It is a known fact that the rate of recombination is (almost) inversely proportionally dependent on the electron temperature, see e.g., Figure 9 in [52] showing the rate of recombination of SH⁺ with electrons on the electron energy. The process of recombination then depletes the lower-energetic part of EEDF resulting in a higher T_eff. This increase of T_eff becomes less pronounced with increasing Q_H2S. The electron effective temperature, T_eff, shows gradual decrease during the plasma pulse for pressure p = 3 Pa, and, moreover, T_eff is significantly smaller in comparison with measurement at lower pressure. That fact is understandable as electrons have to undergo more elastic and inelastic collisions with particles of processing gas.

Very similar behavior of the time evolution of the electron density and electron temperature after discharge switch-on was observed in an argon HiPIMS with a titanium target in a recent work [53]. The authors explained the local minimum of the T_e by cooling of the EEDF due to increasing density of metallic species. Cooler electrons reduced the plasma generation rate and, consequently, the sputtering rate was reduced accompanied by the rise in T_e. Moreover, the authors of [54] observed similar behavior of n_e and T_e in the switch-on phase of the HiPIMS discharge. Since we do not have clear evidence about the sputtering of our iron target, we cannot conclusively determine the reason of the temporal behavior of the T_eff in our experiment.

As we have discussed above, the current waveforms in the presence of H₂S in the plasma qualitatively differ from those in pure argon and the discharge current increases at the end of the pulse in the reactive mode. In any case, this phenomenon is probably connected with the covering of the target surface with an iron sulfide layer as usually happens in case of reactive sputtering (target poisoning). As we can see in Figure 4a, the deposition rate does not decrease for higher Q_H2S but increases instead. Thus, the sputtering rate of the poisoned target with sulfide layer is still high and similar to pure iron target sputtering so the gas rarefaction should still exist in front of the target in the reactive case. This statement is also confirmed by results of emission spectroscopy (see Figure 5) where we can see intensive Fe and Fe⁺ emission lines also for high Q_H2S. Higher ion–electron secondary emission from iron sulfide deposited on the target can be responsible for the discharge current growth during the second part of HiPIMS pulse in the reactive case. It is difficult to provide accurate evidence for this statement because we do not know the exact structure of the sulfide layer created on the target surface, but the work function of pyrite FeS₂ is around ≈4.3 eV as reported in [55] and the work function of pure iron is between 4.7 and 4.8 eV as reported in [56]. These facts can support our discussion of the character of the discharge current in the reactive case.

The time evolution of the ion flux density at the position of the substrate obtained by the RF (350 kHz) Sobolewski method can be seen in each graph in Figure 3 as well. We can see in Figure 3, the dependence of maximum substrate ion flux density in the HiPIMS pulse as a function of H₂S mass flow rate Q_H2S. The pulse power, P_p, is again mentioned at each point in the graph. As can be seen in Figure 3a–d, the substrate ion flux density gradually grows during the HiPIMS pulse with a maximum value very close to the end of the pulse. The substrate ion flux density should be a function of ion concentration, n_i, near the substrate and electron temperature, T_e, approximately according formula:

$$j_{SP}=0.606n_{i}e\sqrt{\frac{k_{B }{T}_e}{M_i}}$$

(8)

where e is elementary charge, k_B is Boltzmann’s constant, and M_i is the mass of ions.

The time evolution of ion flux density, j_SP, during the HiPIMS pulse for different Q_H2S can be correlated with the time evolution of electron concentration, n_e, obtained from Langmuir probe data (see Figure 7). The substantial growth of j_SP during the pulse is in an agreement with the growth of n_e (the plasma is quasineutral, n_e ≈ n_i). Furthermore, we can see in Figure 3 that the maximum achieved j_SP during the pulse decreases with increase of Q_H2S at constant pressure p = 1 Pa. The absolute values of j_SP are quite high in HiPIMS pulse and certainly will influence the growth of films during the pulse.

Finally, we can correlate the measured ionization fraction, r, of depositing particles with the results of Langmuir probe measurements (see Figure 7). It is obvious that an ionization cross section of an electron impact ionization collision that undergo the sputtered particles increases with higher electron kinetic energy in the interval just above the ionization potential of these atoms [57]. Thus, the total rate of ionization collisions of electrons with sputtered particles is proportional to n_e and is also dependent on T_e [33]. These facts clearly show that the increase in the measured plasma parameters (n_e, T_e) should simply increase the ionization fraction r. If we add H₂S gas into plasma Q_H2S = 3.8 sccm the n_e decreased more than three times. If we consider that the change of effective T_e is not significant for this change of Q_H2S, the decrease of ionization fraction from r = 43% to r = 30% is in qualitative agreement with the Langmuir probe data. Further decrease of r for higher Q_H2S is again in agreement with the electron concentration, n_e, decrease for these H₂S flows.

The results of XRD of deposited iron sulfide can be seen in Figure 8 for samples S1–S6 with increasing amount of H₂S gas flow Q_H2S = 3.8–22.8 sccm (Table 1). We can see that for lowest Q_H2S (S1) we could identify the mixture of FeS trollite phase with pure iron Fe. For increasing Q_H2S (S2,S3), we could identify a similar mixture but with gradually decreasing amount of pure iron phase Fe. For S4, only the FeS trollite phase was identified. The sample S5 for further increase of Q_H2S was a mixture of FeS₂ of pyrite and marcasite and FeS with trollite phase. The highest amount of pyrite FeS₂ was identified for sample S6 for the highest-used Q_H2S = 22.8 sccm and this sample was a mixture of pyrite and marcasite.

The results of analysis (electron microprobe) of composition of the deposited iron sulfide films (the samples S1–S4 in Table 2) show that these films contain Fe and S atoms with the atomic ratio slightly higher than unity. The analyzed films also contain a contamination of O with several at.%. Therefore, the film stoichiometry was close to FeS. The SEM photography of the typical part of the surface of the films (S1–S4) can be seen in Figure 9. We can identify a higher amount of macroparticles for the deposition conditions with lower Q_H2S.

4. Conclusions

A reactive HiPIMS system in Ar + H₂S gas mixture with a pure iron target was investigated at suitable conditions for iron sulfide thin film deposition. Plasma parameters such as electron density, n_e, and effective electron temperature, T_eff, were determined by the time-resolved Langmuir probe. The electron concentration reached the value n_e ≈ 10¹⁸ m⁻³ after the middle of the HiPIMS pulse for the non-reactive case of Q_H2S = 0 sccm. As the mass flow rate of H₂S gas gradually increased up to Q_H2S = 19 sccm, the electron concentration gradually decreased down to the value n_e ≈ 10¹⁷ m⁻³. The effective electron temperature, T_eff, was always in the range T_eff ≈ 1–1.6 eV during the active part of HiPIMS pulse for all the values of used Q_H2S and gas pressure p = 1 Pa. These findings correspond to results obtained by the QCM with magnetic filter used for the ionization fraction of sputtered particles r measurement. The value of r ≈ 40% measured for pure Fe HiPIMS sputtering at Q_H2S = 0 sccm, decreased to the value r ≈ 30% for Q_H2S ≈ 5 sccm, and r ≈ 20% was found for Q_H2S = 19 sccm. The deposition rate increased in the reactive sputtering mode in comparison with pure iron sputtering and was nearly constant for all investigated H₂S gas flows up to Q_H2S = 19 sccm at pressure p = 1 Pa. This is in agreement with the results of optical emission spectroscopy since we have found still intensive Fe⁺ ion and Fe neutral lines in the plasma also for high Q_H2S = 19 sccm. The study of current waveform character shows that although the deposition rate increases for higher Q_H2S, the discharge current gradually increases in the second part of the active HiPIMS pulse for Q_H2S > 0 sccm in comparison with the non-reactive case where we observed current decrease due to sputtering wind. Since the gas rarefaction near the target should probably still exist at these reactive conditions, this phenomenon could be explained by the higher yield of ion-induced electron secondary emission from the covered target by iron sulfide compounds. The iron sulfide thin films were deposited at conditions used for plasma diagnostics for different gas flow of H₂S. For lower Q_H2S we obtained rather FeS with the structure of trollite and for higher amount we obtained a mixture of pyrite and marcasite.