Generation of singly charged Ar and , doubly charged and , and of and dimer ions in a high power impulse magnetron sputtering (HiPIMS) discharge with a Ti cathode was investigated. Energy-resolved mass spectrometry was employed. The argon gas pressures varied between 0.5 and 2.0 Pa. Energy spectra of monomer ions are composed of low- and high-energy components. The energetic position of the high-energy component is approximately twice as large for doubly charged ions compared to singly charged ions. Intensities of and dimer ions are considerably smaller during HiPIMS compared to dc magnetron sputtering.
I. INTRODUCTION
Plasma-based processes play an important role in both basic research and industrial applications.1 Deposition of thin solid films is frequently carried out with the help of magnetron discharges.2–4 Pulsed magnetron discharges offering an enhanced plasma density have received much attention in the past years.5–11 Electron densities as large as or 1% of the gas density and even larger ionisation fractions of metal atoms with charge states up to 4+ can be achieved during so-called high power impulse magnetron sputtering (HiPIMS).5,12–15 The large ion fraction impacts the growing film and may enhance film adherence, uniformity, and compactness. Recent observations of moving ionisation zones or spokes have shed new light on the ionisation process in magnetron discharges.16–18
In this paper, we investigate the formation of singly charged and monomer, and dimer, and doubly charged and ions in a high power impulse magnetron sputtering (HiPIMS) discharge with a titanium target. Energy-resolved mass spectrometry has been utilized to investigate the details of the ion formation process. A comparison with results obtained during direct current (dc) magnetron sputtering is included.19
II. EXPERIMENT
The experimental setup has been described in detail elsewhere.19–21 The vacuum chamber with a diameter of 35 cm is pumped by a turbo-molecular pump (pumping speed 220 l/s) to a base pressure of less than . Argon gas (purity 99.999%) is introduced into the chamber with the help of a gas flow controller; the argon flow rate is 18 sccm. The operating pressure is set in the range with the help of a ultrahigh vacuum gate valve between pump and chamber. A commercial (Lesker Torus) planar unbalanced magnetron is attached to the horizontal flange of the vacuum chamber and inserted about 29 cm into the chamber. The magnetron is equipped with a Ti target (diameter 2 in., purity 99.7%). The magnetron discharge is driven by a dc power supply (Advanced Energy MDX–1K) in combination with a homemade power switch.22–25 A commercial pulse-delay generator (Agilent 33250A) was employed to set the repetition frequency and the pulse width . The magnetron discharge is operated in HiPIMS mode with a typical voltage of and a mean discharge power of 150 W. The same setup was used for the dc magnetron sputtering mode.
Energy-resolved mass spectrometry is performed with a commercial Hiden EQP 1000 mass/energy analyzer (Hiden Analytical Ltd., UK). The analyzer is equipped with a grounded entrance orifice with a diameter of . The magnetron is mounted at a distance of 5 cm from the orifice. The mass spectrometer is differentially pumped by a separate turbomolecular pump (pumping speed 60 l/s) to a residual gas pressure of better than inside the analyzer. Measurements are performed in the positive ion mass spectrometry (+SIMS) mode with a dwell time of 0.2 s. Typical ion spectra are averaged over 10 scans.
III. RESULTS
Recorded mass spectra of positive ions are dominated by singly charged and ions. Argon has three stable isotopes at , 38, and 40 with relative abundances of 0.337%, 0.063%, and 99.6%, respectively,26 where and are ion mass and ion charge number, respectively. Titanium has five stable isotopes at –50 with relative abundances of 8.3%, 7.4%, 73.7%, 5.4%, and 5.2%.26 Measurements are reported for the most abundant singly and doubly charged charged ( ), ( ), ( ), and ( ) monomer ions and for ( ) and ( ) dimer ions.
Energy spectra of singly charged and ions are displayed in Fig. 1 for three different gas pressures 0.5 Pa, 1.0 Pa, and 2.0 Pa. The energy distribution of ions consists of a low-energy part extending up to about 10 eV followed by a high-energy tail extending up to more than 50 eV. The intensity of the high-energy tail strongly decreases with increasing pressure. Similar energy spectra for ions have been reported before.4,27–30 The energy distribution of ions also displays a low-energy peak which is most pronounced at the largest pressure of 2.0 Pa followed by a second broad peak at about 12 eV and a pronounced high-energy tail extending beyond 80 eV. Also shown are energy spectra of and ions during dc magnetron sputtering measured under otherwise identical conditions at a gas pressure . In comparison to HiPIMS, the energy distributions obtained during dc magnetron sputtering are less broad and the high-energy component is significantly reduced. A similar observation has been made by Lundin et al.31
Energy spectra of doubly ionised and ions are displayed in Fig. 2. The energy distributions of ions show a narrow low-energy peak followed by a pronounced high-energy tails which, for the lowest gas pressure , extends to more than 80 eV. Similar energy distributions are observed for ions where, at the lowest pressure , the high-energy component extends beyond 120 eV and thus much further compared to singly-charged ions at the same pressure.
Energy spectra of dimer ions are displayed in Fig. 3. ions are produced with very low kinetic energies, indicating that the formation processes differ considerably from that of ions.
The pressure dependence of the energy-integrated ion intensities is shown in Fig. 4. The total intensity of ions shows little variation with pressure. The intensity of slightly increases with pressure. Intensities of doubly charged and ions decrease with pressure as does the and the ratio of singly-to-doubly charged ions. The intensity of dimer ions also decreases with pressure. It should be noted that due to its small count rate, the statistical accuracy of the intensity is comparatively poor (about ) and no energy distributions are presented here. The pronounced increase of the ion intensity with pressure is close to a quadratic dependency.
IV. DISCUSSION
A. Singly charged Ti+ ions
The energy distribution of ions is dominated by a pronounced high-energy component which peaks at about 11 eV and extends up to more than 80 eV. Ti atoms originate from the cathode and are liberated by sputtering. Sputtered Ti atoms are ionised in the plasma region and detected as positively charged ions. The energy distribution of sputtered atoms is described by Thompson’s formula32
where is the kinetic energy of sputtered atoms and . The distribution has a maximum near , where is the surface binding energy. The measured energy distribution of ions (Fig. 1) is considerably steeper than predicted by Eq. (1), however. Only the weak high-energy tail at energies larger than 50 eV is in accordance with Thompson’s formula [Eq. (1)]; it resembles the remaining part of the original sputtering contribution as shown in Fig. 5. It means that sputtered ions have lost much of their initial energy through thermalizing collisions. Attempts have been made to describe the measured ion energy distribution of thermalized ions by a so-called shifted Maxwellian function19,33–35
where is the ion temperature and is the energy by which the Maxwellian distribution is shifted. It turns out that a shifted Maxwellian distribution reasonably well describes the energy distribution of ions above 10 eV (Fig. 5). A second Mawellian distribution is required to describe the low-energy part. A similar approach with two Maxwellian distributions has been utilized by Hecimovic et al. and attributed to different temporal phases of the discharge.36,37
B. Singly charged Ar+ ions
The energy spectrum of is characterised by a pronounced peak at very small energies of about 0 eV followed by a shoulder at and a high-energy tail which is strongly pressure dependent and, at the lowest pressure of , extends beyond 50 eV. At first glance, the energy distribution appears different compared to that of ions. A closer look reveals, however, that the high-energy components of and are rather similar both in shape and in position (Fig. 5). The origin of the high-energy component is not fully understood yet. Mechanisms which have been mentioned in this respect are related to accelerated ions impinging on the negatively biased cathode. Part of the impinging argon ions are backscattered and, if simultaneously become neutralized, can leave the cathode with a large kinetic energy.27,28,30
The second mechanism is related to argon ions which are implanted in the cathode’s material due to ion bombardment. Typical implantation depths under the present conditions are about 1.2 nm.38 Buried argon atoms are subsequently liberated by the continuing ion bombardment; the ejected Ar atoms may become ionized again in the plasma region and are detected as positive ions. It was shown that HiPIMS discharges can be dominated by a large fraction of so-called recycled ions, i.e., Ar atoms which return from the cathode and become ionised again.11 As in the case of ions, the measured energy distribution (Fig. 1) is considerably steeper than predicted by Eq. (1). It means that recycled Ar atoms and ions have lost much of their energy through thermalizing collisions. A shifted Maxwellian distribution [Eq. (2)] reasonably well describes the high-energy component above 20 eV but does not comply with the measured distribution at lower energies. Position and width of this component are, particularly at the lowest pressure , quite similar compared to ions (Table I). The smaller ion intensity of the high-energy component compared to ions is readily explained by the much larger ionisation energy which reduces the ionisation probability. The low-energy component of slow ions was accounted for by a second Maxwellian function. Based on time-resolved measurements, it has been shown that the high-component occurs during the on time of the discharge, i.e., when the discharge current reaches its maximum, while the low-energy component occurs during the off time, i.e., during the afterglow when the discharge current approaches zero.36,37
. | + . | + . | + . | + . |
---|---|---|---|---|
(eV) | 0 | 0 | 0 | 2.5 |
(eV) | 2.1 | 2.8 | 1.65 | 0.9 |
(eV) | 7.25 | 17.6 | 6.9 | 15.6 |
(eV) | 2.5 | 2.9 | 3.05 | 4.2 |
. | + . | + . | + . | + . |
---|---|---|---|---|
(eV) | 0 | 0 | 0 | 2.5 |
(eV) | 2.1 | 2.8 | 1.65 | 0.9 |
(eV) | 7.25 | 17.6 | 6.9 | 15.6 |
(eV) | 2.5 | 2.9 | 3.05 | 4.2 |
In addition, the mean free path of is largely determined by charge changing reactions with neutral Ar atoms from the buffer gas, e.g., + Ar Ar + . The reaction gives rise to the formation of ions which are in thermal equilibrium with gas atoms. The cross section of this resonant process is large39 and the corresponding mean free path of ions becomes small. The estimated mean free path , where is the density of argon atoms, is of the order of a few centimetre (Table II), i.e., sufficiently short to explain the conversion of a large fraction of fast into slow ions. The contribution was accounted with a Gaussian distribution; its inclusion improves the agreement with experiment. We thus have three contributions of which two are related to different phases during the HiPIMS pulse, while the part with the lowest kinetic energy is due to charge changing reactions with gas atoms. The width of this contribution is determined by the energy resolution of the analyser. It also explains the observed negative energies which are an artefact of the measuring process.
Ion . | . | . | Mean free path (cm) . | ||
---|---|---|---|---|---|
. | (eV) . | ( ) . | . | . | . |
1 | 57 | 1.5 | 0.73 | 0.36 | |
10 | 45 | 1.9 | 0.93 | 0.46 | |
100 | 34 | 2.4 | 1.2 | 0.61 | |
1.5 | 31 | 2.7 | 1.3 | 0.67 | |
10 | 26 | 3.2 | 1.6 | 0.80 | |
120 | 21 | 3.9 | 2.0 | 0.99 |
Ion . | . | . | Mean free path (cm) . | ||
---|---|---|---|---|---|
. | (eV) . | ( ) . | . | . | . |
1 | 57 | 1.5 | 0.73 | 0.36 | |
10 | 45 | 1.9 | 0.93 | 0.46 | |
100 | 34 | 2.4 | 1.2 | 0.61 | |
1.5 | 31 | 2.7 | 1.3 | 0.67 | |
10 | 26 | 3.2 | 1.6 | 0.80 | |
120 | 21 | 3.9 | 2.0 | 0.99 |
A remarkable difference between and is noted with respect to the relative intensity of the low-energy and the pressure dependence of the high-energy component. The intensity of the high-energy component strongly decreases with increasing pressure while a much weaker pressure dependence is noted for . A possible explanation can be found in the different mean free path of and ions in the Ar buffer gas. The charge changing reaction is endothermic and the corresponding cross section is, hence, much smaller compared to that for the resonant charge changing + Ar reaction.39 It gives rise to a larger mean free path of ions resulting in a smaller pressure dependency. As a further consequence, the low-energy component of the ion energy distribution is comparatively weak and, in particular, much smaller compared to . As in the case of ions, the agreement between fitted and measured energy distributions improves if two Maxwellian distributions which are related to the different phases during HiPIMS and a Gaussian distribution to account for fully thermalized ions are employed.
C. Comparison of singly and doubly charged ions
The energy distribution of ions displays similar features when compared to , i.e., a low-energy peak followed by a pronounced high-energy component which peaks at about 18–24 eV (Fig. 2). The high-energy component is most pronounced at the smallest pressure and strongly decreases at larger gas pressures. The behaviour is similar to that of singly charged ions. The mean free path of ions is largely determined by the resonant two-electron transfer reaction , while the one-electron transfer reaction is more than one order of magnitude smaller.40 The two-electron transfer cross section is thus comparable to that of the one-electron reaction .39 The estimated mean free path of is of the order of a few centimetres (Table II), which is reasonable to explain the observed pressure dependency.
As for argon ions, the energy distribution of and ions are rather similar. The relative intensities of the low-energy and high-energy components show a much weaker pressure dependence compared to and ions, however, which is readily explained by the smaller reaction rates for the non-resonant charge transfer process of Ti ions with Ar atoms.
A major difference between singly and doubly charged ions concerns the energetic position of the high-energy component which peaks near 12 eV for singly charged ions and at for doubly charged ions. It appears straightforward to attribute this difference to the different charge number.16 Figure 6 displays energy distributions for and ions ( ) as a function of the specific energy , i.e., kinetic energy per charge number . The scaling results in rather similar energy distributions. We would like to emphasize that the measured kinetic energy of plasma ions is influenced by the plasma potential.13,36,41,42 Our results indicate an electric field reversal by a localised positive plasma potential. A positive plasma potential or potential hump was employed to explain the formation of moving ionisation zones or spokes in magnetron discharges.17,18,43 The extra energy provided by the plasma potential depends on the charge number and thus will be twice as large for doubly charged ions compared to singly charged ions ( is the elementary charge). The present observations indicate that the energy gained by the plasma potential is identical to the energy shift of the Maxwellian distribution [Eq. (2)] used to fit the high-energy component of investigated ion energy distributions. The deduced (time averaged) plasma potential relevant for the high-energy component amounts to at ; it decreases with increasing pressure.
D. Ar+2 dimer ions
and dimer ions were also recorded. The total intensity of ions is about four orders of magnitude smaller compared to, e.g., ions. The total intensity of is even smaller and amounts to about of the total intensity at (Fig. 4). Due to poor statistics, no energy distributions of ions are reported here. This is in contrast to dc magnetron sputtering where orders of magnitude larger intensities and and ratios of were observed.19 Other dimer ions like Ar were not detected.
The energy distribution of ions is characterized by a pronounced low-energy peak of fully thermalized ions. It indicates that ions form via gas phase processes in a region where the plasma potential is close to zero, i.e., presumably during the off time of the discharge. No evidence for the ejection of dimers from bursting gas bubbles caused by ion implantation has been found.44 Three-body reactions of ions with two neutral argon atoms and associative ionization of highly excited atoms with ground state Ar atoms are the dominant processes.19,45–49 The mentioned processes depend at least linearly or quadratically on the Ar density which explains the strong increase of the density with gas pressure.
Loss of molecules takes place via electron impact-induced dissociation processes and thus increases with electron density.45–50 The rate constant for dissociative recombination strongly increases with decreasing electron temperature. Recombination processes thus play an important role also in the afterglow.21,51 It appears likely that the much larger electron density during the HiPIMS pulse together with an enhanced dissociative recombination in the afterglow is responsible for a larger loss rate of ions and for its much smaller ion density compared to dc magnetron sputtering. Similar arguments may hold for the loss of ions.
V. CONCLUSIONS
The energy distribution of singly charged and , doubly charged and , and of and dimer ions was investigated during a HiPIMS discharge. Measured ion energy distributions display low- and high-energy components which can be described by a shifted bi-Maxwellian distribution. The high-energy components of singly and doubly charged ions fall together when plotted versus scaled energy , where is the charge number. The observation points to an electric field reversal and supports the existence of a potential hump used to explain the occurrence of moving ionisation zones or spokes. The generation of dimer ions displays a quadratic pressure dependency. The intensity of and dimer ions is 2–3 orders of magnitude smaller during HiPIMS compared to dc magnetron sputtering; it is attributed to loss processes which strongly increase with electron density.
ACKNOWLEDGMENTS
The work was partly supported by Project No. 17-08389S of the Czech Science Foundation, by Operational Programme Research, Development and Education financed by European Structural and Investment Funds, and the Czech Ministry of Education, Youth and Sports (Project Nos. SOLID21—CZ.02.1.01/0.0/0.0/16_019/0000760 and CZ.02.2.69/0.0/0.0/16_027/0008215), and by the German Academic Exchange Service (DAAD).