Contents lists available atScienceDirect
Surface & Coatings Technology
journal homepage:www.elsevier.com/locate/surfcoatBipolar HiPIMS for tailoring ion energies in thin film deposition
Julien Keraudy
a,1, Rommel Paulo B. Viloan
a, Michael A. Raadu
b, Nils Brenning
a,b,c,
Daniel Lundin
c, Ulf Helmersson
a,⁎aDepartment of Physics, Linköping University, Linköping SE-581 83, Sweden
bKTH Royal Institute of Technology, Department of Space and Plasma Physics, EECS, Stockholm SE-100 44, Sweden
cLaboratoire de Physique des Gaz et Plasmas - LPGP, UMR 8578 CNRS, Université Paris-Sud, Université Paris-Saclay, F-91405 Orsay Cedex, France
A R T I C L E I N F O Keywords: HiPIMS Sputtering Plasma A B S T R A C T
The effects of a positive pulse following a high-power impulse magnetron sputtering (HiPIMS) pulse are studied using energy-resolved mass spectrometry. This includes exploring the influence of a 200 μs long positive voltage pulse (Urev= 10–150 V) following a typical HiPIMS pulse on the ion-energy distribution function (IEDF) of the
various ions. We find that a portion of the Ti+flux is affected and gains an energy which corresponds to the
acceleration over the full potential Urev. The Ar+IEDF on the other hand illustrates that a large fraction of the
accelerated Ar+, gain energies corresponding to only a portion of U
rev. The Ti+IEDFs are consistent with the
assumption that practically all the Ti+, that are accelerated during the reverse pulse, originates from a region
adjacent to the target, in which the potential is uniformly increased with the applied potential Urev, while much
of the Ar+originates from a region further away from the target over which the potential drops from U revto a
lower potential consistent with the plasma potential achieved without the application of Urev. The deposition
rate is only slightly affected and decreases with Urev, reaching ~90% at Urev= 150 V. Both the Ti+IEDF and the
small deposition rate change indicate that the potential increase in the region close to the target is uniform and essentially free of electric fields, with the consequence that the motion of ions inside the region is not much influenced by the application of Urev. In this situation, Ti+will flow towards the outer boundary of the
target-adjacent region, with the momentum gained during the HiPIMS discharge pulse, independently of whether the positive pulse is applied or not. The metal ions that cross the boundary in the direction towards the substrate, and do this during the positive pulse, all gain an energy corresponding to the full positive applied potential Urev.
1. Introduction
Low-energy ion-bombardment during film growth has been utilized for many years [1] to increase adatom mobilities and give rise to near-surface mixing [2]. This has been used to increase film density [3], control grain size [4] and preferred orientation [5] at low growth temperatures TS, to grow metastable phases [6], to generate or release
film stresses [7], and, with proper choice of substrate, achieve low-temperature epitaxy [8,9]. For magnetron sputter deposition, where the ions which bombard the substrate are typically inert-gas ions, this re-quires high ion fluxes at low ion energies. Greene and co-workers [10,11] have shown that the flux of the bombarding ions typically needs to be 5–10 times higher than the flux of the neutral growth species. In magnetron sputtering, this requires that magnetron sources and magnetic field arrangements are designed in such a way that a sufficiently high plasma density can be maintained close to the
substrate [12] or that auxiliary ion sources are used [11].
The introduction of high-power impulse magnetron sputtering (HiPIMS), which generates large amounts of ions of the growth material [13], opened up for “self-ion-bombardment”, which has, in many cases, resulted in even better films properties [8,14]. The detailed atomistic reason for the improvement, when using HiPIMS, is still to be resolved, but most likely it is a combination of several effects, including a perfect mass-match between bombarding and surface species, a higher intrinsic ion energy Ekin and a reduction in inert gas incorporation in the
growing film due to the significant metal-ion-flux in relation to the noble-gas-ion flux.
For optimal film growth conditions, not only a high ion-flux is needed, but the energy Esof the bombarding species at the substrate
also has to be in a suitable range. Esis determined by the intrinsic
energy Ekinof the ions, together with their acceleration across the
plasma sheath at the substrate, yielding Es= Ekin+ Ze(Up,s− Us),
https://doi.org/10.1016/j.surfcoat.2018.12.090
Received 10 October 2018; Received in revised form 20 December 2018; Accepted 21 December 2018
⁎Corresponding author.
E-mail address:ulf.helmersson@liu.se(U. Helmersson).
1Present address: Oerlikon Surface Solutions AG, Oerlikon Balzers, Iramali 18, LI-9496, Balzers, Liechtenstein.
Available online 27 December 2018
0257-8972/ © 2018 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/BY-NC-ND/4.0/).
where Up,sis the plasma potential near the substrate, Usis a negative
potential applied to the substrate, Z is the charge state of the ion, and e the elemental charge. Optimal values of Esare, depending on the ion
and film material, within the range 20 to 100 eV [15]. In conventional magnetron sputtering, Ekinis low because the ions are primarily formed
from the process gas and extracted from the plasma near the substrate. The desired energy range can, in that case, only be reached by the application of Us, which has the effect to increase the potential drop
(Up,s− Us) across the plasma sheath at the substrate. In HiPIMS
dis-charges, the ion energy distribution function (IEDF) becomes broa-dened towards higher energies, and a fraction of the self-ion flux al-ready has Ekinin the desired energy range [16]. Still, the majority of the
ions have an energy of only a few eV, and a substrate bias is usually needed for optimal results. However, there are often limitations and complexities in applying a substrate bias, which is especially evident when depositing insulating films or depositing on insulating substrates. The use of a radio frequency substrate bias is possible, but has been shown to be difficult to implement during HiPIMS due to the mo-mentarily high ion fluxes during the pulses [17].
Another approach, as was suggested by Nakano et al. [18,19], is to apply a positive potential to the sputtering target after the negative HiPIMS pulses to raise the plasma potential and in this way accelerate ions towards the growing film. This approach was recently used by Wu et al. [20] who applied positive bias pulses (10–150 V) to the target immediately after the negative HiPIMS pulses and by Britun et al. [21] who applied 200–300 V positive bias pulses, with a short delay, after the negative HiPIMS pulses. Wu et al. [20] reported a 19% increase in the deposition rate of Cu films, reduced tensile stress, and smaller grains. They interpreted the results as being due to a high ion-flux bombarding the substrate during the positive pulse during which ions near the target are accelerated to the substrate. Britun et al. [21] report that the discharge ions undergo significant acceleration towards the substrate, but are well-thermalized perpendicular to that direction.
Here, we use energy-resolved mass spectrometry analyses to in-vestigate the effect of positive pulses on the ion energies and fluxes. In our case, we do not observe an increase in deposition rate, but a small decrease (up to 10%), but find that ion energies can be increased in an exceptionally controllable manner. The Ti-ion energy spectra are con-sistent with the assumption that practically all the Ti-ions that are ac-celerated during the reverse voltage pulse originate from a region ad-jacent to the target. The results indicate that the potential increase in this region is so uniform that there are only small changes in the electric fields inside the region, with the consequence that the motion of ions inside it is not much influenced by the application of the reversed po-tential Urev. The ions in this situation, to a first approximation, flow
towards the boundary of this target-adjacent region in directions (to-wards the target, or to(to-wards the substrate) that are independent of whether the positive pulse is applied or not. The ions that cross the boundary in the direction of the substrate, and do this during the po-sitive pulse, all gain an energy corresponding to Urev.
2. Experimental procedure
The experiments are carried out using a magnetically-unbalanced (type II) magnetron source mounted in a stainless-steel high-vacuum system with a base pressure of ~10−7Torr (~10−5Pa). The target is a
50-mm-diameter Ti disk (99.9% purity) with a thickness of 6 mm. Sputtering is carried out in Ar (99.997% purity) introduced at a flow rate of 50 sccm to provide a pressure of 5 mTorr (0.6 Pa). The cathode source is connected to a pulsing unit, Bipolar HiPSTER (Ionautics AB), fed by two dc power supplies, one that delivers a negative potential of 570 V for initiating classical HiPIMS pulses and a second one used to apply reverse positive potential pulses Urevfrom 0 to 150 V. Pulsing is
controlled using a HiPSTER synchronization unit (Ionautics AB) oper-ated at a frequency of 700 Hz. The negative pulses are 30 μs in length (corresponding to a duty cycle of 2%), and are immediately followed by
200 μs long positive pulses (intended to be long enough to act upon a majority of the ions in the decaying plasma after the HiPIMS pulse). In the following, the term “standard HiPIMS” refers to operating condi-tions in which only negative voltage pulses are applied to the target, while the term “bipolar HiPIMS” refers to conditions in which the ne-gative pulses are immediately followed by positive pulses. In both modes, the time-average target power during the negative pulses is maintained constant at ~110 W, corresponding to a power density of ~5.5 W/cm2(calculated using the full target surface area). The target
current and voltage waveforms are recorded using a Tektronix TDS 520 C digital oscilloscope.
In situ mass- and energy-dependent analyses of ion intensities in-cident at the substrate position are carried out using a Hiden Analytical PSM003 mass spectrometer capable of measuring ion energies up to 100 eV (singly charged ions). The spectrometer is facing the target and located at a distance of 8 cm. Ion energy distribution functions (IEDFs) are obtained for Ti+(48 amu), Ar+(36 amu) and Ti2+ions (48 amu)
for applied Urev up to 70 V. The spectrometer orifice is electrically
grounded during these experiments and the ion energy is scanned from 0 to 100 eV/charge, in 0.1 eV/charge steps. The total acquisition time per data point is 200 ms (corresponding to 140 pulses), while the spectrometer settings are separately tuned for each species, to account for the higher transmission at lower mass [22].
Scanning electron microscopy is used to measure film thicknesses which are used for the calculation of deposition rates. Each film is measured in three places and average values are calculated.
3. Experimental results
Typical discharge voltage UDand discharge current IDwaveforms
for standard and bipolar HiPIMS modes are shown inFig. 1. For both types of discharges, the initial negative target ignition voltage is 570 V which decreases slightly to 560 V at the end of the pulse. The IDcurve
rises rapidly at first, and then more slowly, to reach a maximum value ID, maxof ~9.8 A at 26 μs, before falling close to zero as UDis switched
off. In the bipolar mode, the negative pulse is immediately followed by a 200 μs-long reverse positive pulse Urev, which inFig. 1is 70 V, that
Fig. 1. (a) Discharge voltage UD(t) and (b) current ID(t) waveforms recorded
during standard HiPIMS and bipolar HiPIMS with a reverse 200 μs long positive voltage pulse of 70 V applied immediately following the negative pulse. The insert shows a magnification of the current just after switching off the HiPIMS pulse.
initially drive a small negative current, as seen in the insert ofFig. 1b. Fig. 2shows time-integrated IEDF measurements acquired for Ti+,
Ti2+and Ar+at the substrate position for eight different applied U rev
ranging from 0 to 70 V, in which Urev= 0 corresponds to standard
HiPIMS and the other curves to bipolar HiPIMS. We start by looking at the Ti+ flux energy distribution at U
rev= 0, which exhibits a
pro-nounced peak at low energy, ~3 eV. This is followed by a shoulder at energies up to ~20 eV and a high-energetic tail. This IEDF, as well as the IEDFs for Ti2+and Ar+, are similar to what is generally reported for
HiPIMS of metal targets in noble gases [16,23,24]. The origin of the high-energy tails is complex and has been attributed to ion acceleration in plasma instabilities [25], and/or in electric potential structures that are associated with spokes [26,27]. The IEDF curves for bipolar HiPIMS have a number of characteristic features that are useful in disentangling the plasma physics involved:
I. (Ti+) There is a narrow low-energy peak (below 5 eV) which
de-creases gradually in intensity with increasing Urev. This therefore is
an ion population which is influenced by Urev, but not significantly
accelerated by it.
II. (Ti+) A new equally-narrow peak appears at an energy which lies
slightly above eUrev. To the right of this peak there is a shoulder and
a high energy tail, similar to the standard HiPIMS, but up-shifted in energy by eUrev.
III. (Ti+) Between the two narrow peaks described above, the ion
en-ergy intensity level falls by a factor 2 to 3. This lower level is not significantly affected by the magnitude of Urev.
IV. (Ti2+) Both a narrow low-energy peak, and a broader maximum to
the right of it, are up-shifted in energy by 2eUrev, twice as much as
the corresponding features for singly charged Ti+ions.
V. (Ar+) The Ar+ energy distribution has both similarities and
dif-ferences as compared to the Ti+fluxes. The similarities are an
up-shifted narrow peak, and features to the right of it which are all upshifted in energy with eUrev. The main difference is that, to the
left of these up-shifted narrow peaks, the measured intensity in-creases instead of dein-creases, indicating that a significant fraction of the Ar+is accelerated to a portion of eU
rev.
The integrated intensities of features II and IV indicate that about half of the Ti+and Ti2+ions are accelerated over the full potential of
the reversed pulse. This gives the energy gain eUrevfor singly charged
ions, and 2eUrevfor doubly charged ions. Feature III indicates that, for
the Ti+ions, there is another group, again about half of the population,
which is not significantly influenced by the reversed pulse. Between these extremes we find the features I and V which indicate that there are two smaller groups of ions that are only partially influenced; the largest of these, feature V, consists of Ar+ions which are accelerated
but do not gain the full energy eUrev.
A small decrease of the deposition rates were observed when ap-plying the reversed pulse. In these measurements Urevlevels up to 150 V
was used and the deposition rate decreased with increasing voltage to ~90% of that of standard HiPIMS for the highest Urev.
4. Discussion
A simple model is applied to interpret the observed results. The model separates the discharge volume into three regions determined by the magnetic field topology, seeFig. 3a. The first region is the magnetic trap (MT), defined as the region in which both ends of the magnetic field lines intersect the target. The second region is a transition region (TR) in which one end of the magnetic field lines intersect the target and the other end intersects ground, i.e., either the chamber walls, a grounded substrate or the anode at the magnetron which is held at ground potential. In the third region, the grounded region (GR), both ends of the magnetic field lines intersect a surface held at ground po-tential. Please note that the spatial distribution of these three plasma volumes can vary significantly between different magnetrons, de-pending on magnetic field shape and the radial position of the anode ring. Also, due to imperfections in magnetron geometry, the boundaries between the regions can vary azimuthally.
A proposed potential profile during the time of the reverse pulse is shown inFig. 3b. This profile assumes that when the pulse polarity is reversed, a new potential profile is established in the time it takes to establish Child law sheaths at the boundaries – for typical discharge plasmas less than a microsecond [28]. As a rule, this makes the plasma typically a few V more positive than the most positive boundary that it is in good electric contact with. The wall sheath potential difference is marked ∆UWSinFig. 3b. In a plasma where the gyro radius is small, a
condition which generally holds in sputtering magnetrons, such contact is mainly along the magnetic field lines. Inside the magnetic trap, where the magnetic field lines are at both ends in contact with the target, the plasma potential will therefore be typically a few V more positive than Urev, while the plasma potential in the grounded region will be a few V
above ground potential. The potential in the transition region is more difficult to estimate, since here one end of the magnetic field lines is at the target potential Urevwhile the other is at ground potential. The
situation is complicated by the fact that the electron transport to the target is impeded by a large compression of the electron current to the center region of the target, and also by the magnetic mirror effect [29]. This gives a reduced electric contact with the most positive boundary, Fig. 2. Time-integrated ion energy distribution functions (IEDFs) recorded at
the substrate position during standard HiPIMS (Urev= 0 V) and bipolar HiPIMS
the target, which normally determines the plasma potential. In this si-tuation, we can only conclude that the full potential Urevhas to fall
across the transition region, but that the potential profile along the zRT
axis is uncertain. This is indicated by dashed potential lines and a question mark inFig. 3b.
Now we are equipped to discuss the IEDF data. We start by noting that a discharge with reversed polarity cannot be easily ignited in a sputtering magnetron configuration. The discharge therefore dies out very quickly after the negative pulse. Measurements show that the electron temperature after a HiPIMS pulse decays towards ~0.2 eV [30], and therefore the ionization rate rapidly drops to zero. At the onset of the reversed pulse we are left with a “plasma reservoir”, the ions and electrons that are present at the end of the discharge. The question is the further fate of these ions in the new electric potential profile shown inFig. 3b.
Let us start with Ti+ and with the standard HiPIMS discharge.
During the pulse, the majority (70–90%) of the produced Ti+are
at-tracted back to the target by an electric field Ezthat extends into the
plasma [31,32]. For our 30 μs long HiPIMS pulses we estimate that most of the ions that are produced return to the target this way. At the end of the HiPIMS pulse, a minority has already reached all the way to the substrate position (here, the mass spectrometer). None of these popu-lations can obviously be influenced by the reversed pulse. The question therefore is what happens to the remaining ions (also a minority), those that still are in the volume when the HiPIMS pulse is switched off. At this time the Ezfield disappears, and the number of ions that can leave
the magnetic trap towards the substrate is increased [33]. The effect from the reversed potential on these ions can be explained by where they are, and what velocity they have, at the end of the HiPIMS pulse. Some ions have left the magnetic trap during the HiPIMS pulse, and also passed through the transition region before the reversed pulse is ap-plied. The electric field of the reversed pulse does not reach these ions, which therefore continue towards the mass spectrometer without being accelerated (feature III). Ions with high intrinsic energy are likely to be overrepresented in this group. A second group of ions are inside the magnetic trap (inside the boundary marked A inFig. 3), and are at a position and with a velocity such that they leave the trap during the reversed pulse. These ions gain the full energy ZeUrevas they cross the
transition region (features II and IV). A third group of ions are also
inside the magnetic trap, but have such small velocity in the z direction that they leave the trap only after the end of the reversed pulse. Ions with low intrinsic energy are overrepresented in this group and they also gain no extra energy.
Let us now consider why the narrow low-energy peak, in the IEDFs ofFig. 2, decreases gradually in intensity with increasing Urev(feature
I). At first, this seems hard to reconcile with the picture given in the preceding paragraph, that ions which remain in the magnetic trap during the reversed pulse are little influenced by Urev. However, a small
electric field in the trap region can change the number of ions that remain in the trap – the number of ions belonging to the third group described above. We propose, that the influence of Urevin the magnetic
trap is indeed very limited, but not exactly zero and increases with increasing Urev. This is indicated inFig. 3b by a small slope in the
po-tential to the left of zA. For ions with a low kinetic energy, even a very
small potential difference across the magnetic trap can significantly increase the ion loss rate during the reversed pulse. It seems likely that such a potential difference should be higher for larger Urev, and this
would explain the decrease in the low energy peak with increasing Urev
as more of the thermalized ions are drawn out of the magnetic trap during the reversed pulse.
For the Ar+population, most of the features are analogous to those
discussed above for the Ti+and have the same explanations. However,
in contrast to the case for the Ti+, there is here a population of partially
accelerated ions, feature number V in the list above. This difference between Ti+and Ar+is proposed to be due to a difference in spatial
distribution during a HiPIMS discharge, and therefore also at the end of the first pulse. The Ti atoms are sputtered form the target and ionized in the vicinity of it, and a large fraction of them are attracted back to the target. The Ar atoms on the other hand are depleted by gas rarefaction close to the target [34] and replaced by refill from the surrounding gas reservoir. This leads to a situation where Ti+ions are predominantly
found closer to the target [34], in the magnetic trap, while Ar+are
found both in the trap and in the transition region. It is the latter sub-population that can become partially accelerated in bipolar mode. For example: if they were in the middle of that region, and the potential profile were linear, then they would only become accelerated across half the potential, and therefore get the energy eUrev/2.
Fig. 3. (a) The magnetic topology for an unbalanced magnetron such as ours, defining the magnetic trap MT (cross-hatched), the transition region TR, and the grounded region GR, as described in the text. (b) A proposed plasma potential profile during the reversed pulse, taken along the axis zRTthat is drawn in panel (a).
5. Summary and conclusions
We have experimentally investigated the effect on the ion energy of bipolar-HiPIMS, where the standard HiPIMS-pulse is followed by a re-versed potential applied on the target. It is found that the amplitude of the reversed potential gives excellent control over the ion energies. This should be of great importance for the thin film community, especially in the growth of insulating thin films or when insulating substrates are used, making it impossible to apply a suitable substrate bias voltage for the acceleration of the ions to desired energies.
We propose that the potential distribution during the reversed pulse is mainly determined by the electric contact, along the magnetic field lines, from the plasma to the walls and/or the electrodes. This gives a simple model which is illustrated inFig. 3, and which is found to be consistent with the IEDFs presented inFig. 2. According to this model all ions that are inside the magnetic trap when the HiPIMS pulse is switched off, and then leave in the direction away from the target, will gain the same energy per unit charge when passing across the transition region, over to the grounded region. This is the case for most of the Ti+
and Ti2+(the latter receiving twice the energy). Please notice that this
energy gain is independent of the uncertainty in the potential profile in the transition region. Ions that are in the transition region at the time of application of the reverse pulse, on the other hand, will become par-tially accelerated. This is the case for some of the Ar+. Naturally, the
distribution of energy in the partially accelerated population would depend in a complicated way both on the potential profile in the transition region and on the spatial distribution of the ion species in question.
As mentioned in the introduction, Wu et al. [20] reported a 19% increase in the deposition rate of Cu films, while the deposition rate decreases slightly with Urevin this paper. As one possible reason for this
difference we note that the shape of the transition region (seeFig. 3), in which we propose the ions to be accelerated, depends on both the magnetron's degree of unbalance and on the radial location of the anode ring. Differences in these respects may give ion fluxes with dif-ferent divergences in the two devices.
As is clear to anyone reading this article there are great potentials in re-designing the experimental set-up making it possible to change the ion-energy distributions and the collimation of the ion flux at will. Parameters to play with include, for example the shape and strength of the magnetic arrangement, the anode position and shape, the length and intensity of the HiPIMS-pulse, the time delay between the end of the HiPIMS-pulse and the reversed pulse, and the length and magnitude of the reversed pulse, to mention the most obvious parameters. Acknowledgments
This work has been supported by the Swedish Research Council (grant VR 621-2014-4882) and the Swedish Government Strategic Research Area in Materials Science on Functional materials at Linköping University (Faculty Grant SFO-Mat-LiU No. 2009-00971). We would also like to thank Svensk-Franska Stiftelsen for supporting a re-search visit of N.B. to Université Paris-Sud, and Ionautics AB for sup-plying pulsing equipment that made these experiments possible. Finally, the authors acknowledge Prof. T. Minea and Dr. T. J. Petty at Université Paris-Sud for discussions and work onFig. 3, respectively. References
[1] J.M.E. Harper, J.J. Cuomo, R.J. Gambino, H.R. Kaufman, Nucl. Instrum. Meth. B
886 (1985) 7–8,https://doi.org/10.1016/0168-583X(85)90489-6.
[2] V. Georgieva, A.F. Voter, A. Bogaerts, Cryst. Growth Des. 11 (2011) 2553,https:// doi.org/10.1021/cg200318h.
[3] J.J. Cuomo, J.P. Doyle, J. Bruley, J.C. Liu, Appl. Phys. Lett. 58 (1991) 466,https:// doi.org/10.1063/1.104609.
[4] H.A. Atwater, C.V. Thompson, H.I. Smith, J. Appl. Phys. 64 (1988) 2337,https:// doi.org/10.1063/1.341665.
[5] A. Rodríguez-Navarro, W. Otaño-Rivera, J.M. García-Ruiz, R. Messier, L.J. Pilione, J. Mater. Res. 12 (1997) 1850,https://doi.org/10.1557/JMR.1997.0254. [6] J.E. Greene, S.A. Barnett, K.C. Cadien, M.A. Ray, J. Cryst. Growth 56 (1982) 389,
https://doi.org/10.1016/0022-0248(82)90458-4.
[7] M.M.M. Bilek, D.R. McKenzie, Surf. Coat. Technol. 200 (2006) 4345,https://doi. org/10.1016/j.surfcoat.2005.02.161.
[8] M.M.S. Villamayor, J. Keraudy, T. Shimizu, R.P.B. Viloan, R. Boyd, D. Lundin, J.E. Greene, I. Petrov, U. Helmersson, J. Vac. Sci. Technol. A 36 (2018) 061511,
https://doi.org/10.1116/1.5052702.
[9] T. Ohmi, K. Hashimoto, M. Morita, T. Shibata, J. Appl. Phys. 69 (1991) 2062,
https://doi.org/10.1063/1.348732.
[10] T. Lee, H. Seo, H. Hwang, B. Howe, S. Kodambaka, J.E. Greene, I. Petrov, Thin Solid Films 518 (2010) 5169,https://doi.org/10.1016/j.tsf.2010.04.028.
[11] B.W. Karr, D.G. Cahill, I. Petrov, J.E. Greene, Phys. Rev. B 61 (2000) 16137,
https://doi.org/10.1103/PhysRevB.61.16137.
[12] I. Petrov, F. Adibi, J.E. Greene, J. Vac. Sci. Technol. A 10 (1992) 3283,https://doi. org/10.1116/1.577812.
[13] V. Kouznetsov, K. Macák, J.M. Schneider, U. Helmersson, I. Petrov, Surf. Coat. Technol. 122 (1999) 290,https://doi.org/10.1016/S0257-8972(99)00292-3. [14] M. Muhlbacher, G. Greczynski, B. Sartory, N. Schalk, J. Lu, I. Petrov, J.E. Greene,
L. Hultman, C. Mitterer, Sci. Rep. 8 (5360) (2018),https://doi.org/10.1038/ s41598-018-23782-9.
[15] W. Li, Z.-Q. Ma, Y. Wang, D.-M. Wang, Chin. Phys. Lett. 23 (2006) 178http://cpl. iphy.ac.cn/Y2006/V23/I1/178.
[16] J. Bohlmark, M. Lattemann, J.T. Gudmundsson, A.P. Ehiasarian, Y. Aranda Gonzalvo, N. Brenning, U. Helmersson, Thin Solid Films 515 (2006) 1522,https:// doi.org/10.1016/j.tsf.2006.04.051.
[17] D. Lundin, Unpublished Results, (2018).
[18] T. Nakano, N. Hirukawa, S. Saeki, S. Baba, Vacuum 87 (2013) 109,https://doi.org/ 10.1016/j.vacuum.2012.03.010.
[19] T. Nakano, T. Umahashi, S. Baba, Jpn. J. Appl. Phys. 53 (2014) 028001,https://doi. org/10.7567/JJAP.53.028001.
[20] B. Wu, I. Haehnlein, I. Shchelkanov, J. McLain, D. Patel, J. Uhlig, B. Jurczyk, Y. Leng, D.N. Ruzic, Vacuum 150 (2018) 216,https://doi.org/10.1016/j.vacuum. 2018.01.011.
[21] N. Britun, M. Michiels, T. Godfroid, R. Snyders, Appl. Phys. Lett. 112 (2018) 234103,https://doi.org/10.1063/1.5030697.
[22] G. Greczynski, L. Hultman, Vacuum 84 (2010) 1159,https://doi.org/10.1016/j. vacuum.2010.01.055.
[23] P.-Y. Jouan, L. Le Brizoual, M. Ganciu, C. Cardinaud, S. Tricot, M.-A. Djouadi, IEEE Trans. Plasma Sci. 38 (2010) 3089,https://doi.org/10.1109/TPS.2010.2073688. [24] M. Palmucci, N. Britun, T. Silva, R. Snyders, S. Konstantinidis, J. Phys. D. Appl.
Phys. 46 (2013) 215201http://stacks.iop.org/JPhysD/46/215201.
[25] D. Lundin, P. Larsson, E. Wallin, E. Lattemann, N. Brenning, U. Helmersson, Plasma Sources Sci. Technol. 17 (2008) 035021,https://doi.org/10.1088/0963-0252/17/ 3/035021.
[26] C. Maszl, W. Breilmann, J. Benedikt, A. von Keudell, J. Phys. D. Appl. Phys. 47 (2014) 224002,https://doi.org/10.1088/0022-3727/47/22/224002.
[27] N. Brenning, D. Lundin, T. Minea, C. Costin, C. Vitelaru, J. Phys. D. Appl. Phys. 46 (2013) 084005,https://doi.org/10.1088/0022-3727/46/8/084005.
[28] A. Anders, Surf. Coat. Technol. 183 (2004) 301,https://doi.org/10.1016/j.surfcoat. 2003.09.049.
[29] S. Knight, Planet. Space Sci. 21 (1973) 741, https://doi.org/10.1016/0032-0633(73)90093-7.
[30] P. Poolcharuansin, J.W. Bradley, Plasma Sources Sci. Technol. 19 (2010) 025010,
https://doi.org/10.1088/0963-0252/19/2/025010.
[31] C. Huo, D. Lundin, M.A. Raadu, A. Anders, J.T. Gundmundsson, N. Brenning, Plasma Sources Sci. Technol. 22 (2013) 045005, https://doi.org/10.1088/0963-0252/22/4/045005.
[32] A. Mishra, P.J. Kelly, J.W. Bradley, Plasma Sources Sci. Technol. 19 (2010) 045014,
https://doi.org/10.1088/0963-0252/19/4/045014.
[33] A. Butler, N. Brenning, M.A. Raadu, J.T. Gudmundsson, T. Minea, D. Lundin, Plasma Sources Sci. Technol. 27 (2018) 105005, https://doi.org/10.1088/1361-6595/aae05b.
[34] C. Huo, M.A. Raadu, D. Lundin, J.T. Gudmundsson, A. Anders, N. Brenning, Plasma Sources Sci. Technol. 21 (2012) 045004,https://doi.org/10.1088/0963-0252/21/ 4/045004.