Abstract

Zirconium oxide is an attractive material for a robust top coating on architectural glass. Metallic Zr shows a very high affinity to oxygen, which leads to some challenges in target manufacturing and sputtering. Therefore, electrically conductive sub-oxide targets are an interesting alternative to the metal. In this contribution we investigate the sputtering from ZrOx cylindrical targets, with focus on practical aspects, such as burn-in and arc management requirements to the power supply. The ZrOx targets show a distinct transition from a quasi-metallic mode, in which all oxygen supplied to the process is consumed, to a reactive mode, where excess oxygen is present in the sputtering atmosphere. However, the transition between these modes does not show a hysteresis. Basic sputtering data and film properties as function of oxygen addition to the process are discussed.

Introduction

Zirconium oxide is well established as technical ceramic material, due to its superior mechanical and thermal stability. In large-area coating of architectural glass it has recently gained increasing interest. Desirable properties of ZrO2 coatings reported so far include the improved thermal stability of the layer stack, UV blocking properties which can be tuned by nitrogen admixture, as well as a superior scratch resistance [1-4].

Transition mode sputtering of transparent films from metal targets at high rates has been demonstrated. However, this requires gas flow control, and in oxide mode the deposition rate is very low [5]. Moreover, the manufacturing of metallic Zr targets is difficult and expensive due to the high oxygen affinity of the metal.

As an alternative to reactive sputtering from metallic targets, the use of electrically conductive substoichiometric targets is well established for a range of transition metals. However, the manufacturing of ZrOx targets presents a challenge and has only recently been successful.

In the current study, we show results of target development with experimental data from a lab coater and the development and testing of AC power supply settings on a dual magnetron for industrial glass coating.

Experimental Procedure 

The sputtering of ZrO2 from ZrOx suboxide targets was investigated in two different sputtering chambers. The first is a lab coater with a single rotary magnetron 500mm in length. It is driven by a pulsed DC power supply with 95% on-time at a frequency of 20 kHz. The maximum power was 8 kW, which corresponds to 120 kW on a 3.8 m dual magnetron. Ar flows of 200 and 300 sccm were used, leading to chamber pressures of 2.5 and 3*10-3 mbar respectively. As reactive gas, O2 was used. Test coatings were deposited on 1mm thick soda lime glass slides and characterized by optical transmission measurements. The cooling capability of the material, determining the maximum power density was estimated by determining the power delivered to the cooling water in relation to the sputtering power. It is determined by the thermal conductivity and the target thickness.

The second chamber is a jumbo size test chamber with a dual rotatable magnetron 3.8 m in length. In this chamber, full-size process conditions in a coater for architectural glass can be investigated. The magnetron is equipped with a reactive gas supply with reactive gas flow control loop and which allows tuning of the gas flow in 5 zones for best uniformity. At 400 sccm Ar the pressure was 2.2*10-3 mbar. The power supply used is a TruPlasma MF 7100 (G2) 70kHz. Further details of this test set-up are given in [6]. No samples were deposited in this chamber.

Results

Lab Coater Tests.

Figure 1 shows the target voltage and current as function of the sputtering power in 300 sccm Ar with approx. 2 to 7% O2. For these tests, the power was ramped-up at a constant rate. At very low power, the target voltage or impedance is rather high, and a transition to a state of low impedance is seen at about 1 to 3 kW, depending on the oxygen flow.

Figure 1. Cathode voltage and current in pulsed DC sputtering as function of the applied power. Ar flow 300 sccm with the addition of O2 as indicated.

The power required for this transition increases with oxygen flow. This indicates that the target surface will probably be fully oxidized at low power. This state is also associated with a high arcing rate. At very low power, the oxidized state may also be sustained by outgassing of the chamber. At higher power, the surface is cleaned by sputtering and it reaches its sub-oxide state, when the sputtering rate exceeds the rate of surface oxidation. Dynamic film deposition rates at 4, 5 and 8 kW are shown in Figure 2. As it is to be expected, the rate increases almost linearly with the power and, for the oxygen flows used, it decreases only by up to 20% with respect to sputtering in Ar only. Comparing the deposition conditions with the results in Figure 1, it is seen that all deposition tests were carried out in the suboxide regime where the impedance and the arcing rate is low. Optical transmission data at 4 kW and 200 sccm Ar are shown in Figure 3. Without O2, even a 5nm layer shows about 5% extinction in the visible range. With 5sccm O2, a 119 nm thick layer shows considerable extinction, and by increasing the O2 flow to 10 sccm, transparent films are deposited. Absorption data from a second set of samples at 5 kW and 300 sccm Ar show a similar dependence in Figure 4: At 8 sccm, some absorption is seen in the visible, whereas at 10 and 12 sccm, the films are highly transparent at a thickness of about 130 nm.

Figure 2. Dynamic deposition rate as function of the target power for pure Ar and mixtures with O2.

The refractive index is 2.2 at 500 nm. As shown in Figure 1, these samples were deposited with the target surface in its sub-oxide state at low impedance. The target material composition is a key to reliable operation. Figure 5 compares an early target after testing (top) with a target with optimized stoichiometry (bottom). Whereas before optimization, arcing led to severe damage of the target surface by local melting, the optimized target shows only few and very small marks from operation with less-than-optimum arc management.

Figure 3. Light transmission of deposited films at 4 kW in 200 sccm Ar with and without oxygen addition, examples of thin (5 nm) and thicker (approx. 120 nm) films.

Figure 4. Optical absorption of ZrOx samples sputtered at 5 kW with 300 sccm Ar and different O2 flows.

Figure 5. Top: early ZrOx sample targets after tests. Bottom: target with optimized stoichiometry after testing.

The amount of energy transferred into the cooling water is compared in Table 1 to other well-known ceramic target materials. The remaining energy is expected to be radiated from the hot target surface. These data allow a rough estimate of the maximum power density, the expected temperature rise in the chamber and the possible thickness to which targets may be manufactured. As the data show, ZrOx compares well with other ceramic target materials and no significant restrictions to the power density or thickness are to be expected.

Table 1. Cooling capability of ZrOx compared with other ceramic materials for rotatable targets.

Jumbo-size Dual Magnetron Tests.

Operating the targets in 400 sccm Ar and a varying percentage of O2, reveals the transition between fully oxidized and suboxide state as shown in Figure 6. This is the same behavior as shown in Figure 1 for the lab coater. In the full-size tests, the transition takes place between 5 and 20 kW when adding 20 sccm O2 (5%). It shows no hysteresis but the impedance changes with a time constant of about 10 seconds. The fully oxidized state also leads to arcing. In this graph, the arc count rate of the power supply is plotted; therefore, the absolute number depends on the detection thresholds. In the oxidized mode, there are numerous micro-arcs which are not considered here.

Figure 6. Target voltage and arc rate in AC dual magnetron sputtering of ZrOx as function of the MF power for different oxygen flows. Ar flow = 400 sccm.

Figure 7. Oxygen consumption in AC dual magnetron sputtering of ZrOx as function of the MF power. Conditions as in Figure 6.

The transition is shifted to higher power with increasing O2, as it is also seen in Figure 1. At 100 sccm O2, the sub-oxide state is not reached even at 80 kW. It is important to note that the suboxide state at very low powers (< 2 kW) can only be maintained after extensive sputter-cleaning of the target and chamber out-gassing.

A detailed understanding of the transition between the two states is obtained by analyzing the oxygen consumed by the process as function of oxygen supply and MF power. The oxygen consumption may be determined from the pressure change caused by the O2 addition with the plasma on and the pressure change caused by the oxygen with the plasma off:

Consumption = 1-[∆p(O2)on/∆p(O2)off].

Figure 7 shows that in the fully oxidized state at low power, oxygen consumption is very low. Only when the impedance or MF voltage drops to its low state, almost all oxygen is consumed by the sputtering. Consequently, we can expect high sputter rates in the sub-oxide state as long as the impedance is low, and low sputter rates, when the target surface is fully oxidized and the impedance is high.

The influence of the power supply output frequency on the sputtering behavior was investigated both in the sub-oxide state in pure argon and in the fully oxidized state. As Figure 8 shows, the higher output frequency leads to an increased cathode voltage or impedance. This is a common observation in AC sputtering. In fully oxidized mode at 20 – 22 kHz, sputtering at power levels below 20 kW was not possible, because extremely heavy arcing prevented stable operation. The high arc rate in the fully oxidized mode deserves closer investigation, since this state will necessarily be encountered during target burn-in. When the arc detection current threshold is set to a relatively high level, the arc count decreases to almost zero, but oscilloscope traces covering a large number of MF periods show a very high number of peaks on the current envelope which are caused by micro-arcs.

Figure 8. MF voltage in AC dual magnetron sputtering of ZrOx as function of the MF power for different frequencies.

Figure 9. Histograms of the peak current of the MF half waves in fully oxidized operation at different power settings. The maxima represent the MF peak current with no arcing.

These micro-arcs extinguish during polarity change-over of the power supply output without arc treatment. A histogram of the peak current distribution (per half wave) is shown in Figure 9 for output power levels between 20 and 100 kW. The histogram data are normalized to a total probability of 1. They show a maximum at the MF peak current and an exponentially decreasing occurrence of higher micro arc peak currents. A remarkable property of the micro-arcs is the independence of the peak current from the actual MF current or power. As a result, only few events with the peak current exceeding the MF current are observed at the highest power levels investigated. The effect of moisture on the sputtering behavior was investigated by bleeding a small stream of water vapor through a needle valve from water flask. The water flow was roughly determined from the pressure rise. As Figure 10 shows, the introduced amount of water (approx. 35 sccm) has a very similar effect on the plasma impedance or voltage as 20 sccm of oxygen. In this graph, the rate of micro-arcs is plotted, determined by counting the current peaks in oscilloscope traces. Unlike in reactive sputtering from Si (Al) targets [6], no adverse effect of moisture on the arc rate was observed. After the tests under a wide range of process conditions and arcing occurrence, the targets were inspected for damage. Like for the optimized target in Figure 5 (Bottom), only few very small spots of about 1mm diameter were seen.

Figure 10. MF voltage and rate of micro-arcing in AC dual magnetron sputtering of ZrOx in the presence of water vapor as function of the MF power. For comparison, data for 20 sccm O2 are included.

Target Burn-in

When a new target is burnt-in, the surface is initially in the fully oxidized state due to the environmental exposure. In our tests, the power could only be raised in smalls steps during burn-in and the arc management had to be set in such a way that (self-extinguishing) micro-arcs were not treated, but on the other hand, arc bursts or persistent arcs, which re-ignite directly after arc treatment, were extinguished. When the target state shifted to sub-oxide, this was noticed by the drop of impedance and arc rate. However, when the target was brought up to full power (> 10kW/ m and tube), a pronounced pressure rise was seen, caused by target surface or chamber outgassing. This also led to a temporary return of the surface to the fully oxidized state with high arcing rate. As the outgassing decreased, the target returned to its sub-oxide (clean) state.

Discussion.

ZrOx sub-oxide ceramic targets are well suited for the sputtering of transparent zirconia thin films. The results from both test chambers show a transition from a fully oxidized surface to a sub-oxide state without hysteresis, determined by the amount of oxygen added and the sputtering power density. The sub-oxide state is, for pulsed DC and for MF, characterized by a lower target impedance and low arc rate.

At power densities of 10 kW/m and tube, about 5% oxygen are sufficient to achieve transparent films, while the target surface is still in the same sub-oxide state as in pure argon. Both, the deposition rate and the oxygen consumption by sputtering were shown to be high under these conditions. The fully oxidized state is characterized by a higher target impedance, a high arcing rate and a low sputter rate.

The peak current of micro-arcs was found to be independent of MF current; this behavior shows that arc ignition is determined by the physical state of the target surface rather than power density [7]. The high arcing rate when fully oxidized does present a challenge during target burn-in, since this state can be sustained by outgassing of the target surface and the chamber during burn-in in Ar alone. One option to mitigate arcing is to operate at high frequencies above 40kHz. Occasional bursts of arcing are observed, when the target is poorly conditioned or even when it is operated after longer process breaks. Arc bursts or persistent arcs most likely lead to spots of local melting on the target surface as shown in Figure 5. They are a well-known phenomenon with Zn containing targets and require some care in arc management settings [8].

If after arc treatment, the plasma immediately reignites into the arc state, this must be reliably detected and acted upon. Only then, damage to the target can be avoided. For Zn containing alloys, longer arc treatment times of 500µs have proved helpful to prevent persistent arcs. The similarity between Zn and Zr regarding arc bursts is quite surprising, since the underlying mechanism for Zn is most likely connected to its high vapor pressure. The plume of vaporizing target material provides a medium favoring the local arc reignition. A possible explanation for the similar behavior of Zr could be its high oxygen affinity. This could favor arc reignition by two different mechanisms: (1) Rapid local oxidation provides a place for charge build-up, which leads to arc re-ignition; (2) local surface oxidation leads to extreme heating of arc spots in the presence of oxygen due to the heat of reaction. Such hot-spots also can act as preferred locations for arc re-ignition.

Conclusions

ZrOx ceramic targets show two distinct states: fully oxidized and sub-oxide, in a similar way like it is known from metallic targets. A high oxygen consumption by sputtering is only seen in the suboxide state. The process is affected by heavy micro-arcing in the oxidized state only and the arc rate may be reduced by choice of a high MF frequency. Highly transparent layers can be deposited in the transition regime between fully oxidized mode and sub-oxide mode with relatively small oxygen flows at high deposition rate and low arc rate. The oxidized state of the surface is sustained by outgassing at low power. Consequently, target burn-in will be a challenging procedure. In addition, there are occasional arc bursts (persistent or “hard” arcs), which require correct arc management settings to avoid target damage. The likelihood of target damage by arc burst occurrence was minimized by optimizing the stoichiometry of the target material.

References

[1] D. Severin, K. Sarakinos, O. Kappertz, A. Pflug, M. Wuttig “Tailoring of structure formation and phase composition in reactively sputtered zirconium oxide films using nitrogen as an additional reactive gas," J. Appl. Phys. 103 (2008) 083306 http://dx.doi.org/10.1063/1.2903492

[2] K. Juskevicius M. Audronis A. Subacius, R. Drazdys, R. Juskenas, A. Matthews, A. Leyland, "Highrate reactive magnetron sputtering of zirconia films for laser optics applications" Appl. Phys. A 116 (2014) 1229–1240 http://dx.doi.org/10.1007/s00339-013-8214-1

[3] B. Disteldorf, P.J. Lingle, J. Lao, B. Boyce, U.S. Patent # 8389121B2 “Coated article with low-E coating including zirconium oxide and/or zirconium silicon oxynitride and methods of making same” Priority date June 25, 2008, filing date Dec. 13, 2012, issue date March 5, 2013.

[4] J. Oberste-Berghaus, R. Van Nuffel, K. De Jaeger, A. Das, W. De Bosscher, “Properties of Zirconium Oxide Top Layers from Rotatable Targets”, 58th Annual Technical Conference Proceedings of the Society of Vacuum Coaters, pp. 213-217, 2015. https://doi.org/10.14332/svc15.proc.1944.

[5] J. Rezek, J. Vlček, J. Houška, J. Čapek, P. Baroch "Enhancement of the deposition rate in reactive midfrequency ac magnetron sputtering of hard and optically transparent ZrO2 films" Surface & Coatings Technology 336 (2018) 54–60. http://dx.doi.org/10.1016/j.surfcoat.2017.09.015

[6] M.Heintze, I. Luciu, "Nodule formation on sputtering targets: Causes and their control by MF power supplies" Surface & Coatings Technology 336 (2018) 80–83. http://dx.doi.org/10.1016/j.surfcoat.2017.09.009

[7] A. Anders, “Physics of arcing, and implications to sputter deposition” Thin Solid Films 502 (2006) https://doi.org/10.1016/j.tsf.2005.07.228D.M.

[8] M. Heintze, M. Bannwarth, “Process Benefit of Advanced Arc Management in DMS Power Supplies”, 59th Annual Technical Conference Proceedings of the Society of Vacuum Coaters, pp. 190-195, 2016. https://doi.org/10.14332/svc16.proc.0019