Magnetron Sputtering is a technology of high importance due to a wide variety of applications in industries such as thin film photovoltaics, semiconductor, optics, decorative coating and wear and corrosion protection. It has been observed in the past that in certain cases, when sputtering metal or ceramic materials, nodules are formed on the surface of sputter targets near the racetrack region (redeposit area) and sometimes in the racetrack region (figure 1). Usually they posses a shape of a hillock, cone or pyramid. Nodules tend to grow as the deposition run proceeds. Eventually they can cover target surface areas in excess of 30%. Formation of nodules incurs various effects such as changes in sputtering rate, angular distribution of sputtered atoms, enhanced arcing, process drift and destabilisation, which in turn result in defects and lead to poor quality sputtered films. A coating system has to be shut down frequently for cleaning of target surfaces of nodules and debris. This causes an undesired down time and reduces production rate. Formation of nodules on the target surface is therefore highly undesirable. Despite this being a serious industrial problem, in general there is a lack of understanding as to what results in the growth of nodules, which process parameters are important, and how to tackle this problem. The purpose of this article is therefore to shed some light on the mechanisms of nodule formation on sputtering targets as well as on the critical process parameters and available solutions.
Figure 1. a) SEM image showing morphology of nodules. From Lippens et al.  b) Black nodules on ITO sputtering target (image – courtesy of Gencoa Ltd); c) Nodules on Si sputtering target (image – courtesy of Faradox Energy Storage, Inc).
Nodule growth mechanisms
Observation of cones on ion bombarded cathodes dates back to as early as 1942 . Since then this phenomenon has been a subject of interest to scientists and engineers in both academia and industry. Wehner has conducted extensive research work on cone formation . Based on the experimental evidence Wehner concluded that very small amounts of certain impurity atoms or atoms supplied during sputtering from another source can give rise to seed cone formation on ion-bombarded surfaces.The ratio of seed to main atoms required to induce cone growth can be as low as 1 to 500, respectively, as was demonstrated for Mo – Cu case. Interestingly, it was also shown that the seed atoms material need not exhibit lower sputter yield, but must exhibit a higher melting point. Deposit cones can also appear when a much larger flux of a lower melting point metal is deposited on hot higher melting point metal, which is under ion bombardment. At low ion bombardment energies (< 1keV, i.e. typical for magnetron sputtering applications) an elevated temperature (~ ⅓ of melting point) is essential for seed cone phenomenon to occur. When sputtering metal targets, traces of oxygen or nitrogen atoms impede surface movement of atoms and therefore seed cone formation. Ion bombardment near the sputtering threshold results in growth of straight single crystal whiskers, which when subjected to higher energy bombardment are often converted into cones. Wehner stressed that seed cones are the result of an interplay among whisker growth, adatom surface movement and the effects of sputtering.
The sputter target materials that have been most frequently reported to have the nodule growth problem are indium-tin alloy and indium tin oxide (ITO) [3-9]. This is mostly due to the dominance of ITO in the transparent conducting oxide (TCO) coated glass market. The highest priorities currently in the industrial production of ITO are achievement of high sputtering rates and process stability, both of which are compromised by the nodule growth problem.
Other target materials, such s Si and Ti-W , were also reported to suffer from nodule growth.
ITO sputtering from ceramic targets
Nodules formed on ceramic ITO or Indium-Tin alloy targets are typically black in appearance and exhibit high resistivity [3, 4]. Target density was associated with the nodule growth phenomenon quite early, which was followed by significant efforts (figure 2) to increase the density of ceramic ITO targets [5, 11, 12]. Dense targets were found to improve deposition rates, but the nodule growth problem still persisted. The results published by Omata et al.  imply that targets that posses relative density higher than 99.5% but have inhomogeneous conduction electron density due to local deviation from uniform Sn ion and oxygen vacancy distribution suffer from increased rate of nodule formation.
Figure 2. Hystorycal ITO target density improvement. From K. Utsumi et al. 
Papers by Ishibashi et al.  and Lippens et al.  state that in the ITO sputtering case nodule growth is largely related to the fact that on the target surface In2O3 is dissociated into high resistivity In2O sub-oxide and O2 with Sn acting as a promoter of nodule formation. Nodules are constantly being coated with a redeposited material. The overcoat thickness is of the order of several tens of microns and it can be more oxygen deficient than the target surface , possibly due to the reduction of the deposit by Sn . Unevenness of electrical properties and chemical composition lead to continuous growth (enlargement) of nodules (i.e. sputter target erosion rate nearby is faster) during the course of the target life. Figure 3 shows a schematic cross-section of a nodule on an ITO sputter target. Inhomogeneous chemical composition and electrical properties also lead to charge build up in certain target surface areas and arcing. Nodules can crack under thermal stress or be destroyed by an arc. Nodule breakups are followed by a shower of particles, which nucleate a new generation of nodules, thus explaining the observation that once started nodule growth accelerates rapidly .
Figure 3. Schematic cross-section of a nodule on an ITO sputter target. From Schlott et al. .
Reger et al.  have used 3D SIMS to examine about 100 nodules formed during sputtering of ITO target. They were looking for impurities such as C, Mg, Al, Si, Cu, Ti and Zr. Impurity distribution was found to be inhomogeneous with the exception of Cu. Most of the impurities consisted of C or Al. The 3D depth profiling of nodules revealed that on top of most of them C concentration was much higher as compared to the unaffected target surface areas. Authors of  have therefore concluded that C impurities played the most important role in nodule growth on the target they have studied.
Schlott et al.  have used in-situ video monitoring to study nodule formation during ITO sputtering processes. They have tested two 75mm diameter planar ITO targets one after another without braking the vacuum. Surprisingly, the target which they have sputtered first was covered with nodules after 1 hour of sputtering, while the identical second target remained essentially nodule free for ~ 4 hours. Figure 4 compares the two targets after 10 hour long sputtering run. The difference in nodule numbers of almost two orders of magnitude was observed. This example clearly showed that one major source for nodule nucleation is dust and flaking originating from the sputtering system. Level of chamber cleanliness was improved by the operation of the first target, which allowed much longer nodule-free operation of the second target.
Figure 4. A comparison of two ITO targets sputtered for 10 hours one after another without breaking vacuum. From Schlott et al. .
Redeposited material was shown to be a source of nodule formation in two ways : a) nodule nucleation by a redeposition in the race track area and/or vicinity of the racetrack and subsequent sputtering of the redeposits; b) flaking of the redeposited material. Further sources of nodule formation identified in the same and other studies [6, 10] were: the joints between the (ITO) tiles in segmented targets were the dust/particles are easily collected, voids, crevices, microcracks and contamination with particles (SiO2, Al2O3 TiO2 and C) during the target manufacturing process.
A paper by Nakashima and Kumahara  reports the use of very fine raw powder and modified mixing process that provides significantly more uniform SnO2 distribution. The target manufactured using this technology was shown to be much less prone to nodule formation and arcing (figure 5). A further reduction in nodule formation was achieved by reducing the amount of SnO2 in the target from industry standard 10 wt.% to 9 wt.% .
Figure 5. Appearance of improved target (a) and conventional one (b) after 160 Wh/cm2 sputtering. From Nakashima and Kumahara .
Cho et al.  investigated the effect of ITO target doping with Ca via addition of 0.025 and 0.05 wt.% of CaCO3. A substantial decrease in the density of nodules formed and arcs counted, improved target utilisation, process stability and repeatability were demonstrated (figure 6). Notably, no significant property changes in the deposited ITO films were found to occur as a result of Ca doping.
Figure 6. Ca doped ITO targets: images of nodules formed on the target surfaces for different doping and and target erosion levels. From Cho et al. .
Hot ITO Sputtering process is worth mentioning here, since it is believed that by operating an ITO target at approximately (or above) 500 deg. C formation of nodules is reduced. This process however does not seem to have been accepted widely; its use appears to be rather patchy and any technical information on it is scarce.
Reactive ITO sputtering
Nodule formation during reactive ITO deposition from In-Sn (80-20) target has been investigated by Lippens et al. . Results indicated that the mechanisms of nodule formation and growth are similar to the ceramic ITO sputtering case, but, apparently, there are certain differences too. For example, nodules formed on In-Sn target were often found to contain much more oxygen (In/O ratio corresponds to In2O3) as compared to regions next to them indicating that during reactive sputtering process the nodules nucleate at the locations where the the target is strongly oxidised. In other cases In/O ratio close to 2/1 was measured also associated with a corresponding higher concentration of Sn and/or SnO2  (note: electrical resistivity of SnO2 is at least two orders of magnitude higher than that of ITO target).
Additional experiments with pure In (melting point – 157 deg.C) and Sn (melting point – 232 deg.C) targets were performed. The results showed that nodules grow on both pure In and In-Sn targets. The nodule formation was more pronounced in In-Sn case, which confirms the role of Sn in enhancing the nodule nucleation. These results are in good agreement with Wehner’s work  discussed in earlier sections. Interestingly no nodules were found to grow when sputtering pure Sn target.
Other important observation by Lippens et al.  during reactive ITO deposition is that as the nodule density increases, there is an increasing supply of oxygen from the target due to decomposition of In2O3, which induces process drifts also complicating reactive sputtering process control.
Lo and Draper have studied nodule formation in W-Ti sputtering . They have drawn some useful relationships which have wider applicability and are therefore given in this section. Figures 7 a to c show nodule and density as functions of W-Ti target life and nodule distribution in the radial section a W-Ti target, respectively.
Figure 7. Nodule (a) and density (b) as functions of W-Ti target life and (c) nodule distribution in the radial section a W-Ti target. From Lo and Draper .
Important process parameters
It has been shown in the previous sections that in terms of nodule formation a sputtering target used is one of the major elements in both non-reactive and reactive sputtering processes. Significant materials engineering / development efforts have been spent by sputtering target suppliers in order to eliminate nodule growth. It is, however, also obvious from the above that in addition to the target aspect, processing parameters, as well as the use of appropriate associated equipment, may play an important role in nucleation and growth (rate, density, etc.) of nodules. Below are the important factors:
- Cleanliness of the process chamber – i) presence of foreign particles (e.g. dust, flakes), ii) presence of contaminant reactive gaseous species (e.g. diffusion pump oil vapour, poor base vacuum, leaks, contamination of main sputtering gas or gas supply lines). Clean process environment is required to reduce nodule growth problem. Sputter cathode orientation (i.e. vertical, horizontal) is important with vertical cathode orientation probably being optimal in minimising to some degree contamination of target surfaces (e.g. with particles, flakes) as well as that of the substrates.
- Arc handling – a magnetron sputtering power supply feature – is extremely important for extending the nodule free operation. Arcs adversely affect processes as well as cathode surfaces. Use of power supplies equipped with a fast and reliable arc detection and quenching circuitry (e.g. ) is essential in reducing/eliminating nodule growth. Dual magnetron sputtering combined with medium frequency (MF) AC power has proven to be effective means for reducing number of arc events in reactive sputtering processes.
- Target surface temperature – in general nucleation of nodules is enhanced at elevated target surface temperatures. Hence, the sputtering power density should be selected accordingly and an efficient target cooling is required to minimise nodule growth. An exception to this trend may be the Hot ITO Sputtering process as mentioned earlier.
- Stability of the process (especially in case of reactive sputtering) is beneficial in reducing the growth of nodules. For reactive sputtering processes, active feedback process controllers (e.g. ) are used to keep the target in the ‘transition’ region during the operation, which, in addition to increased deposition rates, reduces the susceptibility of the process to arcing and, therefore, nodule nucleation and growth rate.
Sputtering source design aspects
Magnetron sputtering source design can affect significantly the nodule growth phenomenon. This is mostly related to the relative surface area of redeposition and “slow” sputtering zones where nodule nucleation/growth rates tend to be the highest. Increase in sputtered surface area leads to reduced redeposition zones and, consequently, less nodule growth. The following three magnetron designs (figure 8a-c) can reduce the nodule growth rates by increasing the total target surface sputtering area: i) High Yield, ii) Full Face Erosion and iii) Rotatable Magnetron Sputtering. In terms of reduction of nodule growth rotatable magnetrons offer the best performance. A brief introduction to each of the magnetron types is given below.
Figure 8. a) ‘High Yield’ sputter target; b) ‘Full Face Erosion’ sputter target; c) Rotatable sputter target. Images courtesy of Gencoa Ltd.
- High Yield – the High Yield magnetic arrays (e.g. ) are used in planar cathodes to improve utilization from typically 25% – 35% for standard two pole magnetics to 45-50%. The enhanced target use is created by additional magnetic poles within the magnetic system to deform and flatten the magnetic field structure over the target surface.
- The Full Face Erosion sources in circular and rectangular forms (e.g. ) offer high target use, 45% – 60% due to the scanning nature of the magnetic field. Consequently the redeposition areas are reduced even further.
- Rotatable Magentrons (e.g. ) employ cylindrical sputter targets and increase target utilisation to approximately 80 %. The lifetime is significantly increased too. In contrast to planar cathodes redeposition zones are nearly completely eliminated, which further increases stability of the processes, reduces arcing and nodule formation. A recent paper by Medvedovski et al. reports deposition of good quality ITO from ceramic rotary sputtering targets .
Nodule formation/growth on sputtering targets is a complex phenomenon that occurs on the target surfaces during certain sputter deposition processes. This phenomenon is influenced by a multitude of factors such as sputtering target properties and deposition process parameters, conditions and quality in general. Sensitive processes, such as ITO thin film deposition, require careful consideration of every aspect of materials, process conditions and deposition equipment used.
 A. Guenterschulze and W.V. Tollmien, Z. Phys 119 (1942) 685
 G.K. Wehner J. Vac. Sci. Technol. A 3 (1985) 1821
 S. Ishibashi, Y. Higuchi, Y. Ota, K. Nakamura, J. Vac. Sci. Technol. A 8 (1990) 1403
 P. Lippens, A. Segers, J. Haemers, R. De Gryse, Thin Solid Films 317 (1998) 405
 N. Nadaud, M. Nanot, P. Boch, J. Am. Ceram. Soc. 77 (1994) 843
 M. Schlott, M. Kutzner, B.L. Gehman, N. Reger, F.J. Stadermann, SID 96 DIGEST (1997) 1
 K. Nakashima, Y. Kumahara, Vacuum 66 (2002) 221
 T. Omata, M. Kita, H. Okada, S. Otsuka-Yao-Matsuo, N. Ono, H. Ikawa, Thin Solid Films 503 (2006) 22
 S.H. Cho, Y.M. Kang, J.R. Lee, B.K. Ryu, P.K. Song, J. Kor. Phys. Soc. 54 (2009) 1315
 C.F. Lo, D. Draper, J. Vac. Sci. Technol. A 16 (1998) 2418
 B. L. Gehman,S. Jonsson, T. Rudolph, M. Scherer, M. Weigert, R. Werner, Thin Solid Films 220 (1992) 333
 K. Utsumi, H. Iigusa, R. Tokumaru, P.K. Song, Y. Shigesato, Thin Solid Films, 445 (2003) 229
 V. Bellido-Gonzalez, B.Daniel, J. Counsell, D.Monaghan, Thin Solid Films, 502 (2006) 34
 Rotatable Magnetrons
 E. Medvedovski, C.J. Szepesi, P. Lippens, K. Leitner, R. Linsbod, A. Hellmich, W. Krock, Ceramic Engineering abd Science Proceedings 30 (2010) 93