sputtering materials for vlsi and thin film devices || troubleshooting in sputter deposition
TRANSCRIPT
CHAPTER
8Troubleshooting in SputterDeposition
8.1 IntroductionIn Chapter 3 we discussed those aspects of magnetron sputtering that would influence the perfor-
mance of sputtering targets and as a result the process yield. These can primarily be divided into
two categories i.e., the characteristics of sputtering targets and the process variables specific to
sputtering tools and applications. In the sputtering target industry, the term control parameter is
used for identifying the characteristics of a sputtering target, while in the semiconductor industry
the term recipe is used for defining process variables for sputtering. The major control parameters
for sputtering targets and process variables are listed in Table 8.1. The deposition tool hardware is
also known to have considerable influence on the occurrence of troubleshooting in sputter deposi-
tion. Because of the interdisciplinary nature of the subject, any examination of the attributes of
deposition tool hardware is out of the scope of this discussion. Well-developed quality control sys-
tems in the sputtering target industry and specifications developed in the semiconductor industry
are intended to meet the requirements for trouble-free sputtering of films. However, it is not
uncommon to see occasional failure in meeting requirements and encountering troubleshooting in
semiconductor manufacturing facilities.
Here we briefly review most commonly seen problems as applied to DC magnetron sputtering
and then we expand them in the following sections. One of the major examples of troubleshooting
in the early life of a target is the long burn-in requirement before attaining desired film properties
such as sheet resistance (Rs), thickness uniformity and resistivity. Out of control inhomogeneity of
the sputtering material, inadequate surface finish and contaminations at the sputter surface can lead
to longer target burn-in. As sputtering progresses, sputtering targets may also start to produce
nodules in certain areas of the sputter face because of the re-deposition of the sputtered material.
These nodules can flake under the influence of thermal stresses and can produce particles on the
deposited films. Similarly, re-deposited material on the process kit can produce particles on depos-
ited films. This is a very common phenomenon seen especially in the reactive sputtering of
materials.
Pasting is a high power sputtering step in metal modes that cleans up the sputter surface and
also seals the re-deposited nodules with thin metal layers. If pasting is not done at recommended
intervals of time, flaking of re-deposited materials from the sputtering target and from the process
kit is likely to cause particle formation in the deposited films. Another significant source of in-film
particles is the occurrence of arcing inside the deposition tool. Arc generated at the embedded
dielectric inclusions in the sputtering target can lead to the formation of metal droplets (splats) and
particle formation in the deposited films.
567J. Sarkar: Sputtering Materials for VLSI and Thin Film Devices. DOI: http://dx.doi.org/10.1016/B978-0-8155-1593-7.00008-4
© 2014 Elsevier Inc. All rights reserved.
These examples show that a wide variety of troubleshooting can take place from early-life to
the end-of-life of a target. The consequence of troubleshooting in sputtering is a possible reduction
in the process yields of users’ targets, while failure of delivering satisfactory targets to users is a
quality issue for target suppliers. In the case of troubleshooting that has originated from a defective
target, both the user and the target supplier work closely to resolve the issue with an emphasis on
the root-cause-analysis. Once the reason for failure is known, the necessary changes are brought
into the target manufacturing system to prevent it. Because of the proprietary nature of such trou-
bleshooting information, specific to the deposition tool, process recipe and the type of sputtering
target, we only discuss selective examples here that have already been discussed in the literature.
8.2 Long burn-in of sputtering targetBurn-in of a sputtering target is an essential but unproductive step in the sputter deposition of thin
films. Hence the semiconductor industry places a great deal of emphasis on the consistency of
burn-in performances of targets to keep the down-time to a minimum. There are examples in which
extension of burn-in time has been noticed due to the changes made by the target supplier of the
sputtering materials. Reduced burn-in sputter targets are also in demand for expensive sputtering
materials and advanced applications [1�5]. In the following sections, we discuss the role of indi-
vidual factors in the burn-in performance of a target in greater detail. It is important to recall that
the burn-in performance of a sputtering target is usually evaluated on the basis of time required to
achieve the controlled (within specification) deposition rate, film sheet resistance, film thickness
uniformity, film resistivity and in-film particle count. It should be noted that because of the proprie-
tary nature of these results, normally owned by the original equipment manufacturer (OEM), we
are not in a position to present enough results as a function of target life (in kWh) that would other-
wise illustrate the nature of problems seen at troubleshooting. As a result, primarily we have
addressed those issues that cause troubleshooting and reduce the process yield.
Table 8.1 A Summary of Components that are Known to Influence Sputtering Performance and
Process Yield
Sputteringtarget
Target metallurgy Material purity, microstructure and its homogeneity, crystallographictexture and the distribution of texture components, sputter facecondition (uniformity and depth of deformed layer)
Target design Thickness of target and backing plate materials, roughness of particle-traps (e.g., arc-sprayed region) around sputter face, proper bevel atthe sputter face edge
Target ferromagneticproperties
Pass-through flux and permeability
Processvariables
Recipe Programs for bake-out, burn-in, deposition cycles and pasting
Handling andmaintenance
Sputtering target, process kit
568 CHAPTER 8 Troubleshooting in Sputter Deposition
8.2.1 Inhomogeneous materialIn a strict sense, most sputtering materials do show some degree of microstructural and textural
inhomogeneity with respect to the shape and dimensions of the sputtering targets. These apply to
both high purity metals and compound sputtering materials. Inhomogeneity of the microstructure
and texture seen in sputtering materials primarily depend on the processing history, i.e., material
fabrication methods used and forming methods involved in the manufacturing process.
Inhomogeneities in the three dimensions can be manifested in the plane of the sputter face as well
as in the through-thickness direction.
It is important to remember that while a certain degree of material inhomogeneity can be toler-
ated for less advanced processes, the same degree of material inhomogeneity can be a risk factor
for advanced processes. Hence, based on the process requirements, acceptable control limits for
material inhomogeneity are always established by the target suppliers in conjunction with the users.
Here a specific example is presented in which we distinguish within specification and out-of speci-
fication Rs-uniformity values that resulted from magnetron sputtering of a relatively homogeneous
target and a rather inhomogeneous target.
Rs-uniformity values for deposited films are determined from sheet resistance maps developed
using commercial four-point-probe equipment (see the book’s companion website for details).
Figure 8.1 shows sheet resistance maps of tantalum bi-layers resulting from magnetron sputtering
of these two targets [6]. Identical recipes were used to run these tests up to the same sputtering life.
Note that the contours in Figure 8.1(a) show relatively small variation in sheet resistance from the
center to the edge, while the contours in Figure 8.1(b) show relatively large variation in sheet resis-
tance data. As a result, for more homogeneous sputtering target, the Rs-uniformity value is low
enough to meet the specification. In contrast, for a relatively inhomogeneous sputtering target, the
Rs-uniformity value exceeds the specification limit. The following types of inhomogeneities have
been identified for sputtering materials.
FIGURE 8.1
Sheet resistance maps of tantalum by-layers that show (a) variation in values from the center to the edge of
the wafer leading to “within specification” Rs-uniformity resulted from a relatively homogeneous sputtering
target and (b) large variation in values, reflected in more number of contours, from the center to the edge of
the wafer leading to “out of specification” Rs-uniformity resulted from inhomogeneous sputtering target [6].
5698.2 Long burn-in of sputtering target
8.2.1.1 Inhomogeneous grain structuresInhomogeneity in the grain structure can result from inherent properties of the material and also
because of the deficiencies in the fabrication and processing methods. In general, dilute ductile alloys
in the form of flat products show better homogeneity in grain structure than curved or hollow products.
For example, dilute aluminum alloys (e.g., Al-0.5 wt% Cu, Al-0.2 wt% Si-0.5 wt% Cu) show nearly
homogeneous grain structure after conventional processing using casting and thermo-mechanical pro-
cessing steps, while high purity titanium and tantalum may show inhomogeneous grain structures.
Electron-beam melting of titanium and tantalum are quite common and subsequently these are
thermomechanically processed to achieve the desired microstructure and texture. Figure 8.2 shows
an example of banding of grains in cold-rolled and recrystallized titanium [7]. With careful selec-
tion of thermo-mechanical processing steps and heat treatment schedules, it is possible to produce
titanium without significant inhomogeneity in grain structure.
Apart from material, complex-shaped products may pose challenges in attaining homogeneous
grain structure. Curved and hollow sputtering targets are good examples of this problem. For exam-
ple, hollow cathode magnetron targets are known to have significantly different grain structure in
the round corners of the dome-shaped configuration than the side walls. This is primarily because
the material in the vicinity of the corners is stretched more than the walls during metal-forming
operations and the subsequent recrystallization step often fails to produce the equiaxed microstruc-
ture at the corners. Though inhomogeneity in grain structure is seen in various sputtering materials,
acceptable control limits have been established by sputtering target suppliers in conjunction with
users. Unfortunately, no quantitative relationship between grain structure inhomogeneity and longer
burn-in is available at this point of time.
8.2.1.2 Inhomogeneous texture in through-thickness directionIn addition to the inhomogeneous grain structure, often through-thickness texture variation is also
seen in sputtering materials [8]. In fact, inhomogeneity in grain structure and through-thickness
FIGURE 8.2
Microstructure of titanium deformed to 70% and annealed at 700�C for 15 min showing grain size
inhomogeneity [7].
570 CHAPTER 8 Troubleshooting in Sputter Deposition
texture gradient can co-exist and the origin of the through-thickness texture variation lies in mate-
rial properties such as slip systems, stacking fault energy, deformation history and recrystallization
heat treatments. In a rolled and recrystallized product, it is common to see variation of volume frac-
tions of different crystallographic texture components as a function of depth from the surface to the
plane at half-thickness. A large volume of literature is available on the through-thickness texture
variation in aluminum alloys, tantalum and many other materials [8,9]. For example, Figure 8.3
shows the texture gradient from surface to the mid-plane in a high purity tantalum plate [8]. This
subject has been discussed in greater detail in section 5.6.1 of Chapter 5. Readers are suggested to
refer to crystallographic texture basics given on the book’s companion website for the interpretation
of orientation distribution functions (ODFs).
8.2.1.3 Inhomogeneous texture in the form of bandsTexture inhomogeneity can also be manifested in the form of banding in which similarly oriented
grains align in a particular direction, and these bands are usually separated by grains that have dif-
ferent crystallographic orientations. The alignment of grains originates from differential stretching
of grains and also selective growth of grains in specific orientations during recrystallization heat
treatment. Inherently, grains with certain crystallographic textures have greater resistance to
FIGURE 8.3
Two-dimensional orientation distribution function sections measured at different depths in a 4 mm thick
annealed tantalum plate. Texture gradient in through-thickness direction ranges from (001)[110]. texture at
the surface to (111), uvw. in the mid-plane of the tantalum plate [8].
5718.2 Long burn-in of sputtering target
deformation than neighboring grains. On subsequent recrystallization heat treatment, grains with
specific orientations may align in particular directions as seen in metallographic samples. This leads
to banding of grains and, in recent years, the electron backscattered method of crystallographic tex-
ture measurements has revealed these bands very clearly [8].
Electron-beam melted and thermomechanically processed high purity tantalum is known to have
significant texture banding, which is shown in Figure 8.4 [10]. The effects of microstructure and
texture inhomogeneity can be large enough to cause delay or complete failure in attaining desirable
film properties. This is the case for texturally inhomogeneous tantalum sputtering targets for
advanced chip making (below 65 nm node) and in particular for 300 mm wafer applications in
which tight control on film thickness uniformity is essential.
8.2.2 Surface characteristics8.2.2.1 Surface roughness and sub-surface deformed layerStandard sputtering targets are usually finished with precision machining or grinding followed by light
polishing depending on the nature of the metal or alloy targets. An average roughness (Ra) in the range
of 4�40 µinch is desired for the majority of production sputtering targets, and measurements are usu-
ally made at specific locations at the finished sputtering surface. It should be noted that in various
materials, it is difficult to achieve identical roughness at all parts of the target surface. In such cases
greater roughness is witnessed, with a relatively large Ra number, in the vicinity of the center in a
round target. This is primarily because of the wearing of the cutting tool (flank wear) with elapsed
time required for machining. Typically 5�30 minutes time is required for the final machining step and
the cutting tool starts to wear as it moves from the target periphery towards the center.
In the process of machining and grinding, a relatively thick deformed layer (e.g., as large as
40 µm in tantalum) is generated underneath the machine-finished surface [11]. A direct
FIGURE 8.4
Scanning electron micrograph showing the occurrence of banding of grains in longitudinal section of a
tantalum sample swaged to a true strain of 1:28 and annealed at 900�C for 30 min [10].
572 CHAPTER 8 Troubleshooting in Sputter Deposition
proportionality between the average roughness Ra and the depth of the deformed layer has been
established in materials such as tantalum, titanium, copper and aluminum [2]. Thickness of the
deformed layer was measured from peak broadening using x-rays.
Figure 3.10, in Chapter 3, shows the variation of the depth of the deformed layer as a function
of average roughness in tantalum sputtering targets [2]. The deformed layer retains large cold work
because of the high dislocation density and also residual stresses because of the severe plastic
deformation caused by the machining operations. The crystallographic texture in the deformed layer
is often significantly different than the bulk sputtering material. Hence, the deformed layer pos-
sesses different metallurgical characteristics as compared to the bulk sputtering material. It is
believed that, for certain metals, the deformed layer found at the sputter surface has a negative
influence on the burn-in performance and is an important issue for expensive sputtering material
and processes for advanced chip making. It has been shown that until the entire deformed layer
gets removed in the burn-in process or artificially removed in a specially prepared target, the
desired film properties are not achieved [3].
It is worth noting that the early material removal from the sputtering face of a flat and round
target usually starts from narrow concentric rings, which gradually spread to much wider erosion
grooves (also known as racetracks). This indicates that, for certain processes, the severely
deformed and greater roughness concentric areas found in the vicinity of the center (with
thicker deformed layer) of the sputter surface is first exposed to plasma during burn-in operations,
and this in fact may lead to a situation where burn-in requirement would be even more severe.
This may not be the case for those targets that have deepest racetracks towards the periphery.
For certain types of sputtering targets, it has been shown that if the deformed layer is removed
completely and the roughness is maintained small, it would be possible to achieve reduced
burn-in [3].
A systematic study has been conducted taking into account the surface roughness, the thickness
of the deformed layer, contamination level and their relationships to the burn-in performance of
titanium sputtering targets [2]. Various types of surface finishes were achieved by using a combina-
tion of two different machining methods (ordinary and precision), wet polishing and chemical etch-
ing techniques [3]. Reduction in Rs uniformity and an early saturation of deposition rate were
recorded because of the removal of the deformed layer [3]. Figure 8.5 shows the variation of depo-
sition rate with sputtering life for sputtering targets with different finishes [3]. It is obvious that tar-
get 4, prepared using precision machining, wet polishing and a chemical etch, shows a large
reduction in burn-in, while others do not.
It should be noted that both target 4 and target 5 have no deformed layers and target 5 has high-
er roughness than target 4. Interestingly, target 5 does not show a large reduction in burn-in though
it is free from the deformed layer. This result implies that after removing the entire deformed layer
from the sputter surface if the roughness remains high, a significant reduction in burn-in may not
be achieved. It should also be noted that these types of surface preparations are relatively complex
and in each case the entire process is expected to take several hours. Disposal of acid waste is com-
plicated and expensive. Hence, there have been efforts to prepare reduced burn-in sputtering targets
using dry methods such as ion cleaning, and this led to the development of techniques that are inex-
pensive and environment friendly [5,12,13].
One of the studies on titanium shows that it is not necessary to remove the entire deformed
layer, but the removal of the contaminants from the sputter surface and transforming the deformed
5738.2 Long burn-in of sputtering target
layer into a soft layer can cause an equivalent effect, i.e., a significant reduction in the burn-in with
improved particle performance [12,13]. During preparation of this manuscript it was noted that
only few target manufacturers could supply reduced burn-in titanium and tantalum targets to chip
manufacturers.
8.2.2.2 ContaminationsIn addition to the surface roughness and deformed layer at the sub-surface, sputter surface contami-
nation may play a role in burn-in performance of a sputtering target. In various sputtering targets,
final shiny finish is achieved by applying polishing steps after completing final machining step.
The polishing media can be emery papers as well as abrasive material reinforced polymeric media.
In both cases, abrasive materials are fine ceramic particles either bonded to the paper or to the
polymeric media. During final polishing sputtering targets may retain these fine ceramic particles
and also organic compounds from the polishing media. These ceramic particles and organic com-
pounds are known to cause severe arcing and out-gassing, respectively.
The other source of organic contamination is the plastic bag used in the packaging of the sput-
tering target. The thickness of organic compound layer can be several mono-layers thick and may
lead to degassing problems if a suitable packaging material is not used. In some cases, moisture
pick up by the plastic bags can cause stain generation on the sputter surface as well as on the back-
ing plates. Though some of these issues may not be as critical as the particle contamination, sput-
tering targets with even minor cosmetic differences are rejected by the users. This helps users to
keep the production schedule and quality in control and meet the process yield.
FIGURE 8.5
Variation of deposition rate with sputtering life (power consumed) for sputtering targets with different surface
finishes [3].
574 CHAPTER 8 Troubleshooting in Sputter Deposition
8.3 In-film defect generationDefects such as particles and splats (submicron to tens of micron in size) are known to influence
the process yield in chip, flat panel display and other thin film device manufacturing. Defects are
generated at different steps of a manufacturing process, but we focus on the in-film defect genera-
tion during sputtering steps. Primarily defects are caused by the flaking of re-deposited nodules
from the target itself, flaking of deposited films from the process kit (chamber shield and clamping
ring) and also by arcing activities [14�30].
A variety of sputtering materials such as tungsten�titanium, titanium, indium�tin�oxide and
aluminum alloys show particle problems [14�30]. Figure 8.6 shows a W-Ti particle capable of
causing bridging of metal lines leading to short circuit in an integrated circuit [15]. The severity of
the defect formation depends on the quality of the target material, design of the target and also on
process variables such as deposition rate, maintenance/replacement schedule of the process kit, etc.
In order to keep the interpretations simple, we draw some examples from the studies carried out on
the tungsten�titanium alloys (W-Ti), indium tin oxide (ITO) and aluminum alloys used in the
VLSI and flat panel display applications, respectively [14�30].
8.3.1 Flaking of nodules from the sputtering targetIn-film defect formation due to the flaking of nodules from the sputter surface has the following
steps: (i) nucleation and growth of nodules on the target surface because of the re-deposition of
sputtered materials, (ii) disintegration of nodules because of arcing and (iii) the flaking of nodules
from the sputter surface [25]. Detailed studies on the W-Ti alloy systems have disclosed that nucle-
ation and growth of the nodules preferentially take place in the vicinity of the voids in a target that
is less dense [25].
Figure 8.7 shows less than 1 µm size nodules around a void in a W-10 wt% Ti alloy target that
is 95% dense [26]. In addition, surface imperfections such as crevices and cracks also favor nodule
formation, and shape of the nodules can be remarkably different. These imperfections result from
machining and grinding operations. Evidence of nodule formation have been found on those parts
FIGURE 8.6
Ti-W particle causing bridging of metal lines [15].
5758.3 In-film defect generation
of the sputter surface where magnetic fields are weak. This means that the areas other than erosion
grooves are preferential sites for nodule formation. In round targets, these areas are usually found
close to the periphery and the center of the sputter face. This implies that there would be a differ-
ence in the nodule density and the size distribution along the radial direction from the center to the
periphery.
Figure 8.8 shows the variations of nodule density as a function of radial distance from the center
to the periphery for three W-Ti alloy targets that have different densities (94% for target 3 and
86.4% for target 5), i.e., different void contents [25]. It is clear that the largest densities of nodules
are found close to the periphery of the sputter face followed by the center, while nodule densities
are at their minimum in the erosion grooves. On the contrary, nodule sizes are larger at the bound-
aries of the erosion groove and the smallest at the periphery (Figure 8.9) [25]. These nodules either
undergo fracture or flaking that form particles on the deposited films.
Figure 8.10 (p. 578) shows a nodule that has fractured [25]. Similar observations have been made in
indium�tin�oxide sputtering targets. Figure 8.11 (p. 578) shows two fully grown nodules in which
one of the nodules has the tip intact and the other one has lost the tip because of the fracture [27].
Unlike voids in W-Ti alloys, indium�tin�oxide sputtering targets are sensitive to the distribution of
the oxide phase (SnO2), which has been found to cause arcing during sputtering. With improved metal-
lurgical practices, the distribution of tin oxide can be made uniform and diffusion of Sn into In2O3 can
be completed. Figure 8.12 (p. 579) shows two indium�tin�oxide targets with varying amount of
nodules because of the difference in the target metallurgy. Target (a) has more uniform distribution of
tin oxide than in target (b) and shows 6 times less arcing than the target (b) [28].
8.3.2 Flaking of brittle films from the process kitSputtering materials get accumulated on process kit components such as the chamber shield and the
clamping ring during sputter deposition. In fact, the large surface area of the chamber shield
FIGURE 8.7
SEM image showing W-Ti re-deposited nodules in the vicinity of voids in a sputtered W-10 wt% Ti target [26].
576 CHAPTER 8 Troubleshooting in Sputter Deposition
initially helps capture dust and particles inside the chamber. With continued sputtering, in particular
in reactive sputtering, highly stressed layers of sputtered material are deposited on the chamber
shield. These stressed layers can flake easily unless pasting (applying pure metal coating) is done
at regular intervals. In spite of pasting, certain materials such as TiN, TaN, W-Ti layers may flake
FIGURE 8.9
Variation of nodule size along the radial direction in a W-10 wt% Ti target after 27 kWh sputtering [25].
FIGURE 8.8
Nodule density distributions along the radial direction in W-Ti targets after 27 kWh sputtering [25].
5778.3 In-film defect generation
under the influence of chamber conditions and generate particles on the deposited films. In order to
reduce particle formation from flaking of brittle layers, the chamber shield needs to be replaced at
OEM recommended intervals of time. These clearly show that particle performance of a process is
controlled by both flaking of nodules from the sputtering target itself and the flaking of sputtered
layers from the process kit.
8.3.3 ArcingArcing is a common phenomenon in sputtering, and in this section we first discuss the role of mate-
rial imperfections on the arcing problem, and as a result, defect formation in deposited films.
FIGURE 8.10
SEM image of a broken nodule from W-Ti sputtering target [25].
FIGURE 8.11
Two fully grown nodules at the sputter face of an indium-tin-oxide sputtering target. The left nodule shows a
particulate at the tip (x 250) [27].
578 CHAPTER 8 Troubleshooting in Sputter Deposition
Figure 8.13 shows these imperfections (dielectric inclusions and films, voids, entrapped gases, con-
taminations, etc.) that are of metallurgical origin and can be reduced by improved manufacturing
practices [16]. The other major issue is the prevention and suppression of arcing using advanced
power supply units [17].
Arc is a general term used for representing a high power density short circuit. A possible cause
of arc generation is the rapid accumulation of charge on a small area (often dielectric films and
inclusions) at the sputter surface. An arc resembles a miniature explosion and can cause local
FIGURE 8.12
Indium-tin-oxide sputtering targets showing nodule contents in (a) conventionally processed target and (b)
improved target [28].
FIGURE 8.13
Sputtering target showing undesirable attributes that are of metallurgical origin and may cause arcing [16].
5798.3 In-film defect generation
melting of the sputtering material inside a deposition chamber. The molten material ejected from the
target surface can form particles and splats on the deposited film (Figure 8.14) [23]. Pareto analysis
of chip manufacturers shows that arc-related defects reduce the overall yield of deposited wafers.
Arcs are broadly classified into three groups based on the mechanisms of formation;
namely, microarcs, unipolar arcs and bipolar arcs (Figure 8.15) [30]. Microarcs (also known as
FIGURE 8.14
A splat on the deposited film caused by the arcing at aluminum oxide particles embedded in an aluminum
sputtering target [23].
FIGURE 8.15
Types of arcs that are generated during sputtering: (a) Microarcs, (b) unipolar arc and (c) bipolar arc [30].
580 CHAPTER 8 Troubleshooting in Sputter Deposition
plasma-target/cathodic arcs) are ignited between the plasma and the dielectric parts of the target,
which are often native oxide films (Figure 8.15(a)). Arcs are caused by the electric breakdown of
the dielectric films on the target surface; the mechanism of breakdown is explained in Figure 8.16
[16]. The accumulated positive charge on the front face of dielectric film develops a large voltage
difference across the film because the back of the film is in contact with a negatively charged tar-
get. A breakdown occurs when the voltage across the film (the same as in a capacitor) exceeds the
dielectric strength of the film. As a result, plasma discharges to a point where charge builds up
(inevitably dielectric regions) and electric field is favorably distorted to allow shorter path for
grounding to occur. This is equivalent to the bridging of the cathode dark space. A majority of arc-
ing events, as large as 99%, are caused by the microarcs [30].
Unipolar arcs are ignited between conducting sputtering material and the dielectric parts of the
target (Figure 8.15(b)) [30]. Dielectric materials can be oxide inclusions as well as re-deposited
materials in the vicinity of the periphery and the center of the sputtering face. Formation of AlOx
type amorphous oxide and TiN has been reported in the reactive sputtering of aluminum oxide and
TiN films, respectively. The re-deposition is favored by the weak magnetic fields in these areas. It
should be noted that the discharge takes place between the dielectric and the conducting spots on
the target itself unlike between plasma and non-dielectric films in the microarc case. Unipolar arcs
can initiate microarcs, and in the literature these are often used interchangeably.
Bipolar arcs are generated between the sputtering target and the shield/substrate (Figure 8.15(c))
[30]. Strong electrical instability in bipolar arcs can cause complete shutdown of the system. The
frequency of occurrence of bipolar arcs is relatively low and initiated by the bridging flakes or
improperly centered hardware.
FIGURE 8.16
Electrical breakdown of a dielectric film found on the sputter surface and arc generation [16].
5818.3 In-film defect generation
The probability ratio of arcing for bipolar, unipolar and microarcs is reported to be 1:10:100000
[30]. In the ranking of energy associated with each type of arcing, microarcs are placed on the
lower side, while bipolar arcs are on the higher side (Emicroarcs,Eunipolar,Ebipolar). The above
information provides a basic understanding of the various types of arcs that are encountered in sput-
tering systems. In order to reveal the role of target metallurgy on the formation of arcs, we provide
some examples.
First, we will discuss material imperfections (dielectric inclusions and films, voids, surface pro-
trusion, entrapped gases) that can affect in-film defect formation in the form of particles or arcs.
Subsequently, on a more fundamental basis, the impact of size and shape of dielectric inclusions on
the arc formation will be discussed.
It is a fact that the level of imperfections in the sputtering targets may vary based on the target
manufacturer and even from one lot to the other supplied by the same manufacturer. Fine dielectric
inclusions or films have their origin at skin oxides formed at the surface of the starting material, which
subsequently disintegrate and distribute during melting practices and also metal-forming operations.
Voids and entrapped gases result from the solubility difference of gases in the molten and solid states.
For example, higher solubility of hydrogen in molten aluminum than in solid aluminum may cause
entrapment of hydrogen gas in the form of small voids or porosity in the solid ingots.
Studies on aluminum sputtering targets with different levels of such undesirable metallurgical
attributes reveal that inclusions and voids are the key sources of splats and particles. Splats as large
as 500 µm in size and particles in the range of submicron to tens of micron were recorded. A signif-
icant reduction in in-film defects was achieved by improving the metallurgical practices in target
fabrication processes. Figure 8.17 shows the variation of in-film defects as a function of sputtering
life (in kWh) for a standard target and an improved aluminum target [16].
FIGURE 8.17
Defect-density versus life of sputtering showing reduced defects in improved target as compared to the
standard target [16].
582 CHAPTER 8 Troubleshooting in Sputter Deposition
While we recognize the relative contributions of more than one metallurgical properties to the
in-film defect formation, only a few systematic laboratory scale studies have been carried out to
record the impact of shape and size of the dielectric inclusions on the arc generation and defect for-
mation on the deposited films [22,23]. Attempts have been made to measure the velocity of the
ejected particles, preferred angles of ejection and temperature of the molten droplets. In these stud-
ies either pure aluminum targets or Al-0.5wt% Cu alloy targets have been used and aluminum
oxide inclusions of chosen size and shape were embedded on the racetracks.
Figure 8.18 shows such a target (7.6 cm diameter), which has an aluminum oxide inclusion
embedded in the racetrack [22]. The size of the aluminum oxide inclusion varied between 100 µmand 3000 µm and various power densities (8�56 W/cm2) were used to record the in-film particle
density caused by the arcing [23]. The arc rate was calculated by dividing the number of arcs with
the deposition time (1 min for all the experiments). Figure 8.19 reveals the linear relationship
between the aluminum oxide inclusion size and the arc rate at various power densities [23]. For the
given set of experimental conditions, larger oxide inclusions showed a steady increase in the arc
rate at higher power densities. As a result, in-film particle density rose steadily up to a value of
380 per cm2 for a very coarse oxide inclusion case . For a given particle size (2940 µm), increased
power density enhanced the number of particles in all size categories and more than 60% of the
particles were less than 1.1 µm size (Figure 8.20) [23].
A critical inclusion size for the initiation of arcing was estimated to be 4406 160 µm and also
argued that this size effect was only related to the condition for initiation of arcing and not related
FIGURE 8.18
Schematic of a laboratory scale aluminum alloy sputtering target that has aluminum oxide inclusion
embedded in the racetrack [22].
FIGURE 8.19
Relationship between oxide inclusion size and the arc rate at various power levels [23].
5838.3 In-film defect generation
to the condition required for the ejection of molten metal from the sputter surface to form particles.
It was believed that the force exerted by any arc was strong enough to overcome the surface tension
of the molten aluminum at sputter surface to generate particles. This study also established relation-
ships between arc rates and the in-film particle densities (Figure 8.21) [23].
These results also imply a saturation of particle density with increased arc rate. The role of alu-
minum oxide inclusion area, shape and orientation on the arcing and particle generation was also
reported [23]. The existence of a relationship between the plasma dark-space distance and the diam-
eter of a circle having the same surface area as that of an oxide inclusion was also proposed. It was
reported that below a critical inclusion area, no arcing activity was recorded. When the inclusion
length exceeded twice the dark space distance, the orientation of the inclusion with respect to the
racetrack tangent played an important role in arc generation. Elongated inclusion placed perpendic-
ular to the racetrack tangent showed greater tendency of particle formation than the inclusion
placed parallel to the racetrack tangents.
Though the above experimental results provided a great deal of information about arcing char-
acteristics, the quoted oxide inclusion diameter of 520 µm6 110 µm corresponding to the critical
surface area requirement for the initiation of arc appears unrealistically large and it is believed that
FIGURE 8.21
Relationship between the arc rate and particle density in the deposited films [23].
0.5 0.75 1.1 1.6 2.4 3.6 5.4 8.1 12Particle size (µm)
020406080
100120140160180
Par
ticle
den
sity
(cm
–2) AI2O3 Inclusion:
2940 um SputteringPower density
(W/cm2)8
16 3224
FIGURE 8.20
Particle size distributions in the deposited film after sputtering at various power densities. Particles are
generated by an aluminum oxide inclusion (2940 µm in size) embedded in the sputtering target [23].
584 CHAPTER 8 Troubleshooting in Sputter Deposition
even much smaller inclusions could initiate arcing. For example, in aluminum alloy sputtering tar-
gets for flat panel display applications, 10 µm diameter inclusions were found to be large enough to
cause arcing and particle generation [29,31].
Studies involving high-speed video analysis of arcing revealed some characteristics of ejected
molten droplets such as velocity, trajectory and temperature [22]. Velocities from 5 to over 500 m/
sec were recorded for aluminum droplets, and it was not possible to verify size dependence of
velocity as found in the case of copper [32]. The ejection angles of molten aluminum droplets con-
centrated around 30� from the target surface (Figure 8.22) [22]. The temperature range of the dro-
plets was estimated to be between 1750 K and 3000 K6 1000 K. Other works on copper reported
temperatures between 2000 K and 3200 K [33,34].
It has long been realized that though arcing can be reduced by improving target design, metal-
lurgical practices and also by optimizing the process variables, modifications are required in the
design of power supply. Over the last few decades emphasis has been placed on the detection and
suppression of the arcs and yet improvements have been required with the evolution of the
advanced deposition tools. The subject of arc detection and suppression is out of the scope of this
discussion and details can be found elsewhere [35�38].
8.3.4 Trace elements of sputtering targetHigher trace element content of a sputtering material can cause excessive particle formation during
sputtering and one such example is high purity titanium with higher oxygen content. In a produc-
tion environment, the number of in-film particles continues to increase with sputtering life (kWh).
In order to keep the in-film particle content under control, pasting of chamber shield and process
kit change are required at recommended intervals of time.
It has been shown that by lowering the oxygen content of the titanium, it is possible to maintain an
in-film defect level significantly low up to a large sputtering life [39]. This is primarily because of the
reduced oxygen available to the titanium to form titanium oxide. In various other sputtering materials,
trace elements can be of different chemical nature (e.g., gases, alkali, transition and noble metals) that
may have a tendency to cause arcing in more direct ways or form undesirable inclusions that may lead
to arcing and generation of particles.
0 15 30 45 60 75 90Angle (Degree)
0
5
10
15
20
25
30
No.
of m
acro
-pat
icle
s
90°
0°Target
FIGURE 8.22
Molten metal ejection angle distribution for aluminum alloy sputtering target [22].
5858.3 In-film defect generation
8.3.5 Re-deposited material in hollow cathode magnetron (HCM) targetsThe same as flat sputtering targets, hollow cathode magnetron targets also suffer from particle pro-
blems. The primary source of particles is the re-deposited material. If re-deposited material fails to
stick to the target material properly, re-deposited material will starts to flake under the influence of
cyclic thermal stresses. Figure 8.23 shows an example of a re-deposited area and the phenomenon
of flaking in a titanium HCM target [6]. The shiny and relatively smooth regions are the
re-deposited areas, which in this case extend to the side wall up to a height of one-third of the verti-
cal height of the HCM. The boundaries between the sputtered and re-deposited regions often show
flaking of re-deposited material. Figure 8.24 shows an example of nodule formation because of
re-deposition in a copper HCM target [6]. The structure of re-deposited material in a HCM target
can be drastically different than what we have seen in flat sputtering targets.
FIGURE 8.23
Concentric shiny areas inside a sputtered titanium HCM target showing re-deposited material. Flaking of
re-deposited material is also visible in the inset [6].
FIGURE 8.24
A sputtered copper HCM target showing re-deposited nodules at the edge of the target [6].
586 CHAPTER 8 Troubleshooting in Sputter Deposition
8.3.6 Contaminated particle-traps of the sputtering targetsWe know that various sputtering targets contain particle-trap zones in the form of grit-blasted or
arc-sprayed areas next to the sputter surface [40]. Such rough zones have a large surface area that
helps capture particles from the chamber and improve particle performance of the process.
However, if such regions are contaminated during subsequent finishing steps and handling, these
can be a major source of arcing and particle formation on the deposited films. For example, during
initial cleaning operations arc-sprayed regions may pick up fine cotton threads and dust particles
because of improper handling. Therefore, extra caution is required for the preparation of particle-
trap regions around the sputter face and handling of sputtering targets.
8.4 Bonding-related problemsA large variety of sputtering targets are bonded to appropriate backing plates using resins or elas-
tomers, solders and diffusion bonding methods (see section 4.5 of Chapter 4 for details). While
bond coverage and bond strength values are expected to be satisfactory, in many instances targets
fail to meet the requirements because of incomplete bonding. In the case of solder bonding, if
both target and backing plates are not flat, bond coverage can be significantly low and conse-
quently bond strength would be low. Other reasons include inadequately prepared surfaces, insuffi-
cient pre-coating, applying incorrect bonding temperature and handling problems. Figure 8.25
shows a bond coverage map developed using an ultrasound technique [6]. The central patch
FIGURE 8.25
Ultrasound map showing partial debonding at the target�backing plate interface in a titanium target [6].
5878.4 Bonding-related problems
clearly shows the severity of incomplete bonding. In some cases, defective targets made using sol-
der bonding can be re-bonded with proper measures.
Similarly, bonding defects are also found in diffusion-bonded and resin/elastomer-bonded tar-
gets. In the case of diffusion bonding, meeting surfaces of the target and backing plate are often
prepared using one of the grit-blasting methods to impart adequate roughness. Failure to obtain
desired roughness and sufficiently clean surfaces may lead to incomplete bonding. Minor bowing
of the target and also the backing plate do not pose a severe bonding problem because bonding
takes place under high pressure.
Figure 8.26 shows debonding of a diffusion-bonded titanium target [6]. The sputtering target
appeared to have weakly-bonded regions in the vicinity of the periphery. Separation of the titanium
target and backing plate developed during use under operating power of several kWs. It is worth
noting that such separation in the form of voids stops heat transfer from target to the backing plate.
This leads to abnormal temperature rise at the sputter face. As a result severe grain growth can
occur in target material, which is the case in this example. The shiny regions in the diametrically
opposite positions showed larger grains. Other extreme examples of cone formation because of
selective debonding of target material from the backing plate and the development of cracking
at the sputter face because of creep-related issues are shown in Figure 8.27 and Figure 8.28,
respectively [6].
FIGURE 8.26
A sputtered titanium target showing partial debonding at the target�backing plate interface. Shiny regions
showed large grains because of grain growth driven by heat build-up [6].
588 CHAPTER 8 Troubleshooting in Sputter Deposition
FIGURE 8.27
A tantalum target showing partial debonding at the target�backing plate interface and cone formation at the
sputter surface [6].
FIGURE 8.28
A titanium target showing crack formation in the vicinity of the center during sputtering [6].
5898.4 Bonding-related problems
8.5 Long pump-down time and out-gassingIt is not uncommon to encounter long pump-down time to reach the base pressure in a deposition
tool. Pump-down in sputtering tools usually consists of a rough pumping stage, leak check using
helium, cycle purge using nitrogen, bake-out at high temperature for several hours, cooling and rate
of rise check (air leakage while pump is off) before burn-in of targets start. Evidence shows that
the pump-down time can be long and rate of rise can be large as a result of unclean and out of con-
trol rough o-ring grooves of the sputtering targets. With a smoother and cleaner o-ring groove,
pump-down time has been found to meet specification. It is also recognized that if the density
of the sputtering target is not high enough and the sputter face of a target is contaminated because
of the usage of inappropriate packaging material or mishandling, pump-down time can be longer
than usual.
Once the chamber is ready for burn-in of the sputtering target, power is applied to the sputtering
target in steps to reach the desired power level. At this stage, the sputtering target starts to eject
material from the sputter surface and plasma starts to see the various inhomogeneities such as
inclusions, voids and entrapped gases in the target. Occurrence of prolonged out-gassing has been
reported during burn-in of sputtering targets. In most cases, powder metallurgy processed targets
have shown out-gassing. In the fabrication of sputtering targets, during mixing of powders, con-
taminants can be incorporated into the raw material if the container and the grinding media are not
clean or previously used for a different product. On subsequent processing steps, these contami-
nants may produce gaseous products that would be entrapped into the sputtering target. Similarly,
incomplete removal of the binders in the fabrication of sputtering targets can lead to the generation
of gases that can be entrapped into the sputtering target. During out-gassing these gaseous products
are removed in a low-pressure chamber condition. Though the out-gassing problem is less common,
long pump-down time can be an issue if cleanliness is not maintained at all stages of handling and
installation of the sputtering targets into the deposition chamber [39,41,42].
References[1] Dunlop JA, Goldstein M, Feldewerth GB, Shim C, and Schittny S. Method of reducing sputtering
burn-in time, minimizing sputtered particulate, and target assembly therefor, U.S. Patent No. 6030514, 29
Feb. 2000.
[2] Yamakoshi Y, Miyashita H, Seki K. Sputtering target and method for manufacturing thereof, U.S. Patent
No. 6153315, 28 Nov. 2000.
[3] Takahashi H, Ohhashi T, Seki K. Sputtering target free of surface-deformed layers, U.S. Patent No.
6284111, 4 Sep. 2001.
[4] Joyce JE, Hunt TJ, Gilman PS. Method of manufacturing enhanced finish sputtering targets, U.S. Patent
No. 6309556, 30 Oct. 2001.
[5] McDonald PH. Method of treating sputtering target to reduce burn-in time and sputtering target thereof
and apparatus thereof , U.S. Patent publication no. 2005/0040030 A1, 24 Feb. 2005.
[6] Praxair Inc. USA (r 2013 by Praxair Inc. All rights reserved).
[7] Hayama AOF, Sandim HRZ. Annealing behavior of coarse-grained titanium deformed by cold rolling.
Mater Sci Eng A 2006;418:182.
[8] Michaluk CA, Nowell MM, Witt RA. Quantifying the recrystallization texture of tantalum. JOM 2002;54
(3):51�4.
590 CHAPTER 8 Troubleshooting in Sputter Deposition
[9] Raabe D, Schlenkert G, Weisshaupt H, Lucke K. Texture and microstructure of rolled and annealed
tantalum. Mater Sci Technol 1994;10:299.
[10] Sandin HRZ, Padilah AF, Randle V, Blum W. Grain subdivision and recrystallization in oligocrystalline
tantalum during cold swaging and subsequent annealing, Int. J Refrac Metal & Hard Mater 1999;17:431.
[11] Sarkar J, Gilman PS. Imaging ultrafine grains in machined tantalum subsurface using a focused ion
beam. Scripta Materialia 2008;59:301.
[12] Sarkar J, McDonald PH, Gilman PS. Surface characteristics of titanium targets and their relevance to
sputtering performance. Thin Solid Films 2009;517:1970.
[13] Sarkar J, McDonald PH, Gilman PS. Method and apparatus for treating sputtering target to reduce burn-in
time and sputtering targets made thereby, US patent application no. 2008/0121516, 29 May, 2008.
[14] Turn Jr. JC, Marx DR. The role of tungsten-titanium target density on particulate generation. Material
Research Corporation, Technical note # 1992;1263.
[15] Dunlop J, Waterman E, Brat T. Effect of Ti-W target processing methods on defect generation during
very large scale integrated device fabrication. J Vac Sci Technol A 1992;10(2):305.
[16] Pavate V, Abburi M, Chiang S, Hansen K, Mori G, Narasimhan M, et al. Correlation between aluminum
alloy sputtering target metallurgical characteristics, arc initiation and in-film defects. SPIE
1997;3214:42.
[17] Li X, Narasimhan M, Pavate V, Loo D, Rosenblum S, Trubell L, et al. Integrated arc suppression unit
for defect reduction in PVD applications, 3214. SPIE; 1997.
[18] Waterman E, Dunlop J, Brat T. Tungsten-titanium sputtering target processing effect on particle genera-
tion and thin film properties for VLSI applications. VMIC conference 1990;12�13:329.
[19] Gn FH, Yeap CB, Li HM, Liu EZ, Chew HL. TiW particle control for VLSI manufacturing. SPIE
1995;2635:30.
[20] Wickersham Jr. CE, Poole JE, Mueller JJ. Particle contamination during sputter deposition of W-Ti
films. J Vac Sci Technol A 1992;10(4):1713.
[21] Cheney E, Lazaroff D, Morales D, Yap L, Chiu L. Defect performance for PVD of TiW and TiWN
films. Solid State Tecnol 1997;109(Nov).
[22] Wickersham Jr. CE, Poole JE, Fan JS, Zhu L. Video analysis of inclusion induced macroparticle
emission from aluminum sputtering targets. J Vac Sci Technol A 2001;19(6):2741.
[23] Wickersham Jr. CE, Poole JE, Leybovich A, Zhu. L. Measurements of the critical inclusion size for arcing
and macroparticle ejection from aluminum sputtering targets. J Vac Sci Technol A 2001;19(6):2767.
[24] Wickersham Jr. CE, Poole JE, Fan JS. Arc generation from sputtering plasma-dielectric inclusion inter-
actions. J Vac Sci Technol A 2002;20(3):833.
[25] Lo C, Draper D. Quantitative measurement of nodule formation in W-Ti sputtering. J Vac Sci Technol
A 1998;16(4):2418.
[26] Lo C, Gilman P. Particle generation in W-Ti deposition. J Vac Sci Technol A 1999;17(2):608.
[27] Schlott M, Kutzner M, Hanau DAG, Gehman BL, Reger N, Stadermann FJ. Nodule formation on
indium-oxide tin-oxide sputtering targets. SID Digest 1997;P-31:1.
[28] Nakashima K, Kumahara Y. Effect of tin oxide dispersion on nodule formation in ITO sputtering.
Vacuum 2002;66:221.
[29] Yoshikawa K, Yoneda Y, Koide K. Spray formed aluminium alloys for sputtering targets. Powder
Metallurgy 2000;43:198.
[30] Koski K, Holsa J. and Juliet, P., Surface defects and arc generation in reactive magnetron sputtering of
aluminum oxide thin films. Surf Coat Technol 1999;115:163.
[31] Nishi S, Kusamichi T, Onishi T, Mizuno M, Takahara T, Suemitsu T, Yoshikawa K. Aluminum or
aluminum alloy sputtering target and method for manufacturing the same, US patent application No.
20010004856, June 28, 2001.
[32] Udris YY. Disintegration of materials by an arc cathode spot. Radio Eng Electron Phy 1963;8:1050.
[33] Tanberg R, Berkey WE. On the temperature of cathode in vacuum arc. Phys Rev 1931;38:296.
591References
[34] Boxman RL, Goldsmith SJ. The interaction between plasma and macroparticles in a multi-cathode-spot
vacuum arc. Appl Phys 1981;52:151.
[35] Andres A. Physics of arcing and implications to sputter deposition. Thin Solid Films 2006;502:22.
[36] Andres A. Physics of arcing and implications to sputter deposition. Proc of the 5thICCG, Saarbruecken
2004;1.
[37] Harpold J, Scholl RA. How Advanced Energy MDX products manage arcs, Application notes, Advanced
Energy, 1995.
[38] Schatz D. The MDX as a strategic tool in reducing arcing, Application notes, Advanced Energy, 1983.
[39] Fukuyo H, Shindo Y, Takahashi H. Titanium target for sputtering, US Patent No. 6755948, June 29,
2004.
[40] Mostovoy R, Mori GT. Preventing defect generation from targets through applying metal spray coating
on side walls, US Patent No. 6428663, Aug 6, 2002.
[41] Koski K, Holsa J. and Juliet, P., Voltage controlled reactive sputtering process for aluminum oxide thin
films. Thin Solid Films 1998;326:189.
[42] Mitterer C, Heuze O, Derflinger V-H. Substrate and coating damage by arcing during sputtering. Surf
Coat Technol 1997;89:233.
592 CHAPTER 8 Troubleshooting in Sputter Deposition