Martin P.M. Handbook of Deposition Technologies for Films and Coatings, Third Edition: Science, Applications and Technology

Подождите немного. Документ загружается.

Unﬁltered and Filtered Cathodic Arc Deposition 491

Figure 10.10: Through-ﬁeld geometries used for spot steering.

According to the retrograde motion rule, the spot tends to move in the direction of −j ×B,

where j is the vector of current density, which, per engineering deﬁnition, points from plus to

minus, i.e. into the cathode. B is the vector of the magnetic induction at the location of interest

on the cathode surface.

According to the acute angle rule, which applies in addition to the retrograde motion rule, the

spot tends to move toward the direction given by the acute angle between the magnetic vector

B and the normal of the cathode surface [104].

One may distinguish two basic magnetic ﬁeld conﬁgurations [105]: (1) the through-ﬁeld

conﬁguration, where magnetic ﬁeld lines are essentially normal to the cathode surface (Figure

10.10); and (2) the arched ﬁeld conﬁguration, where ﬁeld lines arch over the cathode, forming

a magnetic tunnel, very much like in sputtering magnetrons (Figure 10.11).

As indicated in Figures 10.10 and 10.11, spot steering leads to preferred regions of arc spot

operation and thus cathode erosion. In both cases, through-ﬁeld and arched ﬁeld, the spots

492 Chapter 10

Figure 10.11: Arched or tunnel ﬁeld geometries used for spot steering.

follow the retrograde motion rule when they have reached the optimum position according to

the acute angle rule. For example, in the arched geometry, the spots tend to follow the tunnel

formed by the arched ﬁeld. Most steered arc sources use the through-ﬁeld conﬁguration

because this type of ﬁeld promotes the transport of plasma to the substrate. The arched ﬁeld, in

contrast, can trap plasma near the cathode. The through-ﬁeld is especially desirable when the

DC arc source is coupled to the ﬁeld of a macroparticle ﬁlter.

Not all arc sources use magnetic spot steering. Some DC systems operate in a so-called

switched mode where subtle electric effects are used to control the direction of spot ignition.

The most prominent examples are systems developed by Vergason Technology, Inc. [106] and

by Vapor Technologies, Inc. [107]. In Vergason’s designs, the cathode is a long rod with spot

sensors at both ends. The sensor picks up information based on the proximity of the spot, e.g.

its magnetic ﬁeld or light emission. The spot appears to travel to the side where the cable is

connected, minimizing the loss in the cathode. When it arrives near the cable connector, the

sensor provides information and a switch box connects the power contact to the other end of

the cathode, thereby reversing the spot’s travel direction. By adding a magnetic ﬁeld [108], the

arc spot appears to travel in a helix around and along the cathode rod.

Unﬁltered and Filtered Cathodic Arc Deposition 493

10.3.2 Pulsed Sources

Pulsed arc sources are rare in the coating industry but are often used in laboratory research.

They offer miniaturization, allowing us to use them in small chambers, typical for

university-style research, with minimal requirements on average power consumption and

cooling. New process parameters appear such as pulse length and duty cycle. At low duty

cycle, temperature-sensitive substrates can be coated such as plastic sheets. One difﬁculty

appears with pulsed operation: the need for frequent and often rapid arc triggering. This

requires trigger concepts other than the mechanical trigger typically used for DC arc sources.

In the case of high current operation, generally exceeding 1 kA with high duty cycle (∼10%),

pulsed sources can reach the deposition rates of DC sources. The average plasma production

scales with the average power, and one can readily derive requirements in terms of arc current

and duty cycle when comparing DC and pulsed arc operation.

Multiple cathode spots form when the arc is operated at high arc current. Taking into account

the self-magnetic ﬁeld produced by the current, simultaneously existing cathode spots tend to

repel each other, which can be understood when considering the motion of one spot in the

magnetic ﬁeld of the other. Therefore, pulsed high current arcs can be considered as a special

case of magnetically steered arc. This kind of mutual spot inﬂuence can be utilized by

implementing a trigger in the center of the cathode [109–112]. The spots repel each other, and

the arc pulse is preferably terminated when the spots reach the outer rim or edge of the

cathode, leading to optimum utilization of the cathode material.

10.4 Macropar ticles and Macroparticle Control

10.4.1 Macroparticle Generation

Macroparticles are liquid or solid debris particles that are produced at cathode spots. Their

presence is the most severe limitation to cathodic arc deposition technology. First, we will

have a closer look what those particles are and how they are produced. Later in this chapter,

mitigation measures such as macroparticle ﬁlters will be discussed.

Macroparticles have the term ‘macro’ in the name because they are indeed massive compared

to ions, atoms, and electrons. Sometimes they are called droplets or microparticles and

nanoparticles. They have a wide size distribution f (d

) that can be best described by a power

law:

f (d

) = Ad

−α

(10.26)

where d

is the diameter or similar characteristic size of the macroparticle, and the

parameters A and α are material dependent. The upper end of the distribution, i.e. the largest

494 Chapter 10

particle, is typically in the 10 ␮m range or even a bit larger for materials of low melting

temperature. The lower limit of the distribution is not well known. Many quote about 0.1 ␮m,

but this is more related to the limitations of the measuring equipment used, such as a scanning

electron microscope (SEM).

When the emission site goes through the stages of life as described in Section 10.2, a highly

localized zone forms between the relatively cold cathode and the hot plasma above it where

the cathode material is in the liquid phase. This liquid is subject to the transient pressure

exerted by the plasma and it responds by yielding. As a result, microscopic droplets or

macroparticles as well as an erosion crater form. The macroparticles preferentially splash to

the side, i.e. under a relatively shallow angle with respect to the cathode surface [50, 113],

although some are formed as the plasma pressure reduces, and so a number are ejected in the

direction normal to the surface.

Macroparticles are generally liquid when they form, and they tend to cool down as they travel

away from the cathode. For refractory metals and graphite, they can be easily seen in ﬂight

owing to their incandescent nature, while macroparticles of low melting-point materials are

not as hot and therefore more difﬁcult to discern visually or photographically. Macroparticle

size distributions and macroparticle interaction with the arc plasma have been studied over the

years by many researchers [50, 54, 113–127] and a detailed review can be found elsewhere [3].

Once they arrive on a surface, such as a substrate to be coated, they may stick and be

incorporated into the growing ﬁlm (Figure 10.12). This, however, is not always the case. In

particular, those macroparticles that have become solid before impact have a high likelihood of

Figure 10.12: Incorporation of a large titanium macroparticle in a titanium nitride ﬁlm: the oval

shape of the large macroparticle is due to the deposition of nitride on top of the approximately

spherical macroparticle; one can also see defects (holes) caused by loss of macroparticles in the

sample fracture process. (Photograph courtesy of Andreas Sch

utze, Liechtenstein.)

Unﬁltered and Filtered Cathodic Arc Deposition 495

Figure 10.13: Computer simulation of ﬁlm growth with incorporation of three macroparticles as

indicated by the arrows. (Adapted from [128].)

bouncing back. Macroparticle reﬂection can still cause a defect in the growing ﬁlm, and

reﬂection is a major issue for macroparticle ﬁlters (see later in this chapter).

The incorporation of a macroparticle into the ﬁlm is likely to cause a defect that is even larger

than the macroparticle because the macroparticle will affect the growth of the ﬁlm after its

arrival: the resulting defect shape is conical (Figure 10.13)

10.4.2 Macroparticle Reduction

As already indicated, macroparticles are highly undesirable, and much effort has been invested

to ﬁnd methods to reduce, or even to eliminate, the macroparticle content in coatings. There

are two main approaches: (1) reduce the macroparticle production in the ﬁrst place; and (2),

separate the plasma ﬂow from the ﬂux of macroparticles. In the following, those approaches

will be examined more closely.

Macroparticle generation is related to the formation of a pool of liquid material, as mentioned

before. Therefore, one idea is to reduce the time for which an emission site is active, thereby

reducing the size of the liquid pool and the amount of material that can be ejected. This can be

done in several ways.

As we know, ignition of a new emission site is promoted in the presence of a dielectric layer,

leading to spots of type 1. Typically, several emission sites are simultaneously active,

electrically in parallel, and the competition for carrying current extinguishes the more resistive

sites, i.e. those that have already formed neutral vapor (choking effect; see Section 10.2).

Therefore, reactive cathodic arc deposition has generally a much smaller macroparticle issue

than the deposition of pure metals, a fortunate side-effect that is used heavily in the coating

496 Chapter 10

Figure 10.14: Comparison of copper macroparticle distribution functions for DC arcs and pulsed

high-current arcs. The number of macroparticles is normalized to size class, deposition area, and

ﬁlm thickness. (Data courtesy of Peter Siemroth and co-workers, Dresden, Germany.)

industry. In fact, the industry may not exist but for that effect: unﬁltered cathodic arc

deposition is acceptable for many applications because the size of the macroparticle is

relatively small and not detrimental to the application.

Another way to reduce the size of macroparticle is through enhancing the apparent spot

velocity, which is just a different way of saying that the average time for which an emission

site is active is reduced. As mentioned before, spot steering can be done magnetically or

through electrical means. Figure 10.14 shows examples of macroparticle size distribution

functions illustrating the spot steering effect, here via using a high current pulsed arc. In all

cases, the power law character of the distribution is maintained, supporting the fractal

approach to process understanding and modeling.

10.4.3 Macroparticle Filtering

Now let us consider the second approach of dealing with macroparticles, the separation of

macroparticles from the plasma, known as macroparticle ﬁltering. The concept is based on the

possibility of guiding a plasma ﬂow along magnetic ﬁeld lines, toward a substrate, while

macroparticles, even if charged, are too massive to be affected by magnetic or electric ﬁelds.

To understand the transport mechanism, we need to recall that charged particles, such as

electrons, can be ‘magnetized’. By this we mean that (1) the gyration radius (Eq. 10.20) is

smaller than a characteristic size like the radius of a ﬁlter duct, and (2) the electron can execute

Unﬁltered and Filtered Cathodic Arc Deposition 497

many gyrations before colliding with another particle. Averaging over the gyration motion one

can state that the electrons move along magnetic ﬁeld lines while their mobility perpendicular

to the ﬁeld line is greatly impeded. Cross-ﬁeld motion is only enabled by collisions and

collective plasma effects (which are beyond the scope of this chapter).

Plasma is a quasi-neutral ensemble of electrons and ions, and so one also needs to consider

what ions do. If we plug the ion mass and a reasonable ion velocity and magnetic ﬁeld strength

in the equation for the gyration radius (Eq. 10.20), we can quickly see that the ion gyration

radius is large, and therefore ions are generally not magnetized. Yet, when electrons are guided

by the ﬁeld, very small deviation from the quasi-neutral charge balance sets up very large

electric ﬁelds, E, as can be calculated by the Poisson equation

∇·E = ρ (10.27)

where ρ = n

−n

is the local net charge and ε

≈8.854 ×10

−12

F/m is the permittivity of free

space. The electric ﬁeld keeps the plasma very close to being quasi-neutral and facilitates the

ﬂow of electrons and ions toward the substrate. Therefore, the transport of plasma in magnetic

macroparticle ﬁlters is achieved with a combined magnetic and electric mechanism. Detailed

descriptions can be made on the basis of hydrodynamic codes or by particle-on-cell

simulations, which have been reviewed recently [3].

There are many different shapes, geometries, and sizes of macroparticle ﬁlters; the review [3]

listed more than 20. Here, we will consider only the most important, illustrating the concept.

That can be best done by ﬁrst looking at the classic macroparticle ﬁlter developed in the

1970s in Kharkov, Ukraine (then the USSR). Taking a quarter torus device, originally

conceived for studying hydrogen plasma transport for Tokomak-related nuclear fusion [129],

Aksenov et al. [130, 131] injected cathodic arc plasma to produce high-quality ﬁlms of

diamond-like carbon (DLC), metals, and nitrides (Figure 10.15). In their apparatus, the ﬁlter

duct is part of the vacuum chamber, and the guiding ﬁeld coils are outside the vacuum and

therefore can easily be cooled. Aksenov et al. isolated the duct wall so it could be positively

biased to reduce plasma transport losses, which was later investigated and utilized by others,

too [132, 133].

This classic design of a ﬁlter has been used and modiﬁed many times, for example, the

bending angle could be less or greater than 90

◦

[134, 135], and the bending could be out of

plane [136, 137] in order to further reduce the likelihood that macroparticles are transported

through the ﬁlter (Figures 10.16 and 10.17).

There are two main mechanisms of macroparticle transport. The most important issue is

macroparticle bouncing from the inner wall of the ﬁlter; and such reﬂections can occur several

times. Therefore, some macroparticles can reach the substrate even if there is no line of sight

between cathode and substrate. The second mechanism could be loosely described as plasma

498 Chapter 10

Figure 10.15: Classic 90



-duct ﬁlter by Aksenov et al. (Adapted from [131].)

wind, i.e. the action of the fast plasma ﬂow on macroparticles. Momentum transfer from ions

could carry small macroparticles with the plasma ﬂow, as wind can transport dust particles.

One can state that ﬁltering is not perfect and what really matters is the ratio of macroparticle

reduction to plasma losses: the quality of a ﬁlter must be judged not only by how well

macroparticles are removed but also by how much plasma is lost. To quantify plasma

transport, one could consider the transport efﬁciency, deﬁned as the ratio of the number of ions

leaving the ﬁlter to the number of ions entering it. Filters are considered efﬁcient when that

ratio exceeds about 25%. While the ion current at the exit can be relatively easily measured, it

is often difﬁcult to determine the ion current at the ﬁlter entrance. Therefore, it is more

common to use the ratio of ion current of the ﬁltered plasma to the arc discharge current:

system

= I

i,ﬁltered

arc

(10.28)

This ratio can be called the system coefﬁcient and is typically in the range 1–4%. The system

coefﬁcient should not be called an ‘efﬁciency’ because it can never reach 100%. The arc

current is mostly carried by electrons and merely serves as a suitable normalization, which

works well because the amount of plasma scales approximately with the arc current.

Unﬁltered and Filtered Cathodic Arc Deposition 499

Figure 10.16: Off-plane double-bend (OPDB) ﬁlter system, originally developed at Nanyang

Technical University, Singapore, and commercialized by NanoFilm Ltd, Singapore. (Courtesy of

NanoFilm Ltd.)

One way of addressing the bouncing issue is to switch from a closed to an open ﬁlter

architecture [137, 138].Byclosed we mean a ﬁlter design where plasma and particle ﬂow are

enclosed, such as by the duct tube that is the part of the vacuum system. In contrast, an

open ﬁlter has openings that allow macroparticles to escape from the ﬁlter volume. In the

simplest case, a curved open coil as shown in Figure 10.18 could be used. An important

element of the open ﬁlter design is the installation of a macroparticle blockage, a

macroparticle ﬁrewall which prevents macroparticles from entering the clean substrate region.

In effect, the macroparticle ﬁrewall separates a plasma source region, where macroparticles are

allowed to ﬂy until they usually come to rest somewhere on the bottom of the chamber, and the

clean substrate region into which plasma is allowed to enter through a designated hole in

the ﬁrewall.

500 Chapter 10

Figure 10.17: 120



-ﬁlter designed for pulsed high-current arc deposition, commercialized by Arc

Precision GmbH, Wildau, Germany. (Courtesy of P. Siemroth.)

Figure 10.18: Open S-ﬁlter illustrating the concept of open-architecture ﬁltering: the

macroparticles can leave the ﬁlter volume; the macroparticle ﬁrewall separates the source region

(right) from the substrate region (left).