Sputtering Targets Adapt to New Materials and Shrinking Architectures

		At a Glance
	Although exotic materials, smaller linewidths and larger wafers are challenging sputtering targets' capabilities, the technology appears prepared, near-term, to meet the demands of next-generation devices.

S puttering is used extensively in silicon technologies. It can deposit a wide range of alloys and compounds and has a moderate capability to cover surface topology. In a sputtering application, the plasma chamber is arranged to enable a high density of ions to strike a target containing the material to be deposited. Thus, it is the target material, not the wafer, that is placed on the electrode with the maximum ion flux (lead photo). To collect as many of these dislodged atoms as possible, the typical sputtering system's cathode and anode are closely spaced at 10 cm or less. An inert gas at a pressure of ~2-3 mTorr is generally used to supply the chamber, resulting in a mean free path on the order of hundreds of microns.

Simple dc sputtering is usually favored for elemental metals. When depositing insulating materials such as silicon dioxide, an rf plasma is used. If the target material is an alloy or a compound, the deposited material's stoichiometry may be slightly different from the target material's. It has been shown, however, that the material with a lower sputter yield will accumulate on the target's surface until the deposited film's composition approximates that of the target's bulk. This makes sputtering very attractive for depositing not only elements, but a wide range of materials (Fig. 1).

The decline of sputtering?

Major challenges facing sputtering technology lie in the fact that as feature sizes shrink, the capability to sputter into the vias becomes increasingly difficult. There have been several innovations over the past several years. One of them, ionized metal plasma, charges the metal, making it possible to accelerate it toward the wafer. This gives it high directionality, which can then make it go into the via holes.

Lead Photo: Increased demands for performance on semiconductor devices are creating a need for new materials beyond the classic aluminum, titanium and copper. The new purity levels being required for these materials by the semiconductor industry will require improved chemistry, mechanical processing and metrology. (Source: Johnson Matthey Electronics)

There is also "long throw" sputtering, where the distance between the target and the wafer is substantially larger (~150-170 mm) than usual. With the greater distance, the angular dispersion becomes more perpendicular to the wafer, accomplishing the same thing as collimation. It still has problems, however, since more metal ends up on the shielding, making it less efficient. Then there are extremely low-pressure applications, in which because there is less collision with other particles, the particles of interest are accelerated to the wafer, not randomly bounced around.

Although some in the industry believe they can hear the death knell of sputtering, most are of the opinion that as the dominant metalization process for so many years, sputtering technology will produce solutions that will extend its life. Currently, with the move toward dual damascene copper architectures for interconnects and the emergence of electroplated copper as the dominant choice (so far) for putting down the bulk copper, the role of PVD has shifted to depositing the critical barrier and seed layer, using IMP Ta/TaN and a copper seed sequence.

Ron Dornseif, principal analyst at Dataquest (San Jose, Calif.), believes that electroplated copper shows the greatest promise. However, with electroplated copper, a copper seed layer is still necessary, so a barrier such as tantalum nitride must be put down to keep the copper from diffusing through the dielectric to the transistor. This means it is necessary to begin with a barrier layer and then put down a copper seed layer, so that when the copper is electroplated it adds thickness to that seed layer. "The seed layer currently is going to be put down with either sputtering or CVD," predicted Dornseif. "In the near term, both of them will be used. Longer term, meaning the next couple of generations, you may see more of a migration to CVD copper as the seed layer."

Fig. 1. The sputtering process enables the user to apply coatings atom by atom, resulting in a highly engineered layer of a desired material with very well-defined properties. (Source: Target Materials Inc.)

As copper interconnect continues to emerge, the metalization world's makeup will begin changing fundamentally. Electroplated copper systems will provide the material fill solution, instead of sputtering targets or CVD tungsten. PVD systems will evolve from Ti/TiN liner tools and Ti/TiN/Al/TiN interconnect tools to TaN/Cu barrier seed layer tools for copper metalization.

Feature size considerations enter into the technology's viability because of particulates. In the wafer fab world, the sputtering systems' business will start to roll over and shift away from aluminum to copper. "A new technology is emerging," said Dornseif. "IBM and Motorola have already announced copper production this year, others will follow and the process will accelerate. By the year 2003 or so, the world will use copper instead of aluminum interconnect. Between now and then a shift will take place, at the expense of sputtering. CVD will pick up some of the lost sputtering business, with electroplate accounting for the bulk of it."

Target materials change

James Gaines, vice president of Target Materials (Columbus, Ohio), does not see sputtering targets going away. "With increased demands for performance on semiconductor devices, there is a need for new materials, beyond the classic aluminum, titanium and copper. We're working with prominent semiconductor companies to develop complex oxides to use in high-performance chips. From our perspective, the market is growing."

This is not to mean there are no difficulties. "Some of the rougher challenges that we see lie in the fact that we're dealing with customers who come from a fairly mature materials base -- the classic semiconductor materials," explained Gaines. "Those materials have had 30 years to be cooked out, optimized and tailored for their function on a chip. The use of complex ceramics is new."

Gaines pointed out that the semiconductor industry is used to getting materials that are 6N (99.9999%) pure chemically. "Those resources are not currently available in ceramics. No other industry asks for such extreme specifications. Much work still remains to be done to develop materials and processes that will turn these complex ceramics into suitable companions for the other materials that are going to be on the chip."

Although Gaines argued that all that is needed to attain these purity levels is better chemistry and mechanical processing, not the development of new technology, there are questions to be answered.

Fig. 2. There is a move in sputtering technology toward IMP plasmas. These use a traditional primary target, as well as a secondary, coil-shaped one, which can be seen as a metal band inside the plasma chamber. (Source: Applied Materials)

PVD's star rises

Murali Narasimhan, global product manager for Applied Materials' (Santa Clara, Calif.) PVD, metal deposition operation, believes PVD will be used well into the future. "Our concept of how its lifetime will be extended and what technologies and changes take place will have changed considerably from what we thought a few years back."

In 1996, Applied introduced its ion metal plasma source (IMP), which is now used in production to deposit titanium and titanium nitride line/barrier films at and below 0.25 µm (250 nm) in high-aspect ratio contact and via structures (Fig. 2). The second-generation IMP source is capable of working with tantalum, tantalum nitride and copper. "This will allow PVD to extend to at least 0.13 µm (130 nm)," predicted Narasimhan, "and enable other film technologies such as electroplating or an integrated PVD/CVD copper scheme." This is because the system is not being used as a traditional film technology, but to enable other film technologies such as electroplating, or a cool copper scheme that uses a subsequent CVD copper seed layer and then PVD on top of it.

As a major player in the sputtering field (Fig. 3), Applied believes that insofar as metalization technology is concerned, PVD will play a different role. "PVD has much to offer, not the least of which is easy maintenance and an overall lower cost-of-ownership," pointed out Narasimhan. "Another advantage lies in high-purity films. This will be PVD's hallmark and why it will continue to exist, although for different applications such as barrier and seed layers."

According to company sources, one of the PVD applications being investigated is using platinum as a potential capacitor electrode for high-dielectric constant barium strontium titinate (BST) oxide capacitors. Ternary alloys for advanced barrier layer applications are also being explored.

An important aspect of PVD relates to its crystallographic orientation capability. PVD makes it possible to control crystal orientation to an extent not possible with CVD or electroplating. This is extremely critical, since film properties can control device properties.

Another major roadblock is in terms of adhesion layer technology. There is much interest in CVD TiN and PVD TiN.

As linewidths shrink, defect densities get more important, with PVD as well as other technologies. Many chip defects are traceable to targets from various sources such as inclusions, porosity, surface finish, surface abnormalities, etc., which cause arcing in the plasma, producing large, die-killing defects. Other defects can be linked with the process chamber. Flaking of deposition from the shielding, as well as the target itself, can be a major contributor. Also, instabilities in the gas flow and the power supply can cause arcing, resulting in defects.

How pure is pure?

Source material purity control is a major headache for target manufacturers today. Sources for the newer materials are limited -- there may be only two or three for a given material. Considerable market pressure must exist for target manufacturers to be able to make their source material providers upgrade a material's base purity. The same is true of chemicals supplied for semiconductor use: higher and higher purity levels are regularly demanded from suppliers who, in turn, must get their suppliers to increase base purity.

The purity question is getting touchier. While a 4N5 (99.995%) purity level is considered adequate for 0.35 µm (350 nm) technology, some chipmakers speculate that 0.18 µm (180 nm) will require 5N or even 6N.

This is creating a monstrous metrology problem. How does one efficiently sample, much less measure, accurately to ensure that what a supplier sells to the end customer is being analyzed close enough for reliability? While it is not impossible to accurately measure (albeit slowly and expensively) such extraordinary purity levels, to do so efficiently will require a whole new magnitude of reliable equipment and sampling techniques. Target manufacturers must be able to make both product and profit, without drowning in ongoing, increasingly complex and expensive time-consuming testing for purity (see "PPT --Time for a Reality Check?" p. 84).

Kazuhiko Hosoya, senior marketing manager for NIMTEC (Chandler, Ariz.), views the quest for purity with equanimity. "We expect the trend for more stringent customer requirements to continue. They want purer films and less defects, with the focus always on materials' cost per wafer. We're seeing joint developments between materials suppliers and equipment manufacturers that sometimes also include the end user." Hosoya added that materials vendors will have to incorporate internal raw material refining -- high purity and high volume -- advanced fabrication techniques for microstructure or texture, and lab capacity for internal target evaluations.

Another need is for non-destructive metrology: in-line techniques that can be used on every target that goes out the door from the manufacturer's. Currently, the only technique available is ultrasonic measurement, which has limitations in how deeply it is capable of scanning the target and in providing adequate resolution. The technique is reliable and offers a low-cost way of testing non-destructively, but presently it cannot detect inclusion particles less than 1 µm.

Conserving source materials

New source materials like tantalum pose another challenge. Because there are not many sources for tantalum, it is necessary to be very conservative in terms of reuse and recycling. At present, a tantalum target costs ~2.5 times more than its titanium counterpart. Target vendors, equipment manufacturers and end users will have to come up with joint strategies on how to drive costs down. Because aluminum and titanium are in sufficiently large supply, the industry has not bothered to recycle used targets. Being more valuable, gold and platinum have been recycled for years, and much of this know-how is applicable to the recovery of tantalum, particularly if the fabs cooperate.

Other improvements are necessary. In traditional PVD, because the erosion grooves were fairly nonuniform on a target, target usage percent was less than 30%. Over the years, better magnetrons have increased this to about 50%. New systems are beginning to get in excess of even that. Work is proceeding to flatten the erosion grooves for better usage, reducing cost.

Although 300 mm designs are essentially a scaling up of existing 200 mm process chambers on the system end, target manufacturers face an entirely different set of problems. They must maintain target metallurgy and consistency across the target face, and materials may have to be sourced differently. Because most targets are sliced horizontally, larger ingots will probably be needed to meet larger diameter requirements. It seems these costs will rise steeply, not linearly, because of the complexities involved if purer, larger ingots are required.

Fig. 3. Shown is the market share of leading sputtering equipment manufacturers, according to Dataquest's 1996 figures, the latest ones available.

Targets and systems

Paul Gilman, director of technology at Materials Research Corp. (MRC, Orangeburg, N.Y.), admitted that there are many challenges in the sputtering target field. "When considering an advanced material, such as a low-permeability cobalt or nickel, you must 'tune' that target for the specific application and particular cathode." By tuning, Gilman means controlling the target's microstructure so that it provides optimal performance in the particular cathode.

"Testing is a critical part of our operation," stated Gilman. "We've a fully equipped applications laboratory with sputtering equipment we've acquired from different vendors. That way, when we develop a new product we can do the actual sputtering as part of our testing. Then, once we translate what we are doing to manufacturing, we can perform our own alpha site testing."

Gilman's view of difficulties facing sputtering is not surprising. "Right now, it is particle reduction, in both aluminum and titanium materials. We work continuously to reduce any kind of inclusions in the material, whether graphite from the actual casting, aluminum oxides or even gaseous species, such as hydrogen."

While MRC does not see sputtering going away any time soon, it recognizes that there are challenges to the technology, such as CVD, particularly in regard to copper, for electroplating. "What we are doing is, essentially, tuning cathodes for electroplating. Phosphorus is added to the copper anodes to control the microstructure. This creates the same type of scenario that you would use for sputtering targets," Gilman pointed out.

As interest in exotic materials grows, the target industry will have  to determine how to make targets out of things such as high-dielectric materials, special precious metal combinations where there might be a combination of two types of precious metals.

Adopting new materials

Johnson Matthey Electronics (JME, Spokane, Wash.) sees as its main goal the integration of the very large change in material sets -- going into materials like copper and tantalum, as well as looking at new exotic materials like ruthenium, platinum and others. Migrating to a larger target size is also a consideration.

"Linewidth geometries will shrink down to possibly the 0.13 µm range, coincidentally with the beginning of 300 mm full production," said Leon Chiu, JME's technical services manager. "That means our customers are looking at even cleaner materials, lower defect levels and higher purities. That is going to be challenging for the very large targets needed for 300 mm, as well as film uniformity issues across such a large diameter."

Another difficulty is the widening range of materials. Before, it was primarily aluminum and titanium. "It now seems we are getting quotes and requests for about every element on the periodic table," commented Janine Kardokus, JME's joint development product manager. "Even with some of the materials that seem to be a certainty, such as copper and tantalum, significant challenges in manufacturing them in high purity will have to be met."

JME ran into the purity metrology issue early on. "We found that our analytical equipment and devices could not keep up with the order-of-magnitude jump in accuracy requirements," said Dr. Sue Strothers, product business unit manager at JME. "We formed a couple of metrology-specific departments and a research group aimed at improving the way we measure things like purity, inclusions in material, etc. Along with that, we're establishing standards for these new types of measurements because although presently some well-recognized standards exist, as we begin considering new sampling tool sets and equipment, the game changes completely."

Fig. 4. A single-wafer, multimodule, high-vacuum processing system designed for precision thin metal film deposition onto silicon wafers up to 200 mm is shown. (Source: Sputtered Films Inc.)

Working together

Andrew Clarke, Sputtered Films' (Santa Barbara, Calif.) senior vice president, believes target and system vendors must work closer together. "As an equipment vendor, we try to understand the material and the target and design our system around using it most effectively. We've had programs where we spent huge amounts of money to develop the cathode water jacket, for example, the target seat inside the machine, and the magnetron to approach 50-60% use of the target material that we put inside." (Fig. 4)

Clarke pointed out that vendors try to work on the texture of the metal in the target to give efficient use. "They've discovered that if grain orientation is a certain way, or the grain size is within a certain range, more of the target material will end on wafers and be a better quality material than if they manufacture it oriented a different way or of a different grain size. They approach it from that standpoint, and we approach it from the equipment standpoint. We tend not to share much of what we learn."

Clarke views target particle emission as an ongoing issue. "The big problem will come up in the sputtering of two-component targets, such as tantalum and silicon. Those will be big problems because in our benchmark studies with SEMATECH, they have shown a lot of interest in ternary alloys. These are the tantalum, silicon, nitrogen alloys. Those particular alloys have to be sputtered from tantalum and silicon targets. Those targets are compounds -- pressed powders. We are working on a project where we actually make the target out of tiles -- tantalum and silicon tiles -- but we are in the very early stages."

Sputtered Films believes that eventually it will have to find a successful way to sputter two materials at the same time, not from a compound target, but from something similar to tile targets. Two targets are cut in "slices," a silicon target and a tantalum target. Then the pieces are put back together, creating a "pie" out of alternated silicon and tantalum pieces. The company has had this in production for more than six years. According to the company, particle performance is good, and for advanced interconnect, it is crucial. "Also, for our work in the X-ray mask program it is crucial," said Clarke, "because there is no reduction in the reticle. When we put down a film for an X-ray mask, a 0.1 µm particle ends up as a 1 µm image. We're going to see if there is some applicability for the really advanced projects, in particle reduction by using these targets."

The cloudy crystal ball

Most industry roadmaps predict further linewidth shrinking; however, some believe there will be more emphasis on flip-chip and system-on-a-chip configurations, which enhance performance without too much shrinking. The argument has been made that copper may be an unproductive pursuit just now, when there is so much that can be realized just through packaging and chip mounting. The consensus is that the move toward smaller geometries is certain to continue. It is just the rate and magnitude that are under question.

It is reasonable to assume that three years from now, 0.18-0.13 µm features will not be uncommon. But the push toward 0.13 µm may slow down as the emphasis on chip-mount technology grows. Very likely there will be work to expand the use of copper interconnects to include a lot more than just interconnect levels on processors.

PVD sputtering should remain strong. Indications are that the limits of what can be done have yet to be reached. There is 0.18 µm and 0.15 µm (150 nm) work being done for   SEMATECH, for the liner and seed layers, which show good performance. Some system manufacturers believe PVD fill will replace electroplated fill. They may be right, but copper fill is still difficult because copper flows too slowly.

All projections indicate that aluminum and titanium use will continue growing. These materials will not be replaced by copper and tantalum any time soon -- certainly not over the next two to three years. Growth will probably continue over the next decade if for no other reason that the overall market growth. As a piece of the pie, certainly copper and tantalum are going to get bigger.

More exotic materials, such as ruthenium, will probably be mainstream in five years. This is because semiconductors are beginning to hit limits in terms of capacitor as well as silicide materials. There is going to be a major shift from titanium silicide to things like cobalt silicide or nickel silicide.

Sputtering targets are not going to vanish from the semiconductor landscape any time soon. They will change and evolve to meet the technology's new demands, but will continue being a familiar linchpin in the production of increasingly complex devices.