Google
WWW
thinfilmmfg.com

27 June 2002

New wet clean systems improve old standby

Next to the engineering marvels responsible for ion implantation or optical lithography, the traditional wet bench looks crude. Fill a row of sinks with concentrated chemicals, dunk a batch of wafers in each one in succession, and you have captured the essence of a wet bench. Yet, despite many shortcomings and continual efforts to displace it, the core wet cleaning technology has been a staple of wafer processing for thirty years.

About 60% of wet cleaning steps take place during the front end transistor and contact formation process. Thirty-eight percent of all wet cleaning steps take place before thermal processes like oxidation. For thirty years, process engineers have depended on the "RCA clean," developed by Werner Kern, for these pre-thermal clean steps.

The RCA clean first uses an H2O-NH4OH-H2O2 solution (standard clean 1; SC1) to remove organic contaminants and particles. After rinsing with dilute HF, a second standard solution (SC2) uses a H2O-HCl-H2O2 mixture to remove metal contaminants. A final rinse prepares the wafer for further processing. Minor variations on this process are common. For example, dilute HF rinses before or after the clean are used to control the formation of silicon's native oxide.

The RCA clean has survived in part because it offers effective contaminant removal in an efficient package. A fully-loaded wet bench can process between 200 and 250 wafers/hour. The large batches and large chemical volumes involved are also responsible for many of the shortcomings of wet bench processing, however.

Back to top

Why not RCA Clean?

High purity cleaning chemicals are expensive. A sink large enough to allow consistent flow around 25 200- or 300-mm wafers requires a large volume of chemicals. To minimize cost, wet bench owners reuse the chemicals for several wafer batches. Doing so increases the risk of cross-contamination: particles from one batch can settle on the next one. Reusing chemicals also constrains the process flow. For example, rinse steps must take place in a separate sink, exposing the wafers to air during the transfer. Particles tend to accumulate at the liquid/air interface, depositing on wafers as they pass through the boundary.

Shrinking device feature sizes require more stringent cleaningspecifications. The traditional RCA clean removes particles by etching away part of the wafer surface, then washing away the particles and etched material together. Even as acceptable particle levels have fallen, so have surface roughness specifications. More stringent specifications reduce the amount of etching allowed. Often, attempts to reduce etching also degrade particle removal.

A third concern facing traditional wet benches is cycle time. A batch can take as much as an hour to make its way through the traditional RCA clean. This long cycle time reduces fab flexibility, delays yield learning, and increases WIP inventory levels. To remain cost efficient, the bench must operate close to full capacity. Any problems that reduce bench throughput will starve downstream processes of material. Smaller batches and single wafer processing reduce cycle time, but also reduce throughput and capital productivity.

The tradeoff between throughput and cycle time varies, depending on the needs of the fab. In general, fabs that run high volumes of a few products are less sensitive to cycle time than fabs that run small batches of many different products. Even high-volume, low-mix fabs may have priority lots or make frequent engineering changes, however.

Other areas of concern for wet bench owners arise from the design of the equipment. Most existing designs use mechanical orifices for flow control. Changing the flow rate, concentration, or chemical composition of a bath requires replacement of one or more orifices. Existing designs typically require large amounts of fab space, and may need complex and expensive facilitization.

Back to top

Rethinking cleaning

Wet bench designers and cleaning experts have proposed a variety of solutions to these problems. In some cleaning steps, extremely dilute chemicals can provide good process performance. Because dilute chemicals are less expensive, they can be replaced after every bath. Thus, these chemistries allow rinsing "in place" in the process sink.

The extensive rinsing needed by traditional RCA cleans is partly due to the highly concentrated chemicals used. Dilute chemicals rinse away more easily.

Megasonics agitation can help dislodge adhered particles, reducing the amount of etching and surface loss. Megasonics can also improve the effectiveness of dilute cleaning chemistries. By helping to distribute chemicals uniformly throughout the bath, megasonics can allow more uniform cleaning in a smaller sink.

Equipment manufacturers offer several alternatives to the traditional wet bench. One of the most successful, the batch spray system, uses spray nozzles to deliver the chemicals to the wafer. A dispense manifold allows precise control of dilution and rapid switching between chemistries. By spraying an entire batch of wafers at once, these systems maintain the throughput advantage of a batch system. By supplying the desired chemical concentration to the wafer surface directly, they achieve better process control.

Batch spray systems like FSI's Zeta account for about 17% of the total wet clean market. The batch spray approach has been particularly successful in post-ash strip and clean applications.

Still, the traditional immersion wet bench dominates the critical clean market, which accounts for the majority of wet clean steps. Two recent product introductions attempt to break the stranglehold.

Back to top

Oasis: Applied Materials enters cleaning market

The more radical of these, the Applied Materials Oasis system, relies on single wafer processing to minimize cycle time. Previous single wafer systems have not been able to match wet bench throughput or process performance. For this reason, single wafer cleaning has been confined to applications with fast process chemistries, and to single-sided applications like backside cleaning

According to Satheesh Kuppurao, technical marketing manager for wet cleaning at Applied Materials, the Oasis system achieves throughput in excess of 100 wafers per hour for pre-thermal cleans, while requiring only 1/3 as much space as a comparable wet bench.

The system achieves such high throughput by combining the SC1 chemistry with chelating agents for metal removal and surfactants for particle removal. The single step Oasis clean is able to reduce process time to about two minutes, Kuppurao said.

The Oasis represents Applied Materials' first foray into the highly fragmented wet clean market. It remains to be seen whether the segment will be receptive to such a radically different cleaning approach from a new entrant.

Back to top

Magellan: FSI builds on batch spray heritage

In contrast, FSI's new Magellan system attempts to combine the advantages of immersion cleaning systems with the company's successful batch spray technology. By using dilute chemistries, the Magellan system compresses an RCA-equivalent clean into just two process modules. The first applies dilute SC1 chemistry, followed by a rapid rinse in the same process sink. The second module uses HF/HCl chemistry for metal removal and surface control, followed by rinse and dry steps. According to Scott Becker, VP of product management for FSI's Surface Conditioning Division, several wet bench OEMs already incorporate FSI's wafer drying technology, which relies on isopropyl alcohol vapor (IPA) to create a surface tension gradient. Becker said cycle time for the two module process is only 28 minutes for a batch of 25 wafers, or only a little more than a minute per wafer.

The Magellan's chemical dispense manifold, proven in the company's spray cleaning tools, can monitor and control chemical concentration to account for temperature variations, water pressure changes, and other process fluctuations.

Back to top

What next for wet benches?

Furnaces, another bastion of large batch processing, have slowly given way to rapid thermal processing systems. RTP reduces both thermal budget and cycle time for dopant activation. Likewise, alternatives to conventional wet bench cleaning must show that they can both improve process capability and enhance fab efficiency.

Back to top

 

This site is Copyright ©2001-2002 by Thin Film Manufacturing. All Rights Reserved