The Environment, Health and Safety Side of Copper Metalization

		At a Glance
	The general EHS concerns of technology development are new chemistries, tools and waste streams. With copper, preliminary effluent treatability studies have shown that a combination of oxidizer removal, micro- or ultrafiltration, and cation exchange can achieve high levels of copper removal. With some additional steps, water recycle from CMP effluent is also possible.

It is widely known in the semiconductor industry that copper will replace tungsten and aluminum metalization for semiconductor devices.¹ Motorola has developed a high-performance logic technology at its state-of-the-art development fab, the Advanced Products Research and Development Laboratory (APRDL), which incorporates six levels of planarized copper interconnects.² During this development, the APRDL environmental group, the group responsible for identifying environment, health and safety (EHS) issues with new technologies, and the site EHS group investigated the potential impacts of bringing copper into Motorola facilities, such as APRDL, for process development and transfer to manufacturing. All potential copper metalization processes were thoroughly evaluated for their EHS impacts. ³

The general EHS concerns of technology development are new chem-istries, tools and waste streams. Any new regulatory issues must also be identified. All materials used in manufacturing must be listed on the U.S. Environmental Protection Agency's (USEPA) Toxic Substance Control Act (TSCA) inventory or have a low volume exemption granted to the supplier (with this exemption, the total amount manufactured cannot exceed 10,000 kg per year for all suppliers collectively). Material hazards, such as toxicity and reactivity, should be identified. New tools should have a complete safety review, such as a third-party risk assessment following the SEMI S2 guidelines. All waste streams must be identified to ensure that they can be accommodated by the site.

Copper metalization, as illustrated in Figure 1, brings four new technologies to a fab: barrier deposition, copper seed and/or fill, copper chemical-mechanical polishing (CMP) and copper wafer reclaim.

Barrier deposition

Since copper readily migrates into the interlevel dielectrics (ILDs -- the insulating layers between metal layers) and thus can easily travel to the transistor level and render the device nonfunctional, a diffusion barrier must be deposited between the metal and ILD layers. Barrier materials, such as titanium nitride, tantalum nitride and tantalum silicon nitride, can be deposited by chemical vapor deposition (CVD) or physical vapor deposition (PVD). For CVD, the TSCA status of the precursors must be determined as well as any material hazards. Since many of the precursors are relatively unknown, industrial hygiene monitoring for byproducts may be conducted during chamber cleaning to determine if there are any worker exposure hazards. PVD processes use solid sources and do not involve any special EHS considerations.

Fig. 1. Copper metalization brings four new technologies to a fab: barrier deposition, copper seed and/or fill, copper CMP and copper wafer reclaim.

Copper seed/fill

Once the barrier layer is deposited, the structures (lines and vias) are filled with copper. Four methods of copper deposition were identified: CVD, PVD, electroplating and electroless plating. If electroplating is used, a copper seed layer must first be deposited, thus involving a combination of these methods.

As with the barrier layer, PVD copper is not a major EHS concern. The primary concern with copper CVD is the reaction chemistry. The same consideration of the precursors and worker exposure used in barrier deposition is warranted. Copper electroplating, or electrolytic plating, is a new technology for the semiconductor industry, involving new chemistries, tools and wastes. Commercially available electroplating chemistries used by the printed wiring board (PWB) industry were evaluated and found to have no particular hazards or worker exposure issues; all materials are TSCA listed. A tool safety review is recommended. The primary concern is the two waste streams generated by the process: a rinsewater and a concentrated bath.

Fig. 2. This model predicts the copper concentration in copper-bearing waste streams based on six levels of metal.

Electroless copper plating is also new for the industry. Again, commercially available plating chemistries were evaluated. Most are formaldehyde-based, which is a carcinogen and a hazardous air pollutant. The Occupational Health and Safety Administration (OSHA) has also set a formaldehyde standard to protect workers from exposure. If an electroless plating process is brought into a fab, initial work space monitoring and emissions testing would be conducted to ensure that respective formaldehyde limits were not exceeded or even approached. Electroless plating, like electrolytic plating, generates a rinsewater and a concentrated bath.

Copper CMP

Copper CMP is a critical process step since no copper etch technology currently exists. Motorola has developed its own copper CMP slurry that is environmentally friendly, having a neutral pH, dilute chemistry and no toxic components. All constituents are TSCA listed. The primary concern is the process effluent, containing all the copper removed from the wafer.

Wafer reclaim

Development and manufacturing fabs generate a large number of test wafers that are reclaimed. When these have a copper film, they cannot be reclaimed using existing processes. Motorola has chosen an ammonium persulfate/sulfuric acid chemistry (used for electroplating anode preparation) for reclaim over other options, such as nitric acid and piranha chemistry (sulfuric acid/hydrogen peroxide). This chemistry is not only more environmentally friendly, but it is also compatible with electroplating waste. Tests on copper removal efficiency were conducted to verify that the chemistry is viable. This process, like plating, generates a concentrated waste and a rinsewater.

Copper waste handling

A model, illustrated in Figure 2, was developed to predict the copper concentration in copper-bearing waste streams based on six levels of metal and the full capacity of the development and manufacturing fabs. As expected, the model predicts that concentrated copper wastes cannot be discharged to the industrial wastewater (IW) treatment system without prior collection and treatment. Rinsewaters from plating operations, wafer reclaim and copper CMP effluent contain relatively low levels of copper and contribute to only minor increases in the total copper concentration of the IW discharge. Although no treatment is necessary for these dilute streams at this time, copper CMP contributes great-er than 80% of the discharged copper on a mass basis and is a likely first candidate if these streams require treatment in the future.

Concentrates

Electrowinning was chosen as the preferred treatment me-thod for the concentrated copper-bearing wastes since it is cost-effective and does not generate other waste streams (e.g., sludge). The process is very similar to the electrolytic deposition accomplished in the production area -- copper is reduced and plated out of solution at the cathode, and water is oxidized (split) to form oxygen at the anode. Because the primary waste streams are themselves plating solutions, and because copper has a relatively high reduction potential, electrowinning is ideally suited for this application.

Table 1. Copper Concentrations Before and After Electrowinning
Bath	Untreated	Treated	Reduction
Reclaim	7,200 ppm	17 ppm	99.8%
Electrolytic	14,300 ppm	7 ppm	99.9%
Electroless	2,259 ppm	30 ppm	98.7%
1:1:1 mix	13,500 ppm	4 ppm	99.9%

To verify treatability of the concentrated waste streams, pilot scale electrowinning tests were performed on commercially available electrolytic and electroless plating solutions and on copper-containing waste from wafer reclaim tests. For a treatment system to be robust enough to meet the needs of both a development and a manufacturing fab, however, it must be able to effectively handle mixtures of these wastes as well as other similar concentrated copper streams. Thus, a 1:1:1 mixture of these wastes was also tested to determine if mixture incompatibility would have any negative impact on collection or treatment. Table 1 gives the initial electrowinning tests results; all waste streams, individually and in a 1:1:1 ratio mixture, can be effectively treated through electrowinning.

Prior to treatment, the mixture exhibited a relatively small temperature change from acid/base reactions and precipitation of a chelating agent. Subsequent treatment of the mixture effected the dissolution or destruction of the precipitate, likely from oxidation at the anode or by dissolved oxygen generated at the anode.

Fig. 3. The change in copper concentration over time during electrowinning treatment is shown.

Fig, 4. Copper concentration drops to zero in the first 20 hours.

Figures 3 and 4 illustrate the copper concentration change over time in the electrowinning process. The plot for electroless plating bath alone is not shown because autocatalytic copper deposition (at ambient temperatures) was nearly complete before the second data point was collected.

Using the results of the compatibility and treatability tests and the modeled maximum volumes of concentrated copper wastes from both fabs, a concentrate collection system and an electrowinning treatment system were de-signed and installed. The system can handle multiple copper-containing chemistries from both the development and the manufacturing fabs, as long as there are no mixture compatibility issues beyond the chelator dropout observed in the pilot studies. The relatively simple infrastructure consists of a segregated PVDF copper concentrates gravity drain line that feeds a large collection tank. Once collected, the copper concentrates can be pumped to an on-site batch electrowinning treatment system or to a tanker truck for off-site reclamation or disposal.

Motorola can currently accommodate all copper metalization processes in development and manufacturing fabs. The calculated copper discharge concentration in the IW stream for full capacity at six levels of metal is well below the current POTW concentration based discharge limit. However, if either the local POTW copper discharge limits (mass or concentration) or the USEPA regulations for the semiconductor industry change, some modifications may have to be made. Each POTW sets copper concentration or mass based discharge limits for the industry based on the stream standards to which it discharges and its ability to remove copper in its treatment processes. If stream standards change or the copper load to the POTW inhibits copper removal or other treatment processes, industrial discharge limits may be lowered. There is currently no regulation for copper discharge for the semiconductor industry categorical standard (40 CFR 469). If one is promulgated, or if these copper processes are regulated under other categorical standards such as the metal finishing standard (40 CFR 433), then further copper removal methods may be required. To prepare for potential POTW discharge limit changes and future regulations, Motorola has extended its copper wastes treatability study to include treatment of copper CMP effluent, since it contributes more than 80% of the copper to the wastewater stream.

CMP effluent

To find the optimum method for CMP effluent treatment, the effluent was first characterized to determine copper distribution and other physical properties that may enhance or interfere with potential treatment schemes. After characterization, various treatment methods were tested to ascertain copper removal performance, treatment system requirements and cost. Samples of Motorola's CMP effluent were submitted to several waste treatment suppliers for the industry to obtain a broad range of options. The criteria for 'best' treatment technology may vary from company to company based on local requirements. Motorola's criteria are the following:

• Design primarily for copper removal -- water recycling is only a secondary consideration.
• Do not generate any other waste streams (sludges) that require subsequent disposal.
• Minimize or eliminate steps that require chemical addition or residence/reaction time.
• Minimize space requirements so that the treatment system can be integrated in the polishing tool, if possible, or placed in the subfab if necessary.

Table 2. Characteristics of Copper CMP Effluent
Parameter	Typical range
Total copper	2.5-5.0 ppm
Dissolved copper(<0.2 µm)	2.5-5.0 ppm
Total organic carbon	50-100 ppm
Total suspended solids(>0.2µm)	350-700 ppm
Oxidation/reduction potential	250-500 mV
pH	7.0-7.5

Based on Motorola's internal CMP effluent analysis and information from the suppliers, the copper in the CMP effluent is almost completely soluble and is not associated with the abrasive. Table 2 summarizes results from the characterization work.

Given this analysis, most of the treatability work has been focused on various combinations of chemical addition, micro- and/or ultrafiltration, reverse osmosis and ion exchange. A typical treatment scheme meeting most of Motorola's requirements (remove the copper and discharge the nonhazardous abrasive solids to the site's industrial wastewater system and ultimately to the POTW) includes the sequence of operations shown in Figure 5. In this scheme, the oxidizer removal may be accomplished by any of several mechanisms, some of which require chemical addition; selection of the mechanism for this step is primarily based on treatment system location and capital/operating cost differences. The cation exchange canisters require periodic replacement and on-site or off-site regeneration. Fortunately, the collection and treatment infrastructure installed for concentrates can be modified to easily accommodate on-site regeneration of the canisters and copper recovery by electrowinning.

Fig. 5. A typical treatment scheme used to remove copper from the CMP effluent, ultimately discharging nonhazardous abrasive solids into the wastewater system.

This treatment scheme can achieve very high levels of copper removal from the relatively dilute CMP effluent. Pilot scale tests indicate copper removal well in excess of 95% can be achieved. Approximate capital equipment costs for this type of system are on the order of $250,000 for a nominal 10 gpm system and $500,000 for a nominal 100 gpm system. Depending on the oxidizer removal mechanism selected,this treatment scheme may have a relatively small footprint and be capable of subfab placement if necessary.

If water recycle is desired, some modifications to this system are necessary. TOC removal must be included, and the cation exchange capability must be expanded to include anion exchange. Several suppliers are working on derivatives of this scheme to eliminate all chemical addition steps and improve the quality of the water available for recycle if required. Development work in this area will continue with more design details and performance disclosure expected before the end of 1998.

Conclusion

As the industry migrates toward copper metalization, it is important to address the environment, health and safety impacts before processes are transferred to manufacturing to reduce the risk of delays resulting from EHS issues. Motorola has thoroughly evaluated the copper metalization processes and has developed a strategy for handling all waste streams and protecting employees and the environment. To handle concentrated copper wastes, an electrowinning system has been designed and installed that is robust enough to effectively treat concentrates from both the development and the manufacturing fabs. Dilute streams, such as rinsewaters, do not require treatment at this time. If discharge limits change or future regulations on copper for the semiconductor industry are promulgated, copper CMP effluent will likely become a candidate for treatment.

Preliminary effluent treatability studies have shown that a combination of oxidizer removal, micro- or ultrafiltration, and cation exchange can achieve high levels of copper removal. With some additional steps, water recycle from CMP effluent is also possible.

Acknowledgments

The authors would like to acknowledge James Legg, Cindy Simpson and Brian Raley for their significant contributions to the information and results presented in this article. The authors would also like to acknowledge Enchem, Kinetico, Nalco, Pall Corp. and US Filters for their work in CMP effluent characterization and treatability studies.

References

1. National Technology Roadmap for Semiconductors, 1997 edition, SIA.

2. S. Ventkatesan, et al., IEDM 97, p. 769.

3. L. Mendicino, et al., Advanced Metalization and Interconnect Systems for ULSI Applications in 1997, p. 549.