Building Copperopolis

Treating wastes generated by copper electroplating tools

Bruce T. Maeda and Robert E. Woodworth, Microbar; and Ken Aitchison, Novellus Systems

A copper recovery system recovers up to 99.99% of total copper from spent plating baths and rinsewater from wafer-plating operations, transforming wastes into high-quality metal.

As next-generation semiconductor devices evolve from aluminum to copper-based metal interconnects, semiconductor fabs will require advanced waste-treatment solutions to address copper-bearing wastes. Several metal treatment technologies are used in the metal-finishing and printed wafer-board industries. However, commercialized waste-treatment equipment from these industries often does not meet the stringent safety, reliability, performance, and automation needs of the semiconductor manufacturing environment.

The integrated copper recovery system, which treats wastes from copper electroplating tools.

This article describes the design and performance of an integrated copper recovery system from Microbar (Sunnyvale, CA) that has been installed at a Novellus Systems facility to treat wastes from copper electroplating tools. The recovery process transforms corrosive, copper-laden, liquid plating wastes into high-quality copper metal, thereby reducing both the volume and toxicity of hazardous wastes that must be transported for off-site disposal. The copper recovery system also treats rinsewater and electrowinning acid residuals to <0.01 mg/L total copper, allowing end-users to meet current and future wastewater discharge limits throughout the United States, Europe, and Asia.

System Design and Safety Features

Copper electroplating tools commonly generate two types of wastes: spent copper sulfate plating baths and rinsewater from wafer-rinsing operations. The spent plating baths typically contain copper sulfate, sulfuric acid, and milligram-per-liter concentrations of proprietary organic and inorganic additives. Copper is present as cupric ions (Cu²⁺) at nominal concentrations of 17,000 mg/L.¹ The sulfuric acid content is typically about 15% by weight.

Plating rinsewater generally consists of deionized (DI) water with a residual plating solution from rinsing operations. This rinsewater contains low concentrations of dissolved cupric ions, typically between 10 and 100 mg/L, with an acidic pH of 2.0 or greater, depending on the volume of rinsewater used. To meet sewer discharge requirements, many facilities must reduce copper in wastewater effluents to below 1.0 mg/L, depending on the pretreatment requirements of the local publicly owned treatment works.

The copper recovery system installed at Novellus's Portland Technology Center in Oregon had to meet the high uptime and reliability requirements of a production environment and Novellus's strict environmental discharge limits for effluents (<0.1 mg/L total copper). It also had to minimize the cost of off-site disposal and liability for wastes and operate in an automated manner with a minimum of operator attention. In addition, to meet facility space requirements, the entire system had to be integrated onto a single skid with a footprint smaller than 9 x 6 ft.

To ensure that the treated effluent in this investigation consistently met copper discharge requirements, conventional copper treatment technologies were combined with an advanced control system, automated process sequencing, copper concentration monitoring, and safety features in accordance with SEMI S2-93 standards.² To meet SEMI's safety standards, the copper recovery system contains a number of features to eliminate single-point failures and minimize worker exposure to chemicals. First, the system is integrated onto a single skid with spill detection and a containment sump that can contain 110% of the greatest tank volume. The system also is surrounded by acid-resistant splash panels that are electronically interlocked or tool accessible to prevent unauthorized entry. Second, the copper recovery system is designed to operate automatically with a minimum of operator attention. Tank-filling, processing, and treatment operations are all automated using a programmable logic controller (PLC) with a color touch screen interface. The system features automated water and clean-dry-air (CDA) purge sequences to remove residual chemicals before maintenance from filter housings, pumps, ion-exchange beds, and piping. The system is also equipped with a manual mode that allows the operator to control all pumps and valves directly from the PLC touch screen without entering the chemical-processing area.

Copper Recovery Operations

As illustrated in Figure 1, the copper recovery system uses three unit operations to achieve its recovery and effluent treatment goals. First, an electrowinning unit recovers elemental copper from spent copper sulfate plating baths. Second, a two-stage neutralization unit combines treated electrowinning residuals and wafer rinsewater and then adjusts the rinsewater pH to optimum values. Third, a selective cation exchange unit removes dissolved cupric (Cu²⁺) ions to parts-per-billion levels before discharge to the facility's main wastewater treatment system.

Figure 1: Schematic of the copper recovery system showing the three unit operations used to achieve recovery and effluent treatment goals.

Figure 2: Schematic of the electrowinning system.

Operation No. 1: Electrowinning. Electrowinning is an electrochemical process that recovers solid copper from high-concentration plating bath solutions by means of electroplating. As shown in Figure 2, plating waste is first stored in a collection tank to permit uninterrupted waste collection from the plating tool during processing. Using an acid-compatible transfer pump, the electrowinning cell is automatically filled with 20 gal of waste solution. The electrowinning cell consists of a polypropylene tank with a set of titanium anodes and high-surface-area cathodes. An acid-compatible pump recirculates fluid in a closed loop, and a direct-current voltage is then placed across the anode/cathode pairs. The imposed voltage may be adjusted to increase the electrowinning current. The arrangement of cathodes and anodes within the electrolytic solution produces an electrochemical cell. When voltage is applied across the cell, electrolytic reactions occur in accordance with Faraday's law, which states that the reaction stoichiometry is directly proportional to the amount of electricity passed.³ Elemental copper is deposited onto a set of disposable high-surface-area cathodes based on electrolytic half reactions. The reduction of copper at the cathode occurs by means of the reaction:

The oxidation of water at the anode occurs by means of the reaction:

Since the plating reaction is electrolytically driven in an aqueous medium, hydrogen ions from the acidic solution can be reduced to hydrogen gas at the cathode.⁴ Cathodic evolution of hydrogen is of great importance in the electrodeposition of some metals, since the reaction competes for electrons at the cathode, depending on the applied cell voltage and plating cell conditions. The competing reduction of hydrogen ions at the cathode occurs by means of the reaction:

Electrowinning cell performance depends on three mass-transport processes: convection in the bulk solution, diffusion from the bulk solution to the electrode surface, and conversion of the adsorbed ion into metal, which is restricted by the limiting current, i^_L, for a particular metal and solution chemistry. The i^_L is defined as follows:

whereby n = the number of electrons involved in the reaction, F = Faraday's constant (96,500 C/mol of electrons), A = the area of the electrode, m = the mass transport and rate coefficient, and c^* = the bulk concentration of electrolyte.³

The optimal metal deposition rate can be controlled by maximizing cathode surface area, bulk mixing, and electrical current efficiency. To maximize copper removal, the recovery system uses reticulated cathodes, which provide a high surface area in a smaller package than conventional wire-mesh cathodes. The theoretical limit to copper removal is based on the cathode surface area and Faraday's law. A current efficiency of 100% implies that only the copper reduction reaction, not the hydrogen ion reduction, is taking place. For Cu²⁺ ions the maximum removal rate at 100% current efficiency is 1.186 g/Ah.

Figure 3: The absorbance spectrum of copper sulfate (CuSO^₄) and sulfuric acid (H^₂SO^₄) illustrating that cupric ions absorb light in the visible/near infrared spectrum of light.

To enable automated system operation and ensure consistent environmental performance, the copper recovery system contains an in-line copper sensor to provide real-time monitoring of cupric concentrations. The sensor consists of a dual-wavelength spectrophotometer that monitors the absorbance of light by the plating solution. Figure 3 illustrates that cupric ions absorb light in the visible/near infrared spectrum of light. By using a spectrophotometer that isolates a specific wavelength of visible/infrared light, the concentration of cupric ions can be quantitatively determined as a function of absorbance, according to the Beer-Lambert law.⁵ This law states that for a fixed optical path length and extinction coefficient, the absorbance of light is directly proportional to the molar concentration of cupric ions:

whereby A = absorbance, = extinction coefficient (L/mol-cm), c = concentration (mol/L), and b = path length (cm).

The direct application of the Beer-Lambert law is complicated by the fact that copper plating baths contain sulfuric acid to increase the conductivity of the solution. Sulfuric acid produces an absorbance spectrum that interferes with cupric ions. To correct for this background absorbance, the in-line copper sensor uses the ratio of copper sulfate and sulfuric acid absorbance as its key monitoring parameter. Using this approach, a more characteristic absorbance spectrum can be generated for cupric ions. As shown in Figure 4, the corrected maximum absorbance for cupric ions occurs in the infrared/visible light range at approximately 810 nm. Therefore, this wavelength of light was selected to establish cupric ion concentrations.

Figure 4: Corrected absorbance spectrum for copper sulfate (CuSO^₄) and sulfuric acid (H^₂SO^₄) showing that the corrected maximum absorbance for cupric ions occurs in the infrared/visible light range at about 810 nm, the wavelength selected to establish cupric ion concentrations.

The in-line copper monitor provides self-diagnostic capabilities for the electrowinning system and provides real-time monitoring of cupric ions at concentrations from 1 to 2000 mg/L based on a fixed optical path length for the copper sensor. By monitoring this concentration range, the electrowinning system can be programmed to operate based on a run time and concentration endpoint. This feature allows hands-free system operation and provides real-time confirmation that residual copper concentration targets are consistently being achieved.

Operation No. 2: Acid Neutralization. The acid neutralization unit combines the treated electrowinning acid residuals and rinsewater and adjusts pH to optimum levels for the cation exchange system. This strategy minimizes heat effects associated with the neutralization of electrowinning residuals, controls the ionic strength of the treated water, and ensures that all copper residuals receive secondary treatment.

The neutralization unit uses a continuous flow-through design that incorporates multiple neutralization stages and an attenuation/pump-out tank. Independent pH transmitters and a proportional integral derivative controller are used to monitor and control the pH during the neutralization process. A target pH of approximately 4.0 pH units was selected to provide the optimum condition for copper removal by the cation exchange system.

Operation No. 3: Selective Cation Exchange. Cation exchange is a treatment process that removes positively charged metal ions (cations) from solution by preferential capture on a polymeric resin that contains charged functional groups. As shown in Figure 5, the cation resin is first conditioned with monovalent hydrogen ions (H⁺). As metal-bearing water is passed through the resin, metal ions of higher valence (charge) are preferentially "exchanged" with the hydrogen ions. This process can reduce effluent metals concentrations to the low-parts-per-billion range, depending on the resin type, waste-stream composition, and pH conditions.

Figure 5: Schematic of the selective cation exchange resin used to reduce effluent metals concentrations to the low-parts-per-billion range.

A selective cation exchange resin is designed to preferentially remove copper from the neutralized wastewater streams. This resin enables copper removal in the presence of competing cations, such as sodium, calcium, potassium, and other metals. This cation exchange resin uses chelating iminodiacetic acid (IDA) groups that demonstrate a high selectivity for copper when conditioned with H⁺ ions.⁶ These IDA-based resins typically have a copper capacity of 1–2 lb per cubic foot of resin.

After neutralization, the wastewater passes through an activated carbon bed to provide filtration and general organic compound removal. In order to achieve maximum loading of the cation exchange resins, the leakage, or breakthrough, of residual metals must be properly controlled. The copper recovery system uses three resin beds in series to allow maximum loading of the primary resin bed and provide excess polishing capacity to achieve low-parts-per-billion levels.

System Performance Testing

The copper recovery system, installed in 1998, has undergone extensive qualification testing to verify system effectiveness and reliability. After initial start-up and commissioning, the system was placed in continuous service to process spent plating baths and wafer rinsewater from process-development and wafer-processing runs. The system demonstrated exceptional reliability, with uptime of more than 98%. The 2% downtime included scheduled maintenance (filter replacement, probe calibration, and cathode replacement). Unscheduled downtime was caused by the fouling of a pH probe by nonprocess materials and a defective pH controller.

To evaluate system performance, representative waste samples were collected from the inlet and outlet of the copper recovery system. These liquid samples were submitted under chain-of-custody record to an analytical laboratory approved by the Environmental Protection Agency (EPA). Samples were then analyzed for total copper using inductively coupled plasma (ICP) techniques (EPA method 200.7). Each of the three unit operations was characterized individually to ensure performance to specification and to identify operating windows. Since the electrowinning unit is a batch operation, it limits the throughput of the waste recovery system. Thus, most of the unit operation testing focused on key parameters to maximize electrowinning performance.

Following the characterization tests of the individual unit operations, the copper recovery system was operated as an integrated system during multiple 5000-wafer processing runs. These tests allowed the recovery system to operate for an extended period with wastes that are representative of both the composition and flow rates of the copper electroplating tools. Adjustments were made to subsystem operating parameters to increase overall treatment efficiency and reduce batch process time.

Results of Treating Spent Plating Baths

A typical electrowinning profile, including in-line copper sensor readings, is shown in Figure 6. Several samples of plating bath waste were taken from the electrowinning cell during the batch process period. These data were analyzed for total copper using ICP. A final sample was taken at the end of each run for laboratory confirmation. Electrowinning cell current, voltage, and copper concentration were monitored from the system PLC and logged by the Chem Manager data acquisition software from Microbar.

Figure 6: A typical electrowinning profile, including in-line copper sensor readings, showing a reduction in total copper versus time for the electrowinning system.

At the beginning of the electrowinning run shown in Figure 6, the voltage of the rectifier was repeatedly increased to maintain maximum rectifier current. After each voltage increase, the current dropped sharply. This is believed to be caused by a change in the conductivity of the plating bath solution. The reduction is evidenced by the oscillations in the current curve in the first 21/2 hours. An increase in applied voltage would maintain a higher current throughout the initial phase of the electrowinning run. The copper concentration data display a distinctive hockey-stick-shaped curve at the end of the run, denoting the transition of copper removal to a diffusion-limited process. As cupric concentrations in the electrolyte approached 1000 mg/L, both the current and copper removal rates dropped dramatically.

Total Copper before Electrowinning (mg/L)	Total Copper after Electrowinning (mg/L)	Electrowinning Period (hr)	Copper Recovery (%)
6,500	0.14	9.4	99.998
8,300	0.50	17.0	99.994
8,900	0.48	14.1	99.995
10,000	14.00	12.0	99.860
11,000	6.20	15.7	99.944
12,000	0.31	18.4	99.997
13,000	8.00	12.0	99.938
14,000	0.47	10.5	99.997

Table I: Typical data from the treatment of spent plating baths.

Analytical results from the treatment of spent plating baths are summarized in Table I. These results demonstrate that the system can recover up to 99.99% of copper from spent plating wastes and generate residuals with <1 mg/L total copper, given sufficient run time. Processing times for 20-gal batches of spent plating wastes ranged from 9 to 18 hours for initial cupric concentrations between 6500 and 14,000 mg/L total copper. The wide range of electrowinning periods was a function of influent copper concentration, end-point copper concentration, and initial rectifier current.

Fluid convection was earlier identified as a variable that can affect electrowinning efficiency. Most of the electrowinning runs during the test program had a recirculation flow rate of 4.5–5 gal/min. Two tests were performed at a recirculation flow rate of 10 gal/min, but they did not result in substantially higher copper removal rates. Higher fluid flow rates and velocities in the electrowinning cell should promote better mixing and enhance convection in the bulk solution. However, no positive impact was observed during testing. The key parameter that had the greatest impact on copper removal rate was rectifier current and voltage.

Rinsewater Treatment Results

The spent plating solution wastes and rinsewater were generated intermittently by the copper electroplating tool. During a representative 14-day period, three electrowinning batches and a total of 3100 gal of rinsewater were treated. Based on a set point of 3.0 for the first stage and 4.0 for the second, the acid neutralization system was able to control pH to average values of 2.99 and 3.91, respectively. The standard deviation of pH was 0.024 for the first stage and 0.035 for the second.

Acid Neutralization Tank Outlet (mg/L)	Cation Bed No. 1 Outlet (mg/L)	Cation Bed No. 2 Outlet (mg/L)	System Effluent (mg/L)
16	<0.010	—	<0.010
42	—	<0.010	<0.010
67	0.010	—	0.010
28	<0.010	<0.010	<0.010
68	<0.010	<0.010	<0.010
81	<0.010	<0.010	<0.010
120	<0.010	<0.010	<0.010
140	<0.010	<0.010	<0.010
180	0.028	<0.010	<0.010

Table II: Typical rinsewater and electrowinning residual treatment data showing total copper concentrations.

Analytical results from the treatment of rinsewater and electrowinning acid residuals are summarized in Table II. During the qualification period the cation exchange system was subjected to influent cupric concentrations ranging from <16 to 180 mg/L. These elevated cupric concentrations were generated by prematurely stopping an electrowinning run to generate higher endpoint concentrations to test the cation exchange system. The acid residuals from the electrowinning batch process were metered into the rinsewater wastes before proceeding to acid neutralization and cation exchange. Results demonstrate that the selective cation exchange system can consistently reduce copper in treated wafer rinsewater and in treated electrowinning residuals below 0.01 mg/L.

Figure 7: Data indicating a typical reduction in copper at various process points in the copper recovery system and the associated pH values.

A typical reduction in copper at various process points in the copper recovery system and the associated pH values are shown in Figure 7. Total copper levels are displayed on a log scale. During the test period the average plating bath waste varied from 6500 to 17,000 mg/L total copper. Acid residuals from the electrowinning process were mixed with rinsewater to produce inlet concentrations flowing to the first cation bed averaging 100 mg/L total copper. The effluent to the system drain was always below 0.01 mg/L total copper. During the cation exchange process the treated water experienced a slight reduction in pH because of the displacement of hydrogen ions from the cation exchange resins.

The cation exchange system operated without trouble throughout the entire test program and consistently exceeded the original performance goal of 0.1 mg/L total copper. Analytical results confirm that the system effluent did not exceed 0.01 mg/L total copper during any of the documented test runs. After the treatment of 22,221 gal of wastewater, the concentration at the outlet of the first cation-exchange bed was observed at 0.028 mg/L total copper. However, because of the multiple-bed design, the system effluent, sampled when it exited the last bed, remained below 0.01 mg/L.

Waste Generation and Closed-Loop Recycling Opportunities

The copper recovery system produces elemental copper (plated on removable cathodes) and a treated water stream that contains nondetectable amounts of total copper, typically below 0.01 mg/L. Because of the relative high purity of semiconductor plating solutions, it is anticipated that the elemental copper recovered by electrowinning will be a high-quality, processed scrap metal that can be used for copper-containing products.

From a mass-loading basis, virtually all of the influent copper is recovered as elemental copper via the electrowinning process. The amount of dissolved copper remaining in the electrowinning acid residuals and influent rinsewater is relatively low by comparison. With the current system design, no attempt has been made to recover the dissolved-phase copper removed by the cation exchange system, in part because of the domestic availability of off-site resin regeneration facilities that are permitted to handle copper wastes. However, in Asia or Europe, off-site regeneration facilities that specialize in copper-bearing wastes may not be available.

Depending on the number of copper electroplating tools installed, the installation of an on-site regeneration system could provide a cost-effective closed-loop solution for the recovery of all copper from electroplating wastes. Cation exchange resins can typically be regenerated using multiple-bed volumes of hydrochloric or sulfuric acid. The electrowinning acid residuals could be used to regenerate the beds. The regenerating acid would lower the bed pH to below the decomplexation pH of the chelating resin, release copper ions, and recondition the resin back to the hydrogen form. The copper-laden regenerant could then be directed back to the spent plating bath collection tank for further treatment by electrowinning.

Conclusion

The copper recovery system used at Novellus's Portland Technology Center recovers up to 99.99% of total copper from spent plating wastes generated from wafer-plating operations. It has processed more than 3000 gal of spent plating bath and more than 50,000 gal of wafer rinsewater from multiple electroplating tools. However, the throughput of the copper recovery system is limited by the electrowinning batch process time. In the course of the investigation it was determined that adjusting the rectifier voltage improves current efficiency and thus reduces processing time. Using real-time copper concentration monitoring of the electrowinning process also reduces run time, tracks cathode copper loading, and allows for automated processing sequences.

The capacity of the system could be increased by current control of the electrowinning power supply. The system could be further automated by the addition of real-time copper sensors to monitor influent concentrations and cation exchange bed breakthrough. By reusing the sulfuric acid from treated plating baths to regenerate the cation resin, closed-loop recovery of all copper could be achieved.

Acknowledgments

The authors would like to thank Elizabeth Ellinger, senior mechanical engineer, and Jay Jung, controls engineering manager, of Microbar (Sunnyvale, CA) and Scott Stoddard, facilities manager, at Novellus Systems (Portland, OR) for their invaluable assistance on this project.

References

1. Enthone-OMI, CuBath SC Technical Data Sheet (Concord, ON, Canada: Enthone-OMI, 1998).

2. Semiconductor Equipment and Materials International, SEMI S2-93A, Safety Guidelines for Semiconductor Manufacturing Equipment (Mountain View, CA: Semiconductor Equipment and Materials International, 1996).

3. AJ Bard and R Faulkner, Electrochemical Methods (New York: Wiley, 1980).

4. JO'M Bockris and AKN Reddy, Modern Electrochemistry, vol. 2 (New York: Plenum Press, 1970).

5. AS Wingrove and RL Caret, Organic Chemistry (New York: Harper & Row, 1981).

6. F Helfferich, Ion Exchange (New York: Dover, 1995).

Bruce T. Maeda is a senior chemical engineer in the R&D group at Microbar (Sunnyvale, CA), where he has been responsible for the design of automated bulk-delivery and waste collection systems, CMP wastewater treatment systems, and the company's integrated copper-waste treatment system. He has spent more than 13 years in the chemical recycling, soil and groundwater remediation, environmental consulting, and semiconductor manufacturing equipment industries. Maeda is a registered chemical engineer in California. He holds a BS in chemical engineering from the University of California, Berkeley. (Maeda can be reached at 408/541-1040 or bmaeda@microbar.com.)

Robert E. Woodworth is a senior mechanical engineer in the R&D group at Microbar where he is responsible for pump development, applications testing, and the design of liquid chemical delivery systems. He has worked for more than 10 years in R&D for the energy, environmental, and semiconductor industries. He received his BS in mechanical engineering from San Jose State University. (Woodworth can be reached at 408/541-1040 or rwoodworth@microbar.com.)

Ken Aitchison, PhD, is the senior director of systems, environmental, safety, and health at Novellus Systems (San Jose), where he is responsible for the life cycle management of chemical processes and products used in semiconductor process equipment. He has more than 20 years of experience in the field of materials processing chemistry including CVD, sputtering, sol-gel, and electrodeposition. He received a BA in chemistry from New York University (New York City) and a PhD in chemistry from the University of London. (Aitchison can be reached at 408/324-3905 or ken.aitchison@novellus.com.)