Advertiser and

Buyer's Guide
Buyers Guide

Chip Shots blog

Greatest Hits of 2005
Greatest Hits of 2005

Featured Series
Featured Series

Web Sightings

Media Kit

Comments? Suggestions? Send us your feedback.

Wet Surface Technologies

Performing advanced post-CMP cleans to reduce copper defectivity and surface roughness

Todd Buley, Yakov Epshteyn, and Mike Kulus, Rohm and Haas Electronic Materials CMP Technologies; and Cuong Tran, Kyle Bartosh, Darryl Peters, and Chris Watts, ATMI Surface Preparation Products

The effectiveness of the copper chemical-mechanical planarization (CMP) process can be improved by establishing an efficient and complete post-CMP cleaning process that reduces the roughness of the polished wafer surface and leaves the polished surface defect- and contamination-free. The cleaning process starts immediately after the barrier polish process is performed on the last platen of the CMP tool. It continues, and typically concludes, in a scrubber module. The chemistry used in these process steps can both reduce the roughness of the polished wafer surface and have a major impact on defectivity levels. This article focuses on studies that were performed to determine the optimal cleaning-chemical treatments for reducing surface roughness and defect levels.

In order to minimize slurry particles, organic contaminants, and overall polished-surface defect counts to acceptable levels, various cleaning processes have been developed for use on the last platen in the CMP tool and the mechanical brush scrubber. This article investigates cleaning processes that use three different mechanisms to remove contamination from copper surfaces and discusses their compatibility with specific barrier slurry systems. It also probes the impact of alkaline post-CMP cleaning formulations on cleaning efficacy. In addition to studying defect-removal mechanisms, the article also highlights the impact of various processes on copper-surface roughness and scratches.

Post-CMP Cleaning Mechanisms

Most barrier slurries contain benzotriazole (BTA), which provides copper corrosion protection during and after the polishing process. In addition, some processes use a BTA rinse after the barrier polish to protect the copper surface while it is being transferred from the barrier polishing platen to the cleaning system. However, if the resulting BTA-copper complex is not removed effectively, high levels of organic residues can form on the copper surface, severely reducing yields.

To remove the BTA-copper complexes and other organic and inorganic defects, three different cleaning mechanisms have been used. The undercut-and-lift-off approach utilizes chemicals that dissolve the copper oxides in the BTA-copper complexes and in the underlying oxides, effectively lifting off defects. The dissolution approach penetrates, swells, and dissolves the organic film. The displacement approach relies on a copper complexing agent to displace the BTA on the copper oxide surface following barrier removal on the last platen. The resulting organocopper complex is more readily removable in subsequent cleaning steps than BTA.

First-Generation Cleaners

First-generation post-CMP cleaners used citric or oxalic acid solutions. Because of their low pH, these compounds rapidly dissolve both CuO (the outermost oxide) and Cu2O (the underlying oxide), undercutting particles or organic defects in the oxide. If the zeta potential of the substrate and particles is the same sign and nearly equivalent, the particles will not be redeposited. However, for typical substrates and typical slurry abrasive particles, that equivalence is difficult to achieve at a low pH (e.g., approximately 4).1 Consequently, first-generation approaches required the use of a mechanical agent such as megasonic agitation or brushes to assist with particle removal.

The removal of passivating oxides using citric acid solutions yielded a very reactive copper surface that sometimes required passivation with BTA to avoid corrosion while the wafers awaited the next process step. To achieve appropriate adhesion and contact resistance and also to prevent potential delamination resulting from the decomposition of buried BTA films after thermal cycling, the BTA film had to be removed before the etch stop or barrier film was applied. First-generation cleaners also provided poor particle-removal efficiency, generated water marks, and caused handling issues because of bacterial growth.

Second-Generation Cleaners

Second-generation post-CMP cleaners were developed to address the weaknesses of simple first-generation organic acid cleaners. These formulations include ESC784 and ESC794 from ATMI (Allentown, PA), alkaline cleaners that dissolve CuO while leaving the Cu2O intact. Figure 1 presents copper oxide etch-rate data for ESC784 that were generated using an electrochemical technique at 23°C.2 The data were obtained by oxidizing a cleaned copper electrode in a dilute cleaning solution and then reducing the oxides to determine the amount of oxide remaining as a function of the time delay between oxidation and reduction (immersion time). The thickness of the oxide remaining over time can be calculated using Faraday’s equation, the magnitude of the reducing current, the length of time the reducing current was applied, and some molar constants.2 As Figure 1 demonstrates, delay time (immersion time), has no impact on the Cu2O thickness; it affects only the CuO thickness.

Figure 1: Copper oxide etch rates for ESC784 cleaning solution at 23°C.

Figure 2 presents copper oxide etch-rate data for ESC794. Similar to ESC784, this cleaner was electrochemically generated at 23°C. Dissolution of the CuO layer resulted in particle removal through undercut and lift off without removing all of the oxide on the copper surface. ESC794’s high pH results in a large negative zeta potential for most substrates and particles, ensuring repulsion and eliminating particle redeposition.1 It also requires less mechanical action than citric acid solutions to achieve high particle-removal efficiency. In addition, the remaining Cu2O passivates the exposed copper, resulting in a less-reactive surface and minimizing corrosion between barrier CMP and post-CMP cleaning steps. The inclusion of reducing agents reduces the copper etch rate. As shown in Figures 1 and 2, second-generation post-CMP cleaners have a controlled, highly selective etch rate for CuO.

Figure 2: Copper oxide etch rates for ESC794 cleaning solutions at 23°C.

Both ESC784 and ESC794 have been shown to eliminate water marks on hydrophobic organosilicate glass (OSG) dielectrics.3, 4 Several hypotheses attempt to explain how water marks form and how to eliminate them. Water marks seen in front-end-of-line processing are likely different from those in back-end-of-line processing because of the different materials exposed and different cleaning chemicals used. It is not clear whether the undercut mechanism also influences the formation of water marks.

Third-Generation Cleaners

In post-CMP cleaning, a balance must be struck between particle and organic defect removal and surface roughness. Both first- and second-generation post-CMP cleaners use the undercut and lift-off cleaning mechanism, which relies on surface etching to undercut residues and lift them off. Surface etching can result in grain-boundary decoration or selective etching of specific copper grain structures, causing an increase in surface roughness. If the surface roughness is too high, wafers will fail optical defect inspection. However, if the acceptable haze level is set too high on the optical inspection tool to accommodate greater surface roughness after cleaning, killer defects may be missed during inspection, resulting in yield loss.

To reduce surface decoration and problems associated with increased surface roughness, third-generation post-CMP cleaners incorporating water-soluble organic solvents were developed. Using a dissolution mechanism to remove organic residues, these solutions control and reduce the surface etch rate, thus decreasing surface roughness while undercutting and lifting off abrasive particles and some organic defects. Because of their alkaline character, they have a large negative zeta potential, resulting in excellent particle-removal efficiency.

ESC797D, ESC797G, SP50, and SP50(A) from ATMI are third-generation alkaline post-CMP cleaning solutions that dissolve organic defects and BTA-copper complexes. The solutions’ dissolution mechanism involves penetration, swelling, and dissolution using water-soluble organic solvents that can wet typical abrasive particles. The attractive forces between the solvents and the abrasive particles can exceed those between the substrate and the abrasive particles, resulting in the removal of particles from the wafer surface. Since the organic solvents are miscible in nearly all concentrations of water, organic defects remain in solution even after the cleaners have been greatly diluted. In addition to reducing copper surface roughness and removing organic defects, the cleaners eliminate water marks.3 Their high pH results in a large negative zeta potential, eliminating redeposition. Figure 3 presents electrochemically derived copper oxide etch rates for ESC797D used at 23°C. Although the solution did not reduce the CuO dissolution rate to zero, it reduced it enough to minimize undercutting of slurry particles.

Figure 3: Copper oxide etch rates for ESC797D cleaning solution at 23°C.

To further reduce surface roughness, a cleaning mechanism was needed that does not rely on surface etching. Hence, a fourth-generation formulation, ATMI SP28, was developed. Based on the displacement mechanism, this alkaline post-CMP cleaner contains a copper complexing agent known as CA 1. Because this method involves virtually no surface etching, the surface roughness resulting from the CMP process remains unaltered after cleaning. The exposed copper surface is passivated so that there is no need to apply BTA after cleaning to prevent corrosion. The CA 1–copper complexing agent is easier to remove than BTA and provides equivalent corrosion protection.

Figure 4 presents electrochemical data indicating the displacement of CA 1 by BTA. A copper electrode was oxidized using H2O2, producing a layer of Cu2O and CuO with a controlled thickness. The oxidized copper electrode was then passivated in an alkaline cleaning solution containing CA 1. The electrode was rinsed and then placed into an alkaline cleaning solution containing BTA. Monitoring showed that the open-circuit potential (OCP) changed from about –0.3 V (for CA 1) to about –0.2 V (for BTA), indicating that BTA displaced CA 1. The same approach can be used to displace BTA with CA 1. The higher concentration of CA 1 in SP28 drives the displacement reaction, although the formation constant for the CA 1–copper complex is less than that for the BTA-copper complex. While the displacement mechanism is rapid and does not require mechanical agitation to achieve a very high particle-removal efficiency, the addition of mechanical agitation can facilitate displacement and enhance cleaning.

Figure 4: CA 1 displacement by BTA, represented by an increase in open-circuit potential.

Like its second- and third-generation predecessors, SP28 is an alkaline cleaner that contains water-soluble organic solvents that solubilize organic defects, a process that can be thought of as “detergency.” Electrochemical tests involving SP28, which were similar to those used to produce the data in Figures 1–3, indicated that after a copper electrode was oxidized in SP28, there was no CuO left to dissolve. But since the Cu2O had complexed with CA 1, it was not as easily reduced (or oxidized). Similar results were observed for copper oxides after BTA (instead of CA 1) was added.

In addition to being used as a platen-three buff-and-rinse solution, SP28 can be used in megasonics tanks, in CMP spray stations, and in the brushes used in CMP cleaning tools. When used for buff treatment immediately after barrier removal, the solution eases wafer cleaning and enables users to employ less-aggressive cleaning processes (i.e., shorter process times and/or a more-dilute chemistry). These advantages can reduce copper-surface roughness and the cost of ownership.

Polishing and Cleaning Experiments

Experimental Setup. The efficacy of the cleaners was evaluated in a CMP and postpolish cleaning study. Polishing was performed using a Mirra CMP tool from Applied Materials (Santa Clara, CA), while post-CMP cleaning was performed using a Synergy scrubber from Lam Research (Fremont, CA). Wafer processing and metrology were performed in a Class 10 cleanroom environment.

The polishing process used a three-platen configuration for copper clearing and barrier removal. Platen one was configured for copper planarization and bulk removal, platen two was configured for copper clearing utilizing a soft-landing approach, and platen three was configured for barrier clearing. The consumables used for copper and barrier removal (CUP4410 pads and EPL2362 slurry for copper removal, Politex Prima pads and LK393-C4 slurry for barrier removal) were from Rohm and Haas Electronic Materials (Phoenix). Soft Politex Prima pads are better than hard IC1010 pads because they result in notably lower scratch and chatter mark counts on the copper surface.5 However, polishing with Politex Prima pads typically results in high organic-containing particle counts because the surface of the pad cannot be regenerated using diamond conditioning. Consequently, determining the proper cleaning solution for the copper polishing process used in this experiment was especially important.

The study used 15,000-Å electroplated blanket copper wafers, 10,000-Å blanket Coral wafers, 15,000-Å TEOS wafers, and patterned 854 uncapped CDO wafers. The processed wafers were inspected using an Applied Materials Compass 300 defect inspection tool to obtain light-point defect (LPD) counts, while an Applied Materials SEM Vision G2 was used for automatic defect review and classification. A 5-mm edge exclusion with a 0.20-µm detection threshold for copper wafers and 0.3-µm detection threshold for CDO wafers was used to make the LPD measurements. Roughened-surface counts from the SEM Vision G2 were used to understand the effects of exposing the copper surface to the various cleaning chemistries.

Experimental Results Involving Second- and Third-Generation Cleaners

Figure 5 illustrates the ability of ESC784, SP50, and SP50A solutions to remove organic by-products from the surfaces of polished copper blanket wafers. Experiments using all three cleaning products demonstrated stable and repeatable low defect counts on all blanket copper wafers, TEOS wafers, and 854 patterned wafers. Neither excessive surface roughening nor corrosion defects were observed on the 854 wafers when the SP50 and SP50A cleaning solutions were used. While very small numbers of shallow dents were observed on all copper wafers when the SP50A cleaning solution was employed, no other types of defects were found.

Figure 5: Numbers and types of defects on copper wafers that were cleaned using three different chemistries.

Based on these results, the investigators concluded that SP50A cleaning performance was promising and needed to be confirmed by establishing baselines to verify short- and long-term cleaning process stability.

Another test was performed to determine the ability of ESC 797D to remove organic by-products while preventing water-mark formation and copper-surface roughening. Figure 6 illustrates the ability of this cleaner to remove and dissolve organic by-products and other organic residues while virtually eliminating copper-surface roughening and water marks on patterned uncapped CDO wafers. The data in Figure 7 demonstrate the ability of ESC 797D to remove defects from polished blanket TEOS wafers. The performance of this cleaning chemistry was confirmed by means of baselines that verified short- and long-term cleaning process stability, and a statistically significant database has been generated.

Figure 6: Organic by-product counts and surface roughening on polished blanket copper wafers cleaned using ESC797 versus ESC797D. The SEM images show corresponding defects and surface roughening.
Figure 7: Particle, organic by-product, and nonvisible defect counts on polished blanket TEOS wafers cleaned using ESC797 versus ESC797D. The SEM images show corresponding defects.

Experimental Results Involving Fourth-Generation Cleaners

Cleaning performance improved further when the buff polishing process was performed on a Synergy cleaner using LK393-CR slurry, Politex Prima pads, and combinations of ESC784 and SP28 cleaning solutions. In the default (control) test, cleaning was performed using ESC784 chemistry alone. The total chemistry cleaning time using the tool’s two brushes was 80 seconds. In addition, four other tests were performed using the following processes:

• Buffing with SP28 for 15 seconds followed by ESC784 cleaning for 80 seconds (C4-SP28-784).

• Buffing with SP28 for 15 seconds followed by ESC784 cleaning for 30 seconds (C4-SP28-784 [30 seconds]).

• Buffing with SP28 for 15 seconds followed by cleaning with DI water only (C4-SP28-DI).

• Rinsing for 10 seconds followed by buffing with SP28 for 15 seconds followed by cleaning with DI water only (C4-rinse-SP28-DI).

The results of using the various cleaning combinations are presented in Figures 8 and 9. As shown in Figure 8, stable and repeatable low-defect-count performance on both copper blanket wafers and patterned 854 wafers was achieved using the SP28 cleaning solution for buffing and ESC784 for cleaning at 80 and 30 seconds. Moreover, defect characterization demonstrated that more than 50% of the defects that were detected because of the oversensitivity of the metrology recipe on the Compass 300 were in fact nonvisible.

Figure 8: Defectivity levels as a function of different buffing and cleaning processes: (a) the default (control) recipe with ESC784 chemistry alone, (b) C4-SP28-784, (c) C4-SP28-784 (30 sec), (d) C4-SP28-DI, and (e) C4-rinse-SP28-DI.

As illustrated in the images in Figure 9, the use of the SP28 cleaning solution for buffing and ESC784 for cleaning at 30 seconds reduced copper-surface roughening significantly on patterned 854 wafers. While the use of DI water as a cleaning agent further reduced copper-surface roughening, the processes involving DI water resulted in unacceptably low cleaning efficiency.

Figure 9: Images from the five cleaning processes illustrated in Figure 8: (a) the default (control) recipe with ESC784 chemistry alone, (b) C4-SP28-784, (c) C4-SP28-784 (30 sec), (d) C4-SP28-DI, and (e) C4-rinse-SP28-DI.

Because all of these experiments were limited, more studies must be performed to further optimize buffing and cleaning recipes to establish a statistically significant database.


Post-CMP cleaning solutions can have a large effect on wafer defectivity levels. This article has shown that optimized cleaning solutions affect the entire post-copper CMP cleaning process and decrease organic contaminant counts. In addition, the desired level of surface roughness can be tuned by selecting different cleaning chemistries and processes.

The article has demonstrated that the use of SP50A or SP28 cleaning solutions in the buffing process, in conjunction with ESC784 cleaner, resulted in significantly lower defect counts (including organic by-products and residues) than other materials. These solutions also prevented excessive copper surface roughening. The use of ESC 797D cleaning chemistry further enhanced defect results by improving organic by-product removal and reducing water-mark formation.


1. JG Park and AA Busnaina, “Adhesion and Removal of Silica and Alumina Slurry Particles during the Cu CMP Process,” in Proceedings of the Electrochemical Society, vol. 2003-26 (Pennington, NJ: Electrochemical Society, 2004), 312–315.

2. M Hughes, S Naghshineh, and D Peters, “Open Circuit and Galvanostatic Behavior of Copper Oxidized and Reduced in Various Solutions,” in Proceedings of the Electrochemical Society, vol. 2003-26 (Pennington, NJ: Electrochemical Society, 2004), 421–426.

3. K Bartosh et al., “Low Carbon Contamination and Water Mark Free Post-CMP Cleaning of Hydrophobic OSG Dielectrics,” in Proceedings of the Electrochemical Society, vol. 2003-26 (Pennington, NJ: Electrochemical Society, 2004), 305–311.

4. DW Peters et al., “Eliminating Carbon and Water Marks during Post-CMP Cleaning,” Solid State Technology 47, no. 11 (2004): 47–50.

5. Y Epshteyn et al., “Optimizing the Cleaning Processes after Cu CMP,” in Proceedings of the CMP-MIC Conference (Tampa, FL: IMIC, 2004), 323–326.

Todd Buley is engineering manager for the copper applications and defectivity group at Rohm and Haas Electronic Materials CMP Technologies (Phoenix), where he has been for six years. He has spent 16 years in the semiconductor industry, holding positions at Micron Semiconductor and Motorola. He received a BS in electrical engineering from DeVry Institute of Technology in Columbus, OH. (Buley can be reached at 602/470-4421 or

Yakov Epshteyn, PhD, is a cleaning and defectivity technologist in the copper/defectivity applications engineering group of Rohm and Haas Electronic Materials CMP Technologies. He is involved in the development and optimization of post-CMP cleaning processes for both copper and oxide applications and is active in the rationalization of defectivity measurement. Previously, he worked at Novellus as a CMP technologist and defectivity group technical manager. Additionally, he held CMP engineering positions at Atmel (Fab 5), where he was instrumental in developing STI/ILD post-CMP cleaning processes. Epshteyn has published extensively and holds several patents in the areas of CMP and post-CMP cleaning techniques. He received a PhD in friction and wear of materials (tribology) from the Don State Technical University in Rostov-on-Don, Russia. (Epshteyn can be reached at 602/470-4429 or

Mike Kulus manages the copper and barrier slurry business at Rohm and Haas Electronic Materials. He has more than 19 years of experience in the electronics industry. In his previous tenure with Rodel (now the CMP Technologies business of Rohm and Haas Electronic Materials), Kulus held a variety of positions, including general manager, international operations; global industry manager, hard-disk drive and electronic glass industries; national sales manager; and sales engineer. Previously, he also worked in market and product development for Westlake Plastics. He received a BS in business from Arizona State University in Tempe. (Kulus can be reached at 602/470-4438 or

Cuong Tran has been a supervisor of applications engineering in the surface preparation group of ATMI (Allentown, PA) since 2000. He provides technical support to product R&D and global assistance to customers for all semiconductor clean applications. Before joining the company, Tran was a process engineer in the R&D center of PQ in Conshohocken, PA. He received a BS in chemical engineering from Lehigh University in Bethlehem, PA. (Tran can be reached at 610/791-6920 or

Kyle Bartosh is an applications engineer in the surface preparation group of ATMI, where he has worked on copper post-CMP cleaning products for five years. In addition to having published several papers, he has one patent pending. He received a BS in chemistry from Pennsylvania State University in State College. (Bartosh can be reached at 610/791-6902 or

Darryl Peters, PhD, is a senior scientist and IP manager of ATMI’s surface preparation products group. He has more than 20 years of technical and managerial experience in the semiconductor industry and has held various positions at Bell Labs, Hampshire Instruments, Lepton, Ashland-ACT, and ATMI. Peters holds numerous patents in the area of semiconductor cleaning solutions. He received a BS in chemistry from San Diego State University and a PhD in physical chemistry from Ohio State University in Columbus, where he was a predoctoral fellow. (Peters can be reached at 610/791-6916 or

Chris Watts is director of business development for ATMI’s surface preparation group. For more than 20 years, he has held various sales and technical marketing positions at several major surface-preparation equipment and materials suppliers. He received a BS in chemical engineering from Lafayette College in Easton, PA, and an MBA from Lehigh University in Bethlehem, PA. (Watts can be reached at 610/791-6922 or

MicroHome | Search | Current Issue | MicroArchives
Buyers Guide | Media Kit

Questions/comments about MICRO Magazine? E-mail us at

© 2007 Tom Cheyney
All rights reserved.