The Challenges Of the Copper CMP Clean
Copper CMP poses a unique set of challenges to the CMP cleaning process.
Diane Hymes, Hugh Li, Eugene Zhao, John de Larios -- Semiconductor International, 6/1/1998
| At a Glance | |||
| |||
Dual damascene processing utilizes chemical-mechanical planarization (CMP) to define the copper interconnect structure. The contacts and interconnects are etched into the interlayer dielectric (ILD) prior to copper deposition. The deposition process must then fill these high-aspect ratio vias and trenches in a single deposition sequence. CMP is then used to polish the copper overburden to the top of the ILD, which acts as a polish stop. The significant challenge with copper CMP is obtaining the planarized surface without dishing of the copper lines and erosion of the field dielectric, especially as the metal linewidth decreases.
Copper CMP poses a unique set of challenges to the CMP cleaning process. Contaminants from the polishing process must be removed. These contaminants include particles from the slurry, particles from the material being polished, chemical contamination from the slurry and cross-contamination from the in-laid metal. Copper is a contaminant that diffuses quickly in silicon and silicon dioxide; therefore, it must be removed from all wafer surfaces (front, back and bevel edge) in order to prevent an ad-verse effect on device performance.
During the copper CMP process, the copper layer is oxidized to form copper oxides (CuO or Cu2O) and copper hydroxides (Cu(OH) 2), depending upon the slurry pH, electrochemical potential and additives. In a basic or neutral pH cleaning environment, these copper oxides and hydroxides do not dissolve and may be easily transferred to the polyvinyl alcohol (PVA) brushes. If the brushes become contaminated, or are "loaded," by the copper oxides, they may transfer the copper contaminants onto subsequently processed wa-fers. This brush loading effect would then cause severe copper cross-contamination.
|
Fig. 1. In a through-the-brush delivery approach, the brush is porous, which permits the chemistry to permeate the brush and be delivered to the surface of the wafer as it is contacted during the scrubbing process. |
Brush loading may also result from the alumina abrasive particles used in the polishing slurry if the proper cleaning chemistry is not used. For example, in a neutral or inorganic acid cleaning environment, there is an electrostatic attraction between the alumina particle and the silicon dioxide surface that makes it difficult to remove the abrasive particle. Because of this electrostatic attractive force, the alumina particles may also adhere to the PVA brush. The outcome of brush loading, whether it be from copper oxides or alumina particles, is the same, cross-contamination to downstream wafers. Therefore, the goal is to prevent both types of brush loading, while removing the slurry from the surface of the wafer and preserving the quality of the copper layer (i.e., inhibiting copper corrosion).
A proprietary cleaning chemistry developed by OnTrak Systems is used to dissolve copper oxides and prevent brush loading during the CMP cleaning process. Additives are used to help control the electrostatic forces between the particles, PVA brushes and substrate to ensure removal of the polishing slurry and its associated contaminants. The chemistry is compatible with the copper lines and helps prevent corrosion during the cleaning process. Also, this formulation has the ability to remove copper contamination from the surface and subsurface regions of the dielectric layer. The analysis of cleaning copper wafers is presented using this new copper cleaning chemistry, simply referred to as MCC solution for the purpose of this article. These results are gathered on both blanket wafers (blanket film stacks of Cu/Ta/TEOS that are polished back to TEOS) and patterned Cu/TEOS wafers. Particle data are gathered using a Tencor 6420. Copper con-tamination is quan-tified using total reflection X-ray fluorescence (TXRF). Surface microroughness is measured using atomic force microscopy (AFM).
|
|
| Fig. 2. In the slurry dip test, clean TEOS wafers are dipped in a freshly prepared Cu slurry (alumina-based) and briefly rinsed with DIW. Half the TEOS wafers are then scrubbed with DIW, and the other half is scrubbed with the MCC solution. The number of particles removed during this process is obtained by the difference in the pre- and post-particle counts. |
Slurry removal and prevention of brush loading
OnTrak's Synergy system, a chemical-compatible, double-sided brush scrubber, was used to clean wafers after copper polishing. The system is cassette-in and cassette-out, with single-wafer processing in-between. Each wafer is processed sequentially in each of two brush stations followed by spin rinse and dry. The chemical is introduced through the core of the PVA brush. The brush is porous, which permits the chemistry to permeate the brush and be delivered to the surface of the wafer as it is contacted during the scrubbing process. A detailed schematic of this through-the-brush chemical delivery approach is presented in Figure 1.
The ability of the MCC solution to remove the polishing slurry is demonstrated with a slurry dip test. In the slurry dip test, clean TEOS wafers are dipped in a freshly prepared copper polishing slurry (alumina-based) and briefly rinsed with DI water. There is far more slurry on the surface of the wafer in a slurry dip test than exists on the wafer after a polishing process. This test represents a worse-case scenario for particle removal.
In the following experiment, half of the TEOS wafers are scrubbed with DI water (known to cause brush loading with alumina-based slurries) and the other half with the proprietary MCC solution. The difference in the number of particles remaining on the surface of the wafer before and after slurry dip scrub is measured by a Tencor 6420 at >0.2µm threshold and presented in Figure 2. If the difference in particle level is zero, all slurry particles have been removed. The alumina particles were effectively removed from the TEOS surface with the use of the solution in the scrubber. The MCC solution was able to prevent brush loading by modifying the zeta potential of the alumina particles.
| Table 1. Prevention of Brush Loading with MCC Solution | |||||||
| DIW Scrub | MCC Scrub | ||||||
|---|---|---|---|---|---|---|---|
| Wafer # | Pre | Post | Adder | Wafer # | Pre | Post | Adder |
| TEOS #1 | 67 | 72 | 5 | TEOS #1 | 77 | 94 | 17 |
| TEOS #2 | 82 | 79 | -3 | TEOS #2 | 69 | 47 | -22 |
| 5 blanket H2 O2-dipped Cu wafers | 5 blanket H2 O2-dipped Cu wafers | ||||||
| TEOS #3 | 79 | 535 | 456 | TEOS #3 | 62 | 48 | -14 |
| TEOS #4 | 52 | 354 | 302 | TEOS #4 | 77 | 78 | 1 |
| TEOS #5 | 38 | 137 | 99 | TEOS #5 | 57 | 49 | -8 |
As noted earlier, copper oxides may also cause brush loading. The copper is oxidized during the CMP process to form copper oxides and hydroxides. Some of the copper oxide may be transferred to the PVA brush during scrubbing and build up in the brush to cause brush loading. In the following experiment, blanket copper wafers were dipped in an abrasive-free solution containing hydrogen peroxide (2% by weight) for 2 min to form copper oxide on the surface. These copper wafers were then scrubbed together with clean TEOS monitor wafers in the order indicated in Table 1. As predicted, the DI water does not prevent brush loading, and the copper wafers contaminate the brush. The TEOS wafers processed after these copper wafers are then contaminated by the brushes, whereas the MCC solution prevents cross-contamination. Similar particle removal performance is observed on the TEOS monitors processed before and after the oxidized copper wafers when the MCC solution is used.
TXRF is used to quantify the concentration of copper that remains on the surface of the TEOS before and after exposure of the brushes to the oxidized copper wafers. Table 2 reveals a concentration of 100x1010 at/cm2 Cu with the DIW process while the same low level of residual Cu, 1x1010 at/cm2, remains with the use of the MCC solution. The MCC solution was able to dissolve the copper oxide in the PVA brush, preventing the cross-contamination to the downstream TEOS wafers.
|
Table 2. Cu
Concentration on TEOS Monitor Wafers by TXRF (unit: 1 x 10 10 atoms/cm2) |
||||||
|---|---|---|---|---|---|---|
| Chemistry | DIW | 0.1% NH4 OH | MCC solution | |||
| TEOS monitor | Before Cu wafers | After Cu wafers | Before Cu wafers | After Cu wafers | Before Cu wafers | After Cu wafers |
| Cu conc. on TEOS | <1 | 100 | <1 | 1300 | <1 | <1 |
Chemical compatibility
It is essential that the cleaning chemistry is compatible with the copper surface. The RMS roughness by AFM is used to understand the effect of exposing the copper surface to various chemistries. As indicated in Table 3, dilute ammonia can etch the surface of the copper, while the MCC solution does not appear to roughen the polished copper surface (Fig. 3). The AFM microroughness is the same for the DI water and MCC cleans.
|
|
(a) |
|
(b) |
|
|
(c) |
Fig. 3. AFM
scans confirm that the grainy Cu surface is smoothed out during
the CMP process, and scrubbing with MCC solution preserves the
smoothed CMP surface. Shown are
(a) PVD Cu, as-deposited;
(b) CMP and MCC scrub; and
(c) CMP and 0.1% NH4OH scrub. |
|
|
Table 3. RMS
Roughness Values in nm, Measured by AFM (scan size: 5µm x 5µm) |
||
| Cleaning chemistry | PVD Cu as deposited | PVD Cu CMP/slurry A |
|---|---|---|
| Spin-rinse-dry | 5.77 | 0.9 |
| DIW scrub | 5.82 | 0.9 |
| MCC solution scrub | 5.68 | 0.9 |
| 0.1% NH4 OH scrub | -- | 4.4 |
| 0.5% NH4 OH scrub | -- | 11 |
To further illustrate the ability of the MCC solution to remove particles from the polished copper surface, AFM is used to scan a 5µm x 5µm area of the surface. For comparison, a sample that was cleaned in a SRD with DI water is provided. A very clean copper surface is observed with the use of the MCC solution (Fig. 4).
|
|
|
| (a) | (b) |
| Fig. 4. Particle removal on CMP wafers using the MCC solution. Shown are (a) a spin-rinse-dry only and (b) MCC scrub. | |
Removal of surface metallic contamination
One of the greatest challenges to the copper CMP cleaning process is the removal of residual copper from the dielectric regions. In this experiment, a blanket film stack of Cu/Ta/TEOS was polished back to TEOS with a Lam Research A-2000 linear polisher using Rodel's grooved IC-1000 pad and commercially available copper polishing slurry from Cabot Corp. The residual metallic contamination was measured by TXRF. Using the DI water process, the residual copper contamination was on the order of 40x1010 at/cm2, while the MCC solution reduced copper contamination to a detection limit <1x1010 at/cm 2 (Table 4). Additionally, MCC solution has the ability to etch a thin layer of oxide containing copper contamination, which remains in the surface of the oxide after polishing.
|
Table 4. Surface Metallic
Contaminants After Cu CMP and Clean (unit: 1 x 1010 atoms/cm2) |
||||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Side | Scrub chemistry | K | Ca | Ti | Cr | Mn | Fe | Ni | Cu | Zn |
| Front | DIW | 200 | 12 | <5 | <3 | <1 | 10 | <1 | 40 | 7 |
| Front | MCC | 30 | <8 | <5 | <3 | <1 | 5 | <1 | <1 | 4 |
| Measured using TXRF at Charles Evans and Associates | ||||||||||
In the case of patterned wafers, the removal of copper between the lines is not straightforward. Higher concentrations of copper have been observed in areas of high pattern density. It is not clear to what extent this effect is polishing related (for example, silicon dioxide in regions of higher pattern density receive effectively longer overpolish than in regions of low pattern density) or cleaning related (local chemical cross-contamination may result from the exposure of different materials in the surface to the cleaning chemistry). This is the topic of further research at On-Trak.
Conclusion
The ongoing copper research at OnTrak has developed a new cleaning formulation, referred to as MCC solution, which permits brush scrubbing to be successfully used for copper CMP cleaning. This formulation makes use of the zeta potential concept to control the removal of slurry from the wafer surface, while preventing the slurry from loading the PVA brushes. The solution provides the same consistent cleaning performance that is currently observed for tungsten CMP, which also makes use of an alumina-based slurry.
The zeta potential is manipulated by using additives to achieve high negative zeta potential in the presence of a low pH cleaning solution. This acidic solution has the ability to dissolve copper oxides to prevent brush loading and cross-contamination of copper. Finally, the MCC solution is able to etch a thin layer of oxide, facilitating the removal of copper from the dielectric regions.
While significant progress has been made toward developing a chemistry to successfully clean copper CMP wafers and prevent corrosion, further research is still required.
Acknowledgments
We would like to acknowledge the contributions of the following individuals who provided support for this article: Pearce Toulson, Rita Holbert, Debbie Finnegan, Eduard C., Steve Jew and Dr. Rajeev Bajaj.
References
2. W.C. Krusell, J.M. de Larios, J. Zhang, Solid State Technology, No.6 (1995) 109.
3. J. de Larios, J. Zhang, E. Zhao, M. Ravkin, Micro , No.5 (1997) 61.
|
Diane Hymes, Ph.D., is director of cleaning
process technology at OnTrak Systems and has been a member of
the company's technical team for almost four years. Before
joining OnTrak, she worked for six years as an applications
research scientist at MEMC Electronic Materials. She received
her master's and doctorate in materials science and engineering
from Brown University in 1984 and 1987, respectively. Phone: (408) 577-1010 Fax: (408) 942-5335 E-mail: dianehymes@ontk.com |
| Hugh Li, Ph.D., is a process engineer at OnTrak Systems. He received his bachelor's degree in chemistry from the University of Science and Technology of China and his doctorate in physical chemistry from Iowa State University. Prior to joining OnTrak, he worked at Trace Analytical as a research and development chemist. |
| Eugene Zhao, Ph.D., received his bachelor's degree in chemistry from Fudan University in Shanghai and his doctorate in physical chemistry from the University of California at Berkeley. From 1996 to 1998, he worked as a senior process development engineer at OnTrak Systems. His responsibilities included CMP defect characterization and post-CMP cleaning process development for oxide, tungsten and copper CMP processes. Currently, he is a senior process engineer in the CMP Division of Lam Research. |
| John M. de Larios, Ph.D. received his bachelor's degree from the University of California at Berkeley and his master's degree from the University of British Columbia, both in metallurgical engineering, and his doctorate in materials science and engineering from Stanford University. He has written and co-authored more than 30 technical papers on the cleaning of semiconductor surfaces. He is vice president of cleaning technology at OnTrak Systems in Milpitas, Calif. |