CO2-Dissolved Water Cleans for 2xnm-Node Silicon Devices in a Single Wafer Megasonic System


Megasonic cleans have been applied to remove defects such as particles and polymer/resist residues in silicon wafer fabrication of IC devices. However, with the shrink of device technology node, megasonic cleans are being challenged to maintain high cleaning efficiency promoted by streaming force of stable cavitation for the smaller particles without producing pattern collapse caused by violent implosions of transient cavities [1]. S. Kumari et al. reported that CO2-dissolved water (CO2 DIW) was potentially able to suppress wafer damage during megasonic exposure by minimizing unrestrained explosion of transient cavities. This is accomplished through the study on Sonoluninescence (SL), the phenomenon of release of light when liquid is irradiated by sound wafers of sufficient intensity, as a sensitive indicator of cavitation events [2, 3]. This paper compares the effects of CO2 dissolution on particle removal efficiency (PRE) and pattern collapse in a range of megasonic power with >100nm-size Si3N4 particles and 2xnm node line/space-pattern, respectively to N2-gasified water (N2 DIW).


Experiments were performed on a 300mm Akrion Systems’ GoldfingerÒ Velocity tool, which provides two different types of megasonic cleans; Front Side (FS) megasonic systems with a quartz rod connected to piezoelectric crystal (1.6MHz) and Back Side (BS) with a plastic-covered piezoelectric material (830kHz), as shown in Figure 1. CO2 (approx. 1000ppm) DIW and N2 (approx. 20ppm) DIW were prepared using each membrane continuously filled with CO2 or N2 at a certain pressure. For the particle removal experiments, 300mm bare silicon wafers were contaminated with Si3N4 particles (>100nm in diameter and around 20,000 particles per wafer). Number of particles on the wafer was counted from 100nm-size by SP1 (KLA-Tencor) before/after contamination and after cleans. Pattern collapse evaluations were conducted on two different kinds of multi-stacked gate poly structures; 25nm-width with 9:1 aspect ratio (AR) and 35nm-width with 10:1 AR.

Results and Discussion

PRE for Si3N4 particles was compared between CO2 DIW (RT) and N2 DIW (RT) in 0~50W range of FS and BS Meg power as shown in Figure 2. GoldfingerÒ BS Meg can remove particles from both the front and back sides at the same time with sufficiently high PRE as FS Meg does for front side only. CO2 DIW showed >50% lower PRE than N2 DIW that would be related to the ability of CO2 to quench SL generation in DIW exposed to megasonic radiation [3]. Acidity of CO2 DIW would be one of the reasons for lower PRE of CO2 DIW; however, spiking diluted ammonia water (1:800 =30% NH4OH:DIW) to the CO2 DIW (no change on CO2 concentration) puddle on the wafer surface during megasonic radiation provides comparable PRE to N2 DIW.

Pattern collapse was compared between CO2 DIW and N2 DIW with 25nm-width (AR=9:1) gate poly wafers in 0~50W range of FS or BS power. As shown in Error! Reference source not found., pattern collapse was greatly improved by CO2 dissolution with zero collapse at 30W Meg power, which has >40% PRE. Wafer damage was evaluated again on a 34nm-width (AR=10:1) gate poly pattern in order to see any loss by pattern collapse when using diluted NH4OH spikes to improve the

PRE of CO2 DIW. According to Table 1, wafer damage was not found even at 40W BS Meg power at which >85% Si3N4 particles are removed from the silicon surface. The results indicate that CO2 suppresses pattern collapse in DIW, and is also able to inhibit wafer damage in the presence of other gases that may cause pattern collapse.


  • Kapila, P.A. Deymier, H. Shende, V. Pandit, S. Raghavan, F.O. Eschbach: Proc SPIE-Int Soc Opt Eng 6283 (2006), 628324/1–628324/12.
  • J. Walton and G. T. Reynolds, Adv. Phys., 33 (1984), p. 595.
  • Kumari, M. Keswani, S. Singh, M. Beck, E. Leibscher, P. Deymier and S. Raghavan, Microelectron Eng., (2010). doi:10.1016/j.mee.2010.10.036

Figure 1: GoldfingerÒ FS and BS Megasonic systems and their schematic diagrams of the sound transmission path

Figure 2: PRE of CO2 DIW (with/without NH4OH Spike) and N2 DIW as functions of FS and BS Meg power

Figure 3: Full wafer scan results for pattern collapse comparison between N2 DIW and CO2 DIW as a function of Meg power @ 25nm (with 9:1 aspect ratio) gate poly structure

Table 1: Full wafer scan results of pattern collapse on 35nm (with 10:1 aspect ratio) gate poly structure and PRE (>100nm Si3N4) after CO2 DIW clean with NH4OH spike as function of BS Meg power

Split Process Condition Pattern Collapse Expected PRE (@ 100 nm)
1 BS Meg 30W 30 sec with NH4OH spike No Damage 75%
2 BS Meg 35W 30 sec with NH4OH spike No Damage 80%
3 BS Meg 40W 30 sec with NH4OH spike No Damage 85%