Laser Solution for Resist Stripping

Ruth DeJule, Associate Ed -- Semiconductor International, 11/1/1998

Laser cleaning has moved beyond particle removal into dry, single-step photoresist removal with a photochemically assisted laser ablation technique developed by Oramir Semiconductor Equipment (Haifa, Israel). As device geometries continue to shrink, contaminants such as particles, metals, photoresist and other organic and non-organic residues have an increasingly negative impact on yield. Similarly, as wafer processing becomes more aggressive and the tolerance for contaminants reduced, traditional cleaning requirements become more stringent. The trend toward all-dry processing can overcome potential drawbacks associated with wet chemistries.

Laser cleaning has effectively been used to remove microcontaminants and particulates from surfaces such as ULSI wafer devices after various production stages. Cleaning of organic and inorganic particles differs. While organic particles undergo an ablation process followed by instantaneous combustion, inorganic particles are removed when the particle and/or Si wafer absorbs UV radiation and is detached by the resulting internal shock wave formed in the wafer. Bulk photoresist removal is similar to that of organics.

Resist stripping is typically accomplished using dry plasma ashing. Aggressive and non-selective, plasma ashing leaves polymeric residues that require acids and/or organic solvents for removal. The need for wet chemistries can mean non-uniformities and incomplete resist removal because of mass transport and surface tension associated with the solutions. This is of particular concern for design rules 0.18 µm and below and for 300 mm wafers.

Fig. 1. Via veils are apparent following plasma ashing (top) but are completely removed with single-step, dry laser stripping (bottom). (Source: Oramir)

Oramir's method, called the L-Stripper, uses a combination of UV excimer laser ablation and reactive chemistry to strip the photoresist. The reactive gases are based on ozone, nitrous oxide and very small amounts of nitrogen trifluoride. These gases become highly reactive only during the short laser pulse (~30 nsec). This makes the process highly selective, attacking the resist and side wall polymers but does not etch the thin gate oxide, said Dr. Menachem Genut, vice president and chief technology officer at Oramir (see Figure). In comparison, conventional plasma asher exposures are on the order of 1 min.

During the L-Stripper process, reactive gases are injected into a low vacuum process chamber. Excimer laser pulses ( l = 248 nm), at a given repetition rate and peak intensity, are incident on the sample using a patented optical system to prevent radiation damage. As the laser beam scans the entire wafer, photoresist and embedded contaminants are removed and volotalized by the photoactive process gas. The reactive products are continuously pumped out of the process chamber through a catalytic converter making the process safe and environmentally friendly.

With this photochemically assisted laser ablation technique, no residual residues remain even under difficult stripping conditions such as post poly, via etch and following high dose implants. Alpha testing was performed to evaluate particle and resist removal. No residues were detected by SEM and Auger analysis. To verify that no metal contamination remained after the process, VPD-TXRF tests were performed. Typical results were 3.3 x 10⁹at/cm² for Fe and 4.6 x 10¹⁰ at/cm² for Al, which meet the SIA Roadmap cleanliness standards and prove that, indeed, no wet chemical follow-up is needed, Genut said. Further Auger analysis indicated that no carbon or other contaminants typical of photoresist residues were detected.

The L-Stripper is currently in beta-testing by a consortia consisting of Siemens, Philips and Alcatel Microelectronics at the Frauhofer Institute in Munich, Germany. Capable of processing wafers at a rate of 3 min/6 in. wafer with one chamber operation, the rate is expected to be reduced to 1.5 min/8 in. wafer when the next-generation tool introduces parallel working capabilities. A corresponding reduction in the current cost of ownership from $3.5/wafer to below $2.5/wafer is expected in early 1999.