The technology behind cleaning with supercritical fluids
07/01/2002
By David J. Mount, Laura B. Rothman, Raymond J. Robey, Mir K. Ali
SC Fluids Inc., Nashua, New Hampshire
Overview
Supercritical carbon dioxide is poised to replace conventional photoresist stripping in applications where via depth or underlying material sensitivity make conventional processing difficult or impossible. Other advantages associated with using supercritical fluids for wafer-cleaning applications include the benign temperature of this all-dry process, its environmental friendliness, and cost savings associated with lower chemical and DI water consumption and smaller footprint.
In current device generations, plasma ashing to strip bulk photoresist and subsequent wet cleaning in acids or solvents to remove residues has been both adequate and effective. But shrinking device features and new, sensitive materials threaten the effectiveness of this process. In addition, on very small features such as deep trenches, surface tension and capillary effects may prevent water and solvents used in wet bench processing from wetting device features and preclude efficient residue removal.
Removing bulk photoresist with plasma ashing systems will ultimately result in damage to new low-k films; one study has shown that it can even raise a low-k value to >4.0 [1]. Plasma ashing is also a hot process where wafer temperatures routinely exceed 400°C. In a typical advanced process with 6-9 layers of interconnect and up to 16 ion implant steps used to produce an implant profile or gradient, plasma ashing can significantly affect a processes thermal budget.
Supercritical CO2
The 2001 International Roadmap for Semiconductor (ITRS) lists supercritical carbon dioxide (SCCO2) as one potential alternative to plasma ashing. In addition, it is identified specifically for cleaning and drying high-aspect-ratio structures. The ITRS also lists SCCO2 in applications for wetting structures, contaminant removal and liquid removal, and strip and clean. With near-zero surface tension, SCCO2 is able to wet and clean the smallest of features.
SCCO2 is also an alternative where aqueous solvents are incompatible with low-k films, can cause DUV-resist poisoning, etc. There are numerous references in the literature to chemically-amplified DUV photoresist being damaged (or poisoned) by very low airborne concentrations of ammonia, di-amine compounds. Today, some of the more efficient aqueous stripping chemicals used to remove post-etch residues contain ammonia or di-amine compounds. It is increasingly difficult to keep vapors from these compounds contained in a wet-bench.
SCCO2 processes are carried out in a benign temperature regime limited to a maximum of 150°C. SCCO2 also provides environmental advantages because it can replace the use of high volumes of DI water associated with conventional wet bench processing. Likewise, there are disposal and cost savings when it is used to replace use of isopropyl alcohol in wafer drying applications.
Compared to the relatively large immersion-style wet benches, systems based on SCCO2 have much smaller footprints. For example, a 200mm SCCO2 tool using ~85ft2 of cleanroom can replace a solvent wet-bench installed ball-room style that occupies ~100ft2.
Drawing from industrial applications
Supercritical fluid (SCF) technology is not new. The ability of a SCF to dissolve low-vapor-pressure solid materials was first reported in 1879 at a meeting of the Royal Society in London [2].
Effectively, a SCF is the fourth state-of-matter. It is not a solid, liquid, or gas, but exists in its own homogeneous state where it exhibits the behavior of both gas and liquid. Almost all liquids and gases can attain a supercritical condition.
One of the most attractive aspects of the technology is that in the supercritical state a fluid behaves as a super solvent. In addition, a SCF has effectively zero surface tension and thus the ability to wet the smallest of features imaginable, even those identified for 2015 on the 2001 ITRS.
SCFs, with a variety of co-solvents, have been used widely in industrial applications for chemical extraction and purification, hazardous waste treatment, and critical parts cleaning. For example, it is quite common to extract flavors and fragrances with SCFs for use as additives in foods. Virtually all de-caffeinated coffees and teas are decaffeinated with SCCO2, demonstrating the selectivity of the process: The caffeine polymer is extracted, leaving color, aroma and flavor polymers, while the coffee bean is intact. This strongly suggests that in photoresist removal applications, resists and residues can be selectively removed from wafers that have substantial amounts of carbon in their compositions, such as low-k films or other polymeric materials.
Attaining the supercritical state
Every fluid has a critical temperature (Tc) and critical pressure (Pc) at which it goes supercritical — its critical point. Tc and Pc must be attained at the same time for a fluid to attain the supercritical phase, but the region of the supercritical phase continues above and beyond Tc and Pc (Fig. 1). Interesting chemical effects that can be very beneficial to stripping photoresist and residues occur when the pressure and temperature of CO2 are increased beyond Tc and Pc.
 Figure 1. Phase diagram of CO2. |
The list of compounds that can go supercritical includes very ordinary substances (see table). Water, for example, is an extremely aggressive solvent in the supercritical state and is environmentally benign. Water is too good, however, and has a tendency to dissolve even stainless-steel hardware used to contain it. The choice of specific SCFs for wafer-processing applications must be made with care.
Supercritical carbon dioxide
As it turns out, CO2 is an excellent candidate for photoresist and residue stripping applications. Its critical point is a relatively low 31.1°C and 1070 psi. CO2 is both nonflammable and noncorrosive, has low toxicity, generates no hazardous waste, is approved by the EPA as a non-ozone-depleting chemical alternative, and is very inexpensive. CO2 has low viscosity and low (near zero) surface tension. SCCO2 has no dipole moment, but has a very large quadrupole moment.
In general, the SCCO2 molecule behaves as a nonpolar solvent and "likes" to dissolve organic material. The fact that it has a large quadrupole moment means that it can be soluble in many types of polymers, thus causing them to swell. This is very important to the de-bonding mechanism of removing resist and residue. SCCO2 has both acidic and basic sites and can look like a Lewis acid or base. The properties of SCCO2 contribute to its ability to dissolve a very wide array of substances. Specifically for photoresist and associated residues, co-solvents dissolved in SCCO2 affect the removal of photoresist and residues by enabling the solubilizing power of SCCO2 to be tuned to be highly selective; the co-solvents carry reducing, chelating, or de-chelating agents to the substrate.
Generally, co-solvent additives are nonpolar chemicals that have solubility limits in SCCO2 of 1-15%, depending on the regime of temperature and pressure and the specific co-solvent chosen.
Beyond this explanation, the subject and the chemistry gets quite complex. Since SCCO2 can dissolve very many compounds, it is rather easy to concoct binary, ternary, etc. co-solvent additives to SCCO2, depending on what polymer one is trying to remove selective to another polymer. There are substantial opportunities for R&D in the co-solvent-polymer interaction area. In microelectronic applications, this is just beginning to be explored.
SCCO2 was first extensively studied as a supercritical CO2 resist remover (SCORR) for microelectronics applications by researchers at Los Alamos National Labs (LANL). They looked only at the lower end of the supercritical region, but this was the seminal work in using environmentally benign co-solvent additives. Subsequently, SC Fluids Inc. acquired a license to use the SCORR technology and engaged LANL to develop the commercialization potential of the technology.
In the early 1990s, researchers at IBM also amassed a body of work and intellectual property about microelectronics applications of SCCO2. Presently SC Fluids is conducting joint development work in this field through a formal program with IBM.
SCCO2 for resist removal
Early experimental data from our work with LANL clearly revealed mechanisms that affect the removal of photoresist and residues from semiconductor substrates:
- diffusion of the SCCO2 plus any dissolved co-solvents into the polymeric films,
- swelling of the polymers and subsequent de-bonding from the surface of the substrate, and
- some form of mechanical action to facilitate removal.
Pieces of polymer residues that are not carried off the substrate surface in an initial rapid decompression can be completely removed from the surface of the substrate by a variety of techniques including pulsed decompression and flushing the substrate with liquid CO2. In tests, we were able to stop the process at key points and observe the removal mechanism (Fig. 2). Starting with a post-etch state of ~4000Å of photoresist on a TEOS oxide 0.25μm via-hole pattern, we observed the swelling and foaming of the sidewall polymer in the via-hole and lifting-up of the resist layer that seemingly emanated from the via-hole.
Typically, the sidewall polymer on a post-etched feature is a fluoropolymer. The literature about SCF technology reveals that these types of polymers are readily attacked by SCCO2. Since SCCO2 has near zero surface tension and is under pressure during processing, there is diffusion into the photoresist layer, but at a slower rate than into the sidewall polymer. There is strong evidence of de-bonding of the photoresist layer that seems to emanate from the interface or joint between the sidewall polymer and the resist layer.
When we allowed our processing to proceed to completion, the photoresist and sidewall polymer were effectively removed with no debris or residue at the bottom of the via-hole, which generally results from over-etching of the etch stop layer (barrier TiN in the example shown). It is our opinion that the insoluble residues were removed mechanically during the liquid CO2 decompression flush cycles in the process we used.
 |
In our work with LANL, we also discovered that sidewall polymer residues and photoresists were more strongly bonded to low-k films than to conventional silicon dioxides. The underlying reason stems from the carbon content in low-k film, which does not exist in oxides. The resultant carbon-carbon affinity between photoresist and sidewall polymers and low-k substrate is strong and requires a different processing strategy and co-solvent selection than that used for oxides.
Our work has also shown that the use of a co-solvent greatly facilitates polymer removal from wafers. The literature suggests that at pressures approaching 10,000 psi, the ability of SCCO2 to dissolve polymers is optimal. Our tests at LANL were done at 2900 psi and 90°C, using the environmentally benign co-solvents butylene and propylene carbonates.
More recently, we have been exploring temperature and pressure regimes, in a 200mm automated Arroyo system, that greatly exceed those used in the LANL work. In addition, our process development work with IBM has been expanded to evaluate the effects of the SCCO2 process on low-k and porous low-k wafers supplied by International Sematech. Initial FTIR analysis showed good results with the low-k films not affected by the process. Ellipsometry data (film thickness) and refractive index measurements reveal only very small anomalies that can be removed when a post cure is done.
Conclusion
While conventional cleaning technologies struggle with these applications, future — and even current — photoresist sidewall polymer residue removal applications can be successfully performed using SCCO2, as recognized by the 2001 ITRS. Further, SCCO2 processing is advantageous because it operates in a benign temperature regime <150°C, is all-dry, and is environmentally friendly. SCCO2 processing also has lower costs when it replaces plasma ashing and acid and solvent wet benches and associated chemical usage and disposal.
Acknowledgments
We acknowledge the work of the "super scrub" staff at Los Alamos National Laboratory; Ken McCullogh, Chuck Taft and John Simons at IBM; Karl Tiefert of Agilent; and the "cleans" group at Sematech.
Arroyo is a trademark of SC Fluids Inc.
References
1. C. Gabriel, "Process-induced Damage in Cu and low-k Dielectrics," Solid State Technology, September, 2000, p. 28.
2. Hannay and Hogarth , Proceedings of the Royal Society of London, 1879.
David Mount received his BS in science education from Widener University. He is VP of strategic development at SC Fluids Inc., 472 Amherst St., Nashua, NH 03063; ph 603/598-0408, fax 603/598-0461, e-mail [email protected].
Laura Rothman received her BS in electrical engineering from New York University and MS in electrical engineering from Syracuse University. She is VP of technology at SC Fluids.
Raymond Robey received his BS in chemical engineering from Worcester Polytechnic Institute. He is the director of process engineering at SC Fluids.
Mir Ali received his bachelors and masters degrees in chemistry from the Indian Institute of Technology. He is senior process engineer at SC Fluids.