Ozonated water for photoresist removal
07/01/1999
Steve Nelson, FSI International, Chaska, Minnesota
Processes using ozonated DI water for resist stripping and surface cleaning can reduce conventional chemical usage and operating costs, and improve process performance and yield. Improvements in ozone-generation equipment have made it possible to prepare supersaturated DIO3, maximizing stripping efficiency. In addition, ozonated water additives can prevent unwanted radical reactions that lead to the decomposition of the ozone; this step is essential for stripping metallized wafers.
There are several drawbacks to the conventional methods of removing organic photoresist from wafers:
* Stripping with sulfuric acid and hydrogen peroxide (SPM)requires high consumption of costly chemicals, and disposal is also expensive. In addition, it has been shown that SPM leaves sulfur residues on wafers [1].
 FIGURE 1. Ozonation reaction pathways where M is an organic to be oxidized, I- is an initiator, R represents the O3 radical, and S is a scavenger. |
* Organic solvents are also expensive, disposal is costly, and they are flammability hazards.
* O2 ashing results in plasma-induced damage to the wafer surface.
Ozonated deionized water (DIO3) is a strong oxidizer that can replace these widely used chemistries. DIO3 can be generated near the point of use from oxygen gas and DI water. Both are easily filtered to exceptionally clean levels, making the DIO3 product very clean. DIO3 has a neutral pH, so it is extremely easy to rinse wafers without residue. And ozone (O3) can be converted back into oxygen for easy disposal.
Resist stripping with DIO3 was first presented as a viable process when high O3 concentrations were produced by lowering the temperature of a wafer in an immersion system to about 5°C [2]. The slow fluid flow of the closed recirculation-immersion system, however, was less than efficient at delivering O3 to the wafer surface.
On the other hand, implementing DIO3 in a spray tool achieves a thin boundary layer of liquid on the wafer that insures rapid delivery of O3 to the surface. Rapid mass transfer of dissolved O3 to the wafer surface is critical to achieving industrially feasible resist removal rates. It has also been shown that the thin liquid boundary layer achieved in a spray tool can enhance the transfer of O3 from the gas phase onto the wafer surface [3].
Figure 1 is a general schematic of ozonation reaction pathways [4]. Organics (M) can be attacked by radicals (R.) through an indirect oxidation pathway. Processes that occur through the indirect pathway are favored by the addition of an initiator (I-) to the reaction solution. Examples of initiators are OH-, HO2-, or H2O2. In addition, radicals react with O3 in a chain reaction that results in most of the oxidizing species being decomposed in a very short time. The resulting depletion of O3 hampers the attack of the organic. Processes that occur through the direct pathway are favored by the addition of a scavenger (S), such as HCO3- that renders radicals inert.
Experimental data suggest that photoresist on bare silicon is oxidized by a direct O3 reaction. But aluminum acts as an initiator for radical decomposition of O3, so use of a radical scavenger is essential for removal of photoresist from metallized wafers.
Mass transfer theory
Spray technology is inherently more efficient than immersion baths because it delivers fresh fluid to the wafer surface. Mass transfer of O3 to the wafer surface [5] and mass transfer of contaminants and resist fragments away from the surface [6] are greatly amplified by the dynamic cleaning. An efficient mass transfer results in substantial reductions in chemical and water consumption as well as reductions in cycle time. All of these factors contribute to reduced manufacturing costs.
The etch rate in DIO3 processes is limited by the rate of mass transfer of O3 to the wafer surface. A clear understanding of the mass transfer dynamics is required for process optimization. Equations have been derived for surface mass transfer rates as a function of controlling process parameters [5]. These relationships have been validated by comparison with experimental data.
Liquid flow over the surface of a wafer can be modeled as a boundary layer flow. For spray systems, the velocity profile at any point on the wafer surface is computed from a balance of centrifugal and viscous forces acting on the fluid. For an immersion system, the velocity profile is modeled by assuming the flow between the wafers is a fully developed, 2-D channel flow. In both systems, the parameters that control mass transfer are the concentration of the reactant, the flow rate, and either the wafer spin speed or the flow area between wafers.
Resist stripping
We performed initial feasibility experiments using a single-wafer spray processor that spun a single test wafer at up to 1000rpm. This system dispensed 40ppm of O3 onto the center of the spinning wafer at 1.2 liters/min. At the same time, we mixed an ammonium bicarbonate scavenger into the DIO3 at 0.12 liters/min.
 FIGURE 2. Photoresist remaining on a silicon wafer and calculated removal rate. |
Figure 2 shows the thickness of photoresist remaining on the wafer after stripping with 40ppm DIO3 and the calculated rate of resist removal. All 5210? of resist were removed in <5 min at a constant etch rate of 1190?/min. The constant rate of removal indicates that no other reactions were occurring in the DIO3 solution and that the resist was oxidized by direct oxidation. Silicon was not an initiator in the formation of radicals.
Next, we tested the single-wafer DIO3 process using silicon wafers coated with low-dose boron-implanted (2 ? 1013 atoms/cm2) photoresist. This ion implantation process had no effect on the resist removal with DIO3. All 27,200? of photoresist were completely removed in <20 min, at a constant removal rate of 1325?/min.
In other tests, high-dose arsenic-implanted (8 ? 1015 atoms/cm2) photoresist was not attacked by either DIO3 or SPM, even after 60 min of exposure. An ashing process is necessary to remove the hardened layer on the photoresist surface. Plasma-exposed resist on SiO2 has also been processed with DIO3 and the oxide plasma etch process had no effect on the resist removal rate.
Stripping metallized wafers
We processed wafers with Al features and measured the amount of photoresist remaining after a 40ppm DIO3, single-wafer process (Fig. 3). The resist could not be completely re moved, even after 60 min. The removal rate (Fig. 3b) was not constant, decreasing to nearly zero as the photoresist strip progressed (Fig. 3b).
We believe that Al initiates the formation of radicals in the O3 solution. With no scavenger present, most of the O3 is decomposed by the radical before it can attack the photoresist. So, we modified the DIO3 stripping process for aluminized wafers to include 0.01M NH4HCO3. In water near pH 7, NH4HCO3 dissociates to form the radical scavenger HCO3-. With the NH4HCO3 solution at pH 7-8 and the DIO3 at pH 5-6, Al features exposed to these solutions in control experiments showed no visible attack even after 3 hr of continuous exposure.
 FIGURE 4. Photoresist remaining and calculated rate of removal from an AI-coated wafer after stripping with 36ppm DIO3 and 0.01M NH4HCO3. |
Figure 4 shows the removal of photoresist from an Al substrate by 36ppm DIO3 and 0.01M NH4HCO3. All 12,150? of resist were removed in <12.5 min at a constant removal rate of 1010?/min. The fact that the addition of a scavenger to the reaction solution increased the rate of photoresist removal indicates that the photoresist is primarily oxidized by the direct O3 reaction pathway, rather than by the radical reaction pathway. It is possible that during the direct O3 reaction, a radical is formed as an intermediary, but this radical, if it exists, is created by the reaction with the organic being oxidized, not by an initiator.
The most significant difference between the resist-stripping process on silicon vs. Al is the presence of the initiator in the reaction on Al wafers. The addition of the scavenger, HCO3-, to the process chemistry has made it possible to use the DIO3 process for back of the line (BEOL) cleaning and stripping. We are continuing to develop this application as a potential replacement for BEOL plasma-ashing resist removal.
Single wafer, multiple wafer
We collected data from wafers processed at 1000, 550, and 100rpm spin speeds using 140ppm O3 concentration and 1.7 liters/min DIO3 flow to determine the capabilities of stripping resist in a single-wafer system. The thickness of the resist remaining across the wafer was measured after 30 sec.
 FIGURE 5. Etch rate by DIO3 as a function of distance from the center of the wafer for three spin speeds. |
These data show that the strip rate is highest in the center of the wafer, where the O3 is not depleted (Fig. 5). The average etch rates at the edge of wafers were 6090?/min, 4970?/min, and 3510?/min, respectively. As predicted by mass transfer theory, the highest rate of resist removal occurred at the highest spin speed.
Transferring this process to a multiple-wafer MERCURY MP Surface Conditioning System capable of processing 100 200mm wafers at once is demonstrated with a series of 49-point wafer maps. This system dispenses DIO3 from a center spray post; the spray moves across wafers held in cassettes, generally progressing across each wafer beginning at the position of the arrow shown with each image in Fig. 6. Beginning with 9800? of hard-baked resist (Fig. 6a), the resist thickness decreased to 1500? near the spray post (i.e., arrow) after 5 min of etching (Fig. 6c). Figure 6c clearly shows how O3 was depleted as the DIO3 moved across the wafer surface; the resist layer was still 7050? thick at the far side of the wafer. After 10 min of total process time, all of the resist near the spray post had been removed (Fig. 6b). Because the O3 in the DIO3 now contacting the far side of the wafer was not depleted, oxidation of the remaining 3600? of resist proceeded quickly; the resist layer was completely removed from the wafer in ~16 min of O3 dispense time.
We achieved the optimal performance in our multiple-wafer-processing system at a DIO3 flow rate of 15 liters/min, an O3 concentration of 60ppm, and a 500rpm spin speed. Although the strip rate of the single-wafer process is high, it is not the system of choice for a manufacturing setting. When the time required to load, dry, and unload wafers is included, the process time is approximately 2 min/wafer, yielding a throughput of ~30 wafers/hr. Although the stripping process in a multiple-wafer system is 16 min, when including a 10-min rinse and dry, the throughput of the multiple- wafer system can be as high as 200 wafers/hr.
Conclusion
The high strip rates of a DIO3 process, developed for spray processor systems, makes it competitive with other resist-stripping and cleaning chemistries. It has the potential to revolutionize BEOL processes, where strong acids cannot be used because they attack metal features. The addition of a radical scavenger to the DIO3 chemistry eliminates the decomposition of O3 by radicals and the consequent loss of O3 reactivity. This has potential for plasma ashing in BEOL applications.
Problems encountered in the past with low strip rates have been eliminated by a new O3 generator design and optimization of O3 mass transfer to the wafer surface. Theoretical analysis shows that mass transfer in spray processors is inherently more effective than in immersion systems. n
Acknowledgments
MERCURY MP is a registered trademark of FSI International. ASTeX and Semozon are registered trademarks of Applied Science and Technology Inc.
References
1. P.J. Clews et al., "Sulfuric Acid/Hydrogen Peroxide Rinsing Study," Proceedings of the Fourth International Symposium on Cleaning Technology in Semiconductor Device Manufacturing, pp. 66-73, eds. R.E. Novak, J. Ruzyllo, Electrochemical Society, Pennington, NJ, 1996.
2. R. Mathews, "A New Aqueous-based Technology Employing Subambient Temperature Deionized Water and O3 for Removing Organics," Proceedings of the Semiconductor Pure Water and Chemicals Conference, ed. M. Balazs, pp. 359-373, 1998.
3. S. De Gendt et al., "A Novel Resist and Post-etch Residue Removal Process Using Ozonated Chemistry," Solid State Phenomena, Vols. 65-66, pp. 165-168, 1999.
4. J. Hoigne, H. Bader, "The Role of Hydroxyl Radical Reactions in Ozonation Processes in Aqueous Solutions," Water Res., Vol. 10, pp. 377-386, 1976.
5. N. Naryanswami, S. Nelson, "Dynamics of Mass Transfer on a Wafer Surface in Ozonated Water Processing for Photoresist Removal," presented at the Ultra Clean Processing for Silicon Surfaces Meeting, Ostend, Belgium, Sept. 21-23, 1998.
6. K. Christenson et al., "Mass Transfer in DIO3 Resist Stripping," presented at the Fifth International Symposium of Cleaning Technology in Semiconductor Device Manufacturing, 192nd Meeting of the Electrochemical Society, Paris, Sept. 1997.
Steve Nelson received his BS in physics and his BS in mathematics from the University of Minnesota. He is responsible for the application of ozonated water in process equipment at FSI International, 322 Lake Hazeltine Dr., Chaska, MN 55318; ph 612/448-1049, fax 612/361-7393, e-mail [email protected].