Favorable results without wet chemicals in dry FEOL resist strip-clean
01/01/2003
Engineers from LSI Logic, Gresham, OR, and Ulvac Technologies, Methuen, MA, have developed a production process eliminating sulfuric-acid hydrogen-peroxide wet cleans for post high-dose ion implantation. It combines oxygen and forming-gas RIE for resist crust removal (without conventional "popping"), fluorine-oxygen downstream plasma for plasma residue removal, and a basic DI-water rinse.
null
Beyond its immediate application, this process provides cost savings, improves safety, reduces environmental impact, and is applicable to other FEOL resist strips and cleans, eliminating piranha cleaning.
The clean process is being done on an Ulvac ENVIRO plasma asher: O2 and N2:H2 pass through a quartz or sapphire applicator tube, receive microwave excitation to create generous concentrations of ion species and neutral free radicals, and are then injected into a chamber. Han Xu, director of process technology at Ulvac, explains, "Using a remote plasma generated away from wafers, the lifetime of charged species is much shorter than neutrals, and thus most charged species recombine before reaching the wafer. This minimizes the plasma charge contribution from the charged species. To activate the neutral oxygen so it ashes the resist, the wafer is heated to ~230°C."
Adding CF4 to the O2-N2:H2 plasma generates free F and HF species that render ashing residues water soluble in the subsequent DI rinse. F can also etch SiO2 underneath the residue that may loosen its attachment to the wafer (Fig. 1). Xu says, "Compared to piranha, phosphorus and carbon densities are reduced more than four times using CF4-O2-N2:H2 plasma and residual sodium is 30% lower."
 Figure 2. Sort and yield data comparison. |
null
Xu says, "Scaling to a fab starting 10,000 200mm wafers/week and running 0.25µm technology, the cost savings from eliminating sulfuric-acid hydrogen-peroxide and reducing DI water can exceed $500,000 a year." His calculations show that this new process eliminates 1,144,000 gallons/yr of DI water, 66,560 gallons/yr of sulfuric acid, and 6240 gallons/yr of peroxide, and thus costs associated with "storing, handling, distributing, and subsequent disposal of these hazardous and corrosive chemicals." For 0.18µm technology and beyond, the savings will be more because the number of frontend layers increases.
In evaluating this process, LSI Logic engineers needed to eliminate any concerns about the fluorine plasma attacking device structure oxide layers, such as field oxides. They first determined that a 90Å or thicker oxide was suitable for monitor wafers to get consistent oxide-loss data for plasma processing. The conventional problem is that when thin oxide wafers are exposed to oxygen plasma, the plasma can penetrate underneath and oxidize the Si-SiO2 interface, artificially increasing the SiO2 thickness.
Sam Gu, staff process development engineer at LSI Logic, says, "Using these monitor wafers in a design of experiment, we found that raising wafer temperature and microwave power slightly increases oxide loss. On the other hand, increasing forming-gas flow slightly decreases oxide loss. Most significant, we found that decreasing CF4 flow reduced oxide loss along with residue-stripping efficiency. Using DOE analysis to determine a proper gas combination for suitable strip rates and oxide loss levels, we established a stable manufacturing process." In the end, this all-dry process achieved sort yield data comparable to piranha cleaning (Fig. 2).
Pieter "Pete" Burggraaf, Senior Technical Editor, Solid State Technology