化学品供应量对湿式清洁槽中颗粒去除能力的影响

We found that a wet clean bath that has an interval chemical supply and a frequent fow change was quite eftective for improvingthe cleaning performance under a constant value of total chemical supply amount. Using an interval chemical supply suggests asaving of chemical consumption which is enyironmentally friendly, and has a high throughput process. When wafers were placedwith a wide spacing and a narrow spacing at the same time in the same wet clean bath, the particle removal efficiency was decreasednoticeably at the narow spacing compared to a water arrangement with eoual spacing, In order to prevent the wet clean process fronrdecreasing the cleaning pertormance, there must be acareful consideration of the water placement in the wet clean bath. nterestingla simulation of the supply flow velocity in a wet clean bath was correlated to experimental data of actual particle removability witvarious wafer spacing. We recognized that the wafer spacing and the chemical supply manner are important factors to controllingwet bath cleaning efficiency.

Maintaining a low defect density despite an increase in the numberof processing steps ( which is itself driven by the increasing complexityof device technologies, process integration, and incorporation of newmaterials) is one ofthe most challenging aspects associated with modern semiconductor processing. The rise in the number and variationof processing tools and processing steps in high-volume manufacturing magnifies the risk of cross-contamination with a subsequentloss of yield and device performance. Wet cleans are increasingly be.ing relied on to maintain device yields despite increasing processingsteps. However, lower defect requirements complicate the cleaningprocesses and also increase chemical consumption. Increasing chemical consumption raises concerns for the environmental effect and pro.cessing cost. 

With these constraints in mind, itis essential to develop aclean process that will offer higher performance, lower cost and environmental friendliness. Many challenges have been reported over theyears to improve the performance of wet cleaning processes on meta.impurities,1-14 particles ,13-21 and organic contamination removal23-26on Si wafer surface since wet clean tools were first used for RCAstandard clean processes based on hot alkaline and acidic hydrogenperoxide solutions.26 Single-step cleaning using surfactants has beenreported to reduce chemical costs.27-29 Ohmi et al. reported that Cu iseficiently removed by dilute HF-H202.2  Dilute chemical clean isanticipated as one of the most suitable candidate to realize both lowcost and high performance clean.2 6,13,14,19,22-25 In these clean pro-cesses, 2 types of wet tools are used. A conventional wet clean bathbatch type) is still used in high volume semiconductor manufacturingdue to advantages of low chemical consumption and high throughputdespite increasing industry adoption of single wafer spin-type toolswhich can clean each wafer repeatedly with high performance.

 In a wet clean bath, it is very important to understand the flow velocitynear each wafer to gauge the performance. Habuka et al. reported onthe water filow velocity30,31and bubble motion31,32 in a wet clean bathThese reports gave important insights into the importance of the filowvelocity in a cleaning system. Despite expectations that the flow ve-locity may be one of the most important factors in controlling particleremovability during cleaning and thereby the cleaning performance itself, it has never been reported. In this report, firstly we show the cleanperformance of a conventional wet clean bath using dilute chemicalswith various wafer placements. Secondly, we try improving the clean-ing performance using a unique chemical supply manner, Finally the simulation data of flflow velocities is compared with the actual particle removal maps.

Experimental

Bare 200 mm Si wafers were used in this experiment to test thebath cleaning efficiency. In order to characterize the performance of aconventional wet clean bath, we have utilized a monitoring techniqueusing particles generated by scrubbing with a brush made of polyvinylalcohol (PVA) in a quick turn-around and low cost method that werecently proposed.33 A Dainippon Screen (DNS) FS-820 L batch wetcleaning tool was used for the surface preparation. The cleaning chem-istry consisted of a 30% ammonia 31% hydrogen peroxide mixture(APM), and a 30% hydrochloric acid 31% hydrogen peroxide mix-ture (HPM). The mixing ratio of chemicals was as follows: a dilutedAPM (DAPM) mixture consisted of the following chemicals and ra-tios NH4OH:H202:H20 = 1:1:100, and a diluted HPM (DHPMmixture consisted of HC1:H202:H20 = 1:1:100. Chemical supplydip time and temperature were 180 sec, 480 sec and 50°C for boththe DAPM and DHPM steps. Diluted hydrofluoric acid (DHF) wasadded following DAPM/DHPM treatments as a last step to preparea hydrogen terminated Si (H-terminated Si) surface. HF concentra-tion, dip time and the temperature of the DHF were 0.5%,180 secand 23°C, respectively. The wafer was rinsed with de-ionized water(DIW) after each DAPM and DHPM treatment. An OnTrak DSS200double-sided scrubber (DSS) was used for the PVA brush scrubbingwith DIW. The scrubbing time was 50 sec. A KLA-Tencor Surfscan SPl was used for the particle measurement after the brush scrubbingIn this experiment, oxide thickness was 1 nm after APM/HPM, andit was below the detection limit after the DHF treatment. The watercontact angle was under the detection limit after APM/HPM, and itwas 70° after the DHF treatment. The zeta potentials of the PVA material. the H-terminated Si and the chemical oxide were -24.8 mV+12.5 mV and -11.3 mV, respectively.33 The test solution used in thezeta potential measurement was NaCl 10 mM which NaCl is added toDIW. The solution showed a pH value around 7. Since the backgroundparticle counts were 20 through 150 in the wet clean bath before theremoval tests, we decided to use the wafers that had initial particlecounts around 30000 which are limited on the measurement tool inorder to make the removal efficiency clear. We used wafers with initialparticle counts which the range from 28000 to 29500. The size distributions of the initial particles on wafers were 40-50% in the rangeof 0.2-0.5 um, 45-55% were 0.5-1.0 um, and 1-5% of the particleswere larger than 1.0 μm.