Ultrapure Fluids

Evaluating the use of a filter/purifier in a spray processor for post-RCA rinsing

Léon Winters, Philips Semiconductors; and Joseph Zahka, Millipore

Placing a filter close to the process chamber can minimize contaminant levels in ultrapure rinsewater and on the wafer surface.

During the manufacture of semiconductor devices, the wafer surface is exposed at numerous process steps to potential sources of contamination, including particles, inorganic ions, and organic residues. To ensure a stable production process, it is necessary to maintain a constant, low baseline surface contamination level. Wet cleans are commonly used to achieve that objective.

The effectiveness of a wet clean is determined by three main parameters: its intrinsic contamination level (i.e., the level of contaminants and roughness that is added by the cleaning sequence itself), its capability to remove contaminants and to smooth out surface roughness, and its robustness. A clean is considered robust when it achieves the same low baseline contamination levels on wafers even if high levels of contaminants occasionally appear in the cleaning sequence. The fact that the RCA cleaning sequence, improved with the introduction of megasonics and dilute chemistry, is still in use in front-end processing after 30 years implies that it adheres to these cleaning parameters.

The peroxide-based RCA technique uses two standard clean (SC) chemical mixtures.¹ The SC-1 chemistry consists of ammonium hydroxide, hydrogen peroxide, and ultrapure water (UPW), while SC-2 consists of hydrochloric acid, hydrogen peroxide, and UPW. The wafers are also rinsed with UPW following both the SC-1 and SC-2 cleaning steps. The optimization of the RCA process has been intensively investigated, and researchers have found that the post-RCA wafer surface levels of several metal contaminants depend strongly on the metal concentration in UPW.^2,3 The study discussed in this article evaluated the effectiveness of a filter/purifier at reducing the metal concentration of UPW and, consequently, the wafer surface metal concentration after an RCA clean. Because the filter/purifier had already been tested in a single-tank wet bench configuration, this study focused on its performance in a spray processor.⁴

Designed by Millipore (Bedford, MA) for the purification of critical rinses in wafer manufacturing, the RinseGard point-of-use filter/purifier that was evaluated contains an ultrahigh-molecular-weight polyethylene membrane with a modified surface. Ion-exchange groups that are attached at the hydrophilic surface are able to extract specific metals from the passing UPW. Earlier test reports showed that the filter reduces the levels of aluminum, chromium, nickel, zinc, iron, and copper on the wafer surface by 57–100% after 24 hours of water flushing in an overflow tank.⁴ The filter/purifier is less effective at removing calcium and ineffective for sodium and potassium. Some specifications of the filter are given in Table I.

Experimental Equipment

In semiconductor production environments, wet cleaning is generally performed in wet benches or spray processors. Wet benches contain several tanks; some are filled with chemical mixtures and others are used only for wafer rinsing. Wafers are transported from tank to tank by a robot, and after the final rinsing step the wafers are dried. In contrast, a spray processor can perform an entire RCA cleaning sequence, including all rinses and the final drying step, while the wafers remain in a single closed process chamber. A typical RCA sequence in a spray processor is depicted in Figure 1.

 

Characteristic Capability Rating   Pore size 0.05 µm   Maximum operating     temperature 60°C   Maximum metal capacity 20 mg   Particles added > 0.2 µm <2 particles/ml after a 5-minute flush at 5 L/min

Table I: Specifications of the filter/purifier.

 

Figure 1: Typical RCA sequence in a spray processor.

The Spray Processor. A standard production spray processor was used for all of the experiments that were conducted to investigate the performance of the filter/purifier. In this system, the wafers, contained in a quartz cassette, are loaded onto a turntable, which rotates them past a stationary spray post located in the center of the process chamber. Chemicals, UPW, and nitrogen can be introduced either simultaneously or sequentially into the process chamber through one or more liquid feed systems. Following a programmed sequence, the appropriate chemistry or UPW is directed uniformly through the spray post at the wafers. For optimal performance, it is also possible to program rotation speeds. The spent liquid is generally drained continuously through the bottom of the process chamber so that wafers always contact fresh chemicals and UPW. Because the process chamber is totally closed, the operator and the cleanroom are never exposed to the chemicals.

Three rinsing functions can be programmed: function 1 rinses the feed system tubing, function 2 rinses the process chamber walls, and function 3 rinses the wafer surfaces. Three drying functions also are available. During this final process step, the separate feed systems as well as the process chamber are purged with high-pressure nitrogen.

Analytical Equipment. During the investigation, contamination levels on Wacker P-type (B), <100>, 150-mm, 100–200-cm test wafers were analyzed using three techniques: minority carrier lifetime (MCLT) analysis, vapor-phase decomposition–total x-ray fluorescence (VPD-TXRF), and light-point defect (LPD) counting.

To roughly estimate the performance of the filter without intensive analysis, MCLT was performed using a Model WT-85 analyzer from Microlab. This technique uses a laser pulse to generate electron-hole pairs in several microns of the front side of the wafer. As the minority carriers diffuse further into the substrate, the decrease in carrier concentration can be determined as a function of time by measuring the microwave reflection. These data are then converted to a minority carrier lifetime. To perform reliable measurements, it is important to diffuse the surface contaminants into the substrate bulk and to passivate the surface. Passivation is necessary because surface traps will also decrease MCLT. Therefore, in all the experiments involving MCLT analysis, wafers were oxidized in a standard production furnace at 950°C for 30 minutes. Both contaminant diffusion and surface passivation occurred during this thermal processing. MCLT analysis is not sensitive for all elements; however, it is known that the technique is especially sensitive to iron contamination.⁵

VPD-TXRF was performed with a homemade VPD-line, a Gemetec Pad Scan (Munich), and a TXRF 8010 spectrometer from Atomika Instruments (Oberschleissheim, Germany). In this technique, the native or thermally grown oxide layer of the silicon surface is converted with HF vapor into fluid droplets that contain the impurities of the oxide layer. With a scanning droplet, the fluid reaction products are collected in one microdroplet. The microdroplet is dried on the wafer under controlled conditions and analyzed with TXRF. The TXRF technique uses x-rays impinging on this spot. X-rays impinging on this spot on the surface at a small angle lead to an excitation of the (sub)surface atoms. By measuring the x-ray fluorescence, many elements can be determined.

LPD counts on wafer surfaces were determined using a Model 2020 wafer surface inspection system (Inspex; Billerica, MA). This system uses a tungsten-xenon lamp to illuminate the wafer surface and a highly sensitive camera to measure light scattering caused by surface particles and pits. Processing of the output signal results in an LPD count. The instrument has a sensitivity of 0.3 µm.

Particle measurements in UPW were also performed, using an HSLIS-M50 high-sensitivity optical particle counter from Particle Measuring Systems (Boulder, CO). This system consists of a laser diode, a measurement cell, a photo detector, and a data-handling system. When particles enter the cell, the laser beam focused there will be reflected or scattered. The photo detector measures the scattered light and transforms this measurement into an electrical impulse, the intensity of which is correlated to the particle size and concentration.

Experimental Setup and Results

To determine the effectiveness of the filter/purifier at reducing the wafer surface metal contamination level after UPW rinsing, filter units were placed at various positions in the spray processor. In the first experiment, one filter/ purifier was installed in the main UPW supply; in the next experiments, a single filter/purifier was installed in each rinsewater feed line sequentially; and in the third set of tests, filter/purifiers were placed in all three of the rinsewater feed lines at the same time. To prevent high-pressure changes on the filter membrane during purging with nitrogen, which is necessary to dry the rinsewater feed lines and the wafers, the entrance points of the nitrogen lines were shifted to a point following the filters for the second and third sets of experiments. The configuration of the spray processor, including its filter placements, is shown in Figure 2. In all cases, a Teflon valve was placed on top of the filter holder to make it possible to bypass the filter without disconnecting the tubing.

 

Figure 2: Schematic of the spray processor showing placement of filters.

The first experiment was based on the assumption that placing a filter/purifier in the main UPW supply of the spray tool would lead to an increase of MCLT and a reduction in surface metal contaminants. After several blank runs, the spray processor was loaded with test wafers and a rinse recipe was run with the filter in operation and then with the filter in bypass mode. Next, MCLT analysis of the processed wafers was performed to evaluate the performance of the filter. These results (116 ± 8 milliseconds versus 126 ± 6 milliseconds, respectively) revealed that there was no significant difference in MCLT values between wafers processed with the filter/purifier and those processed without it, suggesting that the main contamination source was not the incoming UPW but one or more components in the spray processor feed systems (the tubing, valves, couplings, or heater, for example).

In the next set of experiments, a filter was placed consecutively into the three rinsewater feed lines. Two separate rinsing processes were performed using water from all three feed lines simultaneously, even though the filter was placed in only one rinsing feed line at a time. In three separate runs, with the filter/purifier placed in the feed lines for rinse functions 1, 2, and 3, respectively, the MCLT results were as follows: function 1: 140 ± 23 milliseconds, function 2: 131 ± 3 milliseconds, function 3: 106 ± 4 milliseconds. Again, these data showed that there was no improvement in MCLT, in spite of the installation of a filter/purifier in one of the feed lines.

In the next test, the wafer surface was exposed to water only from the respective filtered rinsewater feed lines. Rinsing was performed with each filter in operation and in bypass mode. The MCLT results for these runs, presented in Table II and the left side of Figure 3, clearly indicate that there was a significant improvement of the MCLT when only filtered water was used. It was concluded that to protect against contamination from sources inside the spray processor it is necessary to install filter/purifiers in all three rinsewater feed lines.

 

Rinse without Filter (µs) Rinse with Filter (µs)   Rinse function 1 195 ± 18 832 ± 39   Rinse function 2 160 ± 6 528 ± 60   Rinse function 3 240 ± 27 758 ± 133   Reference 1754 ± 16

Table II: MCLT results for various rinsing steps with and without a filter in separate rinsewater feed lines.

After the filters had been installed in all three lines and several blank runs had been performed, the spray processor was loaded with wafers and a final rinse/dry program was run. As shown in Figure 3, the MCLT results for this final rinse also improved--to 451 ± 94 milliseconds with the filter/purifiers in operation compared with 285 ± 10 milliseconds with the filters in bypass mode. Considered together, the MCLT results from all of the experiments, which are summarized in Figure 4, indicate that placing filters in the feed lines improved the performance of the rinsing steps.

 

Figure 3: MCLT results for various RCA clean steps with and without filters.

Figure 4: Summary of MCLT analysis results for rinsing experiments.

The study was then extended by submitting test wafers to a full RCA cleaning cycle. MCLT analysis, VPD-TXRF, and LPD measurements were performed on these wafers. As shown at the far right in Figure 3, the MCLT was higher with the filters in place (600 ± 10 milliseconds) than when they were bypassed (280 ± 26 milliseconds). The VPD-TXRF results showed a decrease in surface concentrations of nickel, copper, and zinc of approximately 50–70% when the filters were in place, as demonstrated in Table III and Figure 5. Although it is believed that lowering the iron concentration was chiefly responsible for the improvement in MCLT, it was not possible to detect a significant decrease in iron levels by VPD-TXRF because the concentrations were so low. An increase of chromium was observed following the RCA sequence, but this high level could not be reproduced in later experiments, thus implying that it was a result of mishandling the sample or an analysis error. Overall, the TXRF results confirm other researchers' findings that the intrinsic performance of the RCA clean is to a great extent determined by the final rinsing step for the reported metals, with the exception of zinc. The LPD measurements showed that there was no significant difference in particle levels between wafer surfaces rinsed with the filter/purifiers in place and those rinsed without them. In both cases, fewer than 5 LPDs larger than 0.3 µm were counted.

 

Metallic Contaminants Final Rinse without Filters (1010 atoms/cm2) Final Rinse with Filters (1010 atoms/cm2) RCA clean without Filters (1010 atoms/cm2) RCA clean with Filters (1010 atoms/cm2) Calcium 8 ± 5 4 ± 1 5 ± 0.5 4 (–) Chromium <0.2 <0.2 1.7 ± 1 <0.2 Iron 0.4 ± 0.1 0.3 ± 0.15 0.4 ± 0.1 0.3 (–) Cobalt 0.2 ± 0.06 <0.1 0.13 ± 0.02 0.13 ± 0.01 Nickel 3 ± 0.1 0.9 ± 0.1 3 ± 0.4 1 ± 0.1 Copper 0.8 ± 0.1 0.3 ± 0.03 0.8 ± 0.03 0.3 ± 0.1 Zinc 1 ± 0.4 1 ± 0.7 0.6 ± 0.01 0.3 ± 0.2

Table III: VPD-TXRF results after final rinsing step and full RCA clean with and without filters.

Figure 5: VPD-TXRF results after final rinsing step and full RCA clean with and without filters.

In the final sets of tests, particles in the UPW just downstream of the filter for rinsing function 1 were measured using an optical counter. With a continuous flow of cold UPW, particle levels stabilized after only 1 hour at <5 particles/ml >0.05 µm. The same results were achieved when hot UPW (70°C) was used. These initial results demonstrate only the potential particle addition attributable to the filter itself, since the standard prefilters in the spray processor were not removed. In the next test, the prefilters were bypassed to challenge the filter. Measurements taken at the filter inlet and outlet showed that while the UPW entering the filter contained approximately 50 particles/ml >0.05 µm, the UPW at the filter outlet contained <5 particles/ml >0.05 µm. During these experiments, it was discovered that for all of the rinsewater feed lines, large amounts of particles are released when valves are switched from one mode of operation to another (for example, from hot to cold rinse). As Table IV shows, however, the filter/purifier is able to significantly reduce even such high levels of particulate contamination.

Conclusion

The performance of a filter/purifier was tested by placing one or more units in various positions in a standard production spray processor. The results revealed that the position of filters in such equipment can be extremely important. Because feed system components can be the major source of UPW contamination, the filters should be placed as close to the process chamber as possible. When a filter was placed at the UPW inlet of the processor, no improvement was observed in minority carrier lifetime after wafer rinsing. However, when a filter was installed downstream in a rinsewater feed line, MCLT results increased significantly. The best postrinsing results were achieved when filters were installed in all three rinsewater feed lines close to the process chamber. In additional tests with three filters in place, wafers were subject to an RCA clean and to a final rinse/dry step only. In both cases, an improvement in MCLT was observed over wafers processed without filters. It is expected that the benefit of the filter, as discussed in this article, depends on the contamination level to which the filter has been exposed.

VPD-TXRF and light-point defect measurements were also performed on test wafers. The VPD-TXRF results indicated that the effectiveness of the filter depends highly on the nature of the metal contaminants. Improvements in the levels of iron, nickel, copper, and zinc were observed when results for wafers processed with the filters were compared with results for runs without filters. LPDs were the same for the runs with and without filters.

Finally, in particle counts taken in the UPW, fewer than 5 particles/ml >0.05 µm were measured at the outlet of the filter, even at rinsewater temperatures as high as 70°C. Although valve switching during processing was identified as a source of substantial particle addition, the filter was able to remove these contaminants.

Acknowledgments

The authors would like to thank Ingrid Rink, Ronald Wortelboer, Erik Derksen, and Ferdinand Landsheer of Philips Semiconductors and Joop Sluis and Rainer Grosche of Millipore for their valuable contributions to this research. Special thanks also go to Harry Thewissen for his careful reading of the manuscript. The study discussed in this article has been performed in the framework of the European Medea T-612 project.

References

1. W Kern and D Puotinen, "Cleaning Solutions Based on Hydrogen Peroxide for Use in Silicon Semiconductor Technology," RCA Review (June 1970): 187–206.
2. L Winters, "The Impact of Rinse Water Temperature and Rinsing Time on Surface Metal Concentration," Internal Philips Report RNR-R52-99/AZ0032 (Nijmegen, The Netherlands: Philips Semiconductors, 1999).
3. LM Loewenstein and PW Mertens, "The Rinsing Problem: Effect of Solute-Surface Interactions on Wafer Purity," Solid State Phenomena 65–66 (1999): 1–6.
4. M Amari et al., "Point of Use Purification of DI Water," Millipore Microelectronics Div. Technical Document MA073 (Bedford, MA: Millipore, 1999).
5. L Köster and P Blochl, "Element Specific Diagnosis Using Microwave Reflection Photoconductive Decay," Japanese Journal of Applied Physics 34, part I, no. 2B (1995): 932–936.

Léon Winters is an engineer at the process and material development department (PMO) of Philips Semiconductors (Nijmegen, The Netherlands). Specializing in wet chemistry, he joined the analytical-chemical department in 1987 and headed that department from 1990 to 1994. Since 1995 Winters has been a member of the contamination studies group of the process and material development department, where he concentrates on the chemical and physical aspects of wet bench and spray processor cleaning technologies. He received a bachelor's degree from the Hogeschool Eindhoven in 1986. (Winters can be reached at leon.winters@philips.com.)

Joseph Zahka is an engineering fellow in the applications department of Millipore's microelectronics division in Bedford, MA. He has worked at the company for 20 years, applying products to the microelectronics, pharmaceutical, and medical industries. He received a BS in chemical engineering from Rensselaer Polytechnic Institute (Troy, NY) in 1970 and received an MS from MIT (Cambridge, MA) in 1971. (Zahka can be reached at 781/533-5403 or joseph_zahka@millipore.com.)

MicroHome | Search | Current Issue | MicroArchives
Buyers Guide | Media Kit