Issue



WAFER CLEANING: Environmentally friendly single-wafer spin cleaning


11/01/1999







Takeshi Hattori, Sony Corp., Atsugi, Japan

Sony has implemented a new single-wafer spin cleaning technique that alternately cycles between ozonized water and dilute HF at room temperature. Each solution is sequentially applied for only a few seconds onto a rotating wafer through jet nozzles, repeating the sequence to achieve the desired surface cleanliness. Tests show that this process can efficiently remove particles and organic and metallic contaminants in a short time without increasing surface microroughness. This technique meets stricter wafer cleanliness requirements, is applicable to larger diameter wafers, and is environmentally friendly.

Wafer-cleaning chemistry, which is the most repetitively applied processing step in any IC fabrication sequence, has remained essentially unchanged for the past 25 years [1, 2]. The most prevalent method worldwide is still hydrogen peroxide-based — most notably RCA standard cleans [3] — in which wafers are sequentially immersed for minutes in a NH4OH-H2O2-H2O mixture and a HCl-H2O2-H2O mixture at elevated temperatures, as well as in dilute HF at room temperature. In some cases, a hot H2SO4-H2O2 mixture is added at the beginning of the sequence. In such immersion-type wet chemical cleaning, even if ultrapure chemicals are introduced and then disposed of after each wafer cleaning treatment, the contamination removal efficiency is dominated by impurities brought into the fresh solution by the wafers to be cleaned [4, 5].

To meet stricter wafer-cleanliness requirements, new cleaning methods in which fresh chemicals are continuously supplied, such as single-wafer spin cleaning, have to be used. Spin-processing equipment has the advantage of a much smaller footprint compared to a conventional wet bench, but its throughput must be increased and its chemical consumption reduced [6]. Here, the best approach is a technique using alternative cost-effective chemicals, rather than those used conventionally, and fewer chemicals to shorten the cleaning period and reduce chemical consumption. This also reduces the quantity of effluents from wafer cleaning.

Particles, metallic contamination, and trace organic contamination, adsorbed on the surface of silicon wafers, are all increasingly detrimental to semiconductor performance and yield. Consider that when silicon wafers are exposed to the atmosphere in a regular cleanroom, gaseous organic molecules in the air quickly adsorb on the wafer surface. In addition, while wafers are stored in plastic boxes to protect them from airborne contaminants, organic volatiles outgas from the plastic and are adsorbed onto wafer surfaces [7]. Such organic contaminants have deleterious effects not only on gate oxide integrity [8, 9], but also for chemical vapor deposition [10]. It is as important to remove organic contaminants as it is to remove particles and metallic contaminants from silicon surfaces prior to subsequent processes [7, 11].


Figure 1. Thickness of chemical oxide on wafer surface measured, using X-ray photoelectron spectrometry, as a function of ozonized-water application time.
Click here to enlarge image

null

All these challenges are behind a new single-wafer spin-cleaning process. This process is low cost and environmentally friendly, has high performance at room temperature, and requires a short amount of time, alternating only ozonized DI water and dilute HF, for removing particles, metallics, and organics from wafer surfaces [12, 13]. We call the process SCROD — spin cleaning with repetitive use of ozonized water and dilute HF.

Particle removal

To evaluate the SCROD process for particle removal, we used both polystyrene-latex (PSL) spheres and Al2O3 particles on 200mm wafers. We chose Al2O3 particles because they are frequently found in wafer processing lines where equipment components are made from aluminum. Al2O3 particles are much more difficult to remove from a silicon surface than PSL spheres.

We used VLSI Standard's sphere deposition system to deposit PSL spheres onto wafers. To deposit the Al2O3 particles, we immersed the wafers in dilute HF, dipped them in deionized water in which Al2O3 particles were intentionally added, and rinsed them with DI water. We controlled the number of Al2O3 particles and PSL spheres to ~2000/wafer of each type.

Following the preparation of wafers, we subjected them to our SCROD single-wafer spin cleaning process, checking contamination levels and microroughness before and after cleaning. We detected the number of particles on the wafer =>0.16µm with an automated light-scattering inspection system.

SCROD applies ozonized water onto the center of a rotating wafer surface through a jet nozzle followed by the application of dilute HF through another nozzle. We varied the application time to determine the shortest time that provided acceptable results with a high reproducibility for volume production; this turned out to be 10 sec each for ozonized water and dilute HF — a total of 20 sec for one cycle — repeating this sequence as many times as needed. After the last dilute HF treatment, the wafer is rinsed with DI water and spun dried.

The various specifications of the SCROD process included:

  • performing all cleaning and drying procedures at room temperature (23°C) in a nitrogen atmosphere to avoid drying spots;
  • using 20 ppm ozone concentration in DI water, measuring it at the outlet of the ozonized-water generator;
  • using 1% HF concentration;
  • using a maximum 1600 rpm spin speed;
  • applying fresh ozonized water at 20cm3/sec and dilute HF at 25cm3 /sec; and rinsing with DI water at 25cm3/sec.

All solutions were disposed of after use (i.e., the SCROD process does not cycle the chemical mixtures).

We found that particles on the silicon wafer surface were removed very efficiently using this method. For example, our comparisons of 20-sec cleaning cycles (i.e., 10 sec of ozonized DI water and 10 sec of dilute HF) and 2-min cycles (i.e., 1 min for each chemical), each repeated one, two and three times, showed that particles were removed by approximately 79%, 85%, and 87%.

We found virtually no difference in particle-removal efficiency between the 20-sec and 2-min cycles, because during the ozonized-water treatment a chemical oxide film grows very rapidly on the wafer surface, becoming almost saturated at ~0.7nm after 10 sec of ozonized water application (Fig. 1). This chemical oxide can be completely etched from the wafer surface within 10 sec by subsequent application of dilute HF. Particles are removed when the chemical oxide is "lifted off" the wafer during the dilute HF treatment. In addition, there will be no difference in the etched silicon depth from the 20-sec or 2-min cleaning cycles.

In addition, we found that applying the chemicals for a longer time (>10 sec each) with the same number of cycles or fewer does not increase particle-removal efficiency. Therefore, the 20-sec cleaning cycle saves both time and chemical consumption. (If 0.5% dilute HF is used, we recommend a 25-sec cleaning cycle, i.e., 10 sec of ozonized DI wafer followed by 15 sec of dilute HF.)

Particles removed from the wafer surface flow away immediately during spin cleaning, so redeposition, which is often observed in immersion-type wet chemical cleaning, is not observed even after HF treatment.

In our tests, PSL spheres were removed by 98%, 99%, and 99.5%, after one, two, and three repetitions of the 20-sec cycle. Thus, PSLs on the wafer surface were much more easily removed from the wafer surface than Al2O3 particles.

Metal removal

For metal removal evaluation, we used copper, iron, and aluminum contaminated wafers:

  • Test wafers were immersed in a dilute HF solution that was spiked with 10ppm Cu by adding a standard Cu-containing solution originally prepared for atomic adsorption spectroscopic analysis [4].
  • Test wafers were immersed in a contaminated SC-1 solution (NH4OH-H2O2-H2O) that was spiked with 1ppb Fe or Al by adding a standard Fe or Al containing solution.


Cu concentration on a wafer surface before and after spin cleaning as a function of the number of 20-sec cleaning cycles; b) Cu concentration on the wafer surface after one cycle of 2-min cleaning, and after six cycles of 20-sec cleaning.
Click here to enlarge image

null

The contamination levels of Cu, Fe, and Al on the wafer surface were controlled within 1012 to 1014 atoms/cm2 for each metal, and the metal contamination level and wafer surface were compared before and after the cleaning.

We measured metallic contaminants on wafers before and after spin cleaning using flameless atomic absorption spectrometry (FL-AAS) after liquid phase decomposition of the contaminants. The detection limits for this technique are 1.5 x 108 atoms/cm2 for Cu, 4.0 x 108 atoms/cm2 for Fe, and 4.0 x 108 atoms/cm2 for Al.

Fe contaminants on a wafer surface as high as 1012 to 1013 atoms/

cm2 were reduced to =<109 atoms/cm2 with only one repetition of the 20-sec ozonized water and dilute HF treatment. Likewise, with just one repetition of the 20-sec treatment, the Al contamination was reduced to the 4.0 x 108 atoms/cm2 detection limit.

Most Fe and Al atoms on the wafer surface were ionized and dissolved into the ozonized water. Some Fe and Al atoms remained in the chemical oxide grown during the ozonized-water treatment because these atoms have higher oxide generation enthalpy than silicon. However, oxide-trapped atoms are dissolved when the chemical oxide on the wafer surface is removed with dilute HF. Fe and Al ions dissolved in the dilute HF are not re-deposited on the wafer surface because these metals have lower electronegativity than silicon. This accounts for the very high removal efficiencies of both Fe and Al.

Cu contaminants as high as 5 x 1013 atoms/cm2 were reduced to ~5 x 1011 atoms/cm2 after one repetition of the 20-sec treatment. This is a higher level than that achieved for Fe and Al because Cu has a higher electronegativity than silicon. While Cu contaminants were not removed even after one 2-min clean, they were effectively removed after six repetitions of the 20-sec clean. Here the total cleaning time is the same as on a 2-min clean (Fig. 2).

We believe that Cu atoms remaining on the wafer surface after ozonized water treatment are uncovered by etching of the chemical oxide grown during the ozonized-water treatment due to the subsequent dilute HF treatment, and they are easily dissolved by the ozonized water treatment in the next cleaning cycle. Therefore, more ozonized-DI-water and HF cleaning repetitions are effective in removing Cu compared to longer chemical application time. Cu ions dissolved in the solution will not be redeposited onto the wafer surface because dissolved Cu ions flow away immediately from the wafer surface with spin cleaning. Thus, Cu contamination is finally reduced to the 1 x 109 atoms/cm2 level or lower using this repetitive cleaning method.

Organic contamination removal

To evaluate organic-removal efficiency of repetitive single-wafer spin-cleaning using ozonized water and dilute HF, we prepared silicon wafers contaminated with butylhydroxytoluene (BHT) — one of the common antioxidants contained in plastic boxes. We also used silicon wafers stored in a plastic box for a long time to examine the ability to remove organic contaminants.

We analyzed organic contaminants on the wafer using gas chromatography and mass spectrometry following thermodesorption (TD-GC/MS) [7], identifying the resultant molecular structures of the desorbed organic compounds by comparing spectra from the mass spectrometer with library spectra data. To obtain a quantitative estimate of residual organic molecules/cm2 on the wafers, the peak areas for BHT in the resultant chromatograms were compared with those for standard samples of BHT whose concentrations were known.

We detected a large amount of 2,6-di-t-butyl-2,5-cyclohexadiene-1,4-dione (an oxidation product of BHT) and dibutyl phthalate (a plasicizer used in polymer molding) on the wafer stored in a plastic box before cleaning. These organics were not detected after just one application of ozonized water in the spin cleaning.

Intentionally contaminated BHT on wafer surfaces was reduced to <109 molecules/cm2 (the detection limit of the TD-GC/MS) after one application of dilute HF following ozonized-water cleaning. Ozonized water has a sufficiently high oxidation potential to degrade organic contaminants, while dilute HF is capable of removing them by lifting off the native oxide film on which organic contaminants are adsorbed [7].

Surface roughness

To evaluate silicon surface roughness, we prepared highly doped surfaces implanting and annealing As, giving a concentration of ~1020 atoms/cm3. Microroughness occurs more readily on an impurity-doped n+-type surface [14].

We measured microroughness on the silicon surface before and after spin cleaning using an atomic force microscope. The thickness of the chemical oxide film grown on the silicon substrate during ozonized-water application was measured using x-ray photoelectron spectrometry.

The RMS measurements of surface roughness before and after as many as 12 cycles of cleaning were 0.25nm and 0.28nm, virtually no difference. We concluded that repetitive SCROD spin cleaning did not produce surface roughness, compared to simultaneous application of HF and ozonized water in a mixture that causes significant surface-roughness [15].

Sequence modification *** When the initial native oxide film on wafers is >1nm, due to prolonged exposure to the ambient, we recommend the application of dilute HF for more than 10 sec before applying any ozonized-water dilute-HF cycles. This will remove the thick native oxide and expose noble metals on the silicon surface and enhance metal removal efficiency in subsequent cleaning. In our tests, Cu contaminants at ~1014 atoms/cm2 on a wafer with a relatively thick native oxide were removed to the 7 x 1013 atoms/cm2 level by ozonized-water application. On the other hand, the same Cu contamination was reduced to 7 x 1011 atoms/cm2 when we used dilute HF before ozonized-water application.


Figure 3. Comparisons of gate oxide integrity for RCA immersion and SCROD cleaning.
Click here to enlarge image

null

The final step of the SCROD cleaning process, after the last dilute HF treatment, includes two alternatives:

Click here to enlarge image

Figure 4. Liquid chemical consumption for a 25-wafer lot in immersion-type RCA cleaning and repetitive SCROD spin cleaning. Assumptions include 3.5 x 104 cm3 bath volume; traditional RCA cleaning steps of 10-min 1:1:5 SC1, 10-min 1:1:6 SC2, and 1-min 1% dilute HF; and dilute RCA cleaning steps of 10-min 1:1:50 SC1, 10-min 1:1:60 SC2, and 1-min 1% dilute HF.

  • A final DI water rinse renders a wafer with a hydrophobic silicon surface.
  • A final ozonized water rinse renders a hydrophilic silicon surface.

In addition to silicon surfaces, the SCROD single-wafer spin-cleaning process can be successfully applied to thin films on the silicon wafer surface, including silicon dioxide, silicon nitride, and polycrystalline silicon, by appropriately controlling the cleaning conditions.

Gate oxide integrity

Use of this cleaning technology, where fresh chemicals and water are continuously supplied, improves gate oxide integrity of MOS transistors compared to conventional immersion-type RCA cleaning, where metallic contaminants accumulate in the solution. Figure 3 shows the time zero dielectric breakdown (TZDB) and time dependent dielectric breakdown (TDDB) of gate oxides prepared with conventional-type RCA cleaning and SCROD cleaning. For both TZDB and TDDB, SCROD gave better results than immersion-type RCA cleaning. This indicates that fresh chemicals and water is a key factor to achieving high gate-oxide integrity [4].

Chemical and water consumption

SCROD markedly reduces chemical and DI water consumption compared to traditional immersion-type RCA cleaning. As shown in Fig. 4, liquid-chemical consumption for spin cleaning with six cycles of ozonized water and dilute HF is only 5% of that used in a traditional RCA clean and 28% of that used in an immersion-type dilute RCA clean in a single-bath system, in which chemicals must be renewed for each step.

In standard immersion-type wet cleaning, the chemical solutions can be repetitively used to minimize chemical consumption, but metals accumulate in the solution, lowering metal removal efficiency. Chemical solutions must be disposed of after each cleaning cycle to meet future metallic-contamination wafer-cleanliness requirements [4].


Figure 5. DI water consumption/25-wafer lot in immersion-type RCA cleaning and repetitive spin cleaning.***
Click here to enlarge image

SCROD also reduces the volume of DI water required for wafer cleaning, which includes water for rinsing the wafers and for diluting chemicals and making ozonized water. If we assume that the water flow rate for wafer rinsing in immersion-type systems is 25 liters/min, DI water for spin cleaning with six cycle repetitions is 8% of the water required for the traditional immersion-type RCA cleaning and <2% for spin cleaning with only one cycle (Fig. 5). RCA cleaning, using both alkali and acid chemicals, needs thorough DI water rinse for typically 10 min after each chemical treatment. Such rinsing is not necessary between the treatments of ozonized water and dilute HF in spin cleaning.

Effluent control

Hot liquid chemicals, such as NH4OH, H2O2, and HCl, are used in immersion-type RCA cleaning, so a large amount of high-concentration gaseous chemicals are exhausted from large cleaning equipment. On the other hand, in a small spin-cleaning system that uses ozonized water and dilute HF at room temperature, the amount of exhaust is small and contains a low concentration of gaseous chemicals.

The effluents that do occur can be easily controlled in our spin-cleaning system:

  • Disposed ozonized water decomposes spontaneously into oxygen and water.
  • Disposed HF can be used to manufacture pure HF and other fluoride chemicals in the form of fluorite in the chemical industry.

Conclusion

We have refined single-wafer spin cleaning at room temperature using only ozonized water and dilute HF. The spin-cleaning sequence consists of alternately applying ozonized water and dilute HF onto a wafer surface for 10 sec. This short-time cycle cleaning can efficiently remove metallic contaminants and particles as well as organic contamination without increasing the microroughness of the surface. This cleaning cycle can be repeated as many times as needed until the surface cleanliness reaches the required level. Further, the 10-sec cycle can be shortened to a few seconds to save time, chemicals, and water consumption [15].

In the final step of cleaning, after the last dilute HF treatment, DI water is applied to the wafer to obtain a hydrophobic silicon surface or ozonized water to obtain a hydrophilic silicon surface. This low-cost, high-performance, room-temperature treatment, in which fresh liquids are continuously supplied, will meet the requirements for stricter wafer cleanliness, larger-diameter wafer processing, and greater respect for the environment.

References

  1. T. Hattori, Solid State Technology, Vol. 38, No. 5, pp. S7-S10, May 1995.
  2. T. Hattori, "Trends in Wafer Cleaning Technology," Ultra Clean Surface Processing of Silicon Wafers, pp. 437-450, Springer-Verlag, 1998.
  3. W. Kern, D.A. Puotinen, RCA Review, 31, pp. 187-206, June 1970.
  4. T. Osaka, T. Hattori, IEEE Trans. Semiconductor Manufacturing, SM-11, pp. 20-24, 1998.
  5. T. Hattori, "Key Issues in Wet Chemical Cleaning of Silicon Surfaces," Cleaning Technology in Semiconductor Device Manufacturing V, PV 97-35, pp. 3-14, The Electrochemical Society, 1997.
  6. N. Yonekawa, et al., "Contamination Removal by Wafer Spin Cleaning Process with Advanced Chemical Distribution System," Cleaning Technology in Semiconductor Device Manufacturing III, PV94-7, pp. 94-101, The Electrochemical Society, 1994.
  7. K. Saga, T. Hattori, J. Electrochem. Soc., 143, pp. 3279-3284, 1996.
  8. S.R. Kasi, et al., Appl. Phys. Lett., 59, pp. 108-110, 1991.
  9. K. Saga, T. Hattori, Appl. Phys. Lett., 71, pp. 3670-3672, 1997.
  10. K. Saga, T. Hattori, J. Electrochem. Soc., 144, pp. L253-L255, 1997.
  11. K. Saga, T. Hattori, J. Electorchem. Soc., 144, pp. L250-L252, 1997.
  12. T. Osaka, T. Hattori, "Contamination Removal by Single-Wafer Spin Cleaning with Repetitive Use of Ozonized Water and Dilute HF," Cleaning Technology in Semiconductor Device Manufacturing V, PV-97-35, pp. 256-263, The Electrochemical Society, 1997.
  13. T. Hattori, et al., J. Electrochem. Soc., 145, pp. 3278-3284, 1998.
  14. T. Ohmi, et al., J. Electrochem. Soc., 140, pp. 811-818, 1993.
  15. T. Hattori, unpublished data (presented at the Spring Meeting of the Applied Physics Society of Japan, March 1998).

Takeshi Hattori received his BS, MS, and PhD degrees from Sophia University, Tokyo, and his EngD from Stanford University, all in electrical engineering. He is chief research scientist at Sony Corp. and is responsible for development of new wafer processes, ultraclean technologies, and manufacturing innovation methodologies for next-generation devices. Sony Corp., 4-14-1, Asahicho, Atsugi, 243-8585, Japan; ph +81/46230-5461, fax +81/46230-5572, e-mail [email protected].