Originally appeared in the March 2001 issue of Micro Magazine. © Canon Communications 2001 Investigating the Formation of Time-Dependent Haze on Stored Wafers Larry W. Shive, Richard Blank and Karen Lamb MEMC Electronic Materials, Inc. Abstract: The storage life of bare silicon wafers has been historically defined as the elapsed time after packaging before small particles begin to form on the surface. Many silicon wafer users have observed this time-dependent haze formation on wafers that have seen extended storage. We have investigated the surface changes of silicon wafers during 6 months and 18 months of storage and show that although most wafers have a very high potential for surface degradation, strict control of moisture inside the wafer package is the primary key to 18-month storage life. We show that surface organics, ions, oxide thickness, metals and particles remain very stable in a well-controlled package environment. However, we also show that the typical levels of organics and ions that are present on commercially available silicon wafers have the potential to form over one million particles >0.12um diameter if the wrong storage conditions exist. A general mechanism for tdh formation is proposed. Introduction Silicon wafer users expect that the surface of the wafers they receive from wafer makers will meet all requirements for light point defects (LPDs), metals, grown-on film quality and cleanability regardless of the storage history of the wafers. Yet, many users have observed wafers with a very large number of >0.12um LPDs that were not present when the wafers left the manufacturer’s facility. This phenomenon has come to be called “timedependent haze” (TDH). Its appearance may or may not cause a LPD rejection at incoming Quality Assurance or a performance problem in the device line but it will always suggest to the user that something in the wafer maker’s process is out of control. A typical “hazed” wafer map is shown in Figure 1. It is common to see a localized region of high LPDs in a random pattern. LPD counts on a hazed wafer are usually at least several hundred more than on non-hazed wafers. It is usually the presence of a pattern and/or unexpectedly high counts that alerts one to TDH. An AFM image (Figure 2) of the defects in this pattern may reveal up to 108 /cm2 defects that are typically 5- 25 nm high and < 500 nm wide. We have found that TDH is usually removed by rinsing the wafer with water or heating it to 200C. Individual defects also tend to disappear very quickly when exposed x-ray beams so their composition is not easily determined by XPS. These observations suggest that TDH is composed of ionic or polar organic compounds with high vapor pressures at <200C. An observation of TDH brings up the question of what actually changed on the wafer surface during shipping and storage and why it only appeared on a small portion of wafers. It has been clearly shown that organic and inorganic contamination can form particles on the surface during storage to form a “haze”. Munter et al. (1) have artificially created TDH by intentionally contaminating wafers with inorganic ions and organic Originally appeared in the March 2001 issue of Micro Magazine. © Canon Communications 2001 solvents and storing the wafers. Vepa et al. (2) artificially created TDH by exposing wafers to organic compounds that are commonly observed on wafers. The TDH that formed was composed of small particles rather than a continuous film or large residues. In this paper, we look at some key surface characteristics of silicon wafers after 6 & 18 month storage in Grade 1 packaging in a typical warehouse environment and compare them to those of freshly cleaned and packaged wafers. The relationship between these changes and TDH formation is discussed. Experimental Procedures Wafer sampling and storage. For over two years, five prime 200 mm diameter wafers were sampled weekly from production after final cleaning, inspected, packaged in Grade 1 packages and immediately stored in a local warehouse for 6 months or 18 months. Grade 1 packaging consisted of a clean, dry polypropylene box, a polyethylene inner bag, a package of desiccant and an outer laminated polyethylene bag that includes an aluminum layer. Other prime samples were pulled during the same time period, packaged and analyzed within 30 days to establish a “fresh wafer” baseline. The warehouse temperature varied from 40ºF to 100ºF depending on seasonal temperatures. Warehouse humidity was not controlled. At the end of the 6 month or 18 month storage time, the wafers were re-inspected for LPDs and then distributed to the analytical laboratory to measure surface organics, ions, metals, and oxide thickness. The outer bag was inspected for tears or pinholes. The “before” and “after” LPDs were measured using an ADE Corporation CR-80 surface inspection tool. Analytical methods. The five wafers/week recovered from storage were distributed for the following analyses. Since only five wafers were available, all the analyses could not be done every week. Desorbable surface organics were measured by the thermal desorption/gas chromatographic method with mass selective and NPD (before June 1998) or atomic emission (after May 1998) detection described by Sun (3-4). Surface ion analysis was done using a combination of water extraction of the surface followed by analysis of the extract by ion chromatography or capillary electrophoresis described by Sun (5). Average oxide thickness was determined using a variation of the method of Vepa (6). Surface metals were determined using acid drop extraction with the acid drop examined by ICP/MS. A Dimension 5000 Nanoscope III AFM made by Veeco- Digital Instruments was used to image the TDH defects. The accelerated TDH test was done on fresh prime wafers as follows. Test wafers are inspected for LPDs and then placed in slots 11, 13 and 15 of the box insert. One milliliter of water was placed in the box and the box was closed and packaged normally, but without the desiccant. The box was staged in a temperature-controlled chamber for 16 hrs at 17ºC then for 4 hrs at 50ºC then for 4hrs at –10ºC and finally staged for 16 hrs at 17ºC. The relative humidity holds at about 90% for most of this test except the 50ºC step. The wafers were unboxed and re-inspected for LPDs and the change in LPDs was determined. Originally appeared in the March 2001 issue of Micro Magazine. © Canon Communications 2001 Results and Discussion Trend charts for changes in LPDs and whole wafer averages for surface organics, surface ions and oxide thickness after 6 month and 18 month storage are shown in Figures 3, 4, 6 and 8. These data will be discussed one at a time. The trend data for the change in >0.12um LPDs after 6 months and 18 months are shown in Figure 3. The manufacturing date of the wafers is shown on the x-axis. They were reinspected 6 months or 18 months later. In the chart, each data point is the average LPD change for a five-wafer sample. The whole database was used to calculate the upper and lower control limits. Six-month storage data are shown up till late December 1998 and 18-month storage data begin in late December 1998. Data points that exceed the upper control limit are defined as TDH events. The average increase of >0.12um LPDs is only 2.0 LPDs/wafer after 6-18 months of storage and the variability around this average primarily reflects the long-term reproducibility (six to eighteen months) of the inspection tool. These LPD inspection tools were serviced and re-calibrated at least once between the pre-storage and poststorage inspections and long term reproducibility was certainly affected. Prior to mid-July 1998, an organic cleaning agent was used in the process and it was the source of all the TDH events up till that time. No TDH events were observed after this organic was removed from the process. The increase in storage time from 6 months to 18 months had no impact on the change in surface LPDs. There was no shift in the average adders in December 1998 when we switched to 18 months of storage. Surface metals do not increase with storage time. One might expect that surface organic levels would increase very significantly, relative to wafers stored for only a few days, when wafers have been stored for 6 months or more. Some common organic contaminants found on stored wafers are listed in Table 1 and include common plasticizers and antioxidants. Freshly packaged wafers (<30 days) already typically have 0.5- 1.5 X 1014 C atoms/cm2 of these compounds. Sugimoto (7) has investigated organic contamination on wafers stored in new boxes for 30 days and found similar plasticizers and antioxidants. The trend chart for average total desorbable organics on wafers stored for 6 months and 18 months is in Figure 4 and shows that the average, 1.2 X 1014 C atoms/cm2 , is no different that that of freshly packaged wafers. Two typical chromatograms of wafers that were stored for six months are shown in Figure 5. Both show detectable levels of antioxidants, plasticizers and siloxanes but the chromatogram on the right shows >5 X 1014 C atoms/cm2 of the cleaning agent. This wafer exhibited TDH. The organic cleaning agent mentioned earlier was found to be the reason for all the Originally appeared in the March 2001 issue of Micro Magazine. © Canon Communications 2001 TDH events before mid-July 1998 in Figure 3. We used NPD organic detection till June 1998 then switched to AE detection to obtain more stable and quantifiable results for surface organics. It appears that <5 X 1014 C-atoms/cm2 of these common plasticizers and antioxidants that are usually deposited during 6- 18 months of storage do not affect TDH if package integrity is maintained. It should be noted that our method is only detecting desorbable organics. High molecular weight, low-volatility organics could easily be present at high levels but not observed on the wafers. These would more likely be left on the wafer by the wafer-making processes than transferred from the package materials. Surface ion contamination, specifically sulfur-containing ions, have been known to cause severe TDH (1). Inorganic ions commonly found on silicon wafers are listed in Table 1. In our study, sulfur-containing compounds and ions were below the detection limits of the GC/AES and CE methods, 109 S-atoms/cm2 and 5 X 1010 SO4 2--ions/cm2 respectively. Fluoride and nitrate ions were also below the detection limits. However, ammonium and chloride ions were found in abundance. Freshly packaged wafers typically have as much as 3 X 1013 NH4 + ions/cm2 and 1.9 X 1013 Cl- ions/cm2 . The trend data after 6 & 18 month storage are shown in Figure 6. The levels of these ions, >1012 ions/cm2 , place NH4 + and Cl- among the most abundant contaminants on hydrophilic silicon wafers. Their levels did not increase during storage and they did not cause TDH formation when organics were controlled and package integrity was maintained. However, as we will discuss later, the potential for particle formation is significant for this level of contamination. Water is the well-recognized accelerating agent for TDH formation. For example, the quantity of water added to a package can be related to the total area of a wafer that is affected by haze, all other variables being kept constant (Figure 7). Therefore we began to look for a test to indicate whether traces of moisture had leaked into the Grade 1 packaging during the >6 months of storage. The tiny remote probes that are commercially available to measure relative humidity and temperature could not be used because they would contaminate the environment inside the package. However, we realized that an increase in oxide thickness would be a very sensitive indicator of a leak. This is because the chemical oxide from a freshly cleaned wafer is only 0.8- 1.1 nm and the wafer will continue to oxidize if further exposed in humid air (8) for extended periods, reaching a thickness of 1.5 nm. Therefore, if the oxide thickness of the wafer reached 1.5 nm, it would indicate that the moisture barrier had failed. A trend chart for average oxide thickness after 6 & 18 months of storage is shown in Figure 8. It is clear that no significant additional oxidation occurred and therefore the moisture barrier was completely effective. However, the data suggest a trend for which we currently have no explanation. That trend is not echoed in the LPD changes on wafers sampled at the same time but stored for 18 months. So, the rapid TDH test appears to be more sensitive to surface changes. Originally appeared in the March 2001 issue of Micro Magazine. © Canon Communications 2001 Although 6 & 18 month tests clearly demonstrate the satisfactory storage life of silicon wafers, they do not provide rapid feedback to process engineering that could quickly signal a potential change in wafer storage life and allow corrective action to be taken. Therefore, a rapid TDH test that uses water and temperature as accelerants was developed for this purpose. Rapid TDH tests can be designed to produce intense TDH on almost any wafer, but this test was designed to create a weak but significant LPD increase on average wafers but extremely high counts on contaminated wafers. For example, the effect of 1015 C atoms/cm2 of the aforementioned cleaning agent was easily observed by the rapid TDH test (Figure 9). The rapid TDH samples were pulled at the same time as the storage life samples. The results are shown in the trend chart in Figure 10. Small increases in LPD changes that were observed by this test were not detected on the parallel sample set used for 6 & 18 month storage, so the rapid test may be slightly more sensitive than the storage test. Even given the low levels of surface contamination on commercially available silicon wafers, the potential to form millions of 0.12um particles is great. For example, 1012 molecules/cm2 of NH4 Cl could form 1,000 particles/cm2 or 300,000/wafer under ideal conditions. Similarly, 1015 C atoms could form 100,000 0.12um particles/cm2 . However, this usually doesn’t occur without accelerants such as condensed water. We propose the following general mechanism for TDH formation. 1. The wafer is contaminated with water-soluble ions or water-soluble organic molecules. 2. Other organic molecules also deposit on the wafer, making it more hydrophobic. 3. Changes in relative humidity caused by temperature changes and/or changes in total water content in the packaged cause water to condense on the wafer surface. 3. The surface water dissolves the water-soluble contaminants. 4. The hydrophobic surface causes the water to form microscopic droplets. 5. The micro-droplets evaporate and leave individual TDH defects. The diameter of a micro-droplet of NH4 Cl solution could be as small as 2.5X that of the particle from which it forms before the salt begins to precipitate from the droplet. This would inhibit the formation of extended films or residues that would not be detected as LPDs. Conclusions Wafers can be consistently stored for up to 18 months without any detectable degradation in >0.12um LPDs, organics, oxide thickness, surface metals and surface ions relative to freshly-packaged wafers. If proper precautions are taken to minimize humidity in the package and maintain package integrity, TDH will not form. The typical average levels of water-soluble inorganic ions on freshly packaged silicon wafers already provide a huge potential to form TDH. But this potential is seldom realized if humidity in the package is controlled since, for normal contamination levels, condensed water is needed to accelerate the process of TDH formation. Therefore, water solubility of the impurities seems to be an important parameter as noted by other investigators (1). Originally appeared in the March 2001 issue of Micro Magazine. © Canon Communications 2001 Organic or ionic wafer contamination on the wafer surface that exceeds the typical levels shown in this report may cause TDH to form even if package humidity is properly controlled at current levels. For example in our specific case, an organic cleaning agent caused TDH formation even though package integrity (and therefore relative humidity inside the package, also) was maintained. Our proposed mechanism suggests that the formation of micro-droplets of water on partially hydrophobic surfaces leads to highly localized precipitation of water-soluble contaminants on the wafer. The density and size of these micro-droplets will be affected by organic contamination and by the relative humidity in the package. Given that organic molecules from plastic packages and some inorganic ions such as NH4 + and Cl- are ubiquitous, it may be necessary to take further measures to control relative humidity in the shipping package in order to avoid TDH formation as customers begin to inspect for 40- 90 nm LPDs. References 1. N. Munter, B. Kolbesen, W. Storm and T. Muller, “Preparation and characterization of time dependent haze on silicon surfaces”, Proceedings of UCPSS 2000, Oostende, Belgium, pp. 91- 29. 2. K. Vepa, J. D. Dowdy, E. J. Mori and L. W. Shive, Proceedings of ECS Spring Meeting, The Electrochemical Society, 93(1), 1141 (1993). 3. P. Sun, M. Adams and T. Bridges, “Monitoring organics on wafers surfaces using thermal desorption GC-MSD/AED,” MICRO, March 2000. 4. P. Sun, M. Adams, L. Shive and S. Pirooz, “Molecular and Ionic Contamination Monitoring for Cleanroom Air and Wafer Surfaces” in In-Line Characterization Techniques for Performance and Yield Enhancement in Microelectronic Manufacturing, SPIE Vol. 3215, D. K. DeBusk and S. Ajuria, eds., October 1997, pp. 118- 127. 5. P. Sun and M. Adams, “Demonstrating a contamination-free wafer surface extraction system for use with CE and IC”, MICRO, April 2000. 6. K. Vepa, K. Baker and L. Shive, “A Method for Native Oxide Thickness Measurement” in Cleaning Technology in Semiconductor Device Manufacturing IV, The Electrochemical Society, Inc., R. Novak and J. Ruzyllo (Eds.), 95-20, pp. 358- 365 (1995). 7. F. Sugimoto and S. Okamura, J. Electrochem. Soc., 146(7), 2725 (1999). 8. L. Shive, C. Frey and C. Vitus, “A Probe of Chemical Oxide Growth Conditions”, Proceedings of UCPSS 2000, Oostende, Belgium, pp. 127- 128. Originally appeared in the March 2001 issue of Micro Magazine. © Canon Communications 2001 Acknowledgements Special thanks for the analytical methods development and data collection go to Gary Anderson, Marty Adams, Kenny Ruth, Phil Schmidt, Dr. Andrei Stefanescu, Dr. Peng Sun and Dr. Hao Zhang. Additional thanks go to Mike Tyler and Lillian Rose for development of the accelerated storage test methods and to Gianpaolo Mettifogo for his on-line testing of this method. About the Authors Larry W. Shive, PhD, is an MEMC Fellow in the Epi Technology Department of MEMC Electronic Materials (St. Peters, MO). He has a PhD in chemistry from Texas A&M University. (Shive can be reached at (636)474-5370 or lshive@memc.com. ) Richard E. Blank, Ph.D., is a Senior Engineer in the EPI Technology Department of MEMC Electronic Materials Corporation (St. Peters, MO). He has a Ph.D. in physics from Michigan State University. (He can be reached by phone at 636.474.7322 or email at rblank@memc.com. ) Karen Lamb is a Research Technician in the EPI Technology Department of MEMC Electronic Materials Corporation (St. Peters, MO). (She can be reached by phone at 636.474.5534 or email at klamb@memc.com. ) Originally appeared in the March 2001 issue of Micro Magazine. © Canon Communications 2001 A List of Figures Figure 1. A whole wafer LPD map showing a localized pattern of TDH. Originally appeared in the March 2001 issue of Micro Magazine. © Canon Communications 2001 Figure 2. An AFM image of a 4um X 4um region of a wafer that exhibited TDH. Typical defect size is 5- 25 nm height and 100- 500 nm width. The estimated defect density in this region is 108 defects/cm2 . Originally appeared in the March 2001 issue of Micro Magazine. © Canon Communications 2001 Figure 3. Trend chart of the change in >0.12um LPDs after 6 month and 18 month storage. Each point represents the average of five wafers sampled on the manufacturing date shown on the x-axis. The 18-month storage testing began with the late-December ’98 wafers. All points prior to then are for 6-month storage. 100 Change in >= 0.12 mciron LPDs 80 18 month storage 60 UCL 40 20 CL 0 -20 -40 LCL -60 1/20/97 8/8/97 Manufacturing Date 2/24/98 9/12/98 CL = 2.0 UCL = 48.9 3/31/99 LCL = -44.8 Table 1. A list of common surface contaminants found on cleaned & packaged wafers. Organics • TXIB, 2,2,4-trimethyl-1,3-pentane diol Diisobutyrate (plasticizer) • DBP, dibutyl phthalate (plasticizer) • DOP , dioctyl pathalate (plasticizer) • BHT (antioxidant) • 2,6-di-t-butyl-4-methylene-2,5cyclohexadiene-1-one (oxidized BHT) • 2,6-di-t-butyl-1,4-benzoquinone (oxidized form of antioxidant) • cyclic polydimethylsiloxanes ((Si(CH3)2O)n-), n=5-10 Inorganic Ions • NH4+ • Cl• SO4= • NO3• NO2• F- Originally appeared in the March 2001 issue of Micro Magazine. © Canon Communications 2001 Originally appeared in the March 2001 issue of Micro Magazine. © Canon Communications 2001 Figure 4. Trend chart of total desorbable surface organics after 6-month and 18-month storage. Each point represents the average for two wafers sampled after 6 or 18 months of storage. The original manufacturing date is shown on the x-axis. The y-axis is 1014 Catoms/cm2 . The 18-month storage testing began with the late-December ’98 wafers. Freshly packaged wafers (<30 days) typically have 0.5- 1.5 X 1014 C atoms/cm2. 14 E14 C atoms/cm 2 12 18 mo storage 10 8 UCL 6 4 2 Cl 0 06/19/97 01/05/98 Manufacturing date 07/24/98 Cl = 1.6 02/09/99 UCL = 6.8 Originally appeared in the March 2001 issue of Micro Magazine. © Canon Communications 2001 Figure 5. Chromatograms of organics desorbed from wafers after six months of storage. The retention times are on the x-axis and the signal intensity is on the y-axis. The dominant signals in the left chromatogram are a quinone(15.87 minutes), an isobutyrate (24.12 min) and a cyclic methylsiloxane (10.3 min). These are also present in the right chromatogram but are overshadowed by signals from an organic cleaning agent (8.46 min., 12.26 min. and 15.10 min.) that has contaminated the wafer. Originally appeared in the March 2001 issue of Micro Magazine. © Canon Communications 2001 Figure 6. Trend chart for surface ions extracted from wafers after 6 months and 18 months of storage. Each point represents the average of two wafers sampled after 6 or 18 months of storage. The original manufacturing date is shown on the x-axis. The 18-month storage testing began with the late-December ’98 wafers. The trends for ammonia and chloride are shown on the left and right, respectively. Freshly packaged wafers have up to 2.9 X 1013 NH4 + ions/cm2 and 0.8 X 1013 Cl- ions/cm2 . 6000 10 10 ions/cm 2 5000 NH4+ trend 18 month storage 4000 3000 UCl 2000 1000 Cl 0 6/19/97 1/5/98 Manufacturing date 700 7/24/98 2/9/99 Cl = 611 UCl = 2300 Cl- trend 600 10 10 ions/cm 2 18 month storage UCl 500 400 300 200 Cl 100 0 6/19/97 1/5/98 Manufacturing date 7/24/98 2/9/99 Cl = 206 UCl = 618 Originally appeared in the March 2001 issue of Micro Magazine. © Canon Communications 2001 Figure 7. The area of a wafer affected by TDH as a function of water volume added to the wafer package in a rapid TDH test. Five wafers were equally spaced in slots 1- 25 in each box. Originally appeared in the March 2001 issue of Micro Magazine. © Canon Communications 2001 Figure 8. Trend chart for whole-wafer average oxide thickness after 6 month and 18 months of storage. Each point represents the average of two wafers sampled after 6 or 18 months of storage. The original manufacturing date is shown on the x-axis. The 18month storage testing began with the late-December ’98 wafers. Freshly packaged wafers have <10 Angstroms of oxide. Oxide thickness, Angstroms 14 13 UCl 12 11 10 Cl 9 18 month storage 8 7 6 1/5/98 4/15/98 7/24/98 Manufacturing Date 11/1/98 2/9/99 5/20/99 Cl = 9.6 UCL = 12.9 Originally appeared in the March 2001 issue of Micro Magazine. © Canon Communications 2001 Figure 9. Wafer map of >0.12um LPDs after an accelerated TDH test on a wafer contaminated with 1015 C atoms/cm2 of an organic cleaning agent. Originally appeared in the March 2001 issue of Micro Magazine. © Canon Communications 2001 Figure 10. Trend chart for the change in LPDs after an accelerated TDH test. Each points the average of three wafers sampled immediately after the manufacturing date shown on the x-axis. The samples were taken at the same time as those shown in Figure 3. Change in >= 0.12 LPDs 120.0 UCL 100.0 80.0 60.0 40.0 CL 20.0 0.0 1/20/99 2/9/99 3/1/99 Manufacturing Date 3/21/99 4/10/99 4/30/99 CL = 25.5 UCL = 102.8 5/20/99 6/9/99