 |
 |
 |
 |
|
Comments? Suggestions? Send us your feedback.
MicroMagazine.com |
FACILITIES TECHNOLOGIESConducting real-time monitoring of airborne molecular contamination in DUV lithography areasOleg P. Kishkovich, Extraction Systems Wafers are susceptible to molecular base contamination from the moment resist is coated until the postexposure bake step. Elevated concentrations of molecular bases cause image degradation (for example, T-topping), reduce production output, and necessitate wafer rework, as depicted in Figures 1 and 2. Some chemically amplified resists may be affected by impurities in the air at very low levels down to 1 or 2 ppb. Ammonia and other basic volatile compounds can affect the performance of resists used to capture sub-0.3-µm images with deep ultraviolet (DUV) lithography.1
In comparison, typical outdoor levels of ammonia and other basic airborne impurities range from 10 to 50 ppb; the ammonia concentration in human exhalation exceeds 100 ppb for nonsmokers and 1000 ppb for smokers; and the ammonia in body odor may exceed 100 ppb. Thus, air quality monitoring in the semiconductor industry is essential to protect the DUV process from molecular base contamination. This article discusses the critical airborne molecular contaminants found in DUV photolithography facilities. It presents an overview of real-time monitoring methodologies that can detect and measure these contaminants at low concentrations, enabling users to anticipate and resolve airborne molecular contamination (AMC) challenges before production problems arise. Citing examples of how real-time monitoring is being employed in production fabs, the article compares the different monitoring technologies, examines single-point investigative and multipoint process monitoring strategies, and describes critical points in tracks, steppers, cleanrooms, and air filtration systems that require monitoring. Key Criteria for Air Quality Monitors To be effective, instruments used to monitor airborne molecular contamination should meet the following criteria:
Real-Time Monitors in DUV Litho Facilities Three Monitoring Technologies. The use of real-time monitors for indoor air quality (IAQ) control in DUV facilities and the development of low-level real-time pollution monitors for IAQ applications are relatively new.3 When lithographic facilities recognized that basic air contaminants might affect chemically amplified resist performance, the need for a fast, reliable monitor arose.4 Different parties involved in DUV lithography—IC manufacturers, tool suppliers, chemically amplified resist vendors, and chemical air filtration companies—were interested in developing such a monitor. However, no coordinated and centralized effort to develop an industry standard for AMC metrology has been undertaken. As a result, three different core technologies, illustrated in Table I and Figures 3—5, are used for the real-time monitoring of basic air contamination in cleanrooms: ion chromatography (IC), ion mobility spectroscopy (IMS), and chemiluminescence (CL).
Table I: Comparison between the performance capabilities of three real-time monitoring technologies.
While IC offers excellent sensitivity, it does not discriminate gaseous ammonia from particulate ammonium salt, cannot easily separate and quantify amines, does not provide data in real time, requires frequent intensive maintenance, and needs a continuous supply of corrosive eluent. It is a powerful analytical tool for R&D and other applications where parts per-billion sensitivity and amine separation capability are essential. The ion mobility spectroscope is a simple, durable instrument with few moving parts that does not require liquid chemicals, high-voltage sources, or high-temperature zones (over 400°C). However, it contains a radioactive source. Because IMS calibration is affected by temperature, humidity, and sample composition, this technique tends to respond differently to the same concentrations of different amines. It cannot separate and identify mixed samples. Moreover, two separate instruments are needed to measure ammonia and normal-methylpyrrolidone (NMP). Typically used as a "low/high gauge," IMS is an effective troubleshooting technique and leak detector. The most suitable technology for real-time AMC monitoring in DUV photolithography facilities is the chemiluminescent approach. In chemiluminescence, light is produced as a result of a chemical reaction. This method was developed to detect nitric oxide (NO) at very low concentrations.5 CL detects all molecular bases that may affect chemically amplified resist performance. CL responds uniformly to equal concentrations of different analytes and responds linearly to a concentration range over five orders of magnitude from the sub-parts-per-billion level to > 20 ppm. Calibration is simple and may be performed with a built-in autocalibrator. However, CL cannot identify individual molecular bases and requires a vacuum pump. The analyzer exhaust contains a high ozone concentration and requires scrubbing. In addition, the catalytic converter may be poisoned by high concentrations of phosphorus-containing compounds (> 1000 ppb). Its preferred usage is as a stationary multipoint instrument for process monitoring in 248- and 193-nm production facilities. A CL instrument developed by Extraction Systems (Franklin, MA) detects and provides information on the total concentration of basic nitrogen-containing compounds in air samples. This tool detects NO molecules by detecting the light emitted by excited NO2* molecules, which are the product of the chemiluminescent reaction: NO + O3 — NO2* + O2
This occurs in the reaction chamber, in which the sample flow mixes with the ozone flow produced by an ozonator. A nitrogen converter, placed upstream from the reaction chamber, converts all nitrogen- Single- Versus Multipoint Monitors. Regardless of the detection principle in use, molecular base monitors used in DUV facilities are either single- or multipoint instruments. Single-point monitors are usually used as mobile troubleshooting tools that are relocated frequently within a facility or even to remote fabs. They are small, relatively simple, and relatively inexpensive. In contrast, multipoint stationary monitors such as those used with tracks and steppers are ordinarily fixed instruments in a facility. They contain a multiplexing device that allows them to switch automatically between different sample locations, associated control equipment, and data-acquisition systems. In theory, the number of channels (sampling points) on a monitor is unlimited. However, for process monitoring or control it is most practical to have 10 sampling points per monitor. If each point performs sampling for 10 minutes, the monitor returns to the same location every 100 minutes—a theoretically maximum amount of time for a contamination event to occur without detection. Using two such monitors increases the number of sampling points and further minimizes potential yield loss caused by contamination. Calibration Real-time air monitoring instruments are only as accurate as their calibration, which should be performed every time they are unplugged and moved from one location to another, and especially after they are shipped. Neglecting calibration may lead to erroneous results. For example, when IMS was first used to detect ammonia in DUV facilities, it was erroneously thought that IMS monitors did not require calibration after shipment or reinstallation. Inaccurate ammonia threshold readings caused by failure to calibrate monitors convinced manufacturers to stop claiming that their monitors are "calibration-free."6 To perform accurate multipoint calibration, calibration gas mixtures are required that should be certified or at least verified by an appropriate analytical technique. They should cover the range of concentrations expected in the cleanroom and, most importantly, should include zero air to calibrate the instrument's true zero. Bottled gas mixtures at low concentrations are not desirable. Because the highest purity of commercially available bottled zero air is 99.9999%, possible uncontrolled and uncertified impurities may reach 1000 ppb. It is preferable to prepare a zero air source on-site by means of proper filtration technology.7 Three alternative technologies exist for the generation of low-parts-per-billion-level calibration mixtures: dynamic-dilution (DD), permeation-type (PT), and diffusion-type (DT) devices. Dynamic-Dilution and Permeation-Type Devices. Dynamic-dilution modules use a high-level calibrated gas mixture (or even neat gas) diluted to the required level. To achieve a low parts-per-billion level, more than one dilution step is required. In its traditional application, DD uses a bottled gas mixture. However, amines and amides are extremely reactive, and bottled mixtures have a limited shelf life. Permeation-type devices are sealed containers filled with liquid chemicals, which diffuse slowly through a permeable membrane that is typically made of PFA Teflon. At a constant temperature there is a constant amount of pressure inside the container, so that the chemical permeates the membrane at a constant rate. The PT device is placed in a controlled flow of zero air to produce a calibration mixture. PT devices for ammonia, NMP, and an array of organic amines are available from various suppliers. An important advantage of these devices is that they easily produce calibration mixtures in the desired concentration range of 10 to 100 ppb at a flow rate of ~1 L/min, which is typical for the real-time monitors discussed here. Another advantage is that they come with National Institute of Standards and Technology—traceable calibration certificates. These features have made PTs very attractive for calibrating different types of low-level real-time gas monitors for ammonia, NMP, and organic amines.8 Nevertheless, these devices have limitations. The calibration procedure for PT devices is based on two assumptions: the permeation rate must not change over time and there must be no chemical conversion of the compound sealed in the PT. However, for very reactive compounds, such as amines and NMP, both assumptions are incorrect. Bench-top ion chromatography and gas chromatography/mass spectrometry (GC/MS) analysis have been used to validate multiple PTs from different manufacturers. Only for ammonia PTs were the certified permeation rates close enough to be determined by analysis. For other amines and NMP, the results were very different. Table II lists some discrepancies between an actual calibration mixture composition and a composition expected from a PT device calibration certificate. While it is considered an industry standard for generating low-level gas calibration mixtures, PT has turned out to be very tricky, if not useless, for such reactive compounds as amines and amides (NMP).
Table II: Output of certified PT devices, as measured by different types of monitors. Experience shows how challenging it can be to prepare reliable, low-level amine standards. In one case, investigators used PT devices to investigate whether an IMS monitor could detect molecular bases other than amines and ammonia.8 However, no additional analytical work was performed to quantify and qualify the PT output. As a result, an ammonia signal was possibly mistaken for an amine signal, which is expected from the gas mixture (Table II). Without substantial analytical evidence, it was concluded that the instrument detected the "total amine" level. Experienced analytical personnel using proper analytical equipment must qualify calibration mixtures and zero air sources. Without these precautions, erroneous and confusing results may be obtained. Diffusion-Type Devices. The third alternative for generating low-parts-per-billion-level calibration mixtures is diffusion-type devices, small glass or quartz containers filled with a neat chemical and an attached capillary. A DT instrument is placed in an oven and maintained at a constant temperature to provide a constant partial pressure of the analyte inside the device. The capillary acts as a diffusion restrictor, providing a low and constant flow of the chemical from the DT. These devices can generally be used with the same types of calibrators as PT devices and can be refilled with fresh chemicals when necessary. However, the refill procedure requires expert care. The use of DT devices does not completely eliminate problems associated with amine reactivity. Interference All analytical technologies are subject to some form of interference. An instrument that is perfectly stable and reproducible under controlled laboratory conditions may show erroneous results after being moved into a real fab environment. This is especially true of low-level air quality monitors. Depending on the concentration level of an interfering chemical, the monitor may show a reading higher (positive interference) or lower (negative interference) than the actual concentration of measured compounds. Also, the physical mechanisms of different forms of interference may differ from one another, thus producing false signals and altering an instrument's response to a measured chemical. In a typical production fab environment, the concentration levels of multiple chemicals in cleanroom air are sometimes orders of magnitude higher than expected. For example, production fabs have reported that when occasional organic solvent spills take place, real-time monitors detect elevated concentrations of ammonia downstream from the chemical filters protecting the DUV clusters. In most cases, such a high challenge of organic compounds has been considered an indication of premature filter failure. However, organic spills may cause NH3 monitors to produce false positive readings. IC, IMS, and CL monitors are subject to different forms of interference. IC-based molecular base monitors may be considered the most robust and are affected only by ionic compounds. Interference from particulate ammonium salts is also possible. IMS-based monitors, on the other hand, may be affected by multiple classes of chemicals, so that both positive and negative interference is possible. Sample train pressure and flow may also affect instrument response. CL-based analyzers are susceptible to interference from nitrogen-containing compounds. If calibration parameters are not balanced properly, those compounds may cause both positive and negative interference. Pressure and flow changes inside the reaction chamber also may affect monitor response. Thus, it is recommended that an autocalibration option be implemented if a CL instrument is used for process control. With autocalibration, all deviations from ideal calibration status can be detected and corrected automatically. Sampling Line and Sampling Time Problems All three types of molecular base monitors are affected by sampling line problems, in which ammonia and other amines are readily adsorbed by virtually any kind of surface as the sample is transferred through the sampling line by a continuous cycle of adsorption and desorption. Some sampling line materials perform much better than others. Continuous quartz tubing heated to 500°C would be the material of choice if it were not so inflexible. Teflon PFA tubing, while flexible, is porous and permeable, making it one of the least desirable materials for this application. When it is exposed to water vapor, it may generate hydrogen fluoride (HF), which reacts with ammonia and other molecular bases and—especially at low concentrations—may remove the compounds of interest. The chemical properties of the final product strongly depend not only on the grade of the raw material but also on how closely manufacturing process parameters are observed. As shown in Figure 6, the quality of one manufacturer's PFA tubing may vary from acceptable to unusable within a lot and from lot to lot. In practice, it is advisable to test every 100-foot piece of sampling tube for dynamic and static response to different molecular bases. The rejection rate is close to 50%.
Two time-related parameters are used to describe the behavior of real-time air monitors: lag time and response time. Lag time is the period between the actual concentration change (assuming this change is very fast) and the moment the first sign of this change is detected by the monitor. Response time is the period between when a change is detected and when it reaches 90% of its final value. The length of a sampling line has a direct effect on sampling time. It may take more than an hour for a monitor with a 100-ft sampling line to respond to the presence of contaminants when the concentration level changes from 1 to 2 ppb. Even more time may be required for a monitor reading to stabilize when the concentration level changes from a high of tens of parts per billion to a low of < 2 ppb. As a result, in applications in which quick response time is critical (such as in leak detection and troubleshooting), the shortest possible length of sampling line should be used. For stationary multipoint monitors, where time resolution is not critical, it is reasonable to change sample locations every 10 to 15 minutes in a typical fab environment. In general, shortening this time is not recommended. When switching from high to low readings, part of the sampling train, common to all locations, must have enough time to desorb previously absorbed high concentrations of molecular bases. Real-Time Monitor Use in Practice Real-time monitoring is used to characterize AMC in the cleanroom, process tools, and air filters. All lithography process tools used for exposing chemically amplified resists and developing images use some form of chemical air filtration, either integrated into the tool enclosure or mounted externally. The objective is to filter the cleanroom makeup air before it contacts the sensitive resist. If air in the tool is recirculated inside the tool, this air also is filtered to remove internally sourced pollutants and pollutants that were not removed by the makeup filter. It has been shown that some types of chemical filtration, when used alone, may not adequately protect the process. Limited filter removal efficiency and filter degradation require precise monitoring. For example, in one instance an IMS-type ammonia monitor was used to measure concentrations in the cleanroom and inside both the exposure chamber and the development track.7 By means of a trap-sampling and the IMS-type ammonia monitor, a general range of 20—30 ppb of ammonia, NMP, and a momentarily higher challenge because of a spill were found. It was discovered that 88% filter removal efficiency is maintained after 900 hours. Thus, given a recurrence of a 240-ppb spill, a contaminant level of 29 ppb would remain in the tool. At 10-ppb normal ambient, this would be detected as 1.2 ppb. Under such circumstances, it is essential that the monitor be stable and precise.
In another case, Cypress Semiconductor characterized the total base challenge in the cleanroom and identified a small amount of AMC challenge to the tool filtration. In this example, the results of which are illustrated in Figure 7, a chemiluminescence-type monitor with one sample port was used to identify a total base concentration of 1-ppb background concentration and a 7-ppb momentary spike.
Similarly, Hewlett-Packard's fab in Palo Alto, CA, used a CL-type single-point monitor to detect increases in the cleanroom total base concentration and correlate the events to a hexamethyldisilazane refill, as shown in Figure 8. The variation in image stability, as measured with expose-PEB tests in the exposure chamber of a Nikon S201A stepper and a DNS 80B development track, illustrates the value of monitoring multiple locations for diagnostic purposes.10 Using trap sampling with IC and GC/MS analysis, VLSI Technology in San Antonio, TX, conducted initial air sampling tests and identified ammonia, monomethylamine, trimethylamine, diethylaminoethanol, and monoethanolamine in a photolithography bay. At the same time, a single-point portable monitor was used to map the cleanroom, chase, and surrounding areas. Total molecular base concentration ranged from 12 to 23 ppb in various locations in the cleanroom and adjacent chase, as depicted in Figure 9. Figure 10 shows that based on these findings a 10-point CL-type monitor was installed to monitor the cleanroom, exposure tool chamber, and development track enclosure. The monitor was installed to track and ensure acceptable molecular processing conditions and optimize tool filter life. Exposure- and development-tool air quality averaged 0—2 ppb with occasional excursions, as illustrated in Figure 11. The event lasted 2 hours and coincided with tool servicing. Thus, in this case the monitor successfully identified an event while no process was at risk.
CL-type real-time monitors are also of potential use in so-called resist torture chambers, in which the environmental stability of DUV positive tone 193-nm photoresists is studied. This is part of an ongoing project being conducted by International Sematech's Resist Technology Center in coordination with Extraction Systems. Conclusion This article has surveyed the prevalent approaches to monitoring air quality in the DUV photolithography environment. Of the three core technologies—ion chromatography, ion mobility spectroscopy, and chemiluminescence—the last lends itself especially well to production monitoring in DUV facilities. Furthermore, factors such as calibration, interference, sampling line problems, and sampling time influence the effectiveness of real-time AMC monitoring. References 1. SA MacDonald et al., "Airborne Chemical Contamination of a Chemically Amplified Resist," SPIE: Advances in Resist Technology and Processing 1446, no. VIII (1991): 2—12. 2. D Kinkead et al., "Forecast of Airborne Molecular Contamination Limits for the 0.25 µm High Performance Logic-µm Process Generation," Sematech Technology Transfer Document 95052812A-TR (Austin, TX: Sematech, 1995). 3. O Kishkovich and M Joffe, "Real-Time Low-Level Pollution Monitoring for IAQ Applications," in Proceedings of the International Specialty Conference on Measurements of Toxic and Related Air Pollutants (Research Triangle Park, NC: Air and Waste Management Association, 1996), 509—520. 4. K Dean and R Carpio, "Real-Time Detection of Airborne Contaminants in DUV Lithographic Processing Environment," in Proceedings of the Institute of Environmental Sciences (Mount Prospect, IL: Institute of Environmental Sciences, 1995), 9—16. 5. O Kishkovich et al., "Real-Time Methodologies for Monitoring Airborne Molecular Contamination in Modern DUV Photolithography Facilities" (paper presented at SPIE Microlithography '99, Santa Clara, CA, March 14—19, 1999). 6. PN Caugh and BA Thrust, Transactions of the Faraday Society 63, no. 915 (1967). 7. JC Vigil et al., "Contamination Control for Processing DUV Chemically Amplified Photoresists," in Proceedings of SPIE's International Symposium on Microlithography (Bellingham, WA: International Society for Optical Engineering, 1995), 626—643. 8. O Kishkovich et al., "An Accelerated Testing Technique for Evaluating Performance of Chemical Air Filters for DUV Photolithographic Equipment" (paper presented at SPIE Microlithography '99, Santa Clara, CA, March 14—19, 1999). 9. T Bacon et al., "Contamination Monitoring for Ammonia, Amines, and Acid Gases Utilizing Ion Mobility Spectroscopy (IMS)." Molecular Analytics Web page at http://www.ionrpo.com. 10. W Conley et al., "Total Molecular Base Monitoring," Future Fab International 3, no. 5 (1998): 213—215.
Oleg P. Kishkovich, PhD, is the director of technology and principal scientist at Extraction Systems (Franklin, MA), where he oversees method development for molecular contamination monitoring and product analysis for the chemical air filtration of DUV lithography environments. With expertise in chemical kinetics, atmospheric chemistry, trace gas analysis, and magnetic resonance techniques, he has given presentations at international conferences on AMC measurement and control in ultraclean environments and has published several papers on the real-time monitoring, measurement, and control of AMC. He received his PhD in chemical and molecular physics from the Moscow Institute of Chemical Physics and an MS in engineering physics from the Moscow Institute of Physics and Technology. (Kishkovich can be reached at 508/553-3900, ext. 31, or okishkovich@extractionsystemsinc.com.)
MicroHome |
Search | Current Issue | MicroArchives Questions/comments about MICRO Magazine? E-mail us at cheynman@gmail.com. © 2007 Tom Cheyney |
Figure 1: Severely T-topped Apex E test wafer at ambient levels of only an 8—10-ppb concentration of amines.
Figure 2: SEM cross section of resist-coated wafer exposed to a 24-ppb concentration of amines for 11 minutes.
Figure 9: Total molecular base concentrations (in ppb) in a production fab (map indicates sample locations).
Figure 10: Sample locations in a new DUV tool cluster.
Figure 11: Molecular base rapid thermal multiprocessing test detected an 18-ppb molecular base spike for 2 hours inside a DUV production tool. 