 |
 |
 |
 |
|
Comments? Suggestions? Send us your feedback.
MicroMagazine.com |
ULTRAPURE GASESUsing an in-line monitor to obtain real-time moisture measurements(First of two parts)Jian Wei, John E. Pillion, Steven M. King, and Matt Verlinden, Millipore Widely considered the most significant contaminant in semiconductor process gases, moisture can cause particle formation and may result in electrical property defects that adversely affect device yields. The risk of moisture-related electrical damage is steadily increasing as linewidths continue to shrink below 0.5 µm and more process steps are required. In addition, in corrosive and reactive gases, moisture contamination can severely damage gas lines and related components (e.g., mass-flow controllers, valves, and pressure regulators). Moisture in a gas line can also be an indication that a leak has occurred. Several techniques are useful for detecting moisture in process gases. However, many characteristics of these methods—such as bulky instruments, slow measurement speed, and high capital costs—limit their effectiveness for in-line monitoring of gas purity. In sputtering applications, for example, gas flow may last for only 30 seconds, so detection of moisture within that time is a requirement for an in-line device. Therefore, many semiconductor manufacturers have expressed a desire for a moisture detection tool that more directly addresses the industry's emerging requirements for continuous monitoring in process applications with zero contamination tolerance. After briefly reviewing the available measurement techniques, this article describes the development of an in-line monitor that provides continuous, accurate readings of moisture levels in inert gases in real time. The device, which is readily integrated with existing process controls, combines this advanced functionality with such ease-of-use features as rapid start-up, compatibility with existing gas lines, and output signals calibrated in parts per billion. Current Monitoring Techniques The systems used to monitor moisture in process gases range from highly accurate, and expensive, atmospheric pressure ionization mass spectroscopy (APIMS) analyzers to a variety of less accurate, less costly alternatives.1—3 Two widely used detection devices are coulometers and polymer-coated quartz crystal microbalance (QCM) sensors. Other types of non-APIMS analyzers include capacitive and dew-point hygrometers. Measuring approximately 6 x 4 x 3 ft, APIMS units are designed primarily for laboratory use. Typical commercial systems operate by ionizing a sample of gas at atmospheric pressure and then routing the sample to a mass analyzer. These instruments are extremely sensitive, allowing measurement in the parts-per-trillion range for a variety of contaminants. They also have a fast response. However, APIMS is not an absolute measuring technique, as are dew-point hygrometers, and analyzers must be calibrated using a moisture-generating device. The analyzers are also an expensive option for monitoring gas purity. The approximate per-unit cost ranges from $100,000 to $500,000, and analyzer operation usually requires the presence of a specially trained technician. While an APIMS analyzer could conceivably be configured on a process gas line, such an installation would be very time-consuming and costly. Coulometers use a hygroscopic coating of phosphorous pentoxide between two wire electrodes as the sensing element. The electrodes electrolyze the moisture that the coating retains from the gas sample into hydrogen and oxygen. The current formed in this electrolysis process is directly related to the amount of moisture in the gas stream. Coulometric analyzers provide a reasonably fast response to rising moisture levels, but their drydown time can be lengthy, causing monitoring gaps. These instruments also require a constant supply of power to maintain the integrity of the sensor. Some coulometers incorporate an internal moisture source that prevents damage to the sensor cell when the unit is used for long periods in very dry gas. This feature also improves the unit's sensitivity, but it makes this type of analyzer unsuitable for in-line applications. For QCM sensors with a hygroscopic polymeric coating, the sensor crystal is periodically exposed to the sample gas and then to an internally generated reference gas. Moisture in the sample gas is adsorbed by the hygroscopic coating, causing an increase in the mass of the crystal, which, in turn, decreases its oscillation frequency. When the sample gas is turned off and the crystal is exposed to the reference gas, the resulting change in crystal frequency is measured. The cycle is then repeated once each minute. The differences in crystal oscillation frequency between the reference and sample gases are used to compute the moisture in the sample gas. Because most late-model polymer-coated QCM instruments generate their own internal moisture challenge for the reference gas, they are not suitable for in-line applications. Capacitance sensors have a sandwich-type structure: two metal electrodes are separated by a moisture-sensitive dielectric film material, typically alumina or silicon dioxide. Adsorption of moisture by the dielectric film will change its dielectric constant, resulting in a change in the capacitance of the device. This capacitance change is then correlated to the moisture concentration in the gas stream. Flow-insensitive but pressure-dependent, capacitance sensors can be quite small and can be used for in-line applications. However, they respond extremely slowly because moisture in the gas phase and moisture on the dielectric must reach equilibrium before a stable signal is obtained. Dew-point hygrometers measure the temperature at which moisture changes its phase on the surface of a cooled mirror. This parameter is directly related to the concentration of moisture in the gas sample. When moisture in the gas stream forms a dew or frost film on the mirror, light from a high-intensity LED source is scattered by the film, and its intensity is measured by a light detector. As the moisture concentration in the gas stream changes, the temperature of the mirror is varied to maintain a constant dew or frost film thickness and light-scattering intensity. Like capacitive sensors, chilled-mirror hygrometers have a slow response speed because they require that an equilibrium be established, in this case between moisture in the gas phase and the film formed on the mirror. In addition, most, but not all, include a bulky refrigeration system. On the positive side, dew-point hygrometers do not contaminate the gas stream, and they can be used as a secondary standard for moisture concentration accuracy measurements. Gas contamination can also be determined by a variety of optical methods, such as Fourier transform infrared (FTIR), diode laser, and intracavity laser spectroscopy.4 FTIR spectroscopy measures the absorbance of infrared radiation by a contaminant in a gas sample as it passes through a sample cell. The change in absorbance is used to determine the concentration of the contaminant. FTIR spectrometers are relatively large, measuring about 3 ft long by 2 ft wide, with a 1-ft-long sample cell. They require calibration and purging with a dry gas, but can measure a range of contaminants in both corrosive and inert gases. Such systems' need for precise optical alignment makes their use in a fab environment not practical. While the non-APIMS moisture detectors described above are less expensive than APIMS systems, their minimum cost is quite high, approximately $30,000. Like APIMS, most of these devices are large and, when used off-line, provide intermittent monitoring. Originally developed for applications where critical moisture levels are relatively high (parts-per-million), some of the detectors also have significant sensitivity limitations, encountering difficulty providing accurate readings under 10 ppb, with slightly better, but still suboptimal, sensitivity at levels between 10 and 20 ppb. In addition, systems with slower response times may take minutes to several hours to react to the presence of single-digit parts-per-billion moisture. By the time these instruments detect the potential for damage, the damage may already have been done. Therefore, none of the currently used techniques is adequate for achieving the goal of continuous, real-time monitoring needed for contamination-free manufacturing. In-Line Monitor In contrast to other moisture detection devices, the new in-line monitor was specifically designed to measure H2O in inert gas distribution systems continuously at the point of use, ensuring a predictable high-purity gas supply at the tool, where it is essential. Leaks or other causes of contamination occurring within the distribution system can be detected and corrected before serious problems occur. Output signals calibrated in parts per billion can be provided in about 1 minute, allowing rapid response to contamination events. In addition, because the monitor works in the gas stream and recovers immediately after moisture detection, there are no unmonitored intervals when moisture can pass undetected. This capability guards against running critical wafer-processing steps in out-of-spec conditions.
Table I: Performance specifications of the in-line monitor. With a footprint similar to that of a mass-flow controller, the in-line monitor (which measures 4.85 in. high, 1.65 in. deep, and 5 in. long) (see Figure 1) is smaller than other moisture analyzers, which are benchtop devices or larger. Together with its small size, the system's relatively low cost (<$10,000 per unit) makes expanded monitoring of process gases feasible, particularly at the points of use. The principal performance specifications of the new monitor are summarized in Table I; Table II compares its performance, cost, and size characteristics with three existing technologies: APIMS analyzers, QCM sensors with polymer coating, and coulometers.
Table II: Performance, cost, and size comparison of the in-line monitor with other moisture detection systems.
The sensor technology used in the monitor consists of a piezoelectric QCM sensor that has been sputter coated with a reactive-metal thin film and attached to gold electrodes. During operation, an oscillating voltage is applied to the microbalance so that it vibrates. When moisture in the sampled gas stream reacts with the metallized quartz crystal, the result- ing mass change leads to an alteration in the oscillation frequency of the crystal. The device's onboard electronics compare the rate of frequency change with the frequency of a reference crystal to calculate mathematically the moisture level in the gas, and then output a measurement signal to a computer display or to conventional analog process controls. A principal difference between the performance of the in-line monitor, which relies on a metallized QCM sensor, and QCM sensors with polymer coating is that the latter require time for the desorption of moisture from the sensor. The new monitor does not rely on the desorption of H2O from the crystal surface. Its thin reactive-metal sensor coating reacts with moisture irreversibly, resulting in a very fast response.
In addition to the metallized QCM sensor, the in-line monitor integrates three sets of components—electronics, software, and housing—in a single unit (see Figure 2). As described above, the onboard electronics interpret changes in the sensor crystal's oscillation frequency, calculating moisture levels in single-digit parts per billion. The monitor can be interfaced with both digital and analog control systems, and can activate an alarm to alert operators to out-of-spec processing conditions. If desired, the output signal also can be displayed via the Windows-based software (see Figure 3), which simplifies data analysis. Power requirements are nominal (+12 V dc), and available analog signal outputs are 4—20 mA, 0—5 V dc, and RS-232 serial, allowing integration within any fab. The monitor can be installed in-line in a minute or less, and requires no special training to operate. It is calibrated at the factory and does not require recalibration.
APIMS Correlation During the development project, the response of in-line monitor to low levels of moisture was evaluated by connecting it in parallel to an AttoSpec-1 APIMS analyzer (ABB Extrel, Pittsburgh, PA). Dry nitrogen gas for the experiments was prepared using molecular purifiers capable of removing H2O, O2, CO, and CO2 to <100 ppt. Some of the dry gas was passed over a heated water permeation tube to add moisture at a parts-per-million level. The exact moisture level in the nitrogen downstream from the permeation tube was calculated based on the permeation tube emission rate and the gas flow. The moisture level was also checked using a coulometric analyzer. The wet gas was then further diluted with additional purified "zero" nitrogen to form moisture standards that ranged from 0 to 20 ppb by volume, which were used to challenge the in-line monitor. Prior to its transfer to the APIMS system, the moisture standard gas was further diluted by factor of 10 to avoid a saturation/nonlinear response. The gas flow into the in-line monitor was 1 std L/min throughout the study.
Table III: Responses of the in-line monitor (ILM) and an APIMS analyzer to standard moisture challenges. The goal of the experiment was to correlate the response of the calibrated in-line monitor to a series of moisture challenges between 0 and 15 ppbv with the response of an APIMS analyzer to equivalent challenges. The results, which are shown in Table III, indicate that the in-line monitor responded to moisture levels in this range within ±15% and that its response was linear compared with that of the APIMS. This linear response was then used to find the detection limit of the monitor by linear regression analysis, which yielded a limit of <1 ppbv/V with a 95% confidence limit. By comparison, the detection limit of the APIMS system is 60 ppt. It can be concluded from these results that the response of the in-line monitor to a series of moisture standards correlates well with the response of an APIMS analyzer. The accuracy of the calibrated monitor compared with the permeation source was within the specified ±5 ppb or 30% range for the device. Beta Site Testing Beta site testing of the in-line sensor technology was conducted at semiconductor manufacturing fabs and OEMs as well as at gas suppliers. The testing involved metal-sputtering applications, performance validation, and conventional monitoring. It provided invaluable information on the design and performance of the technology to the development team and to the companies that participated. The in-line monitor was bench tested and challenged at these facilities to evaluate its response time and measurement accuracy in comparison with conventional analyzers. In metal-sputtering applications, the monitor were used to detect moisture in ultra-high-purity argon and nitrogen gases at the point of use in the sputtering tools and in bulk gas delivery systems. In addition to verifying the monitor's performance characteristics, the APIMS comparison testing, moisture challenges, and other evaluations yielded data that were helpful to the participating beta sites. Many incidents of higher-than-anticipated moisture levels were detected. For example, when it was used in a metal-sputtering application, the instrument detected unacceptable levels of moisture in the process gas. Point-of-use purification was implemented to solve the problem and the system was successfully qualified. The monitor also provided data that helped engineers to locate leaks in ultra-high-purity process gas lines, allowing repairs to be completed expeditiously. An in-depth analysis of the beta site testing will appear later this year in MICRO. Potential Applications The first generation of the in-line monitoring device was designed for moisture detection in the nitrogen and argon gases used in sputtering applications, where moisture can be extremely detrimental to the process. The next models to be introduced will be capable of detecting moisture in additional inert gases, such as helium, hydrogen, and oxygen, and plans are also under way to develop an in-line device for detecting moisture in corrosive gases. As the applications for in-line monitoring increase, the sensors will likely become standard features on gas lines. In other words, the technology promises to greatly expand the scope of moisture monitoring in process gases, thereby reducing both contamination-related downtime and yield loss. When moisture levels are monitored continuously at multiple locations in a gas distribution system, it will become relatively easy to track down possible leak sites. Since finding leaks can take hours or even days, the potential savings in downtime promises to be significant. Continuous on-line monitoring will also enable the precise quantification of the moisture levels that are acceptable to produce a specific level of device yield and electrical performance. While process engineers have long recognized that lower moisture levels mean better results on the wafer, they were previously unable to correlate the moisture content of the gas that enters a process tool with the quality of the results on the wafer. Once the threshold moisture level for a given process is determined precisely using monitoring data, the in-line monitor will provide real-time measurements indicating whether the moisture in the gas stream is below that threshold. In-line monitors can also be helpful in optimizing the management of gas systems. For example, in a CVD tool that uses a number of gases, the tool is purged at the end of each preventive maintenance cycle to bring down the moisture level, but until now there has been no way to determine the exact moisture level in the purge. Using the in-line monitor in this scenario will provide moisture level data from moment to moment, so operators will know exactly when the purge has reduced the moisture level sufficiently. In some cases, purge times may be reduced, resulting in more uptime for the tool. On the other hand, by ensuring that purge cycles are not ended before moisture levels have returned to specification, use of the monitor will prevent moisture-related wafer damage. Conclusion When the condition of the gas entering a semiconductor process tool is unknown, process control is like the proverbial shot in the dark. The recently developed in-line monitor can provide some needed light in the form of data on moisture contamination in the gas, which is known to cause damage to both in-process devices and equipment. Because it operates continuously and provides immediate output, the monitor will enable manufacturers to identify problems before such damage can occur. In sum, the enhanced control of gas purity provided by the device promises significant improvements in both wafer yields and equipment uptime. References 1. McAndrew JJ, and Boucheron D, "Moisture Analysis in Process Gas Streams," Solid State Technology, 35:55—60, 1992. 2. Bhadha PM, "Control of Moisture and Contaminants in Shielding Gases," Welding Journal, May, pp 57—63, 1994. 3. Carroll DI, Dzidic I, Horning EC, et al., "Atmospheric Pressure Ionization Mass Spectroscopy," Applied Spectroscopy Reviews, 17(3):337—406, 1981. 4. Stallard BR, Espinoza LH, and Niemczyk TM, "Trace Water Determination in Gases by Infrared Spectroscopy," in Proceedings of the Institute of Environmental Sciences' 41st Annual Technical Meeting, Mount Prospect, IL, Institute of Environmental Sciences, pp 1—8, 1995. Jian Wei, PhD, is a consulting scientist at Millipore in Bedford, MA, where she is involved in establishing trace analytical technologies and capabilities for the development of point-of-use gas purification/filtration materials and devices and ultra-high-purity (UHP) gas monitoring systems. She assists Millipore's OEM and semiconductor customers in setting up gas purity requirements programs, and develops programs with OEM suppliers to improve the purity of components for UHP gas distribution systems. She holds a PhD in chemistry from Tufts University, is a member of the American Chemical Society, and is involved in the SEMI technical committee developing standards for UHP gas analysis and point-of-use gas purification. (Wei can be reached at 617/275-9200.) John E. Pillion, PhD, is a senior research scientist at Millipore, where he is responsible for sensor development in the company's microelectronics division. A member of the American Chemical Society and the Institute of Environmental Sciences, he is a graduate of Princeton University, where he obtained a PhD in inorganic chemistry. Steven M. King, in his position as a technical program manager, is responsible for moisture monitor technology development in Millipore's microelectronics division. He has a BS in mechanical engineering from Worcester Polytechnical Institute (MA). Matt Verlinden is a senior program manager for Millipore and is responsible for managing the product development and marketing efforts of the company's in-line monitor team. He received BS and MS degrees in chemical engineering from Iowa State University and has an MBA from the Sloan School of Management at MIT.
MicroHome |
Search | Current Issue | MicroArchives Questions/comments about MICRO Magazine? E-mail us at cheynman@gmail.com. © 2007 Tom Cheyney |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Figure 1: Moisture monitor installed in-line on a gas distribution system. Photo by Ed Shvartzman
Figure 2: Schematic diagram of the in-line monitor, with a detail pullout of its quartz crystal microbalance.
Figure 3: Windows-based display of a real-time H2O measurement in nitrogen taken at a semiconductor fab. Photo by Paul Prescott