Better process protection with automated gas monitoring/purification
09/01/2002
The presence of moisture, oxygen, and other impurities in a process gas stream is highly undesirable. These impurities decrease yields because they form haze and oxides in films, cause dislocation defects by reacting to form particles [1], and increase the corrosiveness of some gases [2, 3]. Impurities in process tools can be a result of poor gas quality, gas distribution leaks, virtual leaks (a trapped contaminant that outgases over time), inadequate purge procedures, and inadvertent system breaches. Process gas purification - whether at the source, in the gas manifold, or at the point of use - is a critical part of preventing impurities from reaching process tools [4].
A smart purifier with the capability to close gas line valves in the event of an impurity intrusion that would exceed a process specification or the purifier's recommended range of operation has been developed by Matheson Tri-Gas. Inputs for the control logic include changes in the purifier bed temperature as well as external signals from on-line analyzers that monitor the concentration of key impurities in the inlet gas stream.
 Figure 1. The NanoShield APPS process gas purification system. |
Two beta versions of the system - the NanoShield Automated Process Protection System (APPS)-were installed for evaluation on house nitrogen and hydrogen lines at Analog Devices Inc. (ADI). These lines supplied an advanced low-temperature epitaxial process using an Applied Materials Centura epi reactor.
System configuration
Briefly described, the APPS system watches for incoming contamination by monitoring purifier-bed temperature (i.e., reactions) with an array of thermistors that are compared to an adjustable setpoint. In characterizing the system, we determined that the thermistors responded within a few seconds to temperature changes in the purifier bed.
 Figure 2. A typical variation of oxygen concentration in the house nitrogen at the point of use. |
Automated valves provide the ability to isolate or bypass the purifier if an abnormal burst of contamination occurs. This system also has the ability to interface with a variety of on-line gas-monitoring devices, including an oxygen sensor, a moisture analyzer, and a pressure switch.
The system's PLC-based controller can be customized to use only internal temperature or rely on external inputs, such as gas line monitors, to control valves on process gas lines, thus protecting the purifier and downstream lines. Additional controller outputs provide the option to interrupt other process tool gas streams for safe shutdown in cases of unacceptable contaminant intrusion in the monitored gas supply system. The controller also has the option to send information to a central system monitor that may be programmed to generate e-mails to notify plant personnel when an "event" has occurred.
Beta-site testing
Often, process requirements are customer and application specific. Process development engineers must determine if gas purification is critical to the quality of a given process being developed. Most processes at ADI require higher gas purity levels than typically available from cylinder or house gas sources. Gas purifier systems are installed to meet high-purity demands.
The two beta versions of APPS were used at ADI to evaluate whether reliable and safe response of gas purifiers can be guaranteed, even in the case of accidental intrusions of incompatible gases.
The beta-site installation for the ADI nitrogen supply line included a Matheson Tri-Gas P-4000 Nanochem OMX purifier equipped with a four-junction thermistor probe, a Delta F FAH0100 high-resolution oxygen monitor, and a pressure switch that provided inputs for the PLC-based valve control box (Fig. 1). The oxygen monitor was mounted ∼50 ft upstream of the purifier to improve controller response in the event of an oxygen intrusion. A second sample line from the purifier outlet to the oxygen monitor was installed for additional monitoring of the process gas purity prior to entering the process tool.
null
Alarm setpoints were adjusted to 30°C for the purifier temperature sensor, 20ppm for the oxygen monitor, and a pressure trip point of 40psi for the pressure switch. Although the user can switch the purifier to a bypass mode in case of an alarm, both beta installations at ADI were configured for complete shutdown, including isolation of the purifier, switching off lamps, and isolation of all reactor chambers. Since the epi reactor is a single-wafer system, loss of one wafer is considered more preferable than contamination of the tool by introducing high levels of oxygen and other impurities.
We used a Yokogawa data logger linked to the fab network as a stand-alone central system monitor to collect output from the oxygen monitor and from thermistor voltages that were converted to temperature by the purifier control box.
The data logger was an early warning system, with control limits of 10ppm oxygen and 27°C purifier temperature, to inform process engineers in the event of a control limit violation. The data logger was programmed to issue a control limit violation e-mail alert to engineering when either or both of the control limits were exceeded and when the oxygen or temperature levels dropped back below control limits. In this way, engineering could respond if necessary and important data such as the duration of the out-of-control period and values of all data channels during the event were documented for any future diagnosis of problems.
Figure 2 shows typical oxygen concentrations in the house nitrogen supply measured at the point of use. Normal oxygen levels range from <5ppb to as high as 1000ppb depending on the overall fab consumption of nitrogen. The increase of oxygen levels to ∼300-500ppb between 4:00 and 7:00 pm (see Fig. 2) is an example of typical concentration levels observed when the nitrogen source is switched over to a liquid source backup tank.
Since oxygen contamination varies greatly, it is important to periodically verify oxygen levels in the purified gas downstream of the purifier. With an incoming oxygen concentration of <20ppm, the output gas quality of the purifier should remain below 1ppb. When the gas quality downstream of the purifier was monitored, none of the incoming variations were detected and the oxygen level was reduced below the 5ppb detection limit of the Delta F monitor.
Process results are proprietary. However, SIMS measurements used to determine background levels of carbon and oxygen within the epitaxial layer and at the c-Si/epitaxial interface indicated that the purification systems were successfully removing oxidizers and contaminants.
Outside normal gas quality fluctuations as discussed above, nontypical events due to accidental contamination can either be severe enough to trigger shutdown or can remain within process limits.
Figure 3 shows temperature and concentration data during an atypical event when oxygen in the nitrogen peaked to ∼3.4ppm and remained >1ppm for ∼40 min. Part way through the event, the house nitrogen system switched to the backup tank, and then back to the nitrogen plant. The oxygen levels measured were not sufficient to reach the 10ppm control or 20ppm alarm limits, and the purifier control maintained gas flow. The purifier thermistors registered ∼0.2°C temperature increases due to the increased level of oxygen introduced to the bed. While this event was not severe enough to result in a shutdown, it does quantify the sensitivity of the purifier temperature monitors. No harmful effects were noted in the process results or oxygen levels in the chambers. The additional information from the data logging of this subcritical event may provide the opportunity to minimize or prevent the impact of more potentially severe intrusions in the future.
During the beta-site tests at ADI, system data from a large spike in the oxygen concentration was also recorded (Fig. 4). Problems with the nitrogen plant caused a temporary switchover to the backup tank. During the switch back to the nitrogen plant, an equipment malfunction resulted in a short intrusion of compressed air into the nitrogen line. The oxygen level instantaneously jumped above the upper detection limit, and remained there for ∼15 sec. Unfortunately, the concentration exceeded the data collection range and the maximum O2 concentration of the spike is not known. The concentration quickly dropped back in range, and only remained >1ppm for ∼3 min.
The peak concentration greatly exceeded both the control and alarm limits for oxygen concentration, and caused a system shutdown as desired. The data logger detected the out-of-control condition and e-mailed engineering that an alarm event had occurred. The purifier controller detected the alarm condition and closed all four pneumatic valves to isolate the purifier and the process tool. Since the Delta F monitors upstream of the purifier valving, it continued flow and monitoring of incoming nitrogen quality, and the data logger continued to collect oxygen data as well as purifier temperatures. The data in Fig. 4 show that little of the air spike reached the purifier bed. Thus, the temperature increase measured by the thermistors was only slightly higher than the increase for the 3.4ppm spike shown in Fig. 3. Since the purifier was isolated from the gas flow, the heat dissipation from the air intrusion was slow, resulting in a 0.3°C increase that slowly spread along the purifier bed by thermal diffusion.
 Figure 4. Response of the oxygen analyzer and four-junction thermistor probe (Th1, 2, 3, 4) to a >20ppm oxygen spike in a nitrogen supply line. |
Without the alarm from the O2 analyzer and rapid system shutoff, compressed air would have been introduced into the OMX purifier canister, likely causing more significant temperature increases than previously observed. This event can be seen as a relatively small system upset, and likely would have been absorbed by the purifier without deleterious effects. A sustained air intrusion, however, could easily have resulted in a more significant temperature rise, depending on its total amount.
The value in monitoring
The NanoShield APPS controller without the additional oxygen analyzer input would have sensed a temperature rise in the purifier medium during the above intrusion and shut off the flow to the purifier to prevent contamination of the process lines downstream. However, the subsequent condition of the purifier, which is dependent on the temperature rise as a result of oxygen ingression, would have to be judged on the maximum temperature observed and replacement would be necessary if the maximum operating temperature specification of 70°C were to be exceeded.
The beta-site system described not only reliably protects the purifier and the process tool from a small-scale intrusion, but can also detect a serious system upset at an early stage, preventing large-scale contamination of the tool. After thorough purging of the lines and verification of the gas purity with the oxygen monitor, the epi system was back on-line within 24 hr with no detectable process impact. The results support using automated process protection through integrated gas purification and in situ monitoring.
Acknowledgments
Jon R. Welchhans at Matheson Tri-Gas Advanced Technology Center is an additional author of this article. We acknowledge the role Matheson Tri-Gas Site Services played in implementing the NanoShield APPS at Analog Devices. NANOCHEM is a registered trademark and NanoShield a trademark of Matheson Tri-Gas Inc.
References
- O. Kermarrec et al., Solid State Technology, 45(3), pp. 55-60, 2002.
- S. Banerjee, Microcontamination 1991 Conference Proceedings, Santa Clara, CA, Cannon Communications Inc., pp. 621-624, Oct. 1991.
- W. Kroll, Solid State Technology, 27, pp. 220-227, 1984.
- J. Newey, Compound Semiconductor, 7 (11), pp. 51-54, 2001.
Lee E. Vestman, Hans H. Funke, Mark W. Raynor, Matheson Tri-Gas Inc., Longmont, Colorado* Larry J. Lowell, David P. Baril, Analog Devices Inc., Wilmington, Massachusetts.
*An additional author is listed in the Acknowledgments.
For more information, contact Lee Vestman, Matheson Tri-Gas Advanced Technology Center, 1861 Lefthand Cr., Longmont, CO 80501; ph 303/678-0700, fax 303/442-0711, e-mail [email protected].