Evolving gas flow, measurement, and control technologies
10/01/1999
While a focus on overall equipment effectiveness has been in place for some time, the new drive will be toward achieving it through system simplification, new throughput enhancements, and increased reliability. System simplification will be applied directly to fluid delivery systems and those logistical systems that support them. This trend is further supported by the move to 300mm wafers, which will cause a focus on simpler delivery systems that are cheaper and carry lower cost of ownership (COO). For this reason, these changes will only occur as fast or slow as the transition to 300mm.
Consider surface-mount components [1]: While the initial cost of these gas-handling components is greater than that of traditional linear components, the industry expects the overall cost of stocking, assembly, and maintenance to be lower. Surface mount allows equipment makers to stock standard components that can be easily and quickly configured onto a base-block without welding. This configurational ease and decrease in component types that need to be stocked result in cost reduction. The end-user also benefits, since equipment can be assembled and shipped with shorter cycle times and will have a smaller footprint compared to linear gas panels.
 |
A potential problem is the proliferation of surface-mount standards that, if not kept in check, will increase costs and defeat the original intent of this new technology. There are currently at least seven surface-mount standards available or under consideration (Table 1). It is expected that a shakedown of standards will occur and be driven by the major original equipment manufacturers (OEMs) so that only one or two remain in five years. This reduction in standards should be encouraged and embraced by the industry if cost benefits are to be realized. Without overall cost improvements, surface-mount technology will vanish.
The drive toward system simplification will bring about functional integration of gas delivery components. This is more involved than the straightforward mechanical combination of components (e.g., a filter inside a regulator). First, it will require a thoughtful combination of components with concern for serviceability and overall system design. At the same time, the true objective is to combine functions. So, the ideal will be to develop an entirely new device that combines the functions of two or more other components without simply being a mere mechanical combination.
MFC improvements
System simplification will also be achieved through improvement in the range and controllability of flow controllers. While most processes typically operate over a fairly small range of mass flow rates (e.g., 2:1), a particular tool may require flows of many gases that vary by two to three decades and are typically chemically different. Traditional analog mass flow controllers (MFCs) have an operating range of 10:1 and are tuned to control gases properly in a particular density range. Tool manufacturers and users are required to maintain spare MFCs for a variety of ranges and gases, which adds considerably to COO.
 |
Digital MFCs can overcome the limitations of range and controllability typical of analog MFCs. In particular, digital devices can be tuned and calibrated for a number of gases, which adds flexibility. In addition, the range of digital MFCs can often be extended to 100:1 with some degradation in performance. Both of these attributes will gain more popularity as the requirement to reduce COO and maintain tool uptime is stressed.
Efforts to minimize COO will put further emphasis on component self-diagnostic techniques to reduce tool downtime. Diagnostics for most current MFCs are crude, and self-diagnostics nonexistent. MFC diagnostics must currently be performed manually, resulting in higher cost.
Many digital MFCs are capable of enhanced diagnostics. The level is device-dependent, but may include raw sensor data, control valve current or voltage, flags for excessive flow oscillation, or built-in process measurements (e.g., gas line pressure and temperature). Self-diagnostics can alert operators to abnormal operating conditions for proactive checking and remedy. Improved diagnostics, coupled with improvements in MFC sensor technology, will allow for extension of existing calibration and maintenance intervals beyond the current nominal one-year period.
Specialty gas changes
One of the formidable tasks in the development and improvement of semiconductor processes will be the judicious selection of gases and chemicals and associated delivery methods that minimize risks to workers and the environment. For example:
- Minimizing cylinder changeouts by switching to bulk delivery decreases worker exposures to possible gas leaks.
- Toxic specialty gases in subatmospheric form are another advance; their use will increase.
In addition, the concern for environmental pollution caused by the release of global-warming gases is increasing, and more environmentally friendly gases will be investigated and transitioned into manufacturing.
Many on the World Semiconductor Council have agreed to reduce their perfluorocarbon (PFC) emissions by 2010, to 10% of their 1995 levels. Most of the gases with relatively high global-warming potential (GWP) are used in the plasma etch processes for chamber cleaning and dielectric etch processing (Table 2); for these, the GWPs range from 6500-12,000. (GWP100 values reflect the quantity of global warming caused by a chemical over a 100-year time relative to CO2 with a GWP of 1.0.)
The industry is exploring several methods to reduce PFC along with HFC emissions, including reducing gas use through process optimization, using more effective abatement procedures, recovering and recycling etchants, and implementing new etch chemistries.
Process optimization
Typically, processes using these gases have pre-determined timed steps for many critical stages; this type of processing is nonadaptive and therefore not optimized for a particular system condition. Alternatively, several measurement techniques currently exist for adaptive control, including residual gas analysis for end-point detection in etch processes, ellipsometry in deposition processes, and various implementations of gas detection for system purging or pump down. Techniques for measurement of gas purity before the chamber and in situ include Fourier transform infrared spectrometry [2], atmospheric pressure ionization mass spectrometry [3], and residual gas analysis [4]. While present use of real-time process control has been minimal due to limitations in instrumentation stability and reliability, broader use will occur over the next several years for selective processes where process optimization can be significantly increased by its implementation.
New etch gases
Chamber-cleaning etch gases will increasingly be replaced with new chemistries as companies curtail their releases of GWP gas emissions. For example, more NF3 is substituted in this application. While this gas still has a relatively large GWP value, its atmospheric lifetime is significantly reduced from PFCs. In addition, new methods to improve the conversion efficiency of NF3 mean less gas is required for a clean.
Among the considerations for introducing new process etch chemistries to replace PFCs, a suitable replacement gas should not only be friendlier to workers and the environment, but its conversion efficiency and decomposition byproducts must also be considered. Naturally from a process standpoint, the gas must also provide the requisite etch rate and selectivity.
Various research facilities are currently evaluating new HFCs and iodofluorocarbons (IFCs), with substantially reduced GWP and atmospheric lifetimes [5]. Here, a two-fold decrease in average bond energy from a C-I to C-F bond (i.e., 213 vs. 485kJ/mol) is thought to be responsible for the shorter atmospheric lifetimes of IFCs. The percent reduction in kilogram of carbon equivalent (KgCE, a measure of PFC emission) for 1-iodoheptafluoropropane (CF2IC2F5) is roughly 95% of the value for C3F8. This gas is capable of favorable etch-process results.
Bulk gas delivery, purification
Bulk delivery of process gases will continue to increase as the industry transitions to new processes and 300mm-wafer production. The elimination of frequent cylinder changes improves gas quality by minimizing accidental introduction of atmospheric impurities into the gas manifold and also minimizes worker exposure to unintentional gas release. Bulk delivery of hydrogen chloride will continue to increase and be joined by other specialty gases, particularly ammonia.
Purification of high-pressure, high-flow bulk-delivered gas presents new contamination control challenges within the industry. While purifiers are now readily available for most bulk and process gases used in semiconductor manufacturing, the chemistries used in these purifiers may not be suitable for exposure to gases at cylinder pressure. Liquefaction of the gas within the purifier could potentially cause it to react with or dissolve the reactive chemistries or support materials used in the gas purifiers. Thus, new chemistries that are tolerant to exposure to condensed gases will likely appear within the next five years.
Subatmospheric delivery
Employee safety is also prompting the development of new gas-packaging and delivery methods. For example, Safe Delivery Source (SDS), which was introduced in 1994 for ion implant processing, reduces the hazard of specialty gas handling by reducing source gas vapor pressure. This technology decreases cylinder pressures to <1atm, thereby diminishing the potential for accidental gas release and resulting worker exposure. The lowered vapor pressure is a result of physisorption of the gas onto a high-surface-area support medium. The gas is subsequently removed from the cylinder using a pressure differential created by a downstream vacuum source. Current SDS ion implant gases include AsH3, PH3, 11BF3, SiF4 and GeF4, with AsF5 and PF5 under development.
The availability of subatmospheric gas sources is also emerging for chemical vapor deposition (CVD) processing, with AsH3, PH3, and GeF4 now available in reduced pressure forms. The larger cylinders for CVD applications permit greater volumes for the higher flows used in CVD relative to ion implant. Additional new CVD gases available for subatmospheric delivery will appear in the marketplace within the next few years.
Subatmospheric delivery of toxic and hazardous specialty gases has improved environmental, safety, and health aspects, however, flow, contamination control, and delivery using this technique have become more complicated. Pressures within subatmospheric cylinders decrease in use to <10torr. Conventional mass flow controllers, which operate at pressures down to 100torr, will only deliver 50-60% of cylinder content. Thus, new technologies are needed for gas flow control to pressures <10torr for maximum content usage. An approach to extracting more gas from SDS cylinders was recently reported [6].
Real-gas MFC calibration
The advent of digital MFCs has enabled wider control ranges and more flexibility in gas selection, but has not solved primary issues such as improved accuracy of the measured flow with real gases. A recent emphasis has been placed on real-gas calibration with hopes of improving the flow accuracy of MFCs.
Real-gas calibration usually means that at some point a MFC or a few samples of a MFC model are calibrated with a process gas and also with a surrogate (noncorrosive and nontoxic) gas to create a correlation between the process and surrogate gases. This correlation is then universally used on all subsequent MFCs, which often have different designs and operate over wide ranges of flow. Real-gas calibration accuracy depends on MFC design robustness, original correlation accuracy, and MFC manufacturing uniformity.
Real-gas calibration solves some calibration-related problems that could not be addressed through previous approaches based on predictive models because of deficiencies in the accuracy of gas properties provided by gas suppliers or available in the general scientific literature.
Flow accuracy issues are further complicated when using vapor controllers. These typically have wider operating temperature ranges and, since the gas correction factor is temperature dependent, the use of universal gas correlations or gas correction factors is not appropriate. Future MFCs will compensate for some of these obstacles by using real-time temperature measurement and compensation techniques. The effectiveness of these techniques will be largely determined by success in the marketplace, as most details of the new capability will be proprietary and not readily validated.
Contamination control requirements
As new gases enter the marketplace, purity is frequently not to the levels required for semiconductor manufacturing. Point-of-use purification of these gases may be essential. However, the industry is moving to delivery of gases of reduced pressures and introduction of new gases (e.g., new fluorinated etchants) with inherently low vapor pressures. Liquefaction of these gases within gas purifiers becomes quite probable. As with the case of bulk gas delivery mentioned above, however, validation of existing purifiers and development of new chemistries will be required. These chemistries should be able to tolerate exposures to condensed gases without subsequent degradation of the purification material and contamination to the gas stream. Filters and purifiers with lower pressure drops across the devices will be needed to help prevent condensation of the gases onto the membranes and support structures.
The 1998 update to the International Technology Roadmap for Semiconductors provides purity requirements for the next several technology nodes. By the 100nm technology node in 2005, the concentration of moisture and oxygen impurities in dielectric etch gases must be 100-500ppt and <500ppt, respectively. Levels for >50nm-sized particles at the point of use must be <2 particles/liter.
The trend to vapor-source materials and the need to control costs will mean that PTFE-membrane Teflon filters are here for the long term. Ten years ago, there appeared to be a trend to eliminate all polymeric media from the gas system on the basis that outgassing from these materials was unacceptable. It was believed that Teflon gas filters would disappear and be replaced by all-metal filters. This revolution was driven in large part by the work of T. Ohmi at Tohoku University.
If we retrospectively examine what Ohmi stated, however, the final conclusion might be different: ". . .contamination resulting from outgassing by some filters, in particular fluorohydrocarbon membrane filters used with cleaning solvents, remains a problem. These filters. . .cannot withstand the high baking temperatures (>200°C) required to remove adsorbed cleaning solvent and moisture molecules" [7]. The fact is that most, if not all, filter manufacturers removed Freon cleaning solvents from their processes years ago.
While metal filters can be baked to higher temperatures to reduce moisture, either by OEMs or users, many metal filters installed are not pre-conditioned and are never baked. So it appears that the outgassing benefits of all metal are not being fully exploited.
The rate of moisture outgassing from metal or ceramic surfaces is slower since they tend to bind moisture more strongly. Moisture is loosely physisorbed onto Teflon PTFE due to its low-energy, inactive surface. The end result is that PTFE filters, if pre-conditioned, can be very dry and tend to stay dry since they do not adsorb much additional moisture. Tests of metal, ceramic, and PTFE filters have shown (see figure) that [8]:
- all filters, from various manufacturers, have rated flows in the 15-30slm range;
- pre-conditioned Teflon filters clearly have the fastest out-of-package, dry-down rate regardless of manufacturer; and
- nonpolymeric filters release moisture more slowly, over longer periods of time, which may delay gas system purge-down in some applications. (Baking these filters would speed this process and result in a very dry system.)
Since PTFE-membrane filters can be made to have very low pressure drop at high particle removal efficiency, they will be needed for low-vapor-pressure gases where moderate temperatures (up to 120°C) are used.
Conclusion
Process gas chemistry and delivery are evolving as new processes come on-line and worker safety and the environment receive more attention. More stringent process requirements are dictating improved methods to deliver gases cleanly and accurately to the tool, while tightened operating budgets are forcing cost savings. The next five years will usher in an era focused on the smart delivery of process fluids.
Acknowledgment
Safe Delivery Source and SDS are registered trademarks of Advanced Technology Materials Inc. (ATMI), and are exclusively licensed to Matheson Electronic Products Group for sale to the ion implant sector of the semiconductor industry. Teflon and Freon are registered trademarks of DuPont Corp.
References
- B. Cullwell, "Modular Gas System Solutions," Workshop on Gas Distribution Systems, Semicon West, p. A-1, July 1997.
- A. Haider, J. McAndrew, R. Inman, Electrochem. Soc., Proc., Vol. 97-22, p. 484, 1992.
- Eiichi Kondoh et al., "Measurements of Trace Gaseous Ambient Impurities on an Atmospheric Pressure Rapid Thermal Processor," J. Vac. Sci. and Tech. A, Vol. A 17, pp. 650-656, 1999.
- I.L. Eisgruber et al., "Intelligent Process Control of Indium Tin Oxide Sputter Deposition Using Optical Emission Spectroscopy," J. Vac. Sci. and Tech. A, Vol. A 17, pp. 190-197, 1999.
- S. Karecki et al., "Use of Novel Hydrofluorocarbon and Iodofluorocarbon Chemistries for a High-Aspect-Ratio Via Etch in a High-Density Plasma Etch Tool," J. Electrochem. Soc., Vol. 145, pp. 4305-4312, 1998.
- N. Urdaneta, J. Krull, B. Brown, "Thermal-based Mass Flow Control for SDS Mass Delivery Systems," Solid State Technology, pp. 59-65, April 1998.
- Y. Kanno, T. Ohmi, "Components Key to Developing Contamination-free Gas Supply, Part III," Microcontamination,, pp. 23-30, Dec. 1988.
- Semaspec 90120397B, "Test Method for Determination of Moisture Contribution by Gas Distribution System Components," February 1993.
 |
Jim Snow received his BS in chemistry from Texas A&M University, and his PhD in organometallic chemistry from the Massachusetts Institute of Technology. Snow is manager of the Microelectronics Gas Research Center at Millipore Corp., 915 Enterprise Blvd., Allen, TX 75013; ph 972/359-4185, fax 972/359-4106, e-mail [email protected].
 |
Stuart Tison received his BS in mechanical engineering from Oklahoma University, and MS in mechanical engineering from Johns Hopkins University. He is chairman of the ASTM committee on vacuum technology. Tison is responsible for metrology and equipment development at Millipore in Allen, TX; ph 972/359-4248, fax 972/359-4111, e-mail [email protected].
 |
Walter Plante received his MS in chemical engineering from Worcester Polytechnic Institute in Massachusetts. Plante is R&D manager for gas contamination control products in the Microelectronics Division at Millipore Corp., 80 Ashby Rd., Bedford, MA 01730; ph 781/533-2220, fax 781/533-3195, e-mail [email protected].