Molecular Filters Advance Gas Purity Control
As contamination control strategies have evolved into integrated systems, the real breakthroughs have come in the area of gas purification.
By Walter Plante and Ramesh Hegde
In-line particle filters have been part of the cleanroom infrastructure in semiconductor fabrication processes for decades. Their use is a necessity for delivering particle-free gases to the point of use. They provide an extremely high level of efficiency–typically greater than 99.9999999 percent for removal of 0.05-µm particles. This exceeds the actual process need. No change in filter membrane technology has ever been necessary to address particle removal effectiveness; instead, technical improvements have focused solely on cleanliness. While there have been some important innovations in particle filter technology with the appearance of all-metal filters around 1987, the real breakthroughs have come in the area of gas purification.
Purifiers act as “molecular filters” by removing trace volatile impurities such as water and oxygen from high-purity gases, while allowing the base gas to pass through unchanged. While particle filters remove hard particles by various fluid-mechanical mechanisms, molecular filters remove molecular impurities by physical or chemical means. This article will describe the current state of gas purification technology, outline the reasons for the evolution of filters to molecular filters and discuss the direction of future trends.
Existing Gas Purification Technologies
Gas purifiers have been used as part of high-purity gas delivery systems since the late 1980s. During this time, the base purification technologies have remained more or less the same, while new developments have occurred in the integration of purifiers and filters and the advancement of purification media chemistry and cleanliness. Point-of-use purifiers have evolved into compact devices where the filter and purifier functions are integrated into one component. This trend towards integration reduces the number of connections in the gas system and reduces the overall footprint of equipment occupying expensive cleanroom space.
Two major classes of ultra-high purity gas purifiers are in common use today: metal-alloy getters and chemisorption purifiers. There are two types of chemisorption purifiers: Reactive Micro Matrix (RMM) and resin-type. These technologies have been described extensively elsewhere.1, 2 The differences lie mainly in the impurities removed, the operating temperature, and compatibility with specialty gases. Getters are typically limited to use in bulk gases and operate at 350-400°C. Getters can also be operated at room temperature with significantly reduced lifetime and efficiency.2 This makes them best suited for high-flow bulk gas installations using special temperature and engineering controls. RMM purifiers are compatible with most semiconductor gases–including reactives and corrosives–operate at room temperature, and are used at all points in the gas system. All purifier types remove that most damaging and insidious of impurities–moisture. Using atmospheric pressure ionization mass spectrometry (APIMS), it has been demonstrated that they can remove water down to the ppt range.3
At one time, all chemisorption-based purifiers were thought to outgas trace hydrocarbons based upon APIMS testing of one specific type of purification media.4 However, recent advancements in chemistry and cleanliness have addressed these issues and have eliminated this as a point of technical differentiation. This has been supported by new studies where hydrocarbon addition was below APIMS detectable levels.3,5 Now the choice between chemisorption purifiers and metal-alloy purifiers depends upon the cleanliness of the entire purifier/filter package, usable lifetime, and the range of gas compatibility.
Is There a Need for Purification?
With the ability to reduce moisture and other impurities to sub-ppb levels at almost any point in the gas system, the value of purification would seem obvious. As with any relatively new technology, however, there are some barriers to acceptance. A common sentiment is “my gases are already clean,” or “we know how to manage our gas systems without making mistakes and causing impurity upsets.” The fact is, in the absence of real-time continuous impurity monitoring of the gas system, no one really knows how effective their gas management procedures are or how often mistakes happen. As long as people are involved in making gas management decisions, there will always be opportunities for impurities to enter and remain in the gas system.
In 1991, we conducted a study of over 40 used filters taken from state-of-the art semiconductor production facilities.6 These filters were studied by SEM/EDS for clues to the type of contamination present in these gas systems. The large majority of the filters were contaminated with stainless steel corrosion products or products of reactions with particle- forming gases such as dichlorosilane. Most of these particles could not have formed without the presence of molecular impurities such as moisture and oxygen. We concluded that particle filters cannot be fully effective at stopping such contamination, because molecular impurities can pass through a particle filter. This was evidenced by a significant number of filters with contamination on the downstream membrane surfaces. Recent data shows this problem to be persistent today.
The electron micrograph (far left) shows an SEM micrograph of a high-efficiency nickel filter removed after less than one year of service in dichlorosilane. The upstream filter`s surface is contaminated with particles which were likely produced in the reaction in dichlorosilane with molecular impurities in the gas line. This is consistent with the fact that EDS analysis of the particles indicated the presence of silicon. In short, there is no evidence today that perfect gas system management is commonplace in full-volume production processes. Process protection will be needed for the foreseeable future. The failure to recognize gas delivery systems as active, dynamic systems rather than static systems operating at a perfect steady state, hour after hour, is dangerous.
The benefits and need for gas purification have been well documented and fall into two major categories:
1) Direct improvement in film and device quality through the reduction of gas impurities reaching the wafer.
2) Impurity removal in the gas distribution system resulting in:
A reduction of cost of ownership through improvements in reliability and uptime.
A reduction in contaminants reaching the wafer as a result of gas-phase reaction or corrosion of the delivery system.
Direct Process Improvements
A number of studies have documented the benefits of point-of-use (POU) purification in the process equipment gas panel and the resulting improvement in film qualities. For some examples of key processes, see “The effects of gas purification on process results,” p.27.
In one study of LPCVD polysilicon, the process was run both with and without POU purifiers over a period of six months. Polysilicon grain structure was compared for each case with a significant increase in grain size and uniformity observed when purifiers were used. A morphology with larger grains can lead to decreased sheet resistance. These results are consistent with the observations of Kamins et al11 who suggest that oxygen impurities can affect crystallization by impeding the surface migration of silicon atoms and can hinder grain grown during recrystalization.
Reducing the Total Cost of Ownership
Some have questioned the value of gas purification by assuming purifiers represent an increased cost of ownership for the gas system.12 While a purifier is clearly more complex, and hence requires more understanding than a particle filter, the cost of ownership taken on a true systems-level will decrease when purifiers are designed into a system. This can occur through defect reduction, through improvements in gas system reliability resulting from improvements in gas purity, or through faster start-up times for new system installations. Also, maintenance operations involving purifiers come at known intervals and can be worked into equipment or system maintenance schedules with little or no additional downtime.
Using a cost-of-ownership model, Kellam et al13 predicted that improvements in silane gas quality would result in significant cost savings for an LPCVD polysilicon process on the order of $300,000 per year. What`s important to note is that their analysis only included particle-induced defects and did not consider the additional savings that could be realized from increased process uptime or improvement in other film properties.
Because capital cost of equipment is typically one of the most significant factors in any cost-of-ownership model, system reliability and uptime are critical. Failures in the gas system due to component failures such as mass flow controllers (MFCs) and diaphragm valves can bring the process equipment to a halt and send maintenance engineers on a costly troubleshooting mission. These failures can be byproducts of corrosion, corrosion-products (caused by the presence of trace water), or reactions with particle-forming gases such as silane (caused by water or oxygen impurities). The most notorious of these failures occurs with MFCs. Wei et al studied MFC reliability in both purified and unpurified HCl lines.14 They found that an MFC protected by a POU purifier maintained its setpoint accuracy, while an unprotected MFC failed to accurately control gas flow after three months of operation (See “MFC reliability”). Such improvements reduce maintenance downtime and improve process control through improved flow control stability. Both factors directly reduce the total system cost of ownership.
When new specialty gas systems are installed, tedious and time-consuming purging procedures must be followed. This can delay the start of the process tool qualification or shut the tool down when the gas system needs to be replaced. Duguid et al15 demonstrated that the purge procedures for a new HCl gas system could be shortened by placing purifiers at appropriate points in the system. With process purifiers in place during system startup in nitrogen, moisture levels were below detection at the POU (See “Startup of new gas system”). When the purifier was switched off-line, moisture levels immediately rose to ppm levels. Although the system was purged with inert gas prior to startup, some additional moisture may have entered during assembly and welding operations. Using a POU purifier during startup, prevents such upsets from reaching the process equipment. This lets purging of the process equipment proceed immediately upon establishing gas flow, even though portions of the delivery system may still be outgassing moisture. This, in turn, improves process equipment utilization and decreases the cost of ownership of the gas system and the tool on a cost-per-wafer basis.
Impurities Generated in the Gas System
While high-quality gas may be placed in the gas cabinet, there is consensus that a pervasive need exists to maintain gas purity all the way to pou. Gas purification works to flatten out upsets that can occur in the system. This run-to-run consistency may be more important than the absolute purity level. Aside from direct contamination of the wafer environment, molecular impurities in the gas system can form secondary contaminants, such as corrosion products or products of reactions with particle-forming gases such as silane and dichlorosilane.
Corrosion in the gas system not only affects component reliability, but also adds contamination to the process in the form of particles or, much worse, metal particles. DePinto16 found that when HCl was used in polysilicon LPCVD for predeposition cleaning, sodium levels were drastically reduced on the wafer, thereby reducing any flat-band voltage shift. On the other hand, iron contamination levels increased when HCl was used, and this was attributed to particles from gas system corrosion. Gas purification is a means to reduce this corrosion by removing moisture from the system at strategic locations.
One approach to corrosion prevention is to use expensive “super alloys” such as Hastelloy C-22 when designing a new system. The corrosion resistance of this alloy is known to be measurably better than that of electropolished stainless steel in aggressive environments such as wet HCl or HBr. It is also well known that moisture is required for corrosion to occur in any of these materials, including stainless steel. In addition, it is often overlooked that the surface quality and availability of Hastelloy components is far below that of stainless steel. A more sensible approach would be to employ purification to remove moisture, while using good quality electropolished stainless steel tubing and components. Designing a system in all Hastelloy recognizes that there may be moisture in the system at some time. Simply removing the moisture alleviates the need to over-design.
In a study of impurity reactions in silane, we found evidence that water reacts slowly with silane to form solid particles.17 This has serious implications for gas system design because unreacted water can pass through a particle filter only to form more particles downstream or in the process chamber. The only solution to the problem is to stop the propagation of water and oxygen by using a purifier compatible with silane gas. It is expected that similar problems exist with other particle-forming gases such as dichlorosilane or boron trichloride.
Future Trends
We expect that the future will lead to the development of purifiers for removal of new impurities and will bring unprecedented interaction with “smart” gas systems as purity monitoring and automation become commonplace.
Recently some questions have been raised as to whether gas impurity levels should continue to be driven lower and lower with each new device generation. Certainly, an emphasis on specialty gas purity over bulk gas purity would provide the most benefit for the industry. For some clues as to where purification technology fits into this purity evolution, one can look to the Semiconductor International Association`s National Technology Roadmap for semiconductors. The Roadmap recognizes this need for focus on specialty gas purity and the use of POU purification to attain these purity levels.18 The Roadmap also calls out high-performance moisture/metals removal in specialty gases as an unmet high priority need for 1995-1996. High performance, long-life purifiers and purifiers for metals removal are both identified as potential solutions. Clearly, there is a large void in volatile metals` removal from specialty gases although metals, as particulates, have been removed by filters for years. Metals` removal should become a development priority for purifier suppliers.
Real-time contamination sensors in fluid lines have also been identified as an important technology need in the SIA Roadmap. This is consistent with the concept of active gas systems having a continuous dynamic which requires monitoring. We believe the in-line monitoring of impurities in inert and specialty gases will become increasingly common in the coming years.19 Such monitors will allow corrective action to be taken when impurity upsets occur and allow for better correlation of impurity levels to process defects. Ultimately, sensor and system integration will lead to automated systems which remove the human factor by taking corrective action when impurity upsets occur.
Will this eliminate the need for purifiers, since the “perfectly controlled” system will become reality? We believe that purifiers will continue to play a complementary and critical role in these automated systems. For instance, when purity upsets are sensed, gas flow can be diverted to strategically placed purifiers to prevent any contamination from reaching the wafer. Also, purifiers will always have a use in speeding purging and flushing during maintenance operations, because the purity monitor can only determine when a low impurity level is reached, it cannot force the system to get there faster.
Numerous studies on high impurity levels in the wafer environment have led some to conclude that controlling gas purity at the inlet to the process chamber is of minimal importance. Undoubtedly, efforts need to continue to address process chamber integrity and process-induced contamination as the contamination control priority; however, the future will bring concern over POU gas purity back to the forefront. Many of the studies on in-situ purity levels have been carried out on large, high-surface area, batch furnaces which are exposed to cleanroom air during each load cycle. Already we have seen a shift from these batch furnaces to single-wafer tools with vacuum load locks and ultra-clean designs. This has lead to orders-of-magnitude improvements in in-situ purity.
As improvements are made in the wafer environment, impurity levels for incoming gases become more significant. We predict that this heightened awareness of in-situ contamination control will eventually shift the balance back to contamination control in the gas system. In reality, the efforts have never stopped in either area, and the process studies cited in “The effects of gas purification on process results” are strong evidence that gas purity control is needed today. Also, gas system reliability remains an issue independent of process chamber purity levels.
The evolution of contamination control strategies from devices which only remove particles to integrated systems which detect contamination and take action by removing both particle and molecular impurities is occurring quickly. By understanding how to best apply these tools, cost, reliability, and process quality can be improved.
The authors would like to thank the following for their input or data used in this paper: Mutsuhiro Amari, Rob Binder, Rebecca Duguid, Isamu Funahashi, and Jian Wei.
A complete list of references cited in this article is available from CleanRooms magazine, (603) 891-9230.n
Walter Plante is Technical Manager of Microelectronics Gas Applications Development for Millipore Corp. (Bedford, MA). He is responsible for overseeing the applications focused technical work in Millipore`s Specialty Gas Research Center, ppb/ppt trace gas analysis laboratories, and submicron particle lab. He directs the efforts of the applications group in solving the gas contamination control problems of the industry. Prior to joining Millipore, Plante was a development engineer with Linde Industrial Gases where he studied particle and homogeneous contamination in semiconductor process equipment using in-situ gas analysis techniques.
Ramesh Hegde is a Senior Technical Consultant for the Microelectronics Division at Millipore. He is also responsible for application development and technical support to Millipore`s Asian customers. During his 18 years with Millipore, Hegde has been extensively involved with many research and development activities and groups. He has a Ph.D. in analytical chemistry.
 |
Electron micrograph of particles captured on a high efficiency nickel filter recently used in dichlorosilane. These reaction products are evidence that contamination problems in gas systems are still common today.
 |
Cut-away view of an integrated purifier-filter. The main housing (shown empty) contains the active RMM purification media for removal of impurities. A high-efficiency metal filter is internally welded on the outlet.
 |
Increased polysilicon grain size and uniformity as a result of POU process gas purification. Far left: without purification average grain size is 285nm. Left: with purification average grain size is 380nm.
 |
 |
MFC reliability improved when POU purification was used in corrosive gases. (MFC #1) the unprotected MFC lost control in three months, while the MFC protected by the purifier (MFC #2) maintains control at 400 sccm.
 |
Moisture levels were below detection during startup of a new gas system when purifiers were used.