Issue



Bulk Silane-- a potential hazard or a potential hazard reducer


01/01/1998







Bulk silane - a potential hazard or a potential hazard reducer?

Lise Laurin, Clear Tech, Newton, New Hampshire

The semiconductor industry has long sought to reduce the hazards involved in silane handling by using very low concentrations in inert carriers or by eliminating silane from processes. A new approach, bulk silane, may reduce both hazards and costs for manufacturers. Spurred on by potential savings, gas suppliers, silane users, and Sematech are investigating how the use of bulk silane can be safely implemented.

Silane (SiH4) is used to produce very pure silicon and silicon-containing films and materials. Composed simply of a silicon atom surrounded by four hydrogen atoms, it readily reacts, typically with pure hydrogen as a byproduct. Its highly reactive nature, however, makes silane difficult to handle safely -which is why the Semiconductor Safety Association regularly sponsors Silane Conferences to improve silane handling.

The pyrophoric nature of silane was evident from its first applications. When a leak occurs, or if a quantity of gas is vented, it tends to ignite spontaneously in air. In the 1970s, a silane gas bottle cabinet exploded violently, prompting gas handling and safety personnel to take a closer look at the gas [1]. In this case, the silane did not ignite spontaneously. Instead, the leaking silane accumulated in the cabinet until something caused it to ignite, resulting in an explosion.

The most recent research, sponsored by Sematech and Factory Mutual [2], begins to explain silane explosions. Based on nearly 300 silane releases, this study shows that silane concentrations below 1.4% are nonflammable. (In specific cases, silane at this concentration will still react with air to form silicon dioxide.) At concentrations between 1.4% and about 4.1%, the silane will not inflame unless an ignition source is present. Although the study calls this concentration "stable," flow rate changes that occur when the silane source is shut off can act as ignition sources. Many safety systems include automatic shutoff as part of the interlock system. If a silane leak is detected, the interlock system shutting off the gas may cause ignition. This "stable" concentration (sometimes called a vapor cloud) should be avoided. At concentrations above 4.5%, the silane will autoignite with exposure to air after some time delay.

If the silane escapes at a high velocity, it may not ignite. Increases in release velocity can extinguish an existing flame. Several theories have been offered to explain this effect, including high flame shear causing quenching and turbulence-induced density differences in the cloud. Although the mechanisms behind this event are still in question, the possibility that high concentrations of silane might build up is clear and must be avoided.

Click here to enlarge image

Figure 1. This simplified schematic of a bulk silane supply shows the backup silane cylinders that add to its reliability.

To address known silane hazard conditions, Factory Mutual recommends a handling system design that will ensure the concentration of silane never exceeds the lower explosive limit (LEL) of 1.4% in the exhaust vent. Factory Mutual also suggests requirements to minimize the effect of pressure shocks. To meet these conditions, the gas-handling designer must look at the maximum silane flow and pressure that could occur in case of a leak; the flow pattern within the enclosure (if any); and the ventilation required to keep the silane below the LEL. For example, if a full bottle of silane leaks in a cabinet, the ventila-tion flow must be sufficient to keep the silane concentration below 1.4% in the exhaust. If a restricting orifice allows a maximumflow rate of 200 liters/min, for example, the minimum ventilation flow rate would be 14,300 liters/min, or 500 ft3/min (cfm). Assuming a cross-sectional cabinet area of 4.5 ft2, this ventilation would require a linear flow rate of about 110 linear ft/min.

The OSHA PEL-TWA exposure limit for silane is set at 5 ppm [3]. However, the limited animal testing done to date indicates silane is slightly toxic, with a four-hour LC50 (lethal concentration killing 50% of test organisms within observation period) in the range of 4000-9600 ppm (as compared with hydrogen cyanide, which is lethal after a 10-min exposure at 191 ppm) [4]. Given the greater hazards from fire and explosion, silane toxicity is not a principal concern.

Bulk distribution configuration

Gas suppliers have carefully deliberated the design of bulk supply systems. In the Praxair [5] setup, for example, each bulk module consists of eight individual supply tubes, with a capacity of 375 kg each. Figure 1 shows how the supply tubes tie into the bulk supply lines. Pressure transducers and flow meters monitor for leaks. Control valves provide for routine control of gas flow, automatic cycle purging, and automatic emergency shutoff. A backup silane supply, made up of traditional silane cylinders, provides redundancy.

Suppliers test tubes on a routine basis, using acoustic emission testing, a helium leak check (accurate to 1 ? 10-7 cc/sec leak rate), and overpressurization to 4000 psi or 2.5? the fill pressure.

A frangible disk, laser-welded to each tube, is designed to rupture in case of a catastrophe. This 1-in. dia. "rupture disk" requires both overtemperature and overpressure to release; it releases when the internal temperature reaches 165?F and the pressure reaches between 3700 and 4000 psi, thus preventing rupture of the vessel itself.

Bulk silane: Lower cost and reduced possibility of error

Changing gas cylinders is one of the riskiest activities in semiconductor manufacturing. The first risk is to the product. A bad bottle or a poor installation can cause contamination-related defects, which, at best, reduce throughput while another bottle is installed, and, at worst, degrade device yield. More importantly, accidents are most likely to happen and errors are most likely to be made during cylinder changes, when the bottle is disconnected from the supply line. If the lines are not purged properly, residual silane can escape, causing a flame or vapor cloud. If the cylinder valves are not closed properly, the contents of the bottle can be released into the cabinet. If an improperly sup-ported cylinder falls, it may break lines or valves, creating a small or large leak of silane. These types of errors lead to explosions.

The installation of silane bottles is costly as well. Each time a bottle is installed, the user "qualifies" it by running a test process to ensure that the product manufactured meets the desired quality and purity. This gas is wasted, as are the test wafers and machine time used to run the qualification.

Click here to enlarge image

Bulk silane appears to reduce both the risk and the costs involved in silane handling. Based on experience going back to the early 1980s, Praxair estimates savings from the use of bulk supply to be $810,000 for the first year of installation, and $240,000/year thereafter. Bulk supply reduces gas container changes from 200/year, or nearly four times/week, down to once/year (see table). This means two orders of magnitude reduction in the probability of an accident. Additionally, experts from the gas supplier will change the bulk supply. Infrequent change-out offers the chance to take additional precautions, including cordoning off the area and calling in emergency personnel. The higher visibility and lower change frequency should combine to reduce the risk of failure. The table, however, fails to show the results of a failure, catastrophic or otherwise.

Bulk installation based on single-cylinder storage hazards

Through the years, good practices have evolved for handling silane coming from single cylinders. Without good data on the results of a release from bulk storage, these practices have been transferred, sometimes imperfectly, from single bottle to bulk supply. For example, if the supply line from the bulkhead, where it enters the building, to the tool area is 1/4 in. in dia. and 20 ft long, at a pressure of 40 psi, this line contains 0.5 standard liters of gas. In the case of a bulk system, the supply line is typically 3/8 in. in dia., and could easily be 50 ft long. At a line pressure of 100 psi, this line contains 30 liters of silane. Leaks from these two lines require different ventilation rates to ensure that the silane concentration never reaches the LEL. "Rules of thumb" (e.g., 200 linear ft/min flow rates are sufficient) that worked with the old installations may no longer apply.

In another example, the Uniform Fire Code allows a maximum of 380 kg of silane/storage area, and thus does not allow for bulk silane storage, which can contain 1000 kg/storage area. Setback distances from structures such as schools, offices, and block walls in the code are based on the lower storage quantity. Actual testing (discussed below) shows that these distances are not adequate for bulk supply. Gas suppliers are experimenting to determine the new setbacks.

CGA testing - setup and preliminary results

In an attempt to quantify the risk of bulk silane supply, as well as determine the necessary safety precautions, the Compressed Gas Association (CGA) has been studying both the theoretical and the simulated results of an accident [6]. Fault tree analysis led the researchers to concentrate on rupture disk failure. To simulate ruptures of this type, they released entire bulk tubes, containing 200 kg of silane, through 1/2-in. and 1-in. orifices, at different ambient temperatures, with and without obstacles in the gas path.

The releases were conducted at the New Mexico Institute of Mining and Technology, Energetic Materials Research and Testing Center. This facility is designed for commercial and military explosives testing and offers a highly controlled environment. The test setup allowed the CGA to perform remote activation of the release and test for overpressure at different distances from the release. Video footage offered the researchers, who were isolated in a bunker, a visual image of the release.

Click here to enlarge image

Figure 2. At 36 ft away, a horizontal release from a typical rupture disk will knock a person over.

Figure 2 shows the overpressure measured at several distances from a horizontal release. The overpressure of 10 psi at the storage site would probably destroy any building there. While this overpressure is below the threshold for fatalities (35 psi), it will rupture eardrums and cause other physiological problems [4]. Between 18 and 32 ft from the release, the overpressure drops to between 1.5 and 3 psi, powerful enough to knock personnel down and partially demolish wood-frame structures. At 72 ft from the release, the overpressure drops below 1 psi, which has enough energy to shatter windows, but perhaps not enough to knock a person down. Video tapes of these releases show flame plumes up to 70 ft long.

Conclusion

Bulk supply of silane can potentially save device manufacturers thousands of dollars/year, and reduces the chance of an accident. At the same time, a catastrophic failure, although much less likely than a failure using single cylinders, can cause tremendous damage. Additional experimentation will determine the new safety mechanisms and practices required to limit the damage to an acceptable level.n

References

1. C. Doolittle, D. Quadrini, Sr., "Historical Perspective -A Comparison of Tests/Models of Silane," SSA Journal, to be published, Winter 1997.

2. F. Tamanini, "Factory Mutual / Sematech Research on Ignition and Reactivity Characteristics of Releases of 100% Silane," Silane Conference Proceedings, Sept. 9, 1997.

3. Silane Gases and Equipment Material Safety Data Sheet, Air Products, MSDS #1052, Sept. 1995.

4. S. Shiban, "Silane: Initial Studies and Lessons Learned," Silane Conference Proceedings, Sept. 9, 1997.

5. J. Kubus, "Design for a Safe and Reliable Silane Distribution System," Silane Conference Proceedings, Sept. 9, 1997.

6. T. Thompson, "Bulk Silane Testing by the Compressed Gas Association," Silane Conference Proceedings, Sept. 9, 1997.

Lise Laurin received her BS degree in physics from Yale University. After 15 years of technical and marketing experience, she founded Clear Tech in 1996, a company that provides market research, strategies, and technical communications support to the semiconductor and other technology industries. Clear Tech, 14 South Main St., Newton, NH 03858, ph/fax 603/382-7682, e-mail [email protected].